Quantified Health

Aletta is making the automated blood draw a reality

2026-10-17T10:02:30.466-07:00

Blood is a rich source of medical information, containing blood cells, electrolytes, hormones, antibodies, infectious organisms, clotting factors and thousands of other biomarkers that provide a invaluable snapshot of the physiological state of the body. The practice of opening a vein (phlebotomy) dates back at least to ancient Greece, although for most of history blood was removed as a treatment rather than collected for diagnostic testing, often with detrimental consequences (Wikipedia). Modern phlebotomy emerged alongside laboratory medicine, and was transformed by the development of the evacuated blood-collection tube with the first commercial Vacutainer tubes being introduced in 1949.

Today, a phlebotomist verifies the patient and test order, applies a tourniquet, locates a suitable vein, disinfects the skin, inserts a sterile needle and fills a series of color-coded vacuum tubes in a prescribed order. The vacuum inside the tube creates a pressure differential that automatically pulls the exact predetermined volume of blood into the closed sterile vial without needing a manual syringe plunger. The needle is then removed, pressure and a bandage are applied, and the labeled tubes are transported to the laboratory. We have all undergone this procedure countless times, and despite enormous advances in how blood is analyzed, the basic process of obtaining it has changed remarkably little over the past 75 years.

That may now be changing. On August 19, 2026, the U.S. Food and Drug Administration authorized Aletta, a robotic phlebotomy device developed by the Dutch company Vitestro (Figure 1). According to the FDA, Aletta is the first standalone robot authorized to draw blood from a patient’s arm without a human operator physically performing the procedure. It received De Novo authorization, the pathway used to establish a new category for a novel low- to moderate-risk medical device. Aletta is authorized for adults in outpatient settings and must remain under the supervision of a trained phlebotomist, although one phlebotomist may oversee as many as three devices simultaneously. In other words, it is an autonomous blood-drawing machine, but not an unsupervised one.

Not surprisingly, automating a blood draw is a major challenge. A vein is not a fixed pipe at a predictable location. Veins differ in diameter, depth and orientation, and they can roll, flatten or collapse as a needle approaches. They may be obscured by fat, edema, scars or darkly pigmented skin, while arteries, nerves and tendons can lie nearby. A robot must determine whether a vessel is actually a vein, select a safe entry point, calculate the depth and angle of insertion, compensate for patient movement and know when not to proceed. It must also reproduce the many less obvious parts of good phlebotomy: tourniquet pressure and duration, skin disinfection, needle speed, tube order, collection volume and tube inversion. Any errors can distort laboratory results and require the patient to return for another blood draw or even worse result in inaccurate information.

Aletta addresses these challenges by combining several imaging methods rather than relying on a camera alone. Near-infrared imaging first creates a model of the patient’s arm and identifies where the ultrasound probe should begin searching. Ultrasound then measures the depth, diameter and direction of candidate vessels, including deeper veins that may not be visible or palpable. Doppler ultrasound analyzes the direction of blood flow to help distinguish a vein from an artery. The software proceeds only if a vessel meets predefined criteria for size, depth and orientation. Aletta then applies a tourniquet, disinfects the skin, inserts a single-use 21-gauge butterfly-style needle at a controlled speed and angle, fills and inverts the required tubes, withdraws and disposes of the needle, and applies an adhesive bandage. A human supervisor still selects the tube order, initiates the session, confirms that the tubes have been adequately filled and cleans the device between patients. If the patient moves excessively, the needle automatically detaches and the procedure stops; other sensors can pause the machine and summon the supervisor when an unsafe condition is detected.

Safety and efficacy data for Aletta were obtained from the ADOPT multicenter trial (link), which was conducted at three outpatient blood-drawing centers in the Netherlands. A total of 1,743 patients completed the robotic screening procedure, and Aletta identified a suitable vein and attempted a draw in 1,633 (93.7%). Among these attempted procedures, it successfully filled all required tubes after a single needle insertion in 94.5% of patients. The success rate remained high among patients who reported difficult venous access (92.7%), patients aged 65 or older (93.4%) and patients with obesity (97.4%). The median time from starting the automated procedure to applying the bandage was 2 minutes and 37 seconds. Published estimates in the U.S. of first-stick success for trained human phlebotomists range from 93% to 97%. All side effects from Aletta were mild and occurred in only 0.6% of cases. Compared with a manual blood draw, 90% of patients said they felt much less pain (19%), less pain (32%), or about the same amount of pain (39%). Overall, 47% preferred Aletta for a future draw, 35% had no preference and 18% preferred a human phlebotomist.

The headline success rate requires an important qualification. The 94.5% figure includes only patients in whom Aletta first identified a suitable vein and decided to insert the needle. It declined to puncture 110 patients, or 6% of those screened. When these patients are included, Aletta successfully collected blood from 88.5% of everyone who entered the robotic screening process. Refusing an unsafe or uncertain attempt is arguably a desirable safety feature, but it means that a human phlebotomist must remain available for a substantial minority of patients.

If Aletta proves reliable and economical in everyday practice, it could alter phlebotomy in much the same way that automation has transformed the clinical laboratory. Starting in the 1970s, the analysis of blood samples began to be automated, and then by the 2000s, the pre-analytical processing steps (e.g. sorting, dividing, centrifuging, and routing blood samples to the appropriate analyzer) became automated. One can imagine longer term, machines like Aletta could be connected to automated sample transport, processing and analysis, creating a continuous system from the patient’s vein to the laboratory result.

There would be high-volume blood-drawing centers operating a combination of automated lanes for routine patients and human-staffed lanes for difficult cases, children, anxious patients and people the machines decline to puncture. Rather than disappearing, phlebotomists could increasingly supervise several devices, reassure patients, respond to adverse reactions and perform the draws requiring human judgment.

In summary, the FDA decision that the Aletta autonomous robot can perform routine blood draws with a reasonable level of safety and effectiveness. Nevertheless, the most likely near-term future is not the wholesale replacement of phlebotomists. It is a hybrid model in which robots handle standardized, repetitive draws while humans supervise the machines and manage the exceptions. The larger impact may be to turn blood collection from a highly operator-dependent craft into a standardized and less labor intensive process. That could improve access to diagnostic testing and reduce waiting times, but the ultimate value of Aletta will depend on whether its reliability survives wider deployment, and whether the benefits justify the cost of putting a sophisticated robot between the patient and the laboratory.

Regarding the last point, Aletta manufacturer Vitestro has not publicly disclosed a purchase price, leasing charge or per-draw fee (which should also include maintenance costs). For comparison, the average U.S. phlebotomist earns approximately $50,000 per year, and including employee benefits the total cost to employers is roughly $75,000. On average, a typical phlebotomist performs about 6 to 10 blood draws per hour. Aletta required a median of 2 minutes and 37 seconds from initiation of the automated procedure to application of the bandage, which theoretically corresponds to more than 20 procedures per hour, but that does not encompass the entire cycle of preparing and loading the tubes, changing patients, cleaning the machine, retrieving samples, resolving interruptions and referring unsuccessful cases to a human phlebotomist. Vitestro itself has said that it is initially targeting high-volume blood-drawing centers serving at least 100 patients per day, where the fixed cost can be spread over many procedures; smaller facilities will probably adopt the technology only after its price declines and its real-world economic benefits have been demonstrated.

Figure 1. The Aletta autonomous robotic phlebotomy device by the company Vitestro (https://vitestro.com/aletta/). The robot is able to draw blood samples on its own without human assistance. It was cleared for use in medical facilities by the FDA in August, 2026.

Evidence that protein restriction may be good for you

2026-10-10T09:05:24.164-07:00

Protein-maxxing is in the news as people deliberately add as much protein as possible to meals and snacks, often through protein powders, shakes, bars and an expanding assortment of fortified foods. Many assume that consuming more protein will help them build muscle, remain healthy and perhaps even live longer, although nearly all Americans already consume more than the recommended daily amount, which is approximately 0.8 g/kg/day (roughly 50g for 140lb individual, Harvard Health).

Dietary protein is digested into constituent amino acids, which the body uses to synthesize new proteins to build and maintain muscle, as well as critical structural and enzymatic components in every cell of the body. 9 out of 20 amino acids are essential in the sense they cannot be synthesized by the body and must be obtained from the diet. For those wanting to lose weight, higher-protein diets can increase satiety and promote short-term weight loss by causing people to eat fewer calories. Additional protein may also be useful for physically active people, patients recovering from injuries and older adults at risk for malnutrition or age-related muscle loss, particularly when it is combined with resistance exercise. Muscle, excluding water, is largely protein (i.e. the proteins actin and myosin make up actomyosin muscle fibers), and so building muscle requires intake of sufficient protein in the diet. However, one meta-analysis found that for healthy adults doing resistance training, gains in fat-free (muscle) mass leveled off at a total intake of about 1.62 grams of protein per kilogram of body weight per day (roughly twice recommended amount). Any excess amino acids are largely oxidized for energy or converted into other compounds instead of being used to make more muscle.

Protein restriction (Scientific American) is basically the opposite of protein-maxxing, but the term can be misleading. According to a new review, it does not mean protein starvation or eliminating protein from the diet. The authors define protein restriction as lowering total dietary protein while still meeting nutritional needs. Protein restriction has been almost exclusively investigated in laboratory animals. For example, control mice receive diets containing about 20% of calories from protein, whereas protein-restricted diets may contain only 5% to 8%. Similarly, humans, on average, obtain approximately 17% of their calories from protein. An equivalent level of protein restriction would be roughly 25% to 50% of the recommended daily protein amount. The research in laboratory animals also suggests that protein quality and amino acid composition may be as important as total quantity. Restricting the essential amino acids methionine, isoleucine or valine can reproduce many of the effects of restricting all dietary protein.

Protein restriction bears some resemblance to the much more mature field of caloric restriction (CR). Caloric restriction is the long-term reduction of calorie intake, roughly 20–40% below usual intake (i.e. level of strict diet), without causing malnutrition or deficiencies in essential nutrients. Unlike ordinary dieting, which is generally a temporary calorie deficit used to lose weight and is relaxed once a target weight is reached, true caloric restriction is a stringent, continuous lifestyle in which reduced intake is maintained for years or potentially for life, even after a person has become lean.

There is a wealth of scientific data demonstrating that caloric restriction can extend lifespan in laboratory organisms from yeast to fruit flies to rodents. In rodents (mice and rats), the lifespan extension can range from 30 to 50% (link). Human studies are problematic from a practical standpoint because remaining chronically hungry is difficult, and most people cannot maintain a substantially calorie-restricted diet for decades. In a two-year short-term study, participants randomly assigned to a 25% calorie reduction achieved an average reduction of about 12%, yet experienced improvements in several cardiometabolic risk factors and a small slowing in one DNA-methylation measure of biological aging. It is thought that protein restriction may provide a way to activate some of the same protective systems as calorie restriction without requiring such a drastic reduction in food intake.

There is considerable support from studies in laboratory animals that protein restriction can extend lifespan, although the total body of evidence is not nearly as extensive as for calorie restriction. In their review, Knopf and Lamming summarize experiments in fish, flies, rats and mice. Depending on the species, strain, sex and diet, restricting protein increased lifespan by approximately 10% to 53% in several rodent experiments (Figure 1), while some fly experiments reported even larger effects. Restriction of individual amino acids has produced similarly striking results: reducing branched-chain amino acids increased male mouse lifespan by about 30%, isoleucine restriction by 33% and valine restriction by 23%. However the data are not uniformly positive. In one study, reducing dietary protein from 21% to 7% increased male mouse lifespan by 34.9% but was associated with a 22% reduction in female lifespan (link).

More specifically, protein restriction seems to improve metabolic health in laboratory animals, independently of whether lifespan is measured. Young rodents placed on low-protein diets generally accumulate less body fat, expend more energy and maintain better glucose control, despite often consuming more food. The intervention can remain effective when started relatively late in a mouse’s life, producing greater energy expenditure and improved metabolic measurements in aged animals.

In mice, the protein fibroblast growth factor 21 (FGF21) appears to be a central mediator of these metabolic effects of protein restriction. Elevated FGF21 communicates low protein availability to the brain, which increases sympathetic nervous-system activity, energy expenditure and heat production in adipose tissue while also improving lipid handling and insulin sensitivity. The evidence for this mechanism comes from genetic experiments in mice: protein restriction induces FGF21, removing FGF21 eliminates many of the benefits (Figure 1), and increasing FGF21 independently reproduces several of them (link). The role of FGF21 in humans is less well-characterized.

Has protein restriction been tested in humans? Yes, but the studies have been short and have measured metabolic outcomes rather than lifespan. In one randomized study, people consuming a low-protein diet for 43 days lost body weight and fat and had lower fasting glucose even though they consumed more calories. Another five-week trial in lean men found improved insulin sensitivity, higher FGF21 levels and increased energy requirements for maintaining body weight. A different 27-day trial involving people with metabolic syndrome reported lower adiposity, improved insulin sensitivity, reduced circulating glucose and lipid levels and less inflammation. Small studies restricting sulfur-containing amino acids such as methionine and cysteine have reported increased FGF21 and modest weight loss. These results show that humans share at least some of the characteristics of protein restriction as mice, especially with respect to metabolic health, but the effect on human lifespan is completely unknown.

The potential healthful benefits of protein restriction in the general population should be distinguished from reducing protein intake for people with certain conditions, most notably chronic kidney disease (CKD). A high-protein diet may be harmful to a person with chronic kidney disease, because metabolizing protein generates urea and other nitrogen-containing wastes normally removed by kidneys and are harmful as they accumulate in the blood. Protein also increases blood flow and filtration pressure within the glomeruli, the kidney’s microscopic filtering units. Healthy kidneys can usually accommodate this rise in filtration as a normal physiological response, and moderately high protein intake has not been shown to damage kidneys that function normally. In CKD, however, the remaining nephrons are already working harder to compensate for lost kidney function, so sustained high protein intake may increase glomerular pressure, worsen protein leakage into the urine, and potentially accelerate further loss of kidney function. This is why people with CKD are generally advised to consume a moderate amount of protein, whereas healthy adults usually do not need to restrict protein solely out of concern for their kidneys.

As a warning, not everyone should undergo protein restriction. Pregnant women, growing children, people recovering from injuries, individuals consuming very few calories and physically active people may have higher protein requirements. Many older adults already eat too little protein because of poor appetite, illness, financial limitations or social isolation; further restriction could worsen sarcopenia and frailty. Protein restriction can reduce lean muscle body mass, which can be partially offset by resistance exercise, but may still be undesirable. The optimal protein amount almost certainly varies from individual to individual, and more generally, the correct protein and amino acid targets for humans remain unknown with the recommended daily amount serving as a broad target.

In summary, protein maxxing is an embodiment of the "more is better" philosophy with respect to a nutrient such as protein, and indeed, higher protein intake can be useful for building or preserving muscle, controlling appetite and supporting recovery, especially when paired with exercise. On the other hand, caloric restriction and protein restriction espouse the "less is better" philosophy regarding lifespan and healthspan extension. From a pragmatic perspective, protein restriction is likely to be easier to achieve than caloric restriction in humans. In laboratory animals, protein restriction consistently improves metabolic health often lengthening lifespan. Mechanistically, the protein FGF21 seems to play a central role in these effects in mice. A small number of short human trials suggest that there may be health benefits to protein restriction. For most healthy people, a range of protein intake values centered around the recommended daily amount is most likely conducive to health. There are not sufficient human data yet to recommend maximizing or minimizing the amount of protein in your diet.

Figure 1. In mice, the beneficial effects of dietary protein restriction on lifespan are reversed by deleting the FGF21 gene. Kaplan-Meier survival curves plotting percent survival as a function of age for wild-type (WT) control diet (solid black line) and wild-type mice under low protein (LP, solid red) diet with the latter showing extended lifespan (left panel). When the FGF21 gene is deleted, then control diet mice (dashed black) live as long or longer than low-protein diet (red dashed) mice suggesting that the FGF21 gene is necessary for LP-induced lifespan extension. n = 30 (mice) per group (Hill et al. Nature Communications, 2022).

Agility drills emphasize changing speed and direction

2026-10-03T09:36:27.680-07:00

YouTube possesses a vast collection of exercise and training videos (e.g. QH). Many athletes post their off-season training routine, and I have noticed that wide receivers in football often incorporate agility ladders into their regimen (Video 1). As ChatGPT helpfully points out, an agility ladder is not a ladder that you climb, it is a lightweight exercise aid laid flat on the ground, with two long sides and a series of crosspieces that form rectangular "squares", which looks like a ladder laid flat on the ground. You move your feet through, into, out of, alongside or across those squares using predetermined patterns; for example, one steps rapidly forward and backward, moving sideways, skipping, hopping on one foot or performing an "in, in, out, out" sequence (see ~2m in Video 1). The ladder provides visible targets for where each foot should land, and its main purpose is to practice precise foot placement, coordination, balance, rhythm and neuromuscular control. Commercial agility ladders generally cost about $10 to $40.

A recent New York Times article describes Scott Delp, a Stanford bioengineering professor, using the same type of agility ladder at age 65 that he first encountered as a high-school football player. While the tool has not changed. his reason for using it has. As a young athlete he wanted faster footwork; now he is interested in coordination and the ability to keep learning unfamiliar movement patterns. This captures the appeal of agility training for the rest of us. Walking, cycling, and many gym exercises move mainly forward and backward. Daily life and recreational sports also ask us to stop, turn, step sideways, avoid an obstacle, and occasionally recover from a bad step.

Agility drills are exercises in which you accelerate, slow down, change direction and reposition your feet while maintaining balance and control. In Physics, acceleration is defined as the change in velocity, and velocity is a vector possessing both a direction and magnitude (speed), and so acceleration is a change in speed and direction, which is what agility drills emphasize.

There is, however, some nuance in the definition of agility drills. Exercise scientists have defined agility as a rapid change in whole-body speed or direction in response to a stimulus. A memorized ladder pattern is therefore more precisely a coordination or change-of-direction drill, but not a true agility drill. Only when, for example, a partner unexpectedly tells you to move left or right does it become an agility drill in the strict sense. This is semantic hair-splitting, and for the purpose of this post, agility drills will refer to exercises that emphasize changes in speed and direction whether in response to a stimulus or memorized.

Standard forms of exercise such as walking, running and cycling mostly move the body in a predictable forward direction. Agility exercises introduce sideways and diagonal steps, starts and stops, changes in tempo, single-leg support and more precise foot placement. In other words, they train the nervous system and muscles to coordinate unfamiliar movements while the body’s center of mass is shifting and accelerating. This may improve balance, coordination, reaction time, leg power and functional mobility -- all capabilities that become particularly important as we age.

One type of agility drill is step training (i.e. repeating a structured stepping pattern). A meta-analysis of step training in older adults (60 or older) showed that it reduced fall risk. Voluntary and reactive step practice were associated with a fall-rate ratio of 0.48 and a risk ratio of 0.51 for becoming a faller (roughly half the risk compared to the control group). The interventions also improved stepping reaction time, single-leg stance, and the Timed Up and Go test (time it takes to stand up from a chair, walk three meters, turn around, walk back, and sit down).

For young athletes at the other end of the age spectrum, agility is essential for high level sports performance with its repeated accelerating, braking, cutting and reacting to other players or a moving ball. As a result, agility tests have become commonplace as part of the evaluation process such as when scouting potential draft picks. These include the three-cone drill plus the 20- and 60-yard shuttles for the NFL football scouting combine. NBA basketball teams test prospects with a lane agility drill, and the reactive shuttle, which involves responding to a changing light stimulus by moving either left or right, which fits the strict definition of an agility drill.

As an aside, it is important to distinguish agility from intensity; they represent two different aspects of exercise. Agility describes the movement challenge, whereas intense training such as high-intensity interval training (HIIT) describes how hard you exercise and how work periods are alternated with recovery. A controlled agility ladder exercise can be moderate or even light exercise, particularly when it is performed slowly to improve balance. On the other hand, repeated 10- to 20-second shuttle sprints with short recovery periods can be both agility training and HIIT.

For a beginner, the New York Times workout provides a sensible progression. Start with "fast feet forward": step both feet onto a very low box or into a taped square and then step backward out of it, using the back of a chair for support if necessary. Next, practice fast feet sideways, shifting laterally across the square. Once these movements feel stable, progress to an "in, in, out, out" pattern while moving forward or sideways through several taped squares. Other useful drills include walking or jogging in a figure eight around two objects, lateral shuffling between two markers, taking several quick steps forward and backward, or raising one knee while passing a tennis ball underneath it. To incorporate true reactive agility, ask a partner to call "left," "right," "forward" or "back." Initially, the goal should be accurate, controlled foot placement, not maximum speed. A drill performed slowly and correctly is more useful than a spectacular-looking sequence during which you repeatedly step on the ladder and nearly fall over.

Agility training should be thought of as a complement to aerobic and strength training. Current federal recommendations call for at least 150 minutes of moderate aerobic activity or 75 minutes of vigorous activity each week, together with muscle-strengthening exercise on at least two days (QH). Adults aged 65 and older are also specifically advised to include balance activities. Agility drills can fill part of that balance and coordination slot, and if done vigorously can contribute to the aerobic work. But a few short ladder bouts do not replace a walk, run, bike ride, or full-body strength program. A practical way to schedule technique-oriented agility is after a 5- to 10-minute warm-up and before hard intervals or heavy lifting, when the movements can still be performed accurately. Slower balance steps can instead be used as a brief stand-alone movement break.

One should be careful when performing agility drills. Although a taped agility ladder may look harmless, the difficulty changes considerably when you add speed, jumping or sharp turns. The faster you are moving, the more momentum your muscles and joints must absorb when you stop or cut, and the more dangerous any spills or falls. Make sure your exercise space is flat, uncluttered, and nonslip, and wear shoes that do not slide. Begin at walking speed even if you already exercise regularly, and increase speed only after you can complete the pattern without catching your feet, losing balance or looking down continuously. The New York Times article appropriately advises people with severe arthritis, significant balance problems, or recent injury to consult a physical therapist or trainer about modifications.

How much agility training should you do each week? There is no official public-health recommendation specifying a minimum number of agility-ladder minutes per week. For the average beginner, a reasonable starting point is one or two sessions per week, each containing approximately 10 to 20 minutes of actual agility work. Select three to five drills, perform two to four sets of each and keep individual work periods relatively brief -- often 10 to 30 seconds -- with enough recovery to preserve accurate movement. The New York Times routine lasts 20 to 25 minutes and uses three or four sets per drill. This can be done once or twice per week.

In summary, agility ladders are not magical devices, nor are agility drills reserved for prospective professional athletes. Their principal value for the average person is that they expand the range of movements the body can execute confidently. Combined with aerobic exercise, strength training and ordinary balance work, a small amount of multidirectional and reactive movement may help preserve the ability to respond when the world quite literally throws something unexpected in your path. That unexpected something may require you to dramatically change speed and direction while maintaining balance.

Video 1. The workout regimen of former Kansas City Chief wide receiver Tyreek Hill heavily involves the use of the agility ladder. In this video, the agility ladder training begins around 1m30s.

Uneven availability of genomic sequencing of advanced cancers

2026-09-26T10:25:03.112-07:00

Cancer is fundamentally a disease arising from altered (mutated) DNA. These mutations (i.e. changes in DNA sequence) can cause a cell to grow uncontrollably, escape the immune system, or spread to other parts of the body (metastasize). Some mutations activate genes called oncogenes which drive tumorigenesis. Other mutations inactivate tumor suppressor genes, which normally prevent cancer (e.g. by repairing DNA). Most cancer-causing mutations arise somatically in your body cells after you are born, although some may be inherited from your parents.

Genomic testing has become standard in modern cancer care. The goal is to identify the various mutations in the cancer cells, especially those that contribute to tumorigenesis. In the past, the tumor was tested one mutation at a time. These days, next generation sequencing (NGS) technology has enabled whole genome sequencing of the tumor cells so that all the mutations can be identified at once. Tumor DNA can be obtained via tissue sample (standard biopsy) or by fragments of tumor DNA circulating in the blood, i.e. via a blood sample ("liquid biopsy").

The mutational profile of a tumor can inform the oncologist on what oncogenes have been activated and what tumor suppressor genes have been inactivated in the tumor. This knowledge can help guide prognosis and treatment because the activation of certain oncogenes, for example, can make a tumor more aggressive, and because many new innovative targeted therapies inhibit particular activated oncogenes that play a central role in driving the cancer. In other words, certain targeted drug treatments are specific for cancers containing certain mutations.

Not surprisingly, the American Society for Clinical Oncology (ASCO) recommends genomic sequencing in metastatic or advanced cancer when alterations have regulatory approval as biomarkers that guide treatment selection or exclusion. Since most major cancers have such regulatory approval for some set of mutations, one can argue that tumor DNA sequencing should be the default option in mody cases of advanced or metastatic cancer. However this ideal is often not met giving rise to the practical question of whether the test is actually ordered, paid for, completed, returned, and interpreted early enough to matter.

A new large cohort study in JAMA Network Open gives a useful, and somewhat sobering, snapshot of how often that process appears to happen in routine US oncology care. The study examined 63,294 patients (median age 68) diagnosed from 2018 through 2022 with one of five common advanced or metastatic cancers: metastatic breast cancer, metastatic prostate cancer, advanced non-small cell lung cancer (NSCLC), metastatic colorectal cancer, or metastatic pancreatic cancer. The data came from the Flatiron Health Research Database, an electronic health record-derived database covering about 280 US cancer clinics and about 800 sites of care. Most patients in the database were treated in community oncology practices, although approximately 13% to 20% of the patients in the different cancer groups were treated at academic practices.

In the paper, time to NGS testing was defined as the date the patient was diagnosed with advanced or metastatic cancer to the date the NGS result was delivered. A time ratio of 1 represented no difference in this time interval from the reference group. A ratio of 1.4 indicated that the estimated time to testing was 40% longer, while a ratio below 1 indicated earlier testing. The reference group was the group with the shortest time to NGS testing. For example, the reference group for socioeconomic status was SES quintile 5, the highest group. Data from patients who did not undergo NGS were treated as censored data which did not occur during a certain observed period and so represented a lower bound.

The investigators then examined whether time to testing was associated with variables such as race and ethnicity, socioeconomic status, insurance, academic versus community practice, and sex. Socioeconomic status (SES) was not based on the patient’s reported income. Instead, it was estimated from Census information about the neighborhood surrounding the patient’s most recently recorded address and divided into five groups from lowest to highest socioeconomic status.

The headline result was that many patients had no recorded next-generation sequencing (NGS). Overall, 34,451 of 63,294 patients, or 54.4%, had NGS after the advanced or metastatic diagnosis. Another 2,714 patients, or 4.3%, had NGS before that diagnosis. But 26,129 patients, or 41.3%, had no recorded NGS at all. Thus, a surprising number of people did not have their advanced cancer sequenced.

The testing rate varied considerably by cancer type with the highest rates for lung and colorectal cancer (Table 1):

Table 1. NGS testing rate after advanced diagnosis by cancer type over time period from 2018 to 2022. A smaller fraction of patients had NGS before diagnosis and represent patients not in either column 2 or 3. Last column shows one year testing rates for years 2018 and 2022 (data from Chehade et al. JAMA Network Open, 2026).

The good news is that the one-year cumulative incidence of testing rose in every cancer category. The increases were particularly striking in colorectal cancer, from 41% among patients diagnosed in 2018 to 78% among those diagnosed in 2022, and in pancreatic cancer, from 30% to 62% (Table 1, last column).

Among patients with a recorded test, the reported median interval from advanced diagnosis to NGS was approximately one month for lung cancer, 1.3 months for pancreatic cancer, 1.5 months for colorectal cancer, 2.9 months for breast cancer, and 11 months for prostate cancer. The relatively long delay sequencing prostate cancer probably reflects the slower progression of advanced prostate cancer (i.e. less urgent), and the fact that genomic results historically guided later-line treatment rather than the initial treatment decision.

There were clear trends between variables and cancers. For example, a strong socioeconomic gradient was associated with advanced lung cancer. Patients in the lowest socioeconomic status (SES) group had a 61% longer time to NGS than lung cancer patients in the highest group. In addition, Black patients had a 38% longer time, and Medicare or other government insurance was associated with a 40% longer time. Similar gradient trends were observed for the other cancers.

In the Discussion, the authors propose several explanations for why people in lower SES groups had lower NGS rates and longer times to NGS:

Lower-SES communities may have less access to health care and advanced diagnostic testing.
NGS can be expensive, and insurance coverage can be inconsistent.
Practices may differ in their infrastructure for ordering, processing, and interpreting NGS.
Physician- and patient-related barriers may affect whether and when testing is pursued.

The physician- and patient-related barriers may include lower referral rates, reduced attendance at genetic-counseling appointments, and limited knowledge about NGS.

Fortunately, the gap is closing, steadily moving towards 100% sequencing of advanced cancers, which would of course ensure everyone no matter what their socioeconomic status would have their advanced tumor sequenced. To speed up this process, the Chehade et al. make the following broad recommendations:

Health care policies that improve access to genomic testing
Greater awareness of NGS among clinicians and patients
Education about targeted therapies against cancers containing specific mutations
Policies and educational efforts aimed at bridging disparities associated with socioeconomic status, race and ethnicity, and insurance coverage
Addressing the broader financial and infrastructure barriers that impede implementation of NGS

In summary, precision medicine tailors disease prevention and treatment to the unique genetic makeup, environment, and lifestyle of each individual. Instead of relying on a traditional "one-size-fits-all" model designed for the average patient, doctors use precision medicine to select the right treatments for the right patients at the right time. Precision oncology is the application of precision medicine to cancer that tailors medical treatment to the specific genetic and molecular characteristics of an individual patient's tumor, and in particular to the specific mutations found in the tumor.

NGS is increasingly the doorway through which patients enter precision oncology. The study described above shows that this doorway opened much wider between 2018 and 2022, especially for lung, colorectal, and pancreatic cancers, but it did not open equally for everyone. The next stage of precision oncology will therefore require more than better sequencing technology. It will require reliable insurance coverage, faster testing workflows, clinician and patient education, access to molecular expertise, and systems that ensure that a useful genomic result leads to an appropriate treatment. In this large US cohort, about 4 in 10 patients with five common advanced or metastatic cancers had no recorded NGS at all; the gap is closing but more work needs to be done.

ChatGPT for Clinicians is an assistant for individual medical professionals

2026-09-19T08:50:31.720-07:00

Previously, I wrote about ChatGPT Health, which is a dedicated consumer health and wellness space within the ChatGPT app that helps patients consolidate, interpret, and better understand their personal health information. Users can connect electronic medical records and compatible wellness apps, such as Apple Health and MyFitnessPal, and use the system to explain laboratory results, identify longer-term lifestyle trends, summarize symptom histories, and prepare useful questions for medical appointments. Its primary value is as a personal "sense-making layer" that converts scattered records, wearable data, and medical terminology into understandable, actionable information, thereby helping patients participate more actively in their care. However, it is not intended to diagnose conditions, recommend treatment, or replace professional medical judgment.

Despite their similar names, ChatGPT Health is distinct from ChatGPT for Healthcare: ChatGPT Health serves individual consumers, whereas ChatGPT for Healthcare is an enterprise-grade platform for hospitals and healthcare organizations, designed for regulated clinical and administrative workflows, evidence-based research, documentation, and institutional system integration.

As if that weren't confusing enough, OpenAI recently released a third chatbot product in the medical/healthcare domain called ChatGPT for Clinicians. It is a specialized, free version of ChatGPT designed to help verified healthcare professionals in the U.S., including physicians, nurse practitioners, physician assistants, and pharmacists with clinical research, documentation, and repetitive administrative work (Figure 1). More specifically, clinicians can use the latest version of ChatGPT to answer clinical questions, search peer-reviewed medical sources with citations (literature search), conduct literature reviews, draft referral and prior-authorization letters, prepare patient instructions, and turn recurring workflows into reusable "skills."

In terms of tasks performed there is substantial overlap with ChatGPT for Healthcare. The main distinction is that ChatGPT for Clinicians is tailored to individual healthcare providers, whereas ChatGPT for Healthcare targets entire healthcare organizations, allowing hospitals and clinics to deploy the platform at scale across clinicians, administrators, and researchers. The advantages of the latter include centralized management, role-based access, audit logs, organizational data connections, enterprise security, and a BAA (Business Associate Agreement for protection patient data) covering the institutional deployment. Thus, ChatGPT for Clinicians is not a renamed or merely smaller version of ChatGPT for Healthcare: the former is designed for individual professional use, whereas the latter is required for organization-wide adoption.

Compared to ChatGPT Health, ChatGPT for Clinicians is serving a completely different clientele, i.e. healthcare providers versus healthcare consumers. In that sense they are complementary products with ChatGPT Health operating as a consumer wellness product for individuals managing their personal information to help patients better understand and participate in their care, whereas ChatGPT for Clinicians acts as a professional assistant for verified healthcare providers to deliver that care.

Because it spans several categories, ChatGPT for Clinicians has no single exact competitor. Its most direct competitors in the clinical reference and evidence space include UpToDate Expert AI, OpenEvidence, Elsevier ClinicalKey AI, Claude for Healthcare, and Doximity's suite of AI tools, which often benefit from deeply curated, expert-authored knowledge bases or established professional networks. Additionally, it competes adjacently with ambient documentation (e.g. AI medical scribe) and workflow automation systems -- such as Microsoft Dragon Copilot, Abridge, Nabla, Suki, and Ambience Healthcare -- which are typically more tightly integrated directly into Electronic Health Record (EHR) environments to capture patient encounters in real-time. Ultimately, OpenAI's primary advantage in this landscape remains its low barrier to adoption as a free tool that consolidates research, writing, and custom workflows into a single, familiar interface.

The public health community generally view ChatGPT for Clinicians as a promising but insufficiently validated clinical assistant whose adoption may be advancing faster than independent evidence, regulation, and health-system governance. They caution that strong benchmark results and physician ratings do not yet demonstrate improved patient outcomes, reduced errors, lower workload, or equitable performance across diverse populations and care settings, particularly because much of the initial evidence was produced by OpenAI itself. Additional concerns include hallucinated or misinterpreted evidence, automation bias caused by polished answers and citations, unclear liability, privacy misunderstandings surrounding HIPAA and Business Associate Agreements, and the possibility that individual clinicians may use the tool outside institutional oversight. Critics also warn that time saved through rapid drafting may be offset by the need to verify sources, correct errors, and reconcile outputs with local workflows. The prevailing critique is therefore not that the tool lacks value, but that it should be deployed with independent evaluation, local validation, equity testing, auditing, incident reporting, and continued clinician oversight rather than being treated as proven clinical infrastructure. OpenAI highlights that physicians rated 99.6% of nearly 7,000 pre-release responses as safe and accurate on its HealthBench Professional evaluation.

Perhaps the greatest potential benefit of ChatGPT for Clinicians is its ability to return time and cognitive attention to healthcare professionals by handling the information-intensive and writing-intensive parts of medical work. By bringing cited clinical search, medical-literature synthesis, documentation assistance, and reusable workflows into one workspace, it can provide a first draft of notes, referral and prior-authorization letters, patient instructions, and evidence summaries that the clinician can then review and refine. Its value therefore extends beyond simply generating text faster: it can reduce the friction between identifying a clinical question and obtaining an actionable, reviewable response, helping clinicians keep pace with expanding medical knowledge, communicate more clearly with patients, and devote more attention to complex decisions and direct patient care. The greatest gains will come from augmentation rather than replacement, with the AI performing preliminary searching, synthesizing, and drafting while the clinician verifies the evidence and retains responsibility for the final judgment.

Because it is free, and because medical professionals already use ChatGPT for some of the tasks described above, one can assume that there will be adoption of a more dedicated deployment. As an assistant there is less concern over errors and hallucinations since the human medical professional user bears the ultimate responsibility. At the same time, the chatbots can excel at repetitive tasks, even those that require some degree of reasoning such as evidence synthesis and question answering. Time will tell regarding the uptake, and how much competition from alternative apps.

Figure 1. ChatGPT for Clinicians homepage (https://chatgpt.com/plans/clinicians/). The product "supports clinical reasoning and document drafting, while clinicians stay in control of care decisions." It is free for verified individual licensed medical professionals in the United States.

New 2026 ACC/AHA Dyslipidemia Guideline for cardiovascular disease risk assessment

2026-09-12T10:18:13.716-07:00

Dyslipidemia is defined as an abnormal pattern of lipids (fats) in the bloodstream, most commonly involving elevated LDL cholesterol, elevated triglycerides, low HDL cholesterol, or other abnormal lipid markers such as non-HDL cholesterol, apoB, or lipoprotein(a). It is a major, often silent, driver of atherosclerotic cardiovascular disease: excess atherogenic lipid particles (i.e. LDL) can enter artery walls, promote inflammation and plaque formation (which clogs the artery), and ultimately increase the risk of coronary artery disease, heart attack, stroke, and peripheral artery disease. Because dyslipidemia usually causes no symptoms until vascular disease is advanced, routine lipid screening is important for early detection. It is also clinically important because it is a modifiable risk factor; lifestyle changes and lipid-lowering medications such as statins can substantially reduce future cardiovascular events.

Traditionally, LDL has been the focus because it is the main atherogenic (atherosclerosis-causing) lipoprotein in the blood. However, other apolipoprotein B (apoB) containing lipoprotein particles can also enter the artery wall contributing to atherosclerotic plaque. This is where the biomarkers non-HDL cholesterol and apoB comes into play and the two terms are somewhat interchangeable; the apoB particles are exactly the same group of lipoprotein particles that carry non-HDL cholesterol. ApoB is a key protein found on the surface of cholesterol-carrying lipoprotein particles in your blood.

Non-HDL cholesterol (Non-HDL-C) is defined as the cholesterol carried by all lipoproteins except HDL. It is calculated from a standard lipid panel: Non-HDL-C = Total cholesterol − HDL-C.

Medically, non-HDL-C is important because it describes the cholesterol carried by all the main atherogenic particles and not just LDL; these include LDL, VLDL, IDL, triglyceride-rich remnant particles, and lipoprotein(a), all of which contain apoB. Physiologically what makes HDL different is that HDL particles collect excess cholesterol from peripheral tissues and the walls of the blood vessels, and carries the cholesterol back to the liver, where it is broken down and excreted from the body via bile. This is essentially the reverse of what all the atherogenic non-HDL lipid particles do.

There are certain scenarios in which non-HDL-C can reflect cardiovascular risk better than LDL-C alone, especially when triglycerides are elevated, as in diabetes, insulin resistance, obesity, metabolic syndrome, or mixed dyslipidemia. In those settings, LDL-C may look only mildly abnormal while VLDL and remnant particles are also contributing significantly to atherosclerosis risk.

Given the need for an update, in March, the American College of Cardiology (ACC), American Heart Association (AHA), and nine other medical organizations issued the new 2026 ACC/AHA Dyslipidemia Guideline that replaced the 2018 guideline on the management of blood cholesterol. The older 2018 U.S. ACC/AHA "blood cholesterol" guideline focused heavily on LDL-C, statin intensity, percentage LDL reduction, and 10-year ASCVD risk estimated with the Pooled Cohort Equations. The newer 2026 guideline broadens the scope to dyslipidemia more generally, including LDL-C, non-HDL-C, triglycerides, remnant particles, apoB, and lipoprotein(a), and replaces the older risk calculator with the PREVENT-ASCVD equations, which estimate both 10-year and 30-year risk in adults ages 30–79. In practical terms, the change is from "estimate 10-year risk and choose statin intensity" to "estimate both short- and long-term risk, personalize risk using factors such as CAC (coronary artery calcium score), Lp(a), apoB, diabetes, kidney disease, and family history, and treat earlier or more intensively when overall lifetime cardiovascular risk is high."

There are several reasons for the change. First, atherosclerosis is increasingly understood as the consequence of cumulative exposure to cholesterol-containing particles over many years. An LDL level that is moderately elevated for 30 or 40 years may cause more damage than a higher level that is present for only a short period. This has led to the "lower for longer" principle: lowering LDL earlier in life should prevent more plaque from forming than waiting until a person becomes old enough to have a high 10-year risk. Second, as described above, LDL cholesterol does not capture every form of lipid-related risk. Third, there are now more effective nonstatin treatments, making it possible to reach lower cholesterol goals when statins alone are insufficient or cannot be tolerated.

From a technical perspective, the new guideline replaces the Pooled Cohort Equations (PCE) with the AHA’s Predicting Risk of Cardiovascular Disease EVENTs, or PREVENT-ASCVD, equations. The older Pooled Cohort Equations estimated the probability of a first heart attack or stroke during the next 10 years for an adult between ages 40 and 79. The relevant variables included age, sex, race, total cholesterol, HDL cholesterol, systolic blood pressure, blood-pressure treatment, smoking, and diabetes. PREVENT-ASCVD applies to adults ages 30 to 79 and calculates both 10-year and 30-year ASCVD risk. It eliminates race from the equation and uses a broader cardiovascular-kidney-metabolic profile that includes age, sex, non-HDL cholesterol, HDL cholesterol, systolic blood pressure, smoking, diabetes, body mass index, kidney function measured by estimated glomerular filtration rate, and the use of blood-pressure and lipid-lowering medications. Optional inputs such as hemoglobin A1c, urine albumin-to-creatinine ratio, and a neighborhood-level Social Deprivation Index can further refine the estimate.

In addition, PCE was built using data from only about 25,000 people from older studies (decades ago), and generated separate equations for Black and White adults. PREVENT was derived from about 3.28 million contemporary adults and externally validated in another 3.33 million. As a result, the PREVENT model possesses superior agreement between predicted and observed risk (calibration). For example, we now know that the PCE 10 year risk was an overestimate because the older data do not reflect the improved health of a modern population.

For the typical patient, the new guideline will refine the treatment regimen rather than replace it. Lifestyle measures -- including a healthy dietary pattern, regular physical activity, weight management, avoidance of tobacco, adequate sleep, and control of blood pressure and diabetes -- remain the foundation of treatment. When medication is appropriate, a statin will usually remain the first drug prescribed. The practical difference is that the clinician will now calculate PREVENT risk, consider a specific LDL-C and non-HDL-C goal, and monitor whether the patient reaches that goal. The routine evaluation may also include a once-in-a-lifetime Lp(a) measurement, while apoB testing may be useful when triglycerides are elevated, diabetes or metabolic syndrome is present, or LDL-C appears satisfactory despite suspected residual risk. If the decision to begin treatment is uncertain, coronary artery calcium imaging may show whether the patient has already developed coronary plaque.

The PREVENT-ASCVD calculator places the patient into one of four 10-year risk categories as shown in Table 1. At a borderline 10-year risk of 3% to <5% (chance of heart attack or stroke), a moderate-intensity statin is reasonable after discussing risk enhancers and preferences. At intermediate risk of 5% to <10%, at least a moderate-intensity statin is recommended, generally aiming for a 30%–49% LDL-C reduction and an LDL-C <100 mg/dL. At high risk of at least 10%, a high-intensity statin is recommended, aiming for at least a 50% reduction and an LDL-C <70 mg/dL.

Table 1. PREVENT-ASCVD 10-Year Risk Categories. Risk score is the probability of a first heart attack or stroke during the next 10 years (ChatGPT).

Patients who do not reach their LDL goal on a tolerated statin may also receive more active treatment than under the older 2018 regimen. The clinician may increase the statin dose or add another drug, commonly ezetimibe, which reduces intestinal cholesterol absorption. Bempedoic acid offers an oral option for some patients who cannot tolerate adequate statin therapy, while injectable PCSK9-directed drugs can produce much larger LDL reductions in patients at high risk or with inherited hypercholesterolemia. People with established ASCVD are most likely to receive intensive combination therapy because many will now have an LDL-C goal below 55 mg/dL. Thus, a hallmark of the new guidelines is more stringent LDL-C goals for higher risk patients that are now achievable through recent LDL lowering medicines.

The 30-year risk assessment tool is explicitly intended for younger adults aged 30 to 59 who present with a low 10-year risk score (<3%). It is possible that their 30-year risk is much higher, and this longer-term perspective can make a major difference for the younger person. For example, consider a 38-year-old with elevated LDL cholesterol, mildly high blood pressure, and a family history of premature heart disease. Because age exerts a large effect on short-term cardiovascular risk, the older model might calculate a low probability of a heart attack or stroke before age 48. That result could support a “watch and wait” approach. The PREVENT model may still calculate a low 10-year risk, but its 30-year estimate can reveal a much greater probability of an event before age 68 e.g. ≥ 10%. In that elevated long-term risk category, a moderate-intensity statin reasonable even when 10-year risk is low (see Table 2).

Table 2. PREVENT-ASCVD 10-Year Risk Categories (ChatGPT).

Importantly, the new guideline does not rely on the calculator alone. It recommends a "calculate, personalize, and reclassify" (CPR) risk refinement framework in which the PREVENT estimate is supplemented by risk-enhancing factors such as family history, elevated Lp(a), chronic inflammatory disease, pregnancy-related risk factors, or chronic kidney disease:

C - Calculate: Run the baseline 10-year PREVENT-ASCVD risk score (followed by 30-year risk score).
P - Personalize: Assess variables excluded from the calculator, such as family history, reproductive risk markers, Lp(a) measurement, and persistently elevated high-sensitivity C-reactive protein (chronic inflammation).
R - Reclassify & Reassess: Selectively utilize Coronary Artery Calcium (CAC) scoring to reclassify risk for borderline or intermediate patients and finalize the treatment plan.

If uncertainty remains after the "C" and "P" steps, coronary artery calcium (CAC) scanning can determine whether calcified atherosclerotic plaque is already present. A CAC score of zero may support deferring medication in an otherwise lower-risk person with no higher-risk condition, whereas any CAC favors treatment, especially at higher scores.

In summary, the overall message of the 2026 guideline is that cholesterol management should begin earlier, consider more than LDL alone, and continue across the lifespan with stringent goals. The older approach was largely concerned with whether a middle-aged patient’s 10-year risk was high enough to justify starting a statin. Now there is a more explicit attempt to match the intensity and timing of treatment to absolute risk, lifetime exposure, additional risk markers, and patient preference. A particular focus is reducing the cumulative burden of atherogenic lipoproteins over a longer time horizon i.e. 30 years. For many patients, the answer will still be healthy behavior plus a statin. The difference is that clinicians now have a more modern risk calculator, clearer treatment targets, additional biomarkers, and a larger selection of drugs with which to pursue the goal of keeping harmful cholesterol particles lower for longer.

Stress testing frontier LLMs on medical benchmarks

2026-09-05T08:56:26.727-07:00

Frontier large language models (LLMs) now routinely achieve top scores on medical exams (QH), diagnostic challenge questions, and health-related visual question-answering tasks. However, acing static benchmarks should not be mistaken with medical competency. For one, a model that chooses the correct answer on a multiple-choice question has not necessarily shown that it used the right evidence, recognized missing information, or would behave safely in a real clinical workflow. In other words, how robust is its understanding of medicine?

In a new Nature Medicine paper, Gu and colleagues dig beneath the surface of the benchmark success stories to evaluate the robustness and readiness of large frontier models in health AI applications. The paper is not a clinical trial, and it does not show whether any model improves diagnosis, treatment, or patient outcomes. It asks a narrower question: When medical benchmarks are perturbed in simple ways, do models still behave as if they understand the clinical evidence?

Over the past few years, cutting-edge AI models have advanced from primarily text-based chatbots to what is termed multimodal models that are able to generate, process, and understand other media formats such as images, audio, and video along with text. This multimodal capability is especially valuable in the medical domain because so much information is in the form of images such as MRI scans.

The researchers evaluated multimodal health AI tasks including visual question answering, radiology report generation, and diagnostic challenge questions from sources such as the journals NEJM (The New England Journal of Medicine) and JAMA (Journal of American Medical Association). The model set included GPT-4o, GPT-5, OpenAI-o3, OpenAI-o4-mini, Gemini 2.5 Pro, Claude 3.5 Sonnet, and several open or medical-oriented models -- models roughly 1 to 2 generations behind the current frontier models.

The researchers wanted to examine whether the models perhaps were memorizing the answer, getting lucky (e.g. right answer for wrong reason), or even cheating (e.g. data contamination of test questions leaking into training dataset) to achieve their impressive benchmark scores. They did so by creating six stress test variants of the benchmarks: remove the image, focus on questions judged to require the image, shuffle answer choices, manipulate distractors, substitute a different clinically plausible image, and audit model-generated reasoning. These tests are artificial by design. If a model's answer depends heavily on answer-option order, memorized patterns, or text cues that remain after the image is removed, the benchmark score may be misleading and the model may not truly possess the medical understanding indicated by the score.

Shuffling the answer choices is something any test maker will do to counter cheating. If the answer key was somehow leaked to the training dataset, then perhaps the model just memorized the multiple choice letter of the answer. In this zeroth-order stress test, the models exhibited no decline in performance after the answers had been shuffled.

A second stress test was image removal on questions that seemed to require an image to answer correctly. On the "visual-required" NEJM data subset, removing the image dropped performance significantly. What was surprising was not the decrease, but the fact that most models were still able to answer correctly about 38-41% of the time without the image, compared with a 20% random baseline (5 option multiple-choice, Figure 1). Several explanations are possible. The text alone may contain substantial clinical clues helping to narrow the answer choice. Another possibility is that even without the image, the models were able to eliminate totally implausible answer choices, a common test-taking strategy. Regardless, the models should have, but didn't, recognize that essential evidence was missing, and instead went ahead and submitted an answer.

An even greater challenge is visual substitution. The researchers took 40 NEJM questions whose answers depended on image interpretation, and then replaced the original image with a clinically plausible alternative image that would change the correct answer. If a model was properly integrating visual evidence, its answer should change when the image changes with no loss in performance. Instead, most models lost roughly 23-33 percentage points in this visual-substitution test; GPT-5 fell from 84% to 53%, a 31-point drop, and Gemini 2.5 Pro fell from 76% to about 53%. Admittedly the sample size is small, and there is no guarantee that the altered question is of the same level of difficulty as the original question, but it is still a useful probe. This decline suggests the models may have relied on memorized associations, or benchmark-specific patterns rather than the current visual evidence.

Finally there was the classic right answer but wrong reason, which is not penalized in a multiple choice exam. First, encouraging reasoning through chain-of-thought prompting did not seem to help. Second, reasoning manual audits (i.e. examining reasoning trace) found fluent but unsupported explanations which included correct answers with incorrect logic, fabricated visual findings, perceptual errors that propagated through later reasoning, and structured but clinically irrelevant rationales. This fits earlier evidence. In a 2024 NPJ Digital Medicine study, GPT-4V reached high multiple-choice accuracy on NEJM Image Challenges but produced flawed rationales in many cases in which the final answer was correct.

Taken together, the Gu et al. paper highlights that benchmark accuracy can hide brittleness: models may answer without essential input, shift under superficial format changes, fail to update when visual evidence changes, and produce explanations that do not faithfully reflect the evidence. However these results do not show that these models are unsafe in every health-related use. It also does not evaluate clinical outcomes in any way such as test prospective deployment, clinician oversight, patient outcomes, workflow constraints, or harm rates. Finally, risk also depends on intended use; a documentation assistant, a patient-facing chatbot, and a diagnostic recommendation system represent three very different use-cases with different criteria for success.

One take home lesson is that medical AI needs to be subjected to as many "stress tests" as possible to reveal possible fragilities in real-world applications. Standard benchmark accuracy is where the evaluation starts, not where it ends. For healthcare, the relevant question is not only "How often is the model right?" It is also: when it is right, is it right for the right reasons, and when the evidence is missing or conflicting, does it know what to do? The path for frontier LLMs becoming useful health tools runs through stress testing, workflow-specific validation, transparent reporting, and ongoing monitoring. The bar needs to be set as high as possible.

Figure 1. Image removal stress test on 197-item NEJM-VS dataset. The questions were deemed by board-certified clinicians to depend on the image for the correct answer. Plotted are the accuracy (%) results on questions (after image had been removed) from five inference runs for each of the 6 models with median line, interquartile range (boxes), and full data range (whiskers). Dashed line indicates 20% chance level (i.e. guessing) which one may expect after critical evidence had been removed (Gu et al. Nature Medicine, 2026).

The Signal Ring blood pressure monitor fits on your finger

2026-08-29T08:51:29.867-07:00

Blood pressure is ordinarily measured with a sphygmomanometer, the familiar inflatable cuff placed around the upper arm. The cuff temporarily compresses the brachial artery, and the device determines the pressure at which blood begins (systolic pressure) and then resumes flowing normally (diastolic). A new generation of wearable blood-pressure devices employ a similar cuff-based approach but in a smaller form factor worn on the wrist (QH). Other wearable monitors adopt a less direct approach using photoplethysmography (PPG) to shine light into the skin and measure the small changes in reflected light produced as each heartbeat expands the arteries and local blood volume.

Several companies are experimenting with reducing the size of the wearable device from a wristband to a ring. One example is the Signal Ring, a new finger-worn blood-pressure monitor developed by the San Francisco startup Vital Signals (Figure 1). The ring is usually worn on the index finger and connects by Bluetooth to an iOS or Android phone. Unlike general-purpose wearable rings that emphasize sleep, activity and recovery (QH), Signal is being marketed primarily as a blood-pressure device. Vital Signals nevertheless describes it as a wellness product rather than a medical device, and says that its readings are informational and should not be used by themselves to diagnose or treat hypertension.

The hardware employs reflective PPG, the same optical technique used by many watches and fitness trackers to measure heart rate. Light-emitting diodes shine into the finger, and a nearby photodetector measures how much light returns. With each heartbeat, arteries in the finger fill and relax, changing the local blood volume and therefore the amount of light absorbed and reflected. The resulting waveform contains information not only about pulse rate, but also the pulse amplitude, rise time, contour, and timing. These properties reflect the interaction among cardiac output, the traveling pressure wave, and the elastic properties of the arterial wall. Counterintuitively, the finger is a sensible location for such a monitor because it provides a strong peripheral pulse signal. Vital Signals says that its sensor samples the pulse four to five times faster than the optical sensors in typical wearables, allowing it to detect details that lower-speed sensors miss.

A fundamental challenge is that these devices are measuring a change in blood volume rather than pressure itself. In simplified terms, pressure change equals volume change divided by arterial compliance, the ease with which the artery expands which depends on vascular tone and elasticity. Compliance varies with age, temperature, stress, posture, exercise and vascular disease. A stiffer or narrower artery can alter the speed and shape of the pulse in a way that resembles higher pressure, even when pressure itself has not changed. Vital Signals says that a proprietary physiological model measures these vascular changes directly from the high-resolution pulse waveform, and then separates the changes in pressure from changes in vascular compliance. In this manner, the PPG measurement of blood volume oscillations can be converted into pressure units.

This can be a tricky conversion because of all the variables and uncertainties involved, and as a result most cuffless devices require regular comparison with a conventional cuff for calibration purposes. According to Vital Signals, Signal Ring never requires calibration, not during setup and not later.

In terms of accuracy, the company has claimed the ring has met ISO 81060-2 requirements based on a 451-person, multisite clinical study. To meet the standard, the average difference from the reference method must be no more than 5 mmHg, with a standard deviation no more than 8 mmHg for both systolic and diastolic pressure. However, the company has not yet publicly released the study’s numerical results, participant characteristics, blood-pressure range, clinical sites, reference-measurement procedure or performance in different skin tones, ages, body positions and real-world activities. In addition, there is no peer-reviewed paper, public trial registration or independent replication of the study. As a result, the Signal Ring has not been reviewed or cleared by the FDA.

As a reality check, a published 2024 study of a different PPG blood pressure ring reported 526 systolic and 513 diastolic pairs from 89 people; average differences from sphygmomanometry, were 0.16 ± 5.90 mmHg systolic and −0.07 ± 4.68 mmHg diastolic, well within the ISO 81060-2 standards. However, the measurements were made under controlled conditions and presumably used calibration. This small study needs to be replicated to build confidence in the capabilities of the ring form factor to measure blood pressure.

For those interested, the Signal Ring is available for preorder from the company for $399, with no subscription fee. The price includes the ring, a charging case, and a sizing kit; sizes 5 through 13 are offered. Purchasers receive a sizing kit before the actual ring because proper fit is important for obtaining a reliable measurement. The product is initially being sold only in the United States. Sizing kits are expected to begin shipping at the end of July 2026 and initial rings are expected to ship in October 2026. The titanium-and-resin ring is rated IP68 (complete dust protection and continuous water immersion), requires a nearby phone for readings, and is advertised for up to five days of use with the charging case. The company advertises a one-year limited warranty and a 30-day return period. The company's own release says Signal Ring is investigational and has not been reviewed or cleared by the FDA as described above.

Signal occupies an interesting niche among wearable blood-pressure devices. The Omron HeartGuide (QH) and Huawei Watch D2 put a miniature inflatable cuff in the watch strap; they are bulkier and the measurement is intermittent, but their method remains close to conventional sphygmomanometry. Certain versions of Samsung's Galaxy Watch use wrist PPG to produce numerical readings (after three initial upper-arm cuff calibration measurements), and requires recalibration every 28 days. The optical Aktiia/Hilo bracelet also uses personal baseline and monthly calibration. In a manufacturer-involved one-month study of 86 adults, it reported differences of 0.46 ± 7.75 mmHg systolic and 0.39 ± 6.86 mmHg diastolic in accepted seated measurements (QH). Apple has taken a more conservative route; its FDA-cleared watch feature looks for patterns associated with hypertension over 30 days but does not display blood pressure values, and instead notifies users that they may have signs of chronic hypertension, which should be confirmed by a medical professional.

Signal Ring would be more convenient and informative than all of the above if it can supply reliable numbers and background trends without either an inflatable strap or calibration. It would be able to provide the most "continuous", "real-time" blood pressure monitoring among all the options. The ring takes background readings throughout the day (the time interval has not been specified by the company) which are combined into a full day blood pressure portrait. It also possesses a user-activated Realtime Tracking mode which takes measurements every 10 seconds.

ChatGPT summed up the Signal Ring as "a promising device with a missing results table." Signal Ring addresses a real limitation of blood pressure monitoring. A comfortable ring could collect information at times when few people would bother to put on a cuff, potentially revealing patterns that an occasional snapshot misses. Their technology which includes high-speed finger PPG plus a model that accounts for vascular tone is plausible, and the reported 451-person study could be convincing if more details are disclosed. The details matter in terms of how far individual readings differed from the reference, who was tested, which readings failed, and whether performance persisted across the conditions in which the ring will actually be worn. In other words, the missing results table needs to be filled in, and that would be the first step towards obtaining FDA authorization and moving from curiosity to health-informing device.

Figure 1. The mysterious and intriguing Signal Ring. How accurately can it measure blood pressure? Its ring form factor is very convenient (https://signalring.com/).

Lifestyle factors can impact cancer relapse

2026-08-22T08:36:06.921-07:00

Cancer relapse, also called cancer recurrence, means that cancer has returned after a period when it was reduced, controlled, or no longer detectable after treatment. The vast majority of tumors respond (i.e. are damaged by) in some fashion to treatment whether surgery, radiation, or drug treatment. This response can be partial, resulting in partial remission, or more complete, resulting in complete remission. The problem is that often the cancer comes back and starts to regrow and re-spread. It does not necessarily imply the original treatment was wrong (or totally ineffective); it usually means that some cancer cells survived, remained hidden at very low levels, and later began growing again.

Cancer recurrence is characterized by when and where the new tumor appears in relation to the original cancer. Early recurrence is if the cancer returns within months or the first few years after treatment concludes. It typically indicates a highly aggressive tumor biology or a pre-existing resistance to the treatment that was used. Late recurrence describes how some cancers (e.g. estrogen-receptor-positive breast cancers or certain melanomas) can return five, ten, or even twenty years after the initial treatment. This is characterized by prolonged cellular dormancy, in which the surviving cells effectively "hibernate" before a biological trigger may cause them to awaken.

In terms of location, recurrence may be local, meaning it returns near the original site; regional, meaning it appears in nearby lymph nodes or tissues; or distant, meaning it spreads to organs such as the liver, lungs, bones, or brain. Distant recurrence, or metastatic recurrence, is especially serious because it is often harder to treat (i.e. found in many more locations) and is a major driver of cancer-related death, although cancer mortality can also involve progressive disease, organ failure, infection, cachexia, blood clots, and treatment complications, and not just metastasis.

Cancers recur because treatment may eliminate all visible or detectable disease without destroying every cancer cell capable of growing again. Tiny numbers of surviving cancer cells, sometimes called minimal residual disease, may remain in the original area, hide in nearby tissues, or spread through the blood or lymphatic system. These cells can be too small to detect on CT, MRI, PET scans, or routine exams, and they may later grow into a recurrent tumor. Some surviving cells are resistant to treatment because tumors are genetically diverse: chemotherapy, radiation, targeted therapy, hormone therapy, or immunotherapy may kill the sensitive cells while leaving behind resistant ones. Other cells may enter a dormant state, dividing very slowly or not at all, which can make them less vulnerable to treatments that mainly attack rapidly dividing cells. Cancer cells can also survive by evading the immune system, adapting to the local tissue environment, or, in some tumors, persisting as stem-like cells that can regenerate the cancer.

The likelihood of recurrence varies widely by cancer type, stage (early or advanced), grade (e.g. high and rapidly growing histological grade), molecular features/subtype (e.g. specific genetic mutations), treatment response, surgical margins, and whether the cancer had already spread microscopically before treatment. Many early-stage cancers have a relatively low risk of recurrence, while some aggressive or advanced cancers -- such as glioblastoma, advanced ovarian cancer, pancreatic cancer, and high-risk bladder cancer -- recur much more often.

More specifically, glioblastoma, a highly infiltrative brain tumor, has a recurrence rate that is considered almost inevitable because its root-like structure makes it practically impossible to eliminate entirely with surgery and radiation. Similarly, cancers that are notoriously difficult to detect early, such as ovarian and pancreatic cancers, experience staggeringly high recurrence rates of 70% to 85% because they often spread on a microscopic level before the primary tumor is ever found or removed. Other cancers tend to recur locally within a damaged tissue field; low-grade, stage Ta non-muscle-invasive bladder cancer, for example, recurs in 10% to 35% of cases within the bladder lining (a much smaller percentage progress into the muscle), requiring lifelong surveillance.

How does one reduce the recurrence risk? First and foremost, get the best possible medical care, and adhere to that care. In addition, it is believed that lifestyle changes can modestly reduce the risk of cancer recurrence by improving the body’s metabolic, hormonal, inflammatory, and immune environment, although they cannot eliminate recurrence risk or replace recommended cancer treatment and surveillance. The most consistently recommended steps are to avoid smoking, limit or avoid alcohol, maintain a healthy body weight, stay physically active, reduce sedentary time, and eat a diet centered on vegetables, fruits, whole grains, beans, and other minimally processed foods. The World Cancer Research Fund advises cancer survivors, when medically able and unless otherwise instructed, to follow general cancer-prevention recommendations after the acute stage of treatment, including healthy weight, regular activity, and a healthier dietary pattern. In addition, the National Cancer Institute (NCI) notes that survivors remain at risk for recurrence and second cancers, and often need support for behaviors such as avoiding smoking and excessive alcohol use and increasing physical activity.

There is a large body of clinical studies that have provided evidence that targeted lifestyle changes can lower the risk of cancer recurrence, although the benefits depend heavily on the specific type of cancer and the intervention used. For example, one observational study of colorectal cancer (CRC) patients demonstrated that patients treated for Stage III colon cancer who engaged in regular physical activity -- specifically, the equivalent of six or more hours of brisk walking per week -- experienced a nearly 50% reduction in the risk of recurrence and overall mortality compared to their inactive peers. Another observational study (the Nurses’ Health Study) of breast cancer survivorship data also supported the importance of activity finding that women who were more physically active after a breast cancer diagnosis had lower risks of breast cancer death and recurrence, with the greatest benefit seen around the equivalent of walking 3 to 5 hours per week. Because these were observational and not randomized controlled trials (RCTs), there is the distinct possibility of confounding, including reverse causation, leading to a spurious association. For example, some may have been more physically active because their tumor was not recurring.

RCTs address these concerns, and several have shown that lifestyle changes can significantly reduce cancer recurrence or recurrence-related outcomes in specific patient groups. One of the strongest examples is the CHALLENGE trial, a randomized phase 3 study of 889 patients with stage III or high-risk stage II colon cancer who had completed surgery and adjuvant chemotherapy; patients assigned to a 3-year structured exercise program had significantly longer disease-free survival than those given health-education materials alone, with reports noting about 28% fewer recurrent or new cancers and 37% fewer deaths over follow-up (NEJM). In breast cancer, the randomized Women’s Intervention Nutrition Study (WINS) found that a low-fat dietary intervention after treatment for early-stage breast cancer was associated with reduced recurrence risk (24% lower relative risk of relapse compared to those receiving standard care).

However, the evidence has not been uniform for every lifestyle intervention; for example, the large WHEL randomized trial found that a very high vegetable, fruit, and fiber diet with lower fat intake did not significantly reduce additional breast cancer events or mortality, showing that lifestyle benefits depend on the cancer type, intervention, and patient population (JAMA Network).

Not surprisingly, the lifestyle modifications that lower risk of cancer relapse (secondary prevention), also lower the risk of cancer occurring in the first place (primary prevention, Figure 1). Whether you are trying to stop a first tumor from forming or a dormant cancer cell from waking up, the underlying biology is remarkably similar. Tobacco exposure, alcohol consumption, excess body fat, physical inactivity, and some dietary patterns can promote both the initial development of cancer and the later growth of residual cancer cells through related mechanisms, including DNA damage, inflammation, altered hormone and insulin levels, impaired immune function, and changes in the tissue environment. However, the evidence for preventing a first cancer is generally broader and stronger than the evidence for preventing recurrence, which remains highly dependent on the particular cancer, stage, treatment, and lifestyle intervention.

In summary, the likelihood of recurrence varies widely depending on specific tumor characteristics but one can minimize the probability of cancer recurrence through a comprehensive, multi-pronged approach that combines rigorous adherence to clinical treatments with targeted lifestyle modifications. Medically, the most critical step is completing all prescribed adjuvant therapies -- such as chemotherapy, radiation, or long-term hormone blockers -- to actively hunt down and eradicate any minimal residual disease, alongside attending strict surveillance appointments to monitor for early signs of relapse. On a lifestyle level, patients can significantly influence their body's internal environment, making it actively inhospitable to returning cancer cells. This is achieved by engaging in consistent, moderate-to-vigorous physical activity, maintaining a healthy body weight to reduce systemic inflammation and circulating hormones, adopting a healthy diet, and eliminating exposure to known carcinogens such as tobacco and excessive alcohol. Together, these clinical and behavioral strategies work synergistically to suppress the biological conditions that dormant cancer cells need to awaken and multiply.

Figure 1. Avoiding smoking, staying fit, and eating healthy are three lifestyle changes that can reduce the risk of cancer occurrence and also the risk of cancer recurrence (ACTC).

What concierge medicine does and doesn’t buy

2026-08-15T10:22:10.221-07:00

Concierge medicine is a healthcare model in which patients pay a membership fee for enhanced access to a physician, often including longer appointments, same-day scheduling, and more personalized care. The patient continues to rely on standard health insurance to cover those services not included in the membership fee. It is most commonly used for primary care, but can also be used by specialists -- such as cardiologists, dermatologists, psychiatrists, or endocrinologists -- for enhanced access, longer appointments, and care coordination.

In addition to the membership fee, a concierge practice may charge separately for self-pay "a la carte" services that are excluded from both the membership and your insurance, such as:

A routine comprehensive physical, which is different from Medicare’s covered Annual Wellness Visit
Elective, expanded, or unusually frequent laboratory tests and screenings that Medicare considers unnecessary or does not cover at that frequency
Optional wellness, cosmetic, massage, or similar noncovered services

Many concierge practices also bill insurance for care covered by that insurance, and so patients may still owe deductibles, copays, or coinsurance. This illustrates how concierge medicine is built on top of health insurance and is not intended to duplicate the latter services, which the concierge doctor may actually use and charge.

The primary appeal is that many people experience standard primary care as rushed, fragmented, and hard to schedule. Under concierge medicine, there are fewer patients per doctor, longer visits, same- or next-day appointments, phone access, and more help navigating specialists or tests. For example, a 2005 study found that concierge-style doctors reported roughly half as many patients as non-concierge doctors, and were also more likely to offer 24-hour direct physician access and house calls.

Prices vary widely, and in the U.S. traditional concierge primary care commonly costs about $2,000–$5,000 per adult per year. Some practices charge $10,000 or more, while highly exclusive or executive-health programs can reach $40,000+ annually (link).

Direct primary care (DPC) is related but not identical to concierge medicine. DPC usually means direct payment for a defined set of primary-care services, often replacing insurance billing for primary care; many patients still carry insurance for specialty care and hospitalization. In other words, direct primary care replaces insurance billing for most routine primary care, whereas concierge medicine adds perks on top of insurance-billed care. Both models generally promise easier access, smaller patient panels, longer visits, and a closer physician relationship.

Thus, the concierge membership fee is not the same as health insurance. For patients, the key distinction is between access and coverage. Concierge membership may make it easier to reach a primary-care doctor. It (typically) does not pay for emergency care, surgery, specialist treatment, hospital admission, advanced imaging, or expensive drugs. Doctors cannot use a concierge-based practice to get around health insurance (e.g. Medicare) payment limits, and tack on extra expenses (charged to insurance) for an insurance-billed office visit; these must be paid for by the patient.

So is concierge healthcare better? The evidence is mixed. A 2025 systematic review in the American Journal of Medicine concluded that evidence confirming concierge medicine's clinical-outcome benefits remains limited, while patient and physician satisfaction appear higher. A 2023 Journal of Health Economics article reported that when physicians switched to concierge medicine, there was stronger evidence of selection by income than by health, large spending increases, and no average mortality effect among affected patients (Figure 1). The latter study does not preclude the possibility of some sort of selection effect of doctors who eventually end up in concierge medicine, but it does argue that there is nothing intrinsically special about concierge medicine versus non-concierge medicine with the same doctor.

Yet there is indirect evidence that suggests that concierge medicine should improve medical care, at least in theory. First, access to primary care matters. In a large U.S. observational study from 2005 to 2015, every 10 additional primary care physicians per 100,000 people was associated with 51.5 additional days of life expectancy and 0.9% to 1.4% lower cardiovascular, cancer, and respiratory mortality.

Second, continuity also appears important. A 2018 systematic review found that 18 of 22 observational studies reported statistically significant lower mortality with greater continuity of doctor care; 16 involved all-cause mortality. In this review, the authors define continuity of care as: "Repeated contact between an individual patient and a doctor." In plain language, this means that the same patient sees the same physician on multiple occasions over time, rather than receiving each episode of care from a different doctor. Continuity is one of the plausible mechanisms by which a smaller, more accessible practice might help some patients.

Finally, concierge practices often emphasize prevention which is a cornerstone of modern preventive medicine. However, the devil lies in the details. Screening can include physical exams, laboratory tests, imaging procedures, and genetic tests, but the National Cancer Institute notes that screening tests have risks: false positives can cause anxiety and lead to more tests and procedures, some cancers found by screening may never have caused symptoms or threatened life, and finding cancer early does not always help a person live longer (link). Good preventive care is not the largest possible test menu. It is the right test, for the right person, at the right interval, with a plan for what to do with the result. Whether or not concierge medicine can do this better than standard medicine is an open question.

In summary, concierge medicine may be worth paying for when the product is clearly defined: faster appointments, longer visits, reliable continuity, and help navigating care. It is most defensible when the practice still follows evidence-based medicine, explains billing clearly, avoids overselling unproven testing, and helps patients understand what the membership does not cover. It is more questionable when the pitch depends on vague "optimization," broad screening packages with uncertain net benefit, difficult cancellation terms, unclear insurance billing, or implied promises of better health outcomes that the evidence does not support.

Put simply, concierge medicine buys physician access, it doesn't buy additional healthcare insurance coverage, and the verdict is still out on whether it improves overall health. Finally, an important consideration is whether you can pay the concierge fee.

Figure 1. Kaplan-Meier survival curves for patients whose physician switches to concierge medicine (solid line) and their matched controls (dotted line). The curves denote the fraction of patients who are alive as a function of the number of years since the physician’s switch to concierge medicine. For the control patients, their physician will switch to concierge medicine at some point after the investigational study period, and so the plot represents years when the doctor was not yet practicing concierge medicine (Leive et al. Journal of Health Economics, 2023).

Destructive MAHA crusade against psychiatric drugs

2026-08-08T10:14:40.731-07:00

Robert F. Kennedy Jr., Secretary of U.S. Department of Health and Human Services (HHS), has made a campaign against psychiatric drugs a federal policy issue and an integral part of his Make America Healthy Again (MAHA) agenda. A February 2025 executive order created the MAHA Commission which was directed to assess the "prevalence of and threat posed by" prescriptions including SSRIs, antipsychotics, mood stabilizers, stimulants, and weight-loss drugs, with an emphasis on childhood chronic disease. Then in May 2025, the commission issued an assessment that framed prescription-drug use and "overmedicalization" as possible contributors to poor child health.

One issue (among many) is that "psychiatric drugs" is a broad clinical category. Antidepressants, stimulants, antipsychotics, mood stabilizers, and sedatives have different indications, evidence, risks, and stopping problems. A useful policy conversation has to separate them. For example, SSRIs, or selective serotonin reuptake inhibitors, are a major class of antidepressants that includes fluoxetine (Prozac), sertraline (Zoloft), paroxetine (Paxil), fluvoxamine (Luvox), citalopram (Celexa), and escitalopram (Lexapro). They are commonly prescribed for depression and several other psychiatric conditions, and their name comes from their main pharmacologic action: inhibiting the serotonin transporter so that more serotonin remains available at neural synapses. Most importantly, there is abundant evidence that SSRIs are effective treatments for major depressive disorder.

A large meta-analysis of 522 randomized trials found that all 21 studied antidepressants were more effective than placebo for acute major depressive disorder in adults, with efficacy odds ratios ranging from 1.37 to 2.13 across drugs (Figure 1), although certainty and tolerability varied. In continuation trials, people who stayed on antidepressants after response had lower relapse rates than those switched to placebo; a Lancet review reported average relapse rates of 18% on continued medication versus 41% on placebo.

On the other side, The New York Times reported that Kennedy had linked SSRIs to mass shootings without evidence and claimed SSRIs can be "harder to quit than heroin". The kernel of truth is that antidepressant withdrawal can be real and clinically significant. The nonsensical overreach is treating that as clinically equivalent to opioid withdrawal, or using it to imply that psychiatric treatment is generally suspect.

Interestingly, the response of the medical community to this attack has been relatively muted. Many psychiatrists agree that deprescribing has been neglected. Starting a medication is often easier than stopping one: prescribing workflows are routine, while slow tapering, small-dose formulations, close follow-up, and reimbursement are less developed. At the same time, psychiatric organizations and clinicians have rightfully objected to broad claims that psychiatric drugs are a main cause of America's mental-health crisis, warning that a blunt reduction agenda could harm patients who need treatment. A prescription count by itself cannot tell whether a person is overtreated, undertreated, or finally receiving care after years without it.

What is the evidence that there is an epidemic of overprescribing psychiatric drugs? In the case of antidepressants, use is common. CDC/NCHS data from 2015-2018 found that 13.2% of U.S. adults had used an antidepressant in the prior 30 days, including 17.7% of women and 8.4% of men. In addition, a 2025 50-state survey published in BMJ Mental Health found that 16.6% of 30,115 U.S. adult respondents reported current antidepressant use. However there is a simple explanation, namely that depression and anxiety are quite common and are being more accurately diagnosed, which is then coupled to the fact mentioned above that antidepressant medications have convincingly demonstrated efficacy treating depression.

One concrete case in which deprescribing may be appropriate is antipsychotic use in dementia for which treatment often continues beyond guideline-recommended durations. A 2026 target-trial emulation in older adults with dementia reported that tapering antipsychotics after 12 weeks, compared with continuation, was associated with 2.46 percentage points lower delirium risk and 2.80 percentage points lower fracture risk over 24 months, without higher mortality, stroke, or pneumonia (link). One can imagine that dementia can trigger (or resemble) a transient psychotic episode that requires an antipsychotic for a defined period of time rather than indefinitely. Importantly these findings do not generalize to schizophrenia, bipolar I disorder, or psychotic depression, where antipsychotics can be central treatment.

The challenge of deprescribing from a psychiatric drug should not be trivialized. The bottom line is that weaning from psychiatric drugs is hard; withdrawal is physiological, not a failure of character. People can have dizziness, insomnia, anxiety, irritability, sensory symptoms, low mood, agitation, or flu-like symptoms after reducing or stopping antidepressants. Moreover, risk is not evenly distributed. A 2023 review by Horowitz and colleagues concluded (albeit from a limited evidence base) that withdrawal risk may be higher after longer treatment duration, higher doses, prior withdrawal difficulty, and with certain drugs such as paroxetine and SNRIs. Guidelines increasingly recognize that tapering may need to be slow. The UK National Institute for Health and Care Excellence (NICE) recommends considering the medication's half-life and duration of use, reducing dose stepwise, using smaller reductions at lower doses, and letting the pace be guided by patient tolerability.

Instead of a crusade against all psychiatric drugs, the goal should be to achieve parsimonious care: the fewest medications that reliably help, combined with non-drug care when useful, reviewed regularly. There should be attention to detail with respect to diagnosis, context, and treatment. It is important to carefully characterize symptom severity, impairment, medical causes, sleep, substance use, trauma, current medications, coexisting conditions, etc. A medication should have a target, such as depression severity, panic frequency, manic relapse prevention, psychosis, attention impairment, insomnia, or obsessive-compulsive symptoms, and treatment should be measured and compared quantitatively to baseline symptom scales to make sure the patient is actually better after treatment. Side effects should be tracked too including sexual dysfunction, metabolic effects, sedation, emotional blunting, movement symptoms, appetite, drug interactions, etc.

In addition, non-drug treatments are valuable. Psychotherapy, behavioral activation, family or school supports, exercise, sleep treatment, substance-use treatment, and social interventions can all help (e.g. link). But "try therapy instead" is a useless slogan if therapy is unaffordable, unavailable, or ineffective.

In summary, we want better outcomes, not fewer prescriptions. The goal isn’t to reduce dependence on anti-depressants; the goal is to improve care for those who are depressed. For some that may mean weaning from anti-depressant drugs in which the costs (i.e. side effects) outweigh the benefits. For others that may mean a clearer path to medication when medication is the right tool. There is a need for more depresecribing research and education such as better informed consent, clinician training, reimbursement for careful tapering, smaller-dose formulations, better access to psychotherapy, etc. Finally and most importantly, a blunt campaign against all psychiatric medication runs counter to this goal of improved care and substitutes a political agenda for scientific and medical evidence such as Figure 1.

Figure 1. Comparison of 21 antidepressants to placebo from meta-analysis of efficacy trials. In total there were 432 double-blind randomized trials involving 102,443 adults with major depressive disorder. "Response" is defined as at least a 50% reduction in depression-rating-scale symptoms, assessed as close to eight weeks as possible. The odds ratio (OR) was calculated as the odds of response with drug divided by odds of response on placebo. For each drug the OR is shown along x-axis with an odds ratio of 1.0 meaning no difference from placebo, and an OR > 1 favoring the drug treatment. Error bars indicate the 95% credible interval (Crl = Bayesian equivalent of 95% confidence interval). Figure is reproduced from Cipriani et al. The Lancet, 2018.

Comparing general-purpose to specialized clinical chatbots

2026-08-01T08:53:33.939-07:00

OpenEvidence and UpToDate Expert AI are two specialized clinical generative AI tools intended to help healthcare professionals retrieve and interpret medical information at the point of care. Clinicians can enter a medical question or patient scenario and receive a response addressing tasks such as reviewing the medical literature and possible diagnoses, comparing treatments, confirming management plans, identifying relevant clinical guidelines, and clarifying medical concepts. OpenEvidence functions broadly as an AI-powered medical search and evidence-synthesis platform, drawing on published research, major journals and clinical-practice guidelines to present supporting references with its answers. UpToDate Expert AI adds a conversational interface to the established UpToDate clinical decision-support service; it answers questions and supports follow-up dialogue using UpToDate’s physician-authored, editorially reviewed content, with links back to the underlying topics. Both tools therefore replace time-consuming manual search (i.e. Google search) and reading of multiple sources with rapid, clinically oriented summaries.

Meanwhile, general-purpose chatbots such as ChatGPT, Gemini and Claude (also known as frontier or foundation models) are increasingly used in healthcare as flexible assistants for both clinical and administrative work (QH). Clinicians may use them to summarize medical literature, answer clinical questions, generate differential diagnoses for consideration, explain guidelines, draft notes and referral letters, prepare patient instructions, translate or simplify medical information and assist with routine documentation; patients also use them for health education, symptom-related questions and navigating healthcare services (QH). These systems are intended to support -- not replace -- professional judgment, because their answers may be incomplete, outdated or confidently incorrect.

The expertise of these general-purpose chatbots in the medical domain are evaluated through several complementary approaches: (1) medical knowledge tests such as licensing-examination questions (e.g. multiple choice questions from the United States Medical Licensing Examination (USMLE); (2) open-ended benchmarks that assess agreement with expert clinical expectations; and (3) blinded clinician review of responses to realistic patient cases or questions arising in practice. Evaluation typically considers clinical correctness, completeness, clarity, appropriate handling of uncertainty, safety, hallucination and refusal rates, while real-world assessments may also examine citation quality, response speed, bias, privacy, usability and effects on clinical workflow.

A reasonable question is whether the specialized AI tools are using in some fashion or have been trained by (distilled) the general-purpose chatbots. We don't know the answer because the training and inference pipelines for many of the specialized clinical AI tools, including OpenEvidence and UpToDate Expert AI, have not been publicly disclosed. Thus, it is unclear how the large language models (LLMs) underlying the tools undergo medical-domain pretraining, fine-tuning or clinician-guided alignment. However, most likely these tools make extensive use of retrieval-augmented generation (RAG), primarily at inference (response) time: when a clinician submits a question, the system searches a curated medical knowledge base (composed of research literature, guidelines, expert-authored clinical content, etc.) selects relevant passages and supplies them to an LLM as context for generating an answer. Briefly, RAG augments the capability of an LLM by identifying additional information (e.g. from a curated database) relevant to a user’s query, and then adding that information to the prompt so that the LLM generates an answer that is more accurate, current, domain-specific, and traceable to source material.

Of course the training of general-purpose chatbots is equally shrouded in mystery. What is known is that their underlying capabilities are built upon massive, highly diverse training corpora that expose them to broad, cross-domain knowledge. These frontier large language models benefit from rapid development cycles, significant financial and computational investment, and rigorous post-training alignment. Because they develop strong general reasoning and sophisticated language-generation abilities from this expansive training approach, their foundational skills transfer effectively into the medical field. As a result, these general-purpose models are able to evaluate complex healthcare scenarios and even outperform purpose-built clinical tools without requiring curated medical references or dedicated domain-specific modifications. The frontier (general-purpose) models also use RAG, but are less dependent on it to produce accurate and topical answers because of extensive reasoning capabilities along with vast parametric memory (i.e. information stored in model weights during training).

There is substantial overlap in the clinical tasks performed by both the specialized AI tools and the general-purpose chatbots such as summarizing medical literature and answering clinical questions supported by references, leading one to ask: "Which is better?" In addition, as the specialized tools have grown in popularity in the medical community, questions have arisen as to the accuracy and clinical effectiveness of the tools. A new paper from researchers at NYU attempts to address these questions and determine whether clinical tools actually provide added value over general-purpose LLMs -- not only on medical examination questions, but also in their agreement with clinician expectations and their responses to questions arising in routine practice.

The researchers employed a three-part evaluation process combining public benchmarks with real-world clinician queries. Three general-purpose frontier large language models (GPT-5.2, Gemini 3.1 Pro and Claude Opus 4.6) were compared to the two specialized clinical AI tools, OpenEvidence and UpToDate Expert AI. For the first comparison, 500 USMLE-style questions from the MedQA benchmark were used to assess general medical knowledge. In the second part, 500 open-ended HealthBench prompts (from realistic medical conversations) were evaluated for response accuracy, completeness, communication, context awareness and instruction following. The HealthBench scoring was performed by a panel of the three frontier models (i.e. GPT, Gemini, and Claude). Finally, to assess performance in routine practice, the investigators also constructed a novel Real Clinical Queries (RCQ) benchmark from 100 de-identified questions submitted by physicians; these questions were answered by all five principal systems plus Google AI Overview. Twelve clinicians, blinded to model identity, independently rated each response (on a 1-4 scale) for clinical correctness, completeness, safety and clarity and flagged harmful content, hallucinations and refusals. All models were allowed to perform internet searches to acquire information for RAG.

Quite surprisingly, across all three evaluations, the general-purpose models outperformed the specialized clinical AI tools (Figure 1). On the 500-question MedQA benchmark, Gemini 3.1 Pro achieved the highest accuracy at 97.4%, followed by GPT-5.2 at 94.2%, while OpenEvidence and UpToDate Expert AI scored 89.6% and 88.4%, respectively. On HealthBench, GPT-5.2 led with a score of 88.0, compared with 62.6 for OpenEvidence and 61.3 for UpToDate Expert AI. The same pattern appeared on 100 real clinical queries evaluated by blinded physicians: Gemini, GPT and Claude formed a statistically indistinguishable top tier, with mean aggregate scores of 3.62, 3.54 and 3.52 on the four-point scale, whereas OpenEvidence, UpToDate Expert AI and Google AI Overview formed a lower tier at 3.24, 3.17 and 3.27. The frontier models performed better across clinical correctness, completeness, safety and clarity, with the largest differences observed in clarity; UpToDate Expert AI also had a notably high refusal rate of 19%. However, harmful-response and hallucination rates were low and did not differ significantly among the systems, suggesting that the principal advantage of the frontier models was in the overall quality, completeness and communication of their answers rather than a measurable reduction in overtly unsafe or fabricated content.

The study concludes that leading general-purpose frontier LLMs currently outperform the specialized clinical AI tools OpenEvidence and UpToDate Expert AI across medical knowledge, alignment with clinical expectations, and responses to real-world physician questions. The authors suggest that the large training corpora, extensive alignment, rapid development cycles and strong cross-domain reasoning of frontier models may, for many medical tasks, outweigh the benefits of domain-specific tuning or curated retrieval. However, they caution that the findings represent a snapshot of a rapidly evolving field and do not establish that general-purpose models will be superior for every clinical application, particularly deeply specialized or institution-specific tasks.

As one might expect, neither OpenEvidence and Wolters Kluwer, the developer of UpToDate Expert AI, were pleased with the findings. OpenEvidence in particular published a forceful rebuttal listing numerous potential shortcomings in their view with the study:

First, OpenEvidence argued that frontier models may have encountered MedQA and HealthBench questions or closely related material during training. Because general-purpose models are trained on enormous web corpora, a high score could partly reflect benchmark familiarity rather than superior clinical reasoning. In other words, there could be data leakage (contamination) from the very large training dataset to the test dataset with very similar questions appearing in both allowing the model to essentially memorize the answer during training to closely related test questions.

Second, they pointed out that the HealthBench benchmark was created by OpenAI, and this benchmark may favor the general-purpose chatbots generally, and ChatGPT (from OpenAI) in particular. For example, the benchmark may reward subjective response characteristics, including formatting, thoroughness and adherence to detailed stylistic rubrics that align with how general-purpose LLMs are trained. In addition, it should be noted that the frontier models (ChatGPT, Gemini, and Claude) served as the graders for HealthBench (during grading they were given the answers which were not available during testing), and their grading could be influenced by familiarity with the linguistic or formatting style, rather than content of their own answers.

Third, the OpenEvidence team raised questions about the lack of transparency in the Real Clinical Queries (RCQ) benchmark. More specifically, OpenEvidence objected that the RCQ questions were not made public, and that the paper provides limited information about their clinical specialties, complexity, intended use, selection criteria and reviewer recruitment. This makes it impossible for outsiders to reproduce the evaluation, and characterize the errors made by each system.

Fourth, the evaluation may not match the products’ intended purpose. OpenEvidence argued that it is primarily an evidence-search, synthesis and citation tool, not a general chatbot optimized for medical-exam questions or lengthy free-form responses. Its claimed value includes finding recent literature, grounding answers in sources and helping clinicians verify evidence. A benchmark emphasizing multiple-choice accuracy, stylistic completeness and general conversational quality may therefore favor frontier chatbots while overlooking citation accuracy, provenance, speed and evidence retrieval.

Fifth, benchmark scores do not establish clinical impact. The proper endpoint is not whether a tool produces the most polished answer in isolation, but whether it improves actual clinical decisions, saves time, reduces errors or benefits patients when incorporated into practice (i.e. practical clinical impact).

The Wolters Kluwer response echoed many of the above criticisms, but in particular emphasized the last point. Their position is that strong performance on exams or response-rating benchmarks does not automatically establish that a system is dependable at the point of care. Licensing-style questions are often too generic, user ratings can reward a convincing answer that omits an essential warning, and narrow case challenges may not test ambiguous questions, malformed prompts, or rare but dangerous failure modes. A model can therefore achieve a high average score while still failing in clinically consequential situations.

Given the technical nature of the dispute, it is unlikely that this brouhaha will be resolved in the near future. Regardless, it is imperative that both general-purpose and specialized clinical chatbots are subjected to rigorous medical evaluation especially after they are widely adopted. The paper from the NYU group provides a first step, and its results should be reproduced and extended. One can even argue that ultimately medical professionals can take advantage of both types of chatbots. If on a particular question, the different chatbots agree, then that provides some degree of confirmation, and if they disagree then further investigation is required to resolve the disagreement and find the best answer to the question.

Figure 1. Evaluation of general-purpose versus specialized clinical chatbots across three benchmarks. The general-purpose models were Gemini 3.1 Pro, GPT-5.2, Claude Opus 4.6, and Google AI Overview; the specialized clinical models were OpenEvidence and UpToDate Expert AI. a) MedQA accuracy (n = 500 questions per model), b) HealthBench score (n = 500 questions per model), and c) RCQ mean aggregate clinician rating (1–4-point scale) for 100 clinical questions. Mean (± 95% CI) percent accuracy is shown on y-axis for the MedQA and HealthBench benchmarks; the mean aggregate score on 4 point scale is shown for the RCQ benchmark (reproduced from Fig. 2 of Vishwanath et al. Nature Medicine, 2026).

An unprecedented Cyclospora outbreak

2026-07-25T16:48:30.878-07:00

In the U.S. we are currently experiencing the largest wave of cyclosporiasis, the intestinal illness caused by the parasite Cyclospora, in history. As of July 20, 2026, the CDC had recorded 4,173 laboratory-confirmed infections acquired in the United States, including 308 hospitalizations and no deaths across 41 states, but according to unofficial tallies from the various states, the number of cases may be approaching 16,000. In the past, we averaged roughly 3000 cases per year.

Most likely there are multiple outbreaks with multiple independent sources, but a clear source has emerged for a five-state subset in Indiana, Kentucky, Michigan, Ohio, and West Virginia. In that outbreak, more than 1,644 cases and 94 hospitalizations have been linked to shredded iceberg lettuce served at Taco Bell restaurants. The apparent culprit was an agricultural company called Taylor Farms which recalled iceberg lettuce sourced from central Mexico, including certain Marketside shredded lettuce and iceberg salad packages sold at Walmart, as well as products sent to Taco Bell and other food-service customers.

Cyclospora cayetanensis is a single-celled parasite that causes cyclosporiasis. It is a protozoan (single-cell eukaryote) -- not a bacterium, virus, or parasitic worm -- and it reproduces inside cells of the small intestine. It cannot replicate, complete its life cycle, or synthesize essential resources outside of a host organism (i.e. obligate intracellular parasite). Specifically, the epithelial layer of the human small intestine are the host cells, and replication of the cyclospora can kill or damage these cells (Figure 1). The degradation of the intestinal lining decreases reabsorption of water and diminishes production of key digestive enzymes, which results in the characteristic watery diarrhea, as well as gas from fermentation of undigested nutrients by the intestinal bacterial flora.

During the infection, the parasite produces environmentally resistant structures called oocysts that are passed in the stool (Figure 1). The oocyst is the parasite's "survival pod" that allows it to leave a human host, survive outdoors in water or soil, and infect the next person who ingests it. After 1 to 2 weeks outside the body under warm, humid conditions, the immature oocyst undergoes sporulation giving rise to a mature oocyst that contains 4 infectious motile sporozoites. When ingested, gastric acid and digestive enzymes in the human digestive tract break open the mature oocyst outer wall, releasing the sporozoites to invade human intestinal cells.

The characteristic symptom of cyclosporiasis is watery diarrhea with frequent and sometimes explosive bowel movements, usually beginning about one week after the contaminated food was eaten, although the incubation period can range from two days to two weeks or longer. Other symptoms include loss of appetite, weight loss, abdominal cramps, bloating, gas, nausea, fatigue, and occasionally vomiting, body aches, headache, or a low-grade fever. Without treatment, symptoms may last for a month or longer (up to 4-8 weeks) and can disappear temporarily before returning. The relatively long latency between exposure (i.e. eating the contaminated food) and symptom appearance is due to the time for the Cyclospora to break out of the oocyst, invade human intestinal cells, and then start replicating. The persistence arises because Cyclospora is an intracellular parasite that can hide out from the humoral immune system so that cytotoxic T-cells are required to fully clear out the infection.

Cyclosporiasis is usually not life-threatening, but prolonged diarrhea can cause substantial dehydration and weight loss; uncommon complications include malabsorption, gallbladder inflammation, and reactive arthritis. Nevertheless, "not usually life-threatening" does not mean harmless. Prolonged diarrhea can cause dehydration and lead to hospital care, while immunocompromised people, older adults, children, and people in poor health may experience a longer or more severe illness. As mentioned above, what makes Cyclospora especially unpleasant is its persistence: without treatment, symptoms may last from a few days to more than a month, improve, and then return. In the current outbreak, roughly 5-7% of reported cases were hospitalized with no deaths.

Transmission occurs via an indirect fecal-food-oral route with human feces containing Cyclospora oocysts from infected person(s) contaminating water, soil, or a food-handling environment (Figure 1). The oocysts mature (sporulate) outside the body in a warm and humid climate, and someone later swallows them in contaminated food or water. Because the oocyst maturation step takes at least one to two weeks (outside of the body), direct transmission from one person to another is considered unlikely.

Because the parasite is believed to infect only humans, contamination ultimately points to human fecal material rather than animal fecal material. Widespread human fecal contamination in agricultural areas can eventually place oocysts on fruits and vegetables. Fresh produce is an effective vehicle because it is often consumed raw and therefore does not undergo a cooking step that would kill the parasite. Common chlorine-based antimicrobial treatments are also not reliably effective against Cyclospora (CDC).

It is interesting how one of the main outbreaks was traced back to Taco Bell, and then ultimately Taylor Farms. The investigation was a two-part exercise in epidemiology and supply-chain detective work. First, state health departments identified people with cyclosporiasis and asked what and where they had eaten during the roughly two weeks before they became ill. Investigators noticed a cluster of people who had eaten at Taco Bell. Second, FDA investigators traced lettuce shipments backward from restaurants where sick people had eaten. Those distribution paths converged on a single supplier, Taylor Farms de Mexico, prompting removal and recall of iceberg lettuce sourced from central Mexico (FDA). There is an important caveat: so far the FDA has no confirmed positive product sample. Thus, the link to the lettuce is supported by strong food-exposure and traceback evidence, but not yet by direct laboratory detection of Cyclospora in a recalled package.

One should never eat the recalled produce. Return or throw it away, and wash containers, utensils, refrigerator surfaces, or counters that touched it with hot soapy water or in a dishwasher. Washing or rinsing the produce, which has a high concentration of Cyclospora, is not sufficient because the oocyst is so hardy. Routine chemical sanitizers are unlikely to kill it. Cooking produce to at least 158°F (70°C) for a prolonged period of time is necessary to kill the parasite. During an active outbreak, monitoring the CDC and FDA recall pages is more useful than avoiding all fresh produce,

Anyone who develops persistent watery diarrhea -- particularly after eating suspect (i.e. recalled) fresh produce -- should contact a health care provider and mention the possible Cyclospora exposure. This is important because routine stool examinations and gastrointestinal test panels do not always include Cyclospora; the clinician may need to request a specific test, and several stool samples collected on different days. The standard treatment is a 7- to 10-day course of the antibiotic combination trimethoprim-sulfamethoxazole (TMP-SMX), commonly sold as Bactrim or Septra, and this treatment is quite effective (link). The drug works by blocking the synthesis of folic acid in bacteria or protozoa; folic acid is an essential building block of DNA bases and hence is required for DNA synthesis in these organisms. In humans folic acid is provided by food and is not synthesized.

Most otherwise healthy people eventually recover without medication, but the untreated illness can be prolonged and relapsing. Drinking enough fluid is also important, and severe diarrhea, inability to stay hydrated, or other evidence of significant dehydration warrants prompt medical evaluation. The practical bottom line is that Cyclospora is rarely deadly but can make a person sick for weeks, and both diagnosis and treatment are easier when the clinician knows to look for this particular parasite.

One final comment is that because a large amount of Cyclospora oocyst infested human fecal material is needed to produce an outbreak, the contamination most likely arises from a systemic issue such as an inadequate sewage system in rural areas rather than from direct deposition by infected individuals. This is more likely in countries possessing less advanced sewage removal and treatment compared to the U.S. Indeed, most recent Cyclospora outbreaks have arisen from farm produce originating from outside the U.S. (link).

Figure 1. Lifecycle of Cyclospora cayetanensis (from CDC). (1) Excretion of immature (unsporulated) oocysts in human stool which is not infective (thus, direct fecal-oral transmission cannot occur). (2) Stool makes it into the environment perhaps through inadequate disposal or treatment of human feces. (3) In a warm (22°C to 32°C) and humid environment, the oocysts sporulate after days or weeks into the mature form each containing 4 sporozoites. (4) The sporulated oocysts can contaminate fresh produce perhaps via the agricultural water supply. (5) The vegetables or fruits are eaten fresh so that the live oocysts are ingested. (6) Digestive enzymes degrade the oocyst outer layers freeing the sporozoites to invade the intestinal epithelial cells. (7) Asexual reproduction lead to formation of merozoites, a form specialized to replicate and then invade neighboring cells. There is also sexual reproduction to create immature (unsporulated) oocysts. These must exit the body in the stool before they can sporulate. Eventually the body immune system can bring the outbreak under control. Treatment with trimethoprim-sulfamethoxazole can greatly slow replication.

The worrisome new Ebola virus outbreak in DR Congo

2026-07-18T08:44:22.954-07:00

Ebola virus disease is a severe, often fatal illness caused by viruses in the orthoebolavirus group (filamentous negative single strand RNA viruses), which are believed to circulate naturally in fruit bats in parts of Central and West Africa. The disease was first recognized in 1976 near the Ebola River in what is now the Democratic Republic of Congo (DR Congo formerly Zaire). Human infections are caused mainly by four species (Zaire, Sudan, Bundibugyo, and Taï Forest viruses) from the six total in the orthoebolavirus genus. Outbreaks usually begin when the virus spills over from an infected animal into a person. Ebola then spreads through direct contact with the blood or other bodily fluids (e.g. vomit, diarrhea, etc.) of a symptomatic patient, contaminated objects, or the body of someone who has died from the disease; it is not transmitted as easily through the air as respiratory viruses such as influenza or SARS-CoV-2.

Symptoms commonly begin with fever, fatigue, muscle pain, headache, and sore throat, followed in some patients by vomiting, diarrhea, bleeding, shock, and multiorgan failure. Fatality rates vary by virus species, outbreak conditions, and access to care, but early diagnosis, intensive fluid and electrolyte replacement, oxygen and blood-pressure support, infection control, contact tracing, safe burials, and, for some virus species, vaccines and monoclonal-antibody treatments can substantially reduce illness and transmission.

Upon entering the body, the Ebola virus systematically attacks the immune system, primarily infecting macrophages and dendritic cells. The rapid viral replication within these cells triggers a massive, uncontrolled release of pro-inflammatory proteins, culminating in a systemic "cytokine storm." This extreme immune response severely damages the endothelial cells that line blood vessels, destroying their structural integrity causing extreme vascular permeability as well as disruption of the coagulation system. As blood vessels leak, patients suffer profound fluid loss, which manifests clinically through paroxysms of severe vomiting and diarrhea, and frequently, uncontrollable hemorrhagic bleeding from the nose, mouth, and other mucous membranes. This catastrophic depletion of intravascular fluid leads to a drastic drop in blood pressure. Without immediate, intensive supportive care to maintain hydration, oxygenation, and hemodynamic stability (blood circulation), the patient rapidly descends into hypovolemic (low blood volume) shock and multi-organ failure, ultimately leading to death.

Ebola virus disease was first identified in 1976 during two nearly simultaneous outbreaks: one caused by Zaire ebolavirus in what was then Zaire (now the Democratic Republic of Congo) which resulted in 280 deaths among 318 reported cases, and another caused by Sudan ebolavirus in Sudan, which caused 151 deaths among 284 cases. By far the largest epidemic historically occurred in West Africa from 2014 to 2016, when Zaire ebolavirus spread extensively through Guinea, Liberia, and Sierra Leone, producing more than 28,000 reported cases and approximately 11,310 deaths. The next-deadliest outbreak took place in eastern Democratic Republic of Congo from 2018 to 2020, causing 3,470 reported cases and 2,287 deaths despite the availability of vaccination and newer treatments.

There is a new severe Ebola outbreak caused by the rare Bundibugyo virus currently spreading across East Africa, primarily impacting the Democratic Republic of Congo (DR Congo) and neighboring Uganda (Figure 1). Declared a Public Health Emergency of International Concern by the World Health Organization, the crisis quickly became the third-largest Ebola outbreak in recorded history shortly after its discovery in mid-May. While exact figures vary across reports due to the difficulties of tracking the disease in an active conflict zone, the outbreak has resulted in at least 2000 confirmed infections and more than 800 deaths. The situation is particularly dangerous because, unlike the strains responsible for recent major epidemics, there are currently no proven vaccines or targeted treatments for the Bundibugyo virus, leaving medical teams to rely entirely on supportive care while racing to launch experimental clinical trials.

The Bundibugyo Ebola virus is not a newly mutated strain, but rather a distinct, previously known species of ebolavirus that was first identified in 2007 during an outbreak in the Bundibugyo District of western Uganda. Like the other (ortho)ebolaviruses, Bundibugyo is zoonotic, meaning it originates from an animal reservoir and likely spills over into the human population through contact with infected wildlife, such as fruit bats or nonhuman primates. Although it causes a clinically similar disease and spreads in the same way (i.e. direct contact with the blood or other bodily fluids of symptomatic patients), it differs genetically from Zaire, Sudan, and Taï Forest ebolaviruses, particularly in the surface glycoprotein targeted by vaccines and monoclonal antibodies. As a result, the highly effective vaccine and antibody treatments developed for Zaire Ebola cannot be assumed to protect against Bundibugyo infection.

Importantly, the Bundibugyo Ebola virus is not considered intrinsically more deadly than the Ebola viruses responsible for the worst previous outbreaks. In fact, it has a lower historical mortality rate compared to the Zaire ebolavirus, which caused the massive 2014–2016 West African epidemic. While untreated Zaire Ebola has sometimes produced fatality rates approaching 90%, the Bundibugyo strain has historically killed about 25 to 40% of recognized patients. The extreme danger of the current crisis stems from its context and scale rather than an increased biological lethality.

More specifically, the recent Bundibugyo Ebola outbreak has been exceptionally difficult to combat because it combines a poorly studied virus with severe medical, logistical, and social obstacles. Unlike Zaire ebolavirus, Bundibugyo virus has no proven, widely available strain-specific vaccine or treatment, forcing clinicians to depend largely on intensive supportive care while experimental therapies are evaluated. The outbreak was already unusually large when detected, suggesting that the virus had circulated unnoticed and established multiple transmission chains before the response began. Containment is further complicated by armed conflict in eastern DR Congo, widespread displacement, porous borders, overstretched hospitals, inadequate laboratory and transportation infrastructure, and shortages of trained personnel and supplies. Effective control also requires identifying and isolating every patient, monitoring contacts for 21 days, and conducting safe burials, but community mistrust and resistance to treatment centers or burial restrictions can undermine these measures. These factors could allow the outbreak to expand into one of the largest and deadliest Ebola epidemics on record.

Zaire ebolavirus is the only Ebola virus species for which both proven virus-specific treatments and a highly effective licensed vaccine are widely established. In addition to early intensive supportive care -- including fluid and electrolyte replacement, oxygen and blood-pressure support, and management of shock and organ failure -- patients can receive either Inmazeb, a combination of three monoclonal antibodies, or Ebanga, a single monoclonal antibody; both bind the viral surface glycoprotein and prevent the virus from entering cells. In the randomized PALM trial, mortality was approximately 33.5 percent with Inmazeb and 35.1 percent with Ebanga, compared with roughly 50 percent or more among patients receiving the less effective treatments ZMapp or remdesivir, demonstrating a substantial but not absolute survival benefit. Prevention is centered on Ervebo (rVSV-ZEBOV) which is a single-dose live recombinant vaccine used primarily in ring-vaccination campaigns involving patients’ contacts, contacts of contacts, and high-risk health workers. A Guinea trial estimated 100 percent efficacy against cases beginning at least 10 days after vaccination, while later real-world studies estimated protection at roughly 84–94 percent. These therapies and the vaccine have transformed the response to Zaire Ebola, but they are species-specific and have not been shown to provide equivalent protection against Bundibugyo, Sudan, or Taï Forest ebolaviruses.

As mentioned above because of evolutionary divergence, there is no guarantee that the Zaire ebolavirus treatments and vaccine will work on the Bundibugyo strain that is spreading in DR Congo. Indeed, there is currently no licensed or clinically proven antiviral or antibody treatment specifically approved against the Bundibugyo virus, and so early, intensive supportive care remains the standard treatment. This includes rapid replacement of fluids and electrolytes lost through vomiting and diarrhea, maintenance of blood pressure and oxygenation, pain and nutritional support, and treatment of bleeding, shock, secondary infections, kidney injury, and multiorgan failure.

Meanwhile, The World Health Organization has also prioritized several experimental virus-directed therapies for clinical evaluation. These include MBP-134 and maftivimab, monoclonal antibodies intended to bind the viral surface glycoprotein and prevent cellular infection, and remdesivir, an intravenous antiviral (developed against the Covid virus SARS-CoV-2) that interferes with viral RNA replication; MBP-134 is being studied both alone and in combination with remdesivir. Obeldesivir, an oral antiviral related to remdesivir, is being investigated primarily as post-exposure prophylaxis for people who have been in close contact with infected patients but have not yet developed symptoms. Although laboratory and animal studies have produced encouraging results, none of these agents has yet been proven effective against Bundibugyo disease in controlled human trials.

For those outside of the Ebola infection region, the danger appears to be minimal; public health officials, including the CDC and WHO, currently consider the threat of the Bundibugyo Ebola virus outbreak spreading widely to be very low (e.g. to United States or Europe). This assessment is firmly grounded in the virus's biology: unlike airborne respiratory pathogens, Ebola transmits only through direct contact with the bodily fluids (e.g. blood, vomit, diarrhea, etc.) of a person who is already visibly ill, making it significantly easier to identify and isolate cases before widespread community transmission can occur. Even in the unlikely event that an infected traveler were to successfully cross international borders, well-resourced Western nations have a proven public health playbook -- rapid identification, strict isolation, aggressive contact tracing, and world-class intensive supportive care -- that successfully prevented sustained domestic outbreaks during the massive 2014 epidemic.

Thus, experts emphasize that the new outbreak does not pose an imminent pandemic threat; instead the true crisis is endemic, deeply concentrated in eastern DR Congo and Uganda (Figure 1), where containment is severely hindered by armed conflict, displacement, and a lack of medical infrastructure. Hopefully, the imminent availability of better treatments followed by the testing of vaccines could be one piece that helps to stem the tide so that the mortality levels do not approach those of the West African Ebola epidemic from 2014 to 2016.

Figure 1. Regions of DR Congo most affected by recent Bundibugyo Ebola virus outbreak (https://wwwnc.cdc.gov/travel/notices/level3/ebola-democratic-republic-of-the-congo).

Biorhythms versus Biological Rhythms

2026-07-11T08:44:12.042-07:00

Biorhythms refer to the idea that human physical, emotional, and intellectual abilities rise and fall in predictable cycles. Popular biorhythm theory proposes three fixed cycles -- a 23-day physical cycle, a 28-day emotional cycle, and a 33-day intellectual cycle -- that supposedly begin at birth and influence performance, mood, judgment, and susceptibility to mistakes or accidents (Figure 1). Although this theory helped popularize the notion that human functioning changes rhythmically, scientific studies have not found reliable evidence for these rigid cycles, and the theory is therefore regarded as pseudoscience. It should be distinguished from genuine biological rhythms, such as circadian, ultradian, and infradian rhythms, which are produced by physiological clocks, influenced by environmental signals, and studied scientifically through the field of chronobiology.

Human biological rhythms are essential, time-dependent physiological cycles categorized primarily by their duration: circadian, ultradian, and infradian rhythms. Circadian rhythms, such as the daily sleep-wake cycle, core body temperature shifts, and metabolic fluctuations, operate on a roughly 24-hour schedule and are masterminded by the brain's suprachiasmatic nucleus (SCN) in response to environmental light. Operating on shorter, higher-frequency loops of less than 24 hours are ultradian rhythms, which include the 90-minute sleep stages cycling between deep and REM sleep, as well as the pulsatile release of hormones like growth hormone and insulin throughout the day. Finally, infradian rhythms take longer than a day to complete; the most prominent human example is the roughly 28-day menstrual cycle governing reproductive hormones, though broader seasonal shifts influencing mood, sleep duration, and immune function also fall into this category. Together, these overlapping internal clocks help the body proactively anticipate and adapt to regular environmental changes to maintain optimal health and functioning.

Circadian rhythms coordinate physical, mental, and behavioral functions with the daily pattern of light and darkness. Governed primarily by the SCN in the hypothalamus, this master biological clock receives information about environmental light from the eyes and synchronizes clocks throughout the body. Circadian rhythms regulate the timing of sleep and wakefulness, melatonin and cortisol secretion, body temperature, blood pressure, digestion, metabolism, immune activity, mood, and cognitive performance. Light is the strongest environmental timing signal: morning light helps align the clock with daytime activity, whereas bright light at night can delay melatonin release and sleep. When circadian rhythms become misaligned because of jet lag, shift work, irregular sleep schedules, or nighttime light exposure, people may experience insomnia, daytime fatigue, impaired concentration, mood disturbances, and metabolic stress.

In humans, the physiological mechanism of circadian rhythms relies on a synchronized, dual-layered system: a central neuroendocrine pacemaker (SCN) and decentralized cellular clocks (in nearly every cell in the body). The SCN master clock is entrained by light via specialized retinal cells containing the photopigment melanopsin to detect environmental light and transmit this information to the suprachiasmatic nucleus in the hypothalamus. The molecular clock mechanism consists of a 24-hour gene–protein negative feedback loop: the proteins CLOCK and BMAL1 activate the PER and CRY genes, whose protein products gradually accumulate, return to the nucleus, and inhibit CLOCK–BMAL1 activity until they are degraded and the cycle begins again. Every cell possesses this 24-hour molecular clock, and the timing of all of these individual clocks are synchronized to the SCN by autonomic nerve signals, chemical (e.g. hormone) signals, as well as temperature and behavioral signals.

Circadian rhythms are important for health because they coordinate the timing of nearly every major body system, ensuring that sleep, hormone release, metabolism, cardiovascular activity, immune function, body temperature, and mental performance occur at biologically appropriate times, promoting daytime alertness/functioning and nighttime restorative sleep. For example, one wants to optimize insulin sensitivity during daylight hours for efficient digestion, while dropping blood pressure at night for cardiovascular rest, and precisely timing the release of hormones like cortisol (wake) and melatonin (sleep). When the internal clock becomes misaligned with the external environment -- as can occur with shift work, jet lag, irregular sleep schedules, nighttime eating, or bright light exposure late at night -- these systems may become desynchronized, contributing to insomnia, daytime fatigue, impaired judgment, metabolic dysfunction, cardiovascular strain, systemic stress, mood disturbances, and altered immunity. This can increase the long-term risk of obesity, diabetes, heart disease, immune dysregulation, and cognitive decline.

In summary, while the terms are often conflated in casual conversation, there is no scientific relationship between popular "biorhythm" theory and actual biological rhythms such as the circadian rhythm. Biorhythm theory is a widely debunked, pseudoscientific concept claiming that human physical, emotional, and intellectual performance is dictated by unchangeable 23-, 28-, and 33-day cycles locked in from the exact date of birth. These proposed cycles have no known physiological mechanism and have not been supported by scientific evidence. People have ups and downs, but if one documents these ups and downs they will not correspond to the fixed biorhythm periods. Circadian rhythms, by contrast, arise from biological clocks in the brain and throughout the body and are continually synchronized by environmental cues, especially light. They have significant health repercussions, the most notable being the wake-sleep cycle, which is necessary for peak performance during the day and revitalizing rest at night.

Figure 1. A typical biorhythm app purporting to show that physical, emotional, and intellectual well-being are fixed to rigid 23-, 28-, and 33-day cycles. Time (day) is on the x-axis, and physical (red) / emotional (blue) / intellectual (green) performance (from 0 to 100%) is on the y-axis.

Daraxonrasib is a big step forward in the fight against pancreatic cancer

2026-07-04T09:15:34.040-07:00

Pancreatic cancer is one of the deadliest cancers because it is usually diagnosed late, behaves aggressively, and has historically responded poorly to many modern cancer therapies. Early pancreatic cancer often causes few or vague symptoms, so many patients are not diagnosed until the disease has already spread beyond the pancreas. It kills approximately 50,000 Americans every year. The incidence rate is lower than that of breast, lung, or prostate cancer, but because of its deadliness, the fatality rate ranks only behind lung cancer (127,000 deaths/year) and colorectal cancer (CRC, 52,000 deaths/year). Given current trends with lung cancer deaths decreasing and CRC deaths relatively stable, the increasing number of pancreatic cancer deaths will soon rank number 2 in cancer deaths, and may eventually supplant lung cancer to become number 1. According to the American Cancer Society, the average lifetime risk of pancreatic cancer is ~1.5%.

Pancreatic Ductal Adenocarcinoma (PDAC) is the most common type of pancreatic cancer representing approximately 90% of all cases. It originates in the pancreatic duct cells that line the tube that connects the pancreas to the small intestine. The prognosis for PDAC is very poor with a 5 year survival rate of less than 10% that has not improved much over the last 50 years. In particular, unlike melanoma, lung cancer, and some blood cancers, pancreatic cancer has generally not responded well to immunotherapy, leaving patients with fewer effective options.

The current standard of care for PDAC depends primarily on the stage of disease and whether the tumor can be surgically removed. For the minority of patients with resectable (removable) or borderline resectable disease, treatment is typically multimodal, combining surgery -- such as a Whipple procedure for tumors in the pancreatic head (near where pancreatic duct empties into small intestine) -- with systemic chemotherapy given before and/or after surgery, often using modified FOLFIRINOX in fit patients. FOLFIRINOX consists of the four chemotherapy drugs folinic acid/leucovorin, 5-fluorouracil/5-FU, irinotecan, and oxaliplatin, and the modified version adjusts dosing to improve tolerability. For locally advanced unresectable disease, systemic chemotherapy is the main treatment, with radiation or chemoradiation sometimes used selectively. For metastatic pancreatic cancer, treatment is generally palliative and focused on prolonging survival, controlling symptoms, and preserving quality of life; standard first-line options include modified FOLFIRINOX and the newer NALIRIFOX regimen, which replaces standard irinotecan with liposomal irinotecan (i.e. irinotecan packaged in lipid particles) for improved delivery to the tumor. In addition, modern care also increasingly takes advantage of tumor molecular profiling to identify the small subset of patients who may benefit from targeted approaches, such as olaparib.

As mentioned above, one obstacle to treatment is that PDACs are largely insensitive (<5% response rate) to immunotherapy and immune checkpoint inhibitors in particular (i.e. anti-PD-1 antibodies), which have proven to be effective against other dangerous cancer types.

This is why the new investigational drug daraxonrasib may be a potential "game changer" for treating metastatic pancreatic cancer; it is a once-daily oral targeted therapy that has ignited optimism across the oncology community. Developed by the biotech company Revolution Medicines, daraxonrasib is a multi-selective inhibitor designed to shut down the active "on-state" of mutated KRAS proteins, which drive more than 90% of all pancreatic tumors. Unlike first-generation inhibitors that are limited to rare sub-variants like KRAS G12C (the 12th amino acid in the KRAS protein is mutated from Glycine (G) to Cysteine (C)), daraxonrasib's novel chemical architecture allows it to target a broad spectrum of the most common pancreatic cancer variants.

KRAS is a gene that encodes the KRAS protein, a member of the RAS family of small signaling proteins that help regulate cell growth, survival, differentiation, and division. In normal cells, KRAS acts like a molecular on/off switch: when bound to GTP it turns on growth signaling, and when it converts GTP to GDP it turns that signaling off. Cancer-associated KRAS mutations can disrupt this switch, leaving KRAS stuck in an active "on" state that continuously drives cell proliferation and tumor growth. KRAS is one of the most frequently mutated oncogenes in human cancer, with common mutation sites at amino acid positions 12, 13, and 61. It is especially important in pancreatic ductal adenocarcinoma, in which KRAS mutations are found in roughly 85–95% of cases, but it is also common in colorectal cancer (40-50%) and lung adenocarcinoma, and occurs at lower frequencies in biliary tract, gastric, endometrial, and certain ovarian cancers. The major KRAS-driven cancers differ in their mutation patterns: pancreatic cancers commonly contain G12D, G12V, or G12R mutations, whereas KRAS G12C is particularly important in lung adenocarcinoma.

For many years, KRAS was considered one of the most important but "undruggable" targets in oncology, because the KRAS protein lacks the kind of deep binding pocket that conventional small-molecule drugs usually exploit to bind and inhibit. That changed with the development of KRAS G12C inhibitors, which take advantage of the Cysteine created by the G12C mutation to bind covalently and lock KRAS in its inactive "off" state. This breakthrough led to the FDA approvals of sotorasib (Lumakras) and adagrasib (Krazati), both of which are primarily utilized for non-small cell lung cancer (NSCLC) and colorectal cancer. Unfortunately these new drugs are completely ineffective against the G12D, G12V, or G12R variants (which lack the Cysteine at position 12) that drive the vast majority of pancreatic cancers.

Daraxonrasib’s special chemical design differs fundamentally from earlier KRAS inhibitors because it is not targeted against one specific mutant amino acid, but rather a broad binding surface region (Figure 1). As described in the previous paragraph, first-generation KRAS inhibitors (e.g. sotorasib and adagrasib) specifically inhibit KRAS G12C by covalently attaching to the mutant Cysteine 12, and thus irreversibly inhibiting the protein, but these drugs are completely ineffective when KRAS has a non-cysteine amino acid at the 12th position. Daraxonrasib, on the other hand, is a noncovalent macrocyclic molecule that acts like a molecular glue. It first binds cyclophilin A (CYPA), a naturally occurring cellular protein, and the resulting drug–cyclophilin A complex then binds active, GTP-bound RAS forming a tri-complex -- cyclophilin A, daraxonrasib, and RAS -- that physically blocks active RAS from interacting with downstream signaling partners such as RAF and PI3K (Figure 1). Cyclophilin A is a generally sticky protein that binds many other proteins, and daraxonrasib is acting as a "glue" (enhancing protein-protein interaction) to help CYPA specifically bind the KRAS mutant proteins at high affinity, which occludes KRAS from performing its function of interacting with other (downstream) signaling proteins like RAF and PI3K. Because the binding relies on conserved features of the molecular surface of active RAS rather than a single mutation-specific chemical handle, daraxonrasib can inhibit multiple KRAS variants, including variants that are especially common in pancreatic cancer.

Many other conceptually promising anti-cancer drugs have failed when put to the test in clinical trials. This was not the case with daraxonrasib; recent clinical results in pancreatic cancer have been striking, especially because they were seen in patients with previously treated metastatic pancreatic ductal adenocarcinoma, a setting where standard therapies have historically produced only modest benefit. In a Phase 3 trial, roughly 500 patients were randomized to receive either once-daily oral daraxonrasib or “standard of care cytotoxic chemotherapy” (most likely FOLFIRINOX or one of its modified versions). Patients treated with daraxonrasib had a median overall survival of 13.2 months, compared with about 6.6 months for those receiving chemotherapy, corresponding to about a 60% reduction in the risk of death. The drug also improved progression-free survival, extending the time before cancer growth from roughly 3.5 months with chemotherapy to about 7.2 months with daraxonrasib, and it produced higher tumor response rates, with objective responses in about one-third of patients. Importantly, daraxonrasib also appeared more tolerable than chemotherapy, with fewer treatment-related discontinuations and delayed worsening of pain and quality of life, although side effects such as rash and stomatitis remained common. These results are considered potentially practice-changing because they suggest that a targeted oral RAS inhibitor (that can be taken at home) can meaningfully extend survival in a cancer type that has long been difficult to treat. Daraxonrasib has not yet been approved by the FDA, but approval remains a formality (most likely later this year) as it is the first treatment in decades to improve survival outcomes for advanced pancreatic cancer patients. In the meantime the FDA authorized an Expanded Access Program for compassionate use.

Unfortunately daraxonrasib is far from a cure, and most tumors are still expected to evolve resistance over time (or not even respond). The emerging evidence suggests that pancreatic cancers most often escape daraxonrasib by reactivating the same growth-signaling networks that the drug initially suppresses, rather than simply developing a new KRAS mutation that prevents drug binding. One major mechanism is amplification or overexpression of the original mutant KRAS gene, allowing the cancer cell to produce so much active KRAS protein that a fixed drug dose can no longer fully suppress signaling. Tumors may also bypass RAS inhibition by activating upstream proteins (e.g. receptor tyrosine kinases such as EGFR, HER2, or MET) or by acquiring mutations in downstream or parallel pathways (e.g. MAPK, PI3K–AKT–mTOR, or MYC pathways). These changes allow cancer cells to restore proliferation and survival signals even while daraxonrasib continues to inhibit part of the RAS pathway. The encouraging news is that resistance to daraxonrasib appears to be less about one simple “escape mutation” on the KRAS protein which has plagued the older first-generation inhibitors, and is more about the cancer rewiring or amplifying its signaling circuitry to overcome RAS pathway blockade. The latter may take time to overcome (QH).

In summary, daraxonrasib is the most promising recent advance in pancreatic cancer treatment because it nearly doubled median survival in patients with previously treated metastatic disease, extending survival from roughly 6.6 months with standard chemotherapy to about 13.2 months. In relative terms, that is a remarkable result for a cancer that has long resisted major therapeutic progress, and the fact that daraxonrasib is an oral targeted therapy (that can be taken at home rather than at a hospital infusion center) with fewer treatment discontinuations than chemotherapy adds to its clinical appeal. At the same time, the absolute benefit remains sobering: for most patients, the improvement is measured in additional months rather than years. The results therefore capture both the hope and the harsh reality of advanced pancreatic cancer -- an unusually large survival gain by the standards of this disease, but still a limited extension of life in a cancer that remains overwhelmingly lethal. Ultimately though, six and a half months of life -- especially with a manageable daily pill that improves quality of life over toxic chemotherapy -- is truly priceless for patients and their families. In addition, the targeted molecular glue approach of daraxonrasib may usher in further gains in the future with a new generation drugs for not only pancreatic cancer, but also other cancers.

Figure 1. The drug daraxonrasib (magenta) acts as a "molecular glue" that links together a "chaperone" protein cyclophilin A (CYPA) with the target protein mutated KRAS (blue RAS). Neither CYPA alone nor daraxonrasib alone can bind KRAS with high affinity, but together they form a ternary (three-way) complex with KRAS. Importantly CYPA helps to occlude an important functional effector surface of KRAS so that it cannot interact with its downstream signaling partners. In addition, the large CYPA-daraxonrasib binding surface is able to bind multiple different constitutively on mutants of KRAS rather than a single mutant (Sang et al. Cell, 2026).

What is a Health Savings Account (HSA)?

2026-06-27T07:26:49.578-07:00

A Health Savings Account (HSA) is a tax-advantaged (exempt/deferred/reduced taxes) financial account that allows people enrolled in a qualifying high-deductible health plan (HDHP) to save money for medical expenses (Figure 1). The two are designed to work as a pair. The HDHP is the insurance plan: it protects you against large medical costs, but it usually has a higher deductible, meaning you pay more out of pocket before the insurance company starts paying its share. Funds contributed to an HSA can be used to pay these (qualified) health costs such as doctor visits, prescription drugs, dental care, vision care, deductibles, copayments, and other out-of-pocket expenses. Note, you can have a High-Deductible Health Plan (HDHP) without ever opening or funding an HSA, but you cannot contribute to an HSA without being enrolled in a qualifying HDHP.

HSAs offer a "triple tax advantage": contributions are tax-deductible or made pre-tax (i.e. not taxed), the money can grow tax-free, and withdrawals are tax-free when used for qualified medical expenses. Unlike a Flexible Spending Account, HSA funds generally roll over from year to year and remain with the account holder even if they change jobs, health plans, or retire, making the HSA both a short-term tool for managing health expenses and a long-term savings vehicle for future medical costs. Annual contributions, however, are capped at $4,400 for individual coverage and $8,750 for families.

So to reiterate, an HSA is not a substitute for health insurance, but rather as mentioned above a tax-advantaged financial companion designed to work alongside a qualifying High-Deductible Health Plan (HDHP). While the health insurance protects you from catastrophic financial risk by covering major medical bills after you pay regular monthly premiums and meet your shared cost responsibilities (i.e. deductibles), an HSA acts strictly as a tax-efficient personal savings vehicle to help you fund those exact out-of-pocket healthcare expenses. The core cost-benefit tradeoff lies in how these two mechanisms balance upfront fees: selecting an HSA-eligible health plan dramatically reduces the amount deducted from your paycheck for monthly premiums, but it requires you to absorb a much higher upfront deductible. For 2026, an HSA-eligible HDHP must have an annual deductible of at least $1,700 for self-only coverage or $3,400 for family coverage, which is offset by lower monthly premiums. Thus, health insurance delivers the actual medical protection and caps your maximum financial risk, whereas the HSA provides unique triple-tax savings and long-term rollover benefits to help you manage those higher out-of-pocket exposures.

Health Savings Accounts work well for those who can comfortably manage higher upfront medical costs in exchange for lower premiums and tax advantages. They are especially attractive for relatively healthy individuals who do not expect frequent medical care, people with enough savings to cover a deductible if needed, higher-income taxpayers who benefit more from the tax deduction, and workers whose employers contribute money to the HSA. HSAs can also be valuable for long-term savers who want to build a tax-advantaged fund for future medical expenses or retirement health costs, since unused funds roll over from year to year and can often be invested. However, HSAs may be less suitable for people with predictable high medical expenses, expensive prescriptions, chronic conditions, or limited cash flow, because a high-deductible plan can require substantial out-of-pocket spending before insurance coverage completely covers expenses (i.e. out-of-pocket limit is $8,500 for self-only coverage or $17,000 for family coverage).

What happens when you turn 65 and go on Medicare (and don't need the HDHP)? The HSA is still there along with all the funds that have accumulated. You can continue to withdraw them tax-free to cover qualified out-of-pocket health costs. Furthermore, you can use these funds tax-free to pay for premiums associated with Medicare Part B, Part D, and Medicare Advantage, although Medicare supplemental (Medigap) premiums are expressly excluded. Finally, reaching age 65 eliminates the standard 20% IRS penalty for non-medical withdrawals, allowing you to use the remaining money for everyday expenses and simply pay ordinary income tax on the withdrawal, making the account function much like a traditional IRA (Individual Retirement Account).

A related financial account is the Flexible Spending Account (FSA) which is an employer-sponsored benefit account that allows workers to set aside pre-tax money to pay for eligible out-of-pocket health expenses, such as copayments, prescriptions, dental care, vision care, and other qualified medical costs just like with HSAs. But they differ in important ways; an FSA is generally tied to an employer, usually has lower contribution limits, cannot typically be invested, and is often subject to a "use it or lose it" rule, meaning unused funds may be forfeited at the end of the plan year except for any allowed carryover or grace period. By contrast, an HSA is personally owned, portable from job to job, available only to people enrolled in an HSA-eligible high-deductible health plan, and allows funds to roll over indefinitely, often with the option to invest the balance.

Another important distinction is that a regular health FSA does not require a high-deductible health plan (HDHP). An FSA is generally an employer-sponsored benefit, and you can often have one with a traditional PPO, HMO, or other non-HDHP plan. Finally, because you personally own your Health Savings Account (HSA), you are permitted to withdraw funds for any reason, but doing so before age 65 triggers the penalty described above on top of standard income taxes. In contrast, an FSA is strictly a reimbursement vehicle, meaning you generally cannot cash it out for non-medical use at all. Thus, FSAs are best suited for predictable near-term medical expenses, while HSAs are more useful as long-term, tax-advantaged accounts for current and future health care costs.

As a side note. a Health Savings Account (HSA) can earn interest much like a traditional bank savings account, with all interest accruing completely tax-free. In addition, the HSA can function as a dedicated investment account (e.g. IRA). Once your cash balance reaches a minimum threshold set by your provider -- often around $1,000 or $2,000 -- you can invest your funds directly into the stock market through mutual funds, exchange-traded funds (ETFs), or individual stocks. Because the funds roll over indefinitely and any capital gains or dividends are entirely shielded from taxes, many individuals choose to pay for current medical expenses out of pocket and leave their HSA fully invested, allowing the account to compound over decades as a tax-free supplemental retirement fund.

For this reason, the IRS regulates how much money can be deposited into a Health Savings Account (HSA) each calendar year, and this annual cap encompasses both your personal paycheck deductions and any contributions made by your employer. For 2026, the maximum allowed contribution is $4,400 for those with self-only coverage and $8,750 for individuals enrolled in a family High-Deductible Health Plan (HDHP). To assist older individuals in preparing for retirement healthcare costs, the IRS also permits anyone age 55 or older to make an additional $1,000 "catch-up" contribution annually.

In summary, the Health Savings Account (HSA) plus HDHP combination may be attractive for healthy individuals without frequent medical care, enough savings to cover the deductible, and are higher-income taxpayers. The HSA can be used to pay for deductibles, copayments, and other out-of-pocket health expenses. It is personally owned, portable from job to job, and rolls over indefinitely (unlike the FSA). As an added bonus Health Savings Accounts can act as a tax-advantaged retirement account by allowing interest and investments.

Figure 1. A health savings account (HSA) is a tax-advantaged (tax-exempt/tax-deferred) investment account (funds over a threshold can be invested) you contribute to that can be paired with an eligible high deductible health plan (HDHP). HSA funds can be withdrawn tax-free to pay qualifying medical expenses such as deductibles (https://www.fool.com/retirement/plans/hsa/). In other words, instead of paying more for a health plan with lower deductibles, etc., you contribute to the HSA.

What is Borderline Personality Disorder (BPD)?

2026-06-20T06:28:59.797-07:00

People with borderline personality disorder (BPD) may exhibit "difficult" behaviors; they are frequently misdiagnosed, and wrongly viewed as possessing a hopeless condition. Affecting an estimated 1.6 percent of the population, BPD is characterized by intense emotional pain, volatile relationships, an unstable sense of identity, and impulsive behaviors. A recent New York Times article relates a personal story, showing how someone can appear successful on the outside while privately struggling with BPD symptoms (e.g. emotional volatility, substance use, and relationship turmoil). At the same time the article highlights that BPD is quite treatable; with evidence-based approaches like Dialectical Behavior Therapy (DBT) and strong community support, individuals with BPD can effectively manage their emotions, navigate interpersonal conflicts, and build stable lives.

The term "borderline" in Borderline Personality Disorder stems from its historical classification and its complex, overlapping symptom profile. Coined in 1938 by psychoanalyst Adolph Stern, the condition was initially described as "borderline" because it appeared to exist on the border of other recognized psychiatric conditions. Today, while modern psychiatry understands BPD as a distinct disorder, the "borderline" label remains highly descriptive because its symptoms so frequently overlap with other conditions like bipolar disorder, major depression, and attention-deficit hyperactivity disorder.

Symptoms of Borderline Personality Disorder (BPD) typically involve a persistent pattern of instability in emotions, relationships, identity, and behavior (Figure 1). People with BPD may have intense mood swings, strong fear of abandonment, volatile relationships that shift between idealization and anger or rejection, chronic feelings of emptiness, and an unstable sense of self. They may also struggle with impulsive or risky behaviors such as substance misuse, reckless spending, unsafe sex, or binge eating, and some experience self-harm or suicidal thoughts or actions. Other symptoms can include intense anger, black-and-white thinking, and stress-related paranoia or dissociation, such as feeling unreal or disconnected during emotional crises. A person does not need every symptom to have BPD; clinicians look for a broader pattern of impairment.

BPD is diagnosed through a comprehensive clinical assessment rather than relying on a blood test, brain scan, or a single online quiz. Mental health professional use nine possible symptom criteria, and a person generally must show five or more as part of a persistent pattern that causes distress or impairment. The nine diagnostic symptom areas are:

Frantic efforts to avoid abandonment, whether the abandonment is real or feared.
Unstable and intense relationships, often shifting between idealizing someone and devaluing them.
Identity disturbance, meaning an unstable self-image or unclear sense of self.
Impulsivity in potentially self-damaging areas, such as spending, sex, substance use, reckless driving, or binge eating.
Recurrent suicidal behavior, threats, gestures, or self-harm.
Emotional instability, such as intense mood shifts that may last hours or days.
Chronic feelings of emptiness.
Inappropriate, intense anger or difficulty controlling anger.
Stress-related paranoia or severe dissociation, such as feeling detached, unreal, or suspicious during intense stress.

A critical part of this process is the differential diagnosis (distinguishing condition from others with similar clinical features) because BPD symptoms frequently overlap with other mental health conditions. It is often confused with disorders involving mood instability, impulsivity, trauma, anxiety, and relationship difficulties. Common related or co-occurring disorders include major depression, bipolar disorder, PTSD or complex PTSD, ADHD, anxiety disorders, substance use disorders, eating disorders, dissociative disorders, and other personality disorders such as narcissistic, histrionic, antisocial, dependent, or avoidant personality disorder. For example, a critical step is distinguishing BPD from bipolar disorder. Clinicians evaluate whether mood changes are rapid and reactive to relationship issues, which is characteristic of BPD, rather than the sustained, energy-altering manic or hypomanic episodes seen in bipolar disorder.

Treatment involves specialized, structured psychotherapy rather than medication, acting as a first-line intervention to help individuals manage intense emotions, reduce impulsive behaviors, and improve relationships. The most common modality is Dialectical Behavior Therapy (DBT), which is a form cognitive-behavioral therapy (CBT) tailored to BPD that focuses on teaching concrete skills to control intense emotions and reduce self-destructive actions by helping patients recognize triggers, and then implement appropriate responses. In addition, Mentalization-Based Treatment (MBT) is a long-term psychotherapy designed to help patients better understand mental states and intentions. While there are no medications that cure the disorder itself, psychiatric drugs are sometimes used as a time-limited adjunct to therapy to target specific, measurable symptoms or to treat co-occurring conditions like depression or severe anxiety. Because the condition carries a significantly elevated risk of self-harm and suicidal behavior, treatment typically requires rigorous safety planning to manage emotional crises.

The overall prognosis for Borderline Personality Disorder (BPD) is optimistic, with many individuals improving substantially over time. Long-term clinical research demonstrates excellent rates of symptom remission; for instance, one ten-year study found that 85% of patients remitted with only 12% relapsing after remission (link), while another study noted that 93% achieved a symptomatic remission lasting at least two years (link). However, clinical experts distinguish between symptom remission -- such as a dramatic decrease in self-harm, emotional crises, and impulsivity -- and functional recovery, noting that building stable relationships and steady employment often happens more slowly. Factors that strongly contribute to a better prognosis include consistent engagement in the therapies like DBT, the effective treatment of co-occurring disorders, and the establishment of stable daily routines.

The risk factors for BPD are an entangled interplay of genetic, developmental, and environmental factors. Hereditary components such as having a close relative with BPD can exist alongside of brain differences that impact impulse control and emotion regulation, and both may be exacerbated by environmental and social stressors, most notably histories of childhood emotional, physical, or sexual abuse, neglect, or early separation from a primary caregiver. Growing up in a highly volatile, invalidating environment marked by frequent interpersonal conflict, or in a household where family members struggled with serious mental illness or severe substance misuse, further elevates a person's risk.

Demographically in the past, approximately 75% of individuals diagnosed with Borderline Personality Disorder were female. This statistic is widely cited, including in diagnostic manuals like the DSM.

However, more recent large-scale epidemiological studies looking at the general population indicate that the prevalence of BPD is actually roughly equal between men and women with males frequently underdiagnosed or misdiagnosed.

In summary, Borderline Personality Disorder (BPD) is a mental health condition marked by persistent instability in emotions, relationships, self-image, and behavior. Despite afflicting an estimated 1.6% of the population, it is perhaps less well known than higher profile mental disorders. Core features include intense fear of abandonment, rapidly shifting moods, volatile relationships, impulsive or risky actions, chronic feelings of emptiness, unstable identity, intense anger, and sometimes self-harm or suicidal behavior. Although BPD can be severe and disruptive, it is treatable, especially with structured psychotherapy such as dialectical behavior therapy.

Figure 1. Symptoms of Borderline Personality Disorder (https://firstlightpsych.com/borderline-personality-disorder/).

Urgent Care versus the Emergency Room

2026-06-13T10:41:52.779-07:00

Urgent care is a type of walk-in medical care for illnesses or injuries that need prompt attention but are not life-threatening emergencies. It serves as a middle ground between a regular primary care visit and the hospital emergency room (Figure 1), offering same-day care for problems such as cold or flu symptoms, sore throat, earache, sinus infections, urinary tract infections, minor cuts or burns, sprains, rashes, mild asthma symptoms, and possible minor fractures. Urgent care centers are often open during evenings or weekends, usually do not require an appointment, and may provide basic services such as rapid testing, X-rays, stitches, splints, and prescriptions. However, urgent care is not appropriate for serious symptoms such as chest pain, trouble breathing, stroke-like symptoms, severe bleeding, loss of consciousness, major trauma, or severe allergic reactions; in those situations, a person should go to the emergency room or call 911. Thus, urgent care is best understood as same-day care for non-life-threatening medical problems when a primary care doctor is unavailable or the issue cannot reasonably wait.

Deciding between urgent care and the emergency room (ER) comes down to a simple question: is your condition life-threatening or limb-threatening? As a general rule, urgent care is for non-critical issues that need attention within 24 to 48 hours -- such as cold and flu symptoms, minor sprains, mild asthma flares, or a small cut needing stitches -- while the emergency room is for situations that could become dangerous if you wait. You should immediately head to the ER or call 911 if you experience severe symptoms like chest pressure, difficulty breathing, sudden face drooping or arm weakness, heavy uncontrollable bleeding, or loss of consciousness. Ultimately, if you are sick or injured and need care today, but your life is not in immediate danger, urgent care is the more appropriate and efficient choice.

A simple rule is that urgent care is for “I need care soon,” whereas the ER is for “this could be dangerous if I wait.”

In general, you can physically go to any urgent care facility, but the best choice is usually one that is in-network with your insurance and able to treat your specific problem. If possible, first check whether your primary care clinic has a same-day appointment, then consider an urgent care affiliated with your health system or insurance plan, followed by a nearby urgent care that accepts your insurance. The main issue is usually cost rather than permission: an out-of-network urgent care may lead to a much higher bill or limited coverage, depending on your plan. Before going, it is wise to confirm that the clinic is open, accepts your insurance, and offers the services you may need, such as X-rays, stitches, urine testing, COVID or flu testing, or basic labs. It is also important to make sure the facility is truly an urgent care center and not a freestanding emergency room, since freestanding ERs can look similar but often bill at much higher emergency-room rates. For example, verifying that an urgent care clinic is a Medicare-approved provider ensures your financial responsibility remains at the standard 20% coinsurance.

Urgent care has surged in popularity over the last decade, with the number of clinics in the U.S. nearly doubling to over 15,000 facilities, driven by a critical need to bridge the gap between routine primary care and the emergency room. As ongoing physician shortages make same-day primary care appointments increasingly difficult to secure, patients are gravitating toward the "on-demand" convenience of urgent care centers, which offer walk-in access, extended weekend and evening hours, and significantly shorter wait times. Beyond convenience, the economic benefits are substantial; treating common acute issues -- such as minor sprains, respiratory infections, or low-grade fevers -- at an urgent care clinic costs a fraction of a typical hospital emergency room visit. Recognizing this efficiency, both major health systems and corporate investors have heavily funded the rapid expansion of these clinics, establishing them as a highly accessible, lower-cost "front door" for everyday healthcare needs.

Although usage patterns rose sharply during the COVID-testing period and then partly normalized, urgent care has become a much more established part of the U.S. healthcare system. In short, urgent care is growing because it is faster than waiting for many primary care appointments, cheaper and less intensive than the ER, well suited for many common acute medical problems, and was popularized as a convenient and accessible venue for COVID testing.

An urgent care visit is completely different from going to the ER. When you go to urgent care, the experience is usually similar to a faster, walk-in doctor’s office visit: you check in with your ID and insurance information, provide your symptoms and medical history, have basic vital signs taken, and are seen by a clinician such as a physician, nurse practitioner, or physician assistant. Urgent care centers can often provide basic testing and treatments, such as rapid strep, flu, or COVID tests, urine tests, simple lab work, X-rays, wound care, splints, stitches, and prescriptions; most visits end with discharge instructions and follow-up advice. By contrast, the emergency room is a hospital-based setting designed for serious or potentially life-threatening problems, and patients are prioritized by triage, meaning the sickest or most urgent cases are treated first rather than simply being seen in arrival order. The ER can provide more advanced testing and treatment, including bloodwork, EKGs, CT scans, ultrasound, IV medications, oxygen, cardiac monitoring, emergency procedures, specialist consultation, and hospital admission if needed. In short, urgent care is usually for stable patients who need same-day treatment and can go home afterward, while the ER is for severe, dangerous, or uncertain conditions that may require intensive evaluation, monitoring, or admission to the hospital.

In the United States, seeking treatment at an urgent care clinic is substantially more affordable than a trip to the emergency room, often costing roughly a tenth of the price for the exact same non-life-threatening issue. A typical walk-in urgent care visit for a minor illness or injury generally ranges from $100 to $300 before insurance, and patients with in-network coverage usually only pay a modest copay of $20 to $75. In stark contrast, an average emergency room visit easily runs between $1,500 and $3,000 or more without insurance -- a staggering difference driven by massive facility fees, complex acuity billing, and the high overhead of operating a 24/7 critical care hospital department, which is necessary and worth the expense for severe, life-threatening conditions.

Practically speaking, in the short term (e.g. readiness for unexpected minor injuries), you should proactively identify a few local, in-network urgent care facilities, and then always call ahead to confirm their current wait times and specific diagnostic capabilities before making the trip. It is also highly beneficial to keep a digital "go-bag" on your phone containing photos of your insurance cards, a current medication list, and the specialist contacts for any elderly parents you might be assisting, which makes the intake process significantly smoother.

From a longer-term perspective, you want to make sure urgent care visits don't disrupt the continuity of your overall medical care, and so it is crucial to establish a reliable primary care physician who can manage your comprehensive health baseline and review any episodic treatments you receive, ensuring your care does not become fragmented. Finally, staying financially organized by understanding your specific coverage nuances -- such as utilizing pre-tax HSA funds or verifying that a clinic accepts Medicare Part B assignment to prevent excess charges -- will shield you from unexpected medical bills and redundant testing down the road.

Figure 1. Primary Care versus Urgent Care versus Emergency Room (azzmedical.com).

From biomarkers to biomolecular profiling

2026-06-06T09:28:15.252-07:00

A recent article in Nature Biotechnology proposes "Biomolecular profiling for noninvasive health monitoring." The central idea is to move personalized medicine away from episodic (i.e. infrequent), invasive diagnostics -- such as blood draws and biopsies that inherently yield delayed results -- toward continuous (i.e. real-time), noninvasive, and context-aware molecular tracking. The key is to integrate two complementary technologies: untargeted mass spectrometry (MS) and wearable chemical biosensors. MS serves as a discovery engine capable of identifying dynamic, functional biomarkers in accessible biofluids like sweat, saliva, and tears, and wearable sensors utilize these specific molecular targets to provide real-time, longitudinal monitoring directly on the body. The hope is that context-aware molecular monitoring can detect physiological change earlier and support more individualized intervention.

Traditionally, biomarkers are defined as measurable biological indicators that provide information about normal physiology, disease processes, or responses to treatment. They can include biomolecules such as proteins, metabolites, lipids, hormones, genes, RNA transcripts, antibodies, electrolytes, or even physiological measurements such as blood pressure, heart rate, or imaging findings. Among other things, these biomarkers are used to detect disease, assess disease risk, classify disease subtype, monitor progression, predict prognosis, guide treatment selection, and evaluate whether a therapy is working. For example, blood glucose is a biomarker used to diagnose and monitor diabetes, troponin helps detect heart muscle injury, prostate-specific antigen can contribute to prostate cancer screening and monitoring, and tumor mutations may guide targeted cancer therapy. More broadly, biomarkers help translate complex biological changes into clinically useful signals, allowing physicians to make more precise, timely, and personalized decisions about diagnosis, treatment, and follow-up.

Mass spectrometry (MS) is widely used as a diagnostic tool to detect and quantify many types of biomolecules -- including metabolites, proteins, lipids, hormones, drugs, toxins, and microbial products. It is a highly sensitive analytical technique that measures the mass-to-charge ratio (𝑚/𝑧) of ions to identify unknown compounds, quantify existing materials, and elucidate the structural properties of complex biological molecules. In practice, a biological sample such as blood, urine, saliva, sweat, tissue, or another body fluid is processed, its molecules are ionized, and the resulting molecular “fingerprints” (mass-to-charge ratios) are analyzed to determine which compounds are present and in what amounts. When ionized a given molecule may be fragmented into smaller pieces but each piece possesses a distinctive mass-to-charge ratio which can identify that piece, and the presence of a given set of pieces indicates presence of original molecule.

One of its major strengths is that MS can measure many molecules at once with high sensitivity and specificity, allowing researchers and clinicians to identify patterns of biological change rather than relying on a single marker. A chromatography step often precedes the MS to partially purify the sample giving rise to liquid chromatography mass spectrometry (LC-MS) or gas chromatography mass spectrometry (GC-MS). Matrix-assisted laser desorption/ionization (MALDI-MS) is a gentler ionization process better suited for biological molecules.

In the new paper, the authors extend the role of mass spectrometry from diagnostic tool to discovery engine capable of identifying new, clinically relevant biomarkers. Rather than relying on a single predefined biomarker, untargeted MS can profile many molecules at once in noninvasive biofluids such as sweat, saliva, tears, interstitial fluid, sebum, and cerumen, revealing molecular patterns that better reflect changing physiological states. Relying on single biomarkers often leads to poor diagnostic specificity, as isolated molecules like lactate or interleukin-6 overlap across multiple distinct physiological and pathological conditions. By measuring multiple biomarkers associated with a disease at the same time, MS can resolve these ambiguities, and hone in on specific pathophysiological pathways. Furthermore, by capturing dynamic kinetic features -- including rise and decay slopes, time-to-peak, and recovery half-life -- MS can successfully distinguish transient physiological shifts from sustained pathophysiology.

After a new group of biomarkers are discovered by MS, then they need to be measured on the subject. The main types of wearable biosensors described in the article are electrochemical, optical, and multimodal integrated sensing platforms (Figure 1). Electrochemical sensors convert chemical interactions into electrical signals and commonly use amperometric, voltammetric, or potentiometric readouts to measure analytes in the body fluid. Optical sensors instead rely on colorimetric, fluorescent, surface-enhanced Raman, or plasmonic signals, making them useful for visually or electronically detecting ions, metabolites, proteins, and other optically responsive molecules. Increasingly, wearable biosensors combine multiple sensing modes with microfluidic sampling, soft materials, wireless electronics, and AI-based signal processing, allowing several biomarkers to be tracked at the same time. The goal is to move biomarker monitoring from occasional laboratory testing to continuous, real-time measurement in everyday life.

Wearable biosensors can assay a wide range of accessible body fluids without the need for conventional blood draws or biopsies (Figure 1). Sweat, saliva, tears, exhaled breath condensate, wound exudate, sebum (hair oil), and cerumen (earwax) are especially attractive noninvasive fluids because they can be sampled at the skin, mouth, eye, breath, wound surface, or ear and can contain clinically useful molecules such as electrolytes, metabolites, hormones, proteins, inflammatory mediators, drug metabolites, lipids, and volatile compounds. Sweat can be analyzed through skin patches or microfluidic collectors; saliva through mouthguards or oral sensors; tears through smart contact lenses or tear-collection interfaces; and breath condensate through mask-like devices that capture respiratory biomarkers. The article also includes interstitial fluid, which is usually considered minimally invasive rather than fully noninvasive because it often requires microneedles or reverse iontophoresis, but it is still far less invasive than blood sampling and is already used for continuous glucose monitoring.

The large number of potential biomarkers necessitates multiplexed sensing which refers to the ability of a wearable biosensor to measure several biomolecules at the same time, rather than tracking only one. As mentioned above, many diseases and physiological states are better defined by patterns of molecules rather than by a single biomarker. For example, a wearable patch might simultaneously measure metabolites, electrolytes, stress hormones, inflammatory proteins, pH, temperature, and sweat rate, allowing the system to distinguish exercise, dehydration, stress, infection, or metabolic dysfunction. Technically, multiplexing can be achieved by placing different chemical sensors on separate electrodes or optical channels, while microfluidics routes small amounts of biofluids across the sensing regions. AI-based analysis will likely be important for interpreting these complex data streams, correcting for noise, dilution, motion artifacts, and sensor drift.

One of the biggest challenges for wearable biosensors is robustness; these devices must operate reliably on a moving, sweating, biologically complex human body rather than in a controlled laboratory setting. In addition, a useful wearable biosensor must collect very small amounts of biofluid, prevent evaporation or contamination, keep fresh sample flowing to the sensing surface, and account for changing secretion rates that can dilute or concentrate biomarkers. The sensors themselves must remain accurate despite motion artifacts, temperature shifts, pH changes, biofouling, skin irritation, adhesive failure, and gradual signal drift during long-term wear. Multiplexed devices add further complexity because measuring several biomarkers at once increases the risk of cross-reactivity, signal crosstalk, and calibration errors. Thus, real-world validation is essential so that they can become dependable clinical tools.

The amount of data will increase dramatically as there is a shift from one to many biomarkers, and also each biomarker is measured not at one time point but at many (continuous) time points. The wealth of biomolecular profiling data will assess dynamic molecular patterns that change with physiology, behavior, environment, and disease state. In traditional diagnostics, a biomarker is often treated as a static concentration -- such as whether one blood value is above or below a clinical threshold. By contrast, continuous wearable sensing makes it possible to view biomarkers as trajectories: how fast a molecule rises, how quickly it recovers, how it fluctuates with circadian rhythms, how it responds to exercise or stress, and how it co-varies with other analytes. For example, clinicians will be able to track multi-analyte signatures that indicate rapidly changing conditions (such as infection, sepsis, or metabolic instability) over a matter of minutes to hours, which would enable a more personalized and context-aware interpretation of health.

In summary, the authors outline their vision of a future in which noninvasive health monitoring relies on the co-evolution of mass spectrometry (MS) and wearable biosensors to shift the clinical paradigm from reactive, episodic testing to continuous "molecular awareness." They believe that to successfully translate these integrated platforms from the laboratory into reliable clinical tools, researchers must address four critical pillars: establishing standardized protocols for biofluid collection and sensor calibration, conducting robust longitudinal clinical validation to link dynamic biomarker trajectories with actionable health outcomes, building secure and interoperable data infrastructures to handle continuous molecular streams, and ensuring equitable device deployment across diverse populations.

If and when these barriers are overcome, then transformative medical applications could emerge. These include detecting presymptomatic molecular deviations before physical symptoms arise, guiding the precise timing of individualized interventions, and continuously tracking conditions like inflammatory flares, metabolic instability, and pharmacological responses, resulting in more proactive and personalized healthcare.

Figure 1. "General architecture of wearable chemical biosensors. Analytes (for example, metabolites/lipids and proteins) are accessed through biofluids collected through skin-interfaced systems such as contact lenses, microneedles, mouthguards or skin patches. Biofluids interface with bioreceptors (enzymes, antibodies, peptides or aptamers) that confer target specificity. Signals are transduced and processed by electronic modules for on-device or wireless readout" (reproduced from Fig. 3 of Kim et al. Nature Biotechnology, 2026).

Bladder cancer is the 6th most common cancer in America

2026-05-29T22:35:06.561-07:00

Bladder cancer is a common malignancy that usually begins in the urothelial cells lining the inside of the bladder and urinary tract. Most cases are urothelial carcinomas (90-95%), while less common types include squamous cell carcinoma and adenocarcinoma. Urothelial cells are specialized epithelial cells that line the bladder making it waterproof, and carcinoma is the most common type of cancer originating in epithelial cells.

Clinically, bladder cancer is divided into non–muscle-invasive disease, which is confined to the inner layers of the bladder wall, and muscle-invasive disease, which has grown into the bladder muscle and generally requires more aggressive treatment (Figure 1). The muscle layer is outside of the urothelial layer, and the danger is that the urothelial carcinoma will punch through the muscle layer and then spread (metastasize) outside of the bladder.

In the United States, bladder cancer is a significant public health burden: for 2026, the American Cancer Society estimates about 84,530 new cases and 17,870 deaths, with a strong male predominance. It is the 6th most common cancer by estimated new U.S. cases (behind prostate, breast, lung, colon, and skin (melanoma) cancers) and the 10th leading cause of cancer death. Age-adjusted data estimate an incidence rate of about 17.9 new cases per 100,000 people per year and a mortality rate of about 4.1 deaths per 100,000 people per year. Bladder cancer is primarily a disease of older adults, with a median age at diagnosis of about 73, and although overall 5-year relative survival is around 79%, prognosis depends strongly on stage at diagnosis.

Bladder cancer risk is driven by a complex interplay of lifestyle, environmental, medical, and demographic factors, with smoking standing as the single most significant contributor, making smokers at least three times more likely to develop the disease and accounting for approximately half of all cases. Occupational exposure to industrial chemicals -- particularly those used in processing paint, dyes, metals, rubber, and petroleum -- also heavily increases risk, as these carcinogens, much like those from tobacco smoke, are filtered by the kidneys and repeatedly expose the bladder lining to toxic damage. Additional environmental and medical hazards include chronic bladder inflammation from recurrent infections, bladder stones, or long-term catheter use; prior pelvic radiation; and a personal history of bladder or other urothelial cancers.

Demographically, the disease disproportionately impacts older adults and men, with White individuals diagnosed at roughly twice the rate of Black or Hispanic populations in the United States. Men have about a 4-fold higher incidence rate of bladder cancer than women in the U.S. For decades, researchers attributed this gap to lifestyle and environmental factors. For example, men smoked at significantly higher rates, smoked more heavily, and started at younger ages than women. In addition, men have vastly outnumbered women in the heavy industries most associated with carcinogenic chemicals, including dye manufacturing, rubber and tire production, leatherworking, metalworking, and commercial painting. However, even when researchers adjusted for smoking and chemical exposure (e.g. comparing male and female non-smokers with similar jobs), men still developed bladder cancer at higher rates. As a result, there has been more focus in recent years to attribute the gender gap to biological and hormonal differences. For example, researchers have found that testosterone receptors are present in the bladder lining and that testosterone activity can promote the development and growth of bladder cancer cells.

The most common and classic warning sign of bladder cancer is hematuria, or blood in the urine, which is typically painless and can be visible to the naked eye -- turning urine pink, red, or cola-colored -- or microscopic, requiring a lab test for detection. Because this bleeding can be intermittent, even a single episode warrants immediate medical evaluation. In addition to hematuria, patients frequently experience irritating changes in urination, such as increased frequency, sudden urgency even when the bladder is not full, pain or burning during urination (dysuria), and waking up often at night to urinate (nocturia). While these urinary symptoms closely mimic benign conditions like urinary tract infections, kidney stones, or an enlarged prostate, they should never be ignored. Therefore, anyone experiencing visible blood in the urine or persistent, unexplained urinary changes should seek prompt care from a healthcare professional for thorough testing and evaluation.

Bladder cancer is diagnosed through a stepwise process that usually begins with evaluation of symptoms such as blood in the urine, urinary frequency, urgency, burning, or unexplained abnormal urinalysis. Initial testing often includes a medical history and physical examination, urinalysis to detect blood or infection, urine culture to rule out urinary tract infection, and sometimes urine cytology or urine biomarker tests to look for abnormal cancer cells or tumor-associated markers. The standard hematuria workup often includes both cystoscopy and imaging of the kidneys/ureters, commonly a CT urogram.

The central diagnostic test is cystoscopy, in which a urologist passes a thin camera through the urethra to directly inspect the bladder lining; enhanced methods such as blue light cystoscopy may help detect subtle lesions like carcinoma in situ. CT urogram evaluates the kidneys, ureters, bladder, lymph nodes, and nearby abdominal/pelvic structures for any metastases. CT can also detect the bladder cancer (i.e. unexplained mass in bladder lining), but it usually does not replace cystoscopy, because small or flat bladder lesions can be missed.

If a suspicious lesion is seen, the diagnosis is confirmed by biopsy, most commonly through transurethral resection of bladder tumor or TURBT, which removes visible tumor tissue for pathological examination. The pathologist determines whether cancer is present, the tumor type, grade, and depth of invasion, which distinguishes non–muscle-invasive from muscle-invasive disease.

Bladder cancer is classified by cell type, stage, and grade, all of which help guide prognosis and treatment. As mentioned above, the most common type is urothelial carcinoma, while less common types include squamous cell carcinoma, adenocarcinoma, small cell carcinoma, and rare sarcomas. Staging describes how deeply the tumor has grown into the bladder wall and whether it has spread (Figure 1). Non–muscle-invasive bladder cancer includes tumors confined to the inner layers, such as Ta noninvasive papillary tumors, Tis carcinoma in situ (Stage 0), and T1 tumors (Stage 1) that invade the connective tissue beneath the lining but not the muscle. Muscle-invasive bladder cancer begins at T2 (Stage 2), when the tumor invades the bladder muscle, and more advanced stages include T3 (Stage 3), with spread into surrounding fatty tissue, and T4 (Stage 4), with invasion into nearby organs or pelvic/abdominal wall. In addition to stage, bladder cancers are graded as low-grade or high-grade, with high-grade tumors appearing more abnormal under the microscope and carrying a greater risk of invasion, recurrence, and spread. Approximately 50% of bladder cancers are low-grade Ta.

Treatment for bladder cancer depends mainly on whether the disease is non–muscle-invasive, muscle-invasive, or advanced/metastatic, as well as the tumor grade, extent of spread, and the patient’s overall health. For non–muscle-invasive bladder cancer (NMIBC), the usual first step is transurethral resection of bladder tumor or TURBT, in which visible tumor is removed through a cystoscope; thus the cytoscope can be used for visualization, obtaining a biopsy, and removing the tumor. This is often followed by intravesical therapy (drug injected directly into the bladder through a catheter), such as BCG immunotherapy or chemotherapy drugs like gemcitabine or mitomycin. Higher-risk or recurrent NMIBC may require repeat TURBT, additional intravesical treatments, newer gene- or immune-based therapies, or sometimes radical cystectomy.

For muscle-invasive bladder cancer, treatment is more aggressive and often includes cisplatin-based chemotherapy before surgery, followed by radical cystectomy (complete removal of the urinary bladder) with lymph node removal and urinary diversion, although selected patients may pursue bladder-preserving trimodality therapy with maximal TURBT, chemotherapy, and radiation. For advanced or metastatic disease, treatment is usually systemic (not confined to bladder since tumor has progressed beyond the bladder) and may include platinum-based chemotherapy, immune checkpoint inhibitors, antibody-drug conjugates, targeted therapy such as erdafitinib for tumors with susceptible FGFR3 alterations, radiation or surgery for symptom control, and clinical trials. Long-term follow-up with cystoscopy, urine tests, imaging, and monitoring for recurrence is an essential part of care.

In terms of last-ditch measures, if the bladder cancer has not spread (metastasized) beyond the bladder, then complete bladder removal (radical cystectomy) is an effective, potentially life-saving treatment. Indeed, radical cystectomy is considered the "gold standard" treatment for muscle-invasive bladder cancer (Stages 2 and 3) or for high-risk, early-stage cancers that continually return despite treatments like BCG immunotherapy. In these scenarios, the cancer poses a severe threat of spreading to other organs. By removing the bladder, surrounding lymph nodes, and nearby reproductive organs, surgeons can often physically eliminate the disease from the body, significantly improving the chances of long-term survival. However, doctors strongly advise against doing it "just to be safe" for early-stage cancers (Stages 0 or 1) because the physical and lifestyle costs of the surgery are immense.

The prognosis for bladder cancer is highly dependent on stage, grade, depth of invasion, response to treatment, and the patient’s overall health. Overall, bladder cancer is often treatable when found early, with an all-stage 5-year relative survival rate of about 77% to 79%, but outcomes vary widely by extent of disease: survival is approximately 97% to 98% for early, non-muscle-invasive (in situ) disease, 71% to 73% for localized disease that has invaded the muscle wall but remains confined inside bladder, 40% to 41% for regional spread to nearby tissues or lymph nodes, and 8% to 9% for distant metastatic disease. Early non–muscle-invasive bladder cancers often have a favorable outlook but are prone to recurrence, requiring long-term cystoscopic surveillance and sometimes repeated local treatments. Muscle-invasive bladder cancer carries a more serious prognosis but can still be potentially curable with aggressive treatment such as chemotherapy, radical cystectomy, or bladder-preserving chemoradiation in selected patients. Metastatic bladder cancer has historically had a poor prognosis, though newer therapies -- especially immunotherapy, antibody-drug conjugates, targeted treatments, and combinations -- are improving outcomes for some patients.

The future of bladder cancer care is moving toward precision (personalized) medicine and organ preservation (bladder-sparing), driven by advancements in molecular biology and artificial intelligence (AI). On the diagnostic side, cystoscopy and biopsy will remain central, but improved urine-based molecular tests (e.g. biomarker panels), enhanced cystoscopy such as blue-light imaging, and AI-assisted cystoscopy, cytology, imaging, and pathology may help detect subtle tumors earlier, reduce missed carcinoma in situ, and better distinguish low-risk from high-risk disease. These technologies may be complemented by non-invasive liquid biopsies, such as circulating tumor DNA (ctDNA) blood tests, which can potentially detect recurrences earlier.

Treatment is expanding beyond traditional TURBT, BCG, chemotherapy, radiation, and cystectomy, with newer intravesical approaches for BCG-unresponsive non–muscle-invasive disease, including FDA-approved nadofaragene firadenovec (IFN alpha-2b) gene therapy, the gemcitabine intravesical system Inlexzo (device continuously administers chemotherapy in bladder), antibody-drug conjugates (ADCs), and systemic immune checkpoint inhibitors, targeted pathway inhibitors, and personalized mRNA cancer vaccines. Some of the latter can be applied intravesically into the bladder or systemically throughout the body if the cancer has spread.

Taken together past and present advances can help explain why bladder cancer incidence ranks 6th among all cancers, whereas mortality ranks only 10th, and the future offers a promising horizon in which treatments are not only more effective at eradicating cancer cells but also far less toxic, ultimately aiming to maximize both long-term survival rates and the patient's everyday quality of life.

Figure 1. Stages of bladder cancer. Note how cancer is confined to bladder at earlier stages but then grows to punch out of the bladder and metastasize at later stages. The urothelial bladder lining is the innermost layer, followed by a connective tissue layer, then the muscle layer, and finally the fat layer.

Ta and Tis are Stage 0 when cancer is confined to the urothelial inner lining. T1 (Stage 1) is when the cancer grows into the connective tissue layer. T2 represents the first muscle-invasive stage, and T3 is when cancer invades the fat layer. In T4 (Stage 4), the cancer has spread outside of the bladder (diagram from https://www.cancerresearchuk.org/about-cancer/bladder-cancer/types-stages-grades/stages).

Judge ruling has halted changes to childhood vaccine schedule

2026-05-23T07:27:48.146-07:00

I wrote previously about a controversial directive from the Department of Health and Human Services (HHS) and CDC reducing the universally recommended U.S. childhood vaccine schedule from 17 diseases down to 11, shifting the remaining immunizations to an optional "shared clinical decision-making" model. While proponents claim this change empowers parental choice and aligns America with peer nations like Denmark, it has triggered fierce, unified backlash from major medical organizations such as the American Academy of Pediatrics (AAP) and the American Medical Association (AMA). Using Denmark as a benchmark fundamentally ignores the stark reality of the U.S. healthcare system, where millions of uninsured or underinsured children rely heavily on baseline immunizations for basic health protection. Ultimately, this directive bypassed traditional scientific review to make vaccination against some highly contagious diseases optional, which threatens to depress vaccination coverage, compromise community immunity, and place the health of innumerable vulnerable children at risk.

Although the directive did not eliminate the full pre-January slate of 17 childhood vaccines, it likely made obtaining all of them harder for some children by adding practical barriers to access. Vaccines moved out of the universal "routine" category and into shared clinical decision-making or high-risk use now require extra discussion and documentation between parents and clinicians, which can slow or discourage vaccination, especially for families who rely on brief clinic visits, walk-in settings, or limited appointment time. The changes also complicate the use of combination vaccines and disrupt standard pediatric workflows, since some shots that used to be bundled together now straddle the line between "routine" and "non-routine." Further, the directive places unexpected financial burdens on families seeking full, multi-dose regimens -- such as the HPV vaccine -- due to newly restricted insurance coverage limits. Taken together, these changes do not create a universal loss of access, but they do make it more difficult and uneven for some children to receive the full 17-vaccine schedule.

In addition, the directive may make it easier for many families to avoid certain vaccines or bypass mandates. By shifting six previously routine vaccines -- including those for Hepatitis A, Hepatitis B, and Meningococcal disease -- into "shared clinical decision-making" categories that explicitly factor in family beliefs, the federal government removed the universal default, significantly lowering the clinical friction of opting out. In states that automatically align their school entry requirements with the CDC schedule, this reclassification means parents may simply avoid these shots without needing to formally apply for an exemption at all. Furthermore, the sweeping federal rollback has catalyzed legislative momentum in states like Florida and Idaho to broaden legal loopholes, such as establishing new "conscience-based" nonmedical exemptions.

So far, the directive does not appear to have produced a clear, nationally measurable drop in childhood vaccination rates, mainly because it's too early to tell. Even so, there are already noticeable warning signs at the state and clinical level: pediatricians have reported greater confusion and skepticism among parents, the new shared-clinical-decision-making framework has made some routine vaccine visits harder to carry out efficiently, and some local data, such as a decline in Michigan’s toddler seven-vaccine-series completion rate, suggest that federal messaging may already be contributing to lower uptake in some places. The most accurate conclusion is that the directive has had noticeable early effects on trust, practice patterns, and some local vaccination rates, but it has not yet been shown to have caused a definitive nationwide decline in childhood vaccination overall.

In an attempt to overturn the HHS directive, a coalition of the nation's leading medical and public health organizations including the American Academy of Pediatrics (AAP), the American College of Physicians (ACP), the American Public Health Association (APHA), the Infectious Diseases Society of America (IDSA), the Society for Maternal-Fetal Medicine (SMFM), and the Massachusetts Public Health Alliance sued HHS and Secretary Robert F. Kennedy Jr. under the Administrative Procedure Act (APA). They alleged that the abrupt reduction of the routine childhood vaccine schedule from 17 to 11 vaccines was unlawful, arbitrary, and capricious because the government intentionally bypassed the statutory role of the Advisory Committee on Immunization Practices (ACIP). Furthermore, they argued that the administration violated the Federal Advisory Committee Act (FACA) by abruptly purging existing ACIP members and replacing them with individuals who lack the required scientific, medical, and immunological expertise.

On March 16, 2026, U.S. District Judge Brian Murphy (Figure 1) overseeing the AAP lawsuit issued a preliminary injunction that temporarily halted the January 2026 HHS/CDC childhood vaccination directive while the lawsuit proceeds. A preliminary injunction is a temporary court order made early in a lawsuit that forces a party to stop doing something until the court can make a final, permanent decision on the case. The court agreed with the plaintiffs that the government likely acted unlawfully and arbitrarily by bypassing the ACIP's statutory role to unilaterally rewrite the schedule, and noted that the rushed, reconstituted committee likely violated the Federal Advisory Committee Act by lacking fair balance and scientific expertise. Consequently, until the courts reach a final verdict is reached, the previously established pediatric immunization schedule is legally back in effect, and the newly formed ACIP is completely sidelined from advancing further policy changes.

The new judge ruling likely makes it somewhat easier for parents to get vaccines, but it does not directly make it harder for children to obtain exemptions. By issuing a preliminary injunction and temporarily restoring the pre-January 2026 childhood immunization schedule, the court removed some of the confusion and added friction created by the directive. The good news is that the preliminary injunction reestablishes a clearer default path for parents and providers seeking the broader pre-2026 slate of recommended vaccines. However, exemption rules for school-entry vaccination are still governed mainly by state law, not by the federal schedule, so the court order did not itself eliminate or tighten any existing medical, religious, or personal-belief exemptions. At most, the ruling may indirectly make broad exemption efforts harder in some places by restoring the older federal schedule that states and institutions often use as a reference point.

In summary, the January HHS/CDC directive currently remains legally frozen by the March 16 preliminary injunction, effectively restoring the previous routine childhood vaccine schedule and sidelining the recently overhauled Advisory Committee on Immunization Practices (ACIP). Meanwhile the underlying lawsuit is progressing into the merits phase in Judge Murphy's Boston district court as the administrative record is assembled. As a counter, in late March, HHS issued a revised ACIP charter designed to broaden membership criteria, a strategic move that could allow them to legally reconstitute the committee but will likely trigger a fresh wave of legal challenges.

Figure 1. Judge Brian Murphy overseeing the lawsuit by medical organizations suing HHS issued a preliminary injunvtion halting the HHS directive reducing the universally recommended childhood vaccine schedule from 17 to 11 diseases.

The healthful benefits of Tai Chi

2026-05-16T10:58:12.924-07:00

Tai Chi, formally known as Tai Chi Chuan, is an ancient Chinese martial art that has evolved into a globally embraced mind-body practice often described as "meditation in motion." Fundamentally, it is a gentle, low-impact exercise system that seamlessly blends slow, deliberate, and continuous movements with deep breathing and mental focus. Practitioners flow through a graceful sequence of postures known as a "form," cultivating a harmony between body and mind. Tai Chi is known for being easy on the joints and highly adaptable for all ages and fitness levels.

Tai Chi possesses a wide range of healthful properties. Medical evidence suggests its strongest benefits are in improving balance, stability, and fall prevention, especially in older adults, while also helping with joint pain, stiffness, and physical function in conditions such as osteoarthritis. Regular practice may also ease certain chronic pain conditions, including fibromyalgia (widespread chronic pain), and can contribute to better flexibility, mobility, and overall body awareness. In addition, Tai Chi’s combination of slow movement, deep breathing, and focused attention may reduce stress, support psychological well-being, and provide modest cognitive benefits, making it a broadly accessible practice that promotes health without placing excessive strain on the body.

According to a comprehensive Cochrane review analyzing over 100 randomized trials, the practice significantly enhances stability in older adults, reducing the overall rate of falls by 19% and decreasing the number of individuals who experience a fall by 20%. Beyond balance, the American College of Rheumatology officially issues a strong recommendation for Tai Chi as a legitimate treatment for knee and hip osteoarthritis, noting its efficacy in alleviating pain and stiffness while restoring physical function. In addition, studies demonstrate that Tai Chi improves postural stability for those with Parkinson’s disease and provides symptom relief for fibromyalgia that equals or exceeds the benefits of standard aerobic exercise. While its impacts on lowering blood pressure and improving cognitive function show positive but more modest results, the scientific consensus firmly establishes Tai Chi as a safe, highly effective, and low-impact adjunct to conventional medical care.

The foundation of Tai Chi rests on the seamless integration of traditional Chinese philosophy -- particularly the interplay of Yin and Yang -- and precise, mindful physical mechanics. Philosophically, the practice emphasizes the harmony of opposites, mental calm, and the principle of yielding to overcome force, prioritizing relaxation and efficiency over brute strength. These ideals are physically manifested through continuous, flowing sequences called "forms," which link classic postures like "Wave Hands Like Clouds" and "Parting the Wild Horse's Mane" without abrupt stops. During these sequences, practitioners maintain an upright, lengthened spine and relaxed shoulders, generating soft, sweeping arm motions that are driven entirely by the turning of the waist and torso. By synchronizing deep, natural breathing with deliberate weight shifts that clearly distinguish between "empty" (unweighted) and "full" (weighted) steps, Tai Chi cultivates a grounded, stable connection to the earth, effectively transforming abstract Daoist concepts into a practical meditation in motion.

Learning Tai Chi typically begins with watching, imitating, and slowly repeating basic movements under the guidance of a beginner-friendly instructor or class. Because the practice depends on subtle details such as posture, balance, relaxed movement, weight shifting, and breathing, feedback from a teacher is especially helpful early on, even though videos and online courses can be useful supplements (Video 1). Most beginners do not start by memorizing a long routine; instead, they first learn foundational skills such as standing posture, coordinated arm and leg motion, and steady breathing, then practice a few core movements until they feel natural before progressing to a short form, often in the Yang style. Regular, patient practice is more important than speed or perfection, and even a few minutes a day can gradually make the movements feel smoother, more balanced, and more intuitive over time.

Writing about Tai Chai reminds me of the post on yoga, which is another holistic mind-body practice that significantly enhances overall well-being and may even contribute to a longer, healthier life. On a physical level, yoga's challenging poses build muscular strength, expand flexibility for daily tasks, and critically improve balance and proprioception, which helps prevent debilitating falls in older age. Mentally, yoga serves as a powerful tool for stress reduction, helping to lower blood pressure, improve sleep quality, and foster emotional serenity. Finally, emerging research suggests yoga may offer modest cognitive benefits for aging individuals.

Both are highly beneficial mind-body practices that combine mindful movement with deep breathing; however, their clinical strengths cater to different primary health goals. Tai Chi is characterized by continuous, slow weight-shifting and an upright posture, making it exceptionally gentle on the body; consequently, it boasts the strongest medical evidence for improving dynamic balance, significantly preventing falls in older adults, and managing knee and hip osteoarthritis. In contrast, yoga encompasses a wider variety of physical demands, ranging from gentle stretching to deep, static poses that actively load the joints and muscles. This makes yoga particularly effective for enhancing flexibility (and strength), relieving chronic back and neck pain, and providing a broader range of psychological benefits, such as stress reduction, improved sleep, and mood regulation. Ultimately, choosing between the two depends on your specific physical needs: Tai Chi is suited for gentle joint mobility and stability, whereas yoga is better for deep flexibility and comprehensive stress relief,

Indeed it is possible to do both: Tai Chi and yoga are complementary practices that when combined create a comprehensive strategy for long-term physical and mental well-being. Together, they address different physiological needs to form a well-rounded routine. Tai Chi excels in cultivating dynamic balance, steadiness, and controlled weight-shifting, making it the superior choice for fall prevention and joint-friendly movement, particularly for those managing osteoarthritis. Conversely, yoga emphasizes static flexibility, pose-based strength, and deep muscular endurance, offering a broader clinical impact for managing chronic back and neck pain alongside stress reduction. By strategically integrating both, practitioners can synergize the unique benefits of each, ensuring that both flexibility and the neurological control to safely utilize that range of motion are continuously developed.

In summary, Tai Chi is a low-impact "meditation in motion" that blends slow, continuous movements, mindful breathing, and Daoist philosophy. It serves as an effective intervention for improving dynamic balance, preventing falls in older adults, and alleviating joint pain associated with osteoarthritis and fibromyalgia. The practice emphasizes upright posture, precise weight-shifting, and gentle joint mobility making it physically distinct from the static, flexibility-focused poses of yoga. Ultimately, Tai Chi offers a safe way to cultivate physical stability and mental calm, and it can act as a perfect complement to yoga by bridging the gap between passive flexibility and active, functional movement.

Video 1. A short video introducing some basic Tai Chi movements and philosophy.

Samsung Hearapy treats motion sickness with sound tone

2026-05-09T08:03:00.000-07:00

I am relatively sensitive to motion sickness, and so when I ride airplanes or boats I take Dramamine which is an over-the-counter (OTC) first-generation antihistamine. Dimenhydrinate (Dramamine's active ingredient) can act as both a H1 histamine receptor and a muscarinic acetylcholine receptor antagonist. Blocking these pathways reduces signaling from the vestibular system (inner ear's balance center) to the brain including to the “vomiting center”. I am always on the lookout for new motion sickness treatments.

Recently Samsung announced its new Hearapy app, which is a free Android application that offers a drug-free approach to managing motion sickness through targeted sound therapy. Built upon early vestibular research from Japan's Nagoya University, the app works by playing a deep, 100 Hz low-frequency tone for 60 seconds prior to travel. When listened to through high-quality earbuds -- most notably Samsung's recommended Galaxy Buds4 Pro (Figure 1) -- this specific acoustic stimulation is designed to physically engage and "reset" the inner ear's balance system. Hearapy aims to provide up to two hours of relief from the nausea, dizziness, and discomfort commonly triggered by cars, trains, and airplanes.

Currently, there is no iPhone version of Hearapy. It is exclusively available for Android devices through the Google Play Store, and Samsung has not announced any plans for an iOS release.

From a physiological standpoint, motion sickness is triggered by actual or perceived movement, such as traveling in vehicles, riding amusement park rides, or experiencing virtual reality simulators. According to the commonly accepted sensory conflict theory, the condition occurs when the brain receives contradictory information from the inner ear (the vestibular system), the eyes, and the body's proprioceptive physical position sensors. For example, your inner ear feels the physical acceleration, bumps, and sway of the vehicle (boat or car), but because your eyes are focused on a stationary book or the interior cabin, your visual system tells your brain, "We are sitting perfectly still." In other words the

motion is felt, but not seen.

This inconsistency confuses the brain's spatial orientation and triggers a cascade of symptoms that often begin subtly with stomach awareness, yawning, warmth, and sweating, before progressing to dizziness, nausea, and vomiting. Susceptibility to motion sickness varies widely; it peaks in children between the ages of 2 and 12, and is higher among pregnant individuals or those with a history of migraines and vestibular disorders.

Treatments for motion sickness generally fall into two categories: non-drug strategies and medications, with prevention usually working better than trying to relieve symptoms after they begin. Behavioral measures focus on reducing the sensory mismatch that triggers motion sickness, such as looking at the horizon or out a window, avoiding reading or phone use, keeping the head still, choosing a seat with less motion, staying hydrated, eating lightly, and getting fresh air. For people who need medication, common over-the-counter options include meclizine and dimenhydrinate (Dramamine), which are most effective when taken before travel but can cause drowsiness. For longer trips or stronger motion triggers, scopolamine -- often used as a patch behind the ear -- can provide longer-lasting prevention, while promethazine may be used in more severe cases but is even more sedating. Overall, the best treatment depends on the severity of symptoms, the type of travel, and the individual’s tolerance for side effects.

Samsung’s Hearapy app uses brief sound stimulation to prepare the inner ear’s balance system before travel. The app plays a controlled 100 Hz low-frequency tone through earbuds or headphones for about 60 seconds at a recommended loudness between 75 and 85 decibels (dB), with the goal of stimulating the vestibular system, which helps regulate balance and motion perception. This approach is intended to reduce the sensory mismatch described above before it leads to nausea, dizziness, or discomfort. The app is simple to use: users connect headphones, press start, and listen to the tone, with the playback length adjustable from roughly 40 to 120 seconds, and Samsung says the effect may last for up to two hours. The company also emphasizes the importance of accurate low-frequency sound delivery, which is why it recommends the Galaxy Buds4 Pro, though other headphones may also work. The loudness of the tone is set by your phone's or headset's volume control and so the desired 75 to 85 decibel range is only approximate.

A 100 Hz tone is considered low-pitched, falling within the bass range (roughly 80–160Hz), often described as a thumping, booming, or a warm sound. It may be felt as a physical vibration, as human ears are less sensitive to frequencies below 100Hz compared to higher frequencies. An intensity level of 75 to 85 decibels (dB) is considered loud and is comparable to a garbage disposal, alarm clock, or a noisy restaurant, and so when setting the volume of the tone, be careful not to set it too loud.

The medical evidence supporting Samsung's Hearapy app is quite preliminary, rooted primarily in a single 2025 peer-reviewed study from Nagoya University (Kato Lab). Researchers found that exposing mice and humans to a 100 Hz pure tone at roughly 80 to 85 dBZ for one minute prior to motion exposure significantly improved both objective and subjective measures of motion sickness. In human trials, the acoustic stimulation objectively reduced physical sway (measured by posturography) and improved autonomic regulation (ECG/heart rate variability) during swing tests, driving simulators, and real vehicle motion, while also lowering subjective discomfort scores on the Motion Sickness Assessment Questionnaire. After the motion, the posturography was performed by having the participant stand with eyes closed, arms at the sides, and legs and feet together, and then the envelope area of center-of-gravity movement was measured for 1 minute. More swaying resulted in a worse posturography score. Samsung's app serves as a direct consumer implementation of this protocol by playing a 100 Hz bass sine tone for 60 seconds, which the company claims can provide up to two hours of relief.

Two previous paper provided additional support. First, Xu et al. (2018) showed improvements in balance that was isolated to the 100 Hz sound component in a piece of music. Second, in earlier work, Kato's lab at Nagoya University demonstrated that symptoms arising from Meniere’s disease (chronic disorder of the inner ear that affects balance and hearing) was alleviated by exposure to a pure tone at 100 Hz.

The hypothesized physiological mechanism is that the sound tone stimulates the otoconia (inner ear structure that senses acceleration) thus priming the balance system prior to travel. This pre-stimulation is thought to improve the body's physical balance, making the vestibular system more resilient to adjust to the physical shaking motions of a vehicle.

However, experts caution that while the underlying biological concept is legitimate, there is not yet a robust body of independent clinical trials specifically validating the app's real-world efficacy across all users and diverse travel conditions, making it an early, low-risk experiment rather than a universally proven medical cure. They also warn that translating these laboratory results into a consumer app presents significant challenges. There could be a "hardware gap," noting that real-world effectiveness depends heavily on whether a user's earbuds can consistently replicate the precise frequency and loudness required without interference from outside noise. For now health officials do not recommend Hearapy as a replacement for proven treatments in severe or high-stakes scenarios.

Clearly more research especially in human subjects is needed; consumer feedback should also provide additional data albeit not scientifically validated as in a trial or study. One question as mentioned in the previous paragraph is how much does the treatment efficacy depend on the quality of the headphones. Regardless I plan to try this free app which is quite easy to test out.

Figure 1. Advertisement for Galaxy Buds4 Pro ear buds using Hearapy as a potential sound treatment for motion sickness.

Who is Peter Attia and why has he been in the news?

2026-05-02T14:26:00.000-07:00

Dr. Peter Attia is a physician, bestselling author, and prominent media figure widely recognized for his proactive approach to longevity and preventive medicine. After earning his medical degree from Stanford University and training in general surgery at Johns Hopkins, with additional oncology research at the National Cancer Institute, he shifted his career to focus on extending human "healthspan" -- maximizing the years of life spent in peak physical and cognitive health. Through his private medical practice (Early Medical, which is a highly exclusive, concierge-style practice), his #1 bestselling book "Outlive: The Science and Art of Longevity", and his highly popular "The Peter Attia Drive" podcast, he champions a data-driven framework that emphasizes early intervention, aggressive cardiovascular screening, structured exercise, and nutritional biochemistry to delay the onset of chronic diseases.

He was recently in the news because of the fallout from revelations about his past relationship with Jeffrey Epstein. In early 2026, the U.S. Department of Justice released a large batch of Epstein-related documents showing that Attia had maintained a close and friendly relationship with Epstein from 2014 to 2019, long after Epstein’s 2008 conviction for procuring a minor for prostitution. The files included emails about health, longevity, and lab data, but also exposed vulgar and inappropriate jokes by Attia, prompting public backlash. The controversy quickly led to professional consequences, including his resignation as a CBS News medical contributor and his departure from David Protein, where he had served as Chief Science Officer. In response, Attia issued a public apology, calling the emails "embarrassing, tasteless, and indefensible," while denying any involvement in criminal activity or knowledge of underage abuse.

Dr. Attia has achieved fame and fortune by bringing his innovative blend of ideas on longevity and preventive health to a mass audience. His ability to translate complex metabolic and aging science for a general audience has earned him widespread mainstream recognition, including a feature segment on 60 Minutes in 2025 and a spot on the inaugural TIME100 Health list in 2024. Taken together, his achievements lie less in scientific achievements than in his success at translating complex medical ideas about aging, exercise, metabolic health, and disease prevention into a widely accessible public platform.

In Attia’s framing, Medicine 1.0 is the premodern era of medicine, before major scientific breakthroughs like germ theory and the modern understanding of the cell transformed medical practice. It can be colloquially termed "The Era of Guesswork" which lasted up until the late 19th century. Medicine 2.0 is modern scientific medicine: it became very good at things like infections, trauma, and acute care, but Attia argues it is still too reactive when it comes to chronic diseases because it often waits until symptoms or obvious disease appear ("The Era of Acute Care").

Medicine 3.0 is his proposed next step: a more proactive, preventive, and personalized model that tries to identify risk earlier and uses tools like lifestyle change, screening, and biomarker tracking to both lengthen lifespan and improve healthspan, ensuring people remain physically and cognitively capable in their later years. He argues that the traditional medical system is dangerously reactive, waiting for disease to strike rather than aggressively preventing the "Four Horsemen" of aging: cardiovascular disease, cancer, neurodegenerative disease, and metabolic dysfunction. To defend against these conditions and avoid the physical decline of the "marginal decade" (last 10 years of life), Attia advocates for a comprehensive toolkit that prioritizes structured exercise (which he frames as training for the "Centenarian Decathlon"), precise nutritional biochemistry, optimized sleep, and robust emotional well-being.

In his view, exercise is perhaps the most powerful intervention for healthy aging. For example, Peter Attia’s Centenarian Decathlon is the idea that you should train now for the physical tasks you want to still be able to do at age 100. Instead of exercising only for appearance or near-term performance, he suggests “backcasting” from late life: first decide what activities would make life satisfying in old age, then build your training around preserving the capacity to do them. On his site, examples include things like lifting a suitcase into an overhead bin, playing with future grandchildren, and handling everyday activities of daily living.

Dr. Attia’s success and popularity stem from his unique ability to perfectly time and package medical science for an increasingly health-conscious public. This clever packaging is most evident in his inventive terminology like the "Centenarian Decathlon" in which he takes an Olympic event that captures the imagination, and transforms it into something quite relevant for all of us. Arriving at a cultural moment when aging demographics are highly motivated to actively extend their healthspan, Attia set himself apart from typical wellness influencers by leaning into his elite credentials and refusing to "dumb down" complex biological ideas. He effectively bridged the gap between evidence-based medicine and popular optimization culture by translating abstract scientific concepts into highly actionable, memorable frameworks, such as "Medicine 3.0". His efforts have been amplified by strategic networking with other high-profile podcasters and the massive reach of his #1 bestselling book Outlive (Figure 1). In a nutshell, Attia transformed proactive longevity from a niche academic topic into a mainstream, urgent, and highly successful media empire.

Not surprisingly given his success and iconoclastic views, Dr. Attia has provoked highly polarized and conflicted responses from the medical community. On the one hand, many physicians and researchers praise his elite medical credentials and credit him for elevating health discourse, challenging the reactive "sick care" model, and successfully championing evidence-based lifestyle interventions like structured exercise and metabolic management. On the other hand, mainstream medical institutions frequently push back against his Medicine 3.0 protocols -- which advocate for aggressive, expensive interventions like whole-body MRIs and continuous glucose monitors for asymptomatic individuals --arguing that they risk over-medicalization, rely on mechanistic theories ahead of long-term clinical data, and promote an elitist tier of healthcare. As a result, Attia is often seen as a compelling public communicator of preventive health whose general principles win respect, even as his more intensive, expensive, and screening-heavy protocols draw significant caution from parts of the medical establishment.

From a health and medicine perspective, one should separate the person (who has now fallen into some disrepute because of association with Epstein) from the ideas, many of which medical experts agree are interesting, important, and evidence-based. More specifically, it would be a worthy endeavor to try to bridge the gap between Medicine 2.0 and Attia's Medicine 3.0.

Figure 1. Outlive: The Science and Art of Longevity is Peter Attia's #1 best-selling book published in 2023. Overall, the book presents his Medicine 3.0 framework in which longevity is not a search for a miracle cure, but a disciplined, preventive strategy for avoiding frailty and preserving quality of life into old age.