Here is the spreadsheet of 5,455 tweets starting on 4/18 at 4:21PM EST through 4/28 at 6:34PM EST: link. This is obviously more days than the conference, but if anyone wants to do any analysis you can cut out what you wish. There was one unfortunate period of about an hour on 4/23 between ~10:30AM and 11:30AM EST when my computer decided to BSOD and therefore tweets weren’t collected.

I was able to figure out how to find and download images in the tweets and automatically upload them to a tumblr here. There were 271 photos so it is a cool way to view the conference through others’ eyes. I will post my code eventually when I clean it up.

This is what the tweet frequency looked like over time:

Some people include their location in their profiles, so I ran those through google maps and retrieved lat/long coordinates. Here is a USA map of the 628 that google could interpret (darker = multiple people from that location):

The mean number of tweets for each person tweeting to the hashtag was 6.3 (SD = 21). A few people tweeting a lot skewed the average (median = 1):

Here are the top 19 tweeters. Impressive numbers:

And the most popular tweets:

I am not sure how useful this is but here is a word cloud of words appearing in tweets at least 100 times:

This plots each person’s tweets over time and orders them by the first tweet so we can see the rate of people joining over time. It is fairly steady but clearly people who tweeted first to the hashtag tended to tweet the most.

I may do some deeper analyses of the tweets in the future, let me know if you have requests in the comments.

As a short summary of the study, Canada Post employees were screened and if they had the metabolic syndrome they were randomly assigned to either remain under the care of a primary care physician, or remain under primary care and add naturopathic treatment on top of that. The naturopathic group received 7 visits at naturopathic centers over 1 year.

Now, the Canadian College of Naturopathic Medicine themselves say naturopathic medicine includes traditional Chinese medicine (accupuncture, accupressure, massage, herbs etc), homeopathy, “physical medicine” (like chiropracticic, hydrotherapy, stretching), bontanical medicine, along with nutrition and lifestyle counseling. As poignantly stated by Timothy Caulfield and Christen Rachul, with the exception of evidence-based nutrition and lifestyle changes these therapies do “not support the proposition that naturopathic medicine is a science and evidence-based practice.” Their review of naturopathic practices in Canada found disturbing treatments such as colon cleansing, hair analysis, detoxification or biotherapeautic drainage, IV therapies, chelation therapy and many others lacking scientific substantiation, and these are used to treat a wide variety of diseases. The basis of naturopathy is a mystical and unscientific belief in “vitalism”, and the practice has homeopathy at the core, which is literally using water to treat disease.

In contrast, here is the “naturopathic” treatment used in the new study:

Though you could debate a couple of them, these are hardly controversial interventions!

Not surprisingly, in the 207 who completed the study, the proportion of individuals in the control (only primary care) group was 48% and in the naturopathic group it was 32%. Cardiovascular risk, as assessed by the Framingham risk score suggested a ~3% reduction in risk reduction in the naturopathic group. It is difficult to estimate if baseline differences had an influence on this as they don’t report p-values:

“The groups were similar at baseline, although the naturopathic group had a nonsignificantly higher prevalence of metabolic syndrome, larger waist and hip circumference, and greater weight. The naturopathic group reported significantly more weekly minutes of moderate exercise.”

They did not document treatments by the primary care physicians, and the added treatment may have influenced the primary care:

“naturopathic doctors measured risk factors 3 times during the course of the study, and participants were encouraged to report the results to their family physicians.”

And there was no blinding.

In an accompanying editorial, the journal’s (CMAJ) deputy editor, Matthew Stanbrook provides this warning ahead of inevitable naturopathic claims:

“We can learn nothing new from this trial about supplements or any other individual component of care, because the trial was not designed to allow their evaluation.”

Yet for some reason also writes:

“The results of Seely and colleagues provide proof of principle that some aspects of cardiovascular prevention could feasibly and effectively be delegated to naturopaths.”

In my opinion that is a dangerous path to suggest, as it is clear that the standardized and limited treatments in this study do not reflect real naturopathic practice.

It is good to have further support that diet and lifestyle changes can reduce cardiovascular risk. But this study was clearly designed to use non-controversial treatments to legitimize naturopathy. Diet and lifestyle counseling can be provided from dietitians, and their colleges and associations do not promote the non science-based treatments as do naturopaths.

From www.hapilabs.com/press-release.asp (underlining mine)

I was surprised at these claims, given that I have read some of the research on eating speed and weight and found it mixed, and overall unconvincing. My post on that is here. The research spans back to the 1960s, but curiously, the company only cites research since 2002:

Taken from www.hapilabs.com/faq.asp

• Ok, let’s review these 3. The 2004 “demonstration” by the North American Association for the Study of Obesity was apparently a conference presentation, and doesn’t seem to have been published after. At least, that is what I can garner from the lead author’s CV. So that doesn’t help us much.
• The “2006″ study was published in 2008, and it was only on women, and it wasn’t the University of Pennsylvania (University of Rhode Island)!. A press release came out in 2006 when it was presented at a conference. Interestingly, this is reported incorrectly in lay articles on the web (example that also includes the 2004 reference from above), so it makes me think the company didn’t read the actual research. Perhaps the company misread the comment from Barbara Rolls in this news report on the study who is from Pennsylvania State and wrote University of Pennsylvania? Also note that the meals were ad libitum.
• The “2011″ cross sectional study was actually published in 2008. It is what it is, but prospective observational and interventional trials are needed to support causality. The big risk here is that everything is self-reported- eating speed might be pretty subjective. Maybe the company got the 2011 date by getting the information from this Daily Mail article which was published in 2011.

So, not a good start for usefulness/accuracy. They link the slowcontrol.com website so I went there to see what else they cite. Note that there are references there for eating speed and glycemia and metabolic risk, though I only critique their selections for energy intake and weight below. The selected references for the former are as weak and unconvincing as those chosen to support the weight claims.

I first clicked under research posted for “Eating Slowly & Energy Intake”, where they only list 2 references:

Taken from www.slowcontrol.com/index.php/energy-behavior/39-eating-rate-during-a-fixed-portion-meal-does-not-affect-postprandial-appetite-and-gut-peptides-or-energy-intake-during-a-subsequent-meal

• The first reference links to a study (PDF) that uses a fixed-portion meal instead of ad libitum and found that eating speed does not influence satiety or calorie intake- it directly refutes the product’s claims!. This suggests the product would be ineffective in these circumstances (like dieting perhaps)! At least they cite conflicting research, but given that this is the only conflicting study they cite and they don’t comment on it, it seems more like an accident.
• The second study they cite used a vibrating pager to reduce eating speed in 3 autistic participants who have a history of very rapid food consumption. Perhaps this product could be beneficial in this subset of people, for as the study says, “[t]his behavior can lead to serious health problems, such as vomiting and aspiration, and may be socially stigmatizing”. But the company is marketing this to the general population so this is not relevant to the vast majority.

Next, I clicked on “Eating Slowly & Overweight”, which contained 5 references:

From www.slowcontrol.com/index.php/eat-slow-overweight

• The first reference is a systematic review of eating behaviors and overweight from 2012. Here is a direct quote from the paper:

“Finally, no longitudinal studies were found regarding the influence on body weight of irregular meals, eating until full (in children) and eating quickly (in adults). Overall, the small evidence on these three eating behaviours does not allow to establish their effect on obesity, even though this relation is biologically plausible (e.g. eating quickly or until full may lead to increased energy intake).

The authors note that behaviors need to be studied simultaneously a framework and existing research is limited in this manner. It is unlikely in my opinion that only targeting one behavior like eating speed will be effective as a general recommendation.

• The second reference is to a position paper (PDF) by the Committee on Nutrition of the European Society for Paediatric Gastroenterology. There is no mention of eating speed in this paper.
• The third reference is to the cross sectional study I discussed above, which is preliminary research.
• The fourth is to a WebMD article discussing that “2004″ study used above that was a conference presentation and doesn’t seem to be published. Of note, we learn there were only 6 participants, greatly limiting interpretation.
• The final reference is to the 2008 trial in women that was discussed above. This is the only trial so far that seems to support the product, though note it was only women, a small sample size, and the test conditions were not exactly “real-world”:

“Under the quick condition, subjects used a large spoon (soup spoon) and were told to consume the meal as fast as possible with no pauses between bites. However, they were instructed to not eat so fast that it was uncomfortable for them. During the slow condition they were instructed to take small bites, put down the spoon between each bite and chew each bite 20 to 30 times. A small spoon (teaspoon) was provided with these meals.”

And, this only tell us that “combined techniques of taking small bites, pausing between bites, and chewing thoroughly can decrease the rate of food ingestion, and enhance effects on satiation, decreasing energy intake.”

The discussion of this paper is really good in showing how contradictory the research is:

“Several explanations are possible for these observed relationships between slow eating and reduced food intake, all of which need additional investigation, and none of which are mutually exclusive. First, prolonged meal duration can allow more time for physiological satiety signals to develop before too much energy has been consumed. However, empirical evidence must be provided before this reason can be concluded. Secondly, this protocol used combined strategies to slow eating pace, including small bites, thorough chewing, and pausing between bites. Any of these factors, or synergy between them, might have helped the women to consume less and feel more satiated. For example, the process of chewing itself can stimulate physiological satiety signals  ^{[28] , [29]} ). In addition, eating slowly allows time for consuming water along with the meal. Indeed, the women drank more water under the slow condition. Although it is possible that this may have increased stomach distension, and thus induced satiety ( ^[3] , [7] ), not all studies have shown that water consumed with a meal reduces energy intake (30). Thus, research is underway in our laboratory to clarify this. Finally, eating slowly allows more time for enjoying food, as supported by the pleasantness ratings in this study. It has been suggested that by slowing down and savoring the sight, smell, taste, flavors, texture, and mouthfeel of food, and sensing hunger being suppressed, more satisfaction can result from fewer calories ( ^{[31] , [32]} ). However, Kaplan (9) reported that subjects gave higher taste ratings after meals eaten at faster rates as compared with slower rates.”

 

What they don’t cite

Among some of the conflicting studies against the eating speed/weight hypothesis that I discussed in my post from 2009, one small study from 1997 counter-intuitively found that participants in an ad libitum context actually eat more when made to pause between bites, and rated hunger and fullness as less satisfied compared to no pauses. The authors suggested that the pauses increase frustration and the increased calorie intake is the result of this.

Additionally, as I noted in my last post, there may be differences in efficacy of reducing eating speed (if there is efficacy) by gender and body weight (example, example).

Importantly, even if reducing eating speed reduced calorie intake, there is no data supporting long-term weight loss. Many interventions can acutely reduce calories, but long-term may not lead to any change. There is even one small trial that showed a reduced calorie intake with a slowed rate of eating but they couldn’t maintain this long-term.

In many papers I looked at, authors suggest that it is unlikely that targeting a single behavior like eating speed is a good idea. Instead, other characteristics of food likely confound or influence the relationship with calorie intake. For a good lay article, see this by Barbara Rolls.

It is pretty clear that this company did not carefully review eating speed research and simply cherry picked some studies they think support their product by skimming media reports.

The company’s press release had this headline:

“World’s First Smart Fork That Helps You Lose Weight by Eating at the Right Time and Pace”

I would suggest this instead for accuracy:

“World’s First Smart Fork that won’t help if you are on a diet, might work a bit in circumstances when you can eat as much as you want though there is conflicting data and there may be differences in efficacy by gender and body weight and also we don’t have data on actual long-term weight loss. Additionally, there is no evidence that this fork consistently leads to slower eating nor changes your habits in the long-run”

But I guess that wouldn’t sell very well.

I also want to take the opportunity to plead to tech blogs: new gadgets may seem cool, but please do the work to see if a company’s claims go beyond the data and don’t just regurgitate the press release. This level of egregious misrepresentation of references is on par with many companies in the dietary supplement industry.

After removing 6 people who tweeted more than 3,200 times in the last year and therefore I couldn’t collect a full year’s worth of tweets, here is a plot of tweet counts per week of the 28 others. It seems there is a slight summer spike:

There were 20,383 links among 41,479 tweets. I wrote code to expand the links that were shortened, and incorporated the number of times the link was retweeted.

So here are the most popular links of 2012 from these 34 people. I added notes to the top 10:

So, it over-represents some stories but for a really simple method it isn’t bad for showing a sampling of popular links.

Happy new year!

Methods

The tweets were collected and added to a CSV file by some Python code that I wrote, using Twitter’s streaming API. The hashtags/keywords that were followed were: #prop37, #noon37, #yeson37, GMO, prop 37, #labelgmo, and #righttoknow. I collected the: tweet, ID, datestamp, username, if a RT, user Bio from profile, profile link, follower/following/listed counts, total tweet count, and location if listed in the profile. The total dataset consisted of 253,861 tweets from October 25 to December 6. It is uploaded here if others want to use it. I decided to limit analyses to only tweets containing prop37, #prop37, #yeson37, or #noon37. This yielded 55,537 tweets that I played around with in R.

Describing the Tweeters

Here is a graph of the frequency of tweets with these hashtags by date. As you can see my internet went out several times for significant periods, but I don’t expect it to change much. I also collected the locations of each person tweeting if they set one in their profile. With these, I geocoded them to lat/long coordinates using the google maps API and did a couple plots (includes retweets). These are not adjusted for population density, but still they show much of the activity in California as expected: Here is a density map of the world, of about 60% of the locations (for whatever reason R couldn’t handle more on my computer). The log of the tweet count makes it a little easier to distinguish:

Frequencies

The main reason I collected the data is to try to attempt to see who is driving Prop37 activism on twitter. So I ran a number of frequencies to describe the population. On average, people tweeting with the prop37 hashtag tweeted almost 3 times, and the person who tweeted most did so 339 times.

If we run a frequency of the tweets, we see that most tweets were original or were not retweeted much. One tweet was retweeted 442 times.

As the 2 hashtags promoted by each campaign were #yeson37 and #noon37, I extracted the tweet count that contained these. The result was surprisingly 1-sided (includes all retweets). However as I show later, the #yeson37 count is artificially inflated by fake accounts.

And if you look through the #noon37 tweets, many of them are clearly in favor of Prop 37 but just adding both hashtags. It seems like this dataset is almost completely representative of people in support of Prop37.  To get a more accurate picture, I wrote a script to randomly poll a sample of the people who added #prop37 to their tweet to see which side they supported, but my account was quickly banned by twitter. I also wanted to poll out of the overall twitter feed those people who list California in their location to see if this could predict the election results. If anyone knows if twitter makes exceptions to do things like this let me know, but I assume not.

So now lets look at the most frequent Prop 37 tweeters. Here are the top 20 after removing fake accounts (read further to see how I determined this). For the top 5, I went through some of the tweets and picked an example of poor information. This is cherry picking, but they are so egregious that it suggests a pattern.

We can also look at the top retweeted tweets. Here are the top 20. Most are from dubious sources, some which I’ve noted in the right-hand column.

Tweet	Count	Notes
RT @ninadobrev: If youre a Mom DEMAND GMO labels for you &amp amp; your family. #VoteYesProp37 (&amp amp;RT if youre a parent who demands it too …	442	Celebrity
RT @dannymasterson: BTW. #monsanto is spending a million $ a day to fight #prop37 . Chemical companies should not control our food	267	Celebrity
RT @mercola: It breaks my heart to hear that #Prop37 is losing the vote t.co/IvvOgcnR	159	Quack & majority funder of Prop 37
RT @charliesheen: GO NINA! Winning!! RT @ninadobrev If youre a Parent DEMAND GMO labels for you &amp amp; your family. RT #VoteYesProp37 htt …	142	Celebrity
RT @MotherJones: “In short, #Prop37 got crushed under fat stacks of cash: its supporters raised $8.7 mil vs. $45.6 mil for opponents”: h …	139	Many anti-GMO articles
RT @mouselink: And meanwhile in #California, it looks like @MonsantoCo has won out over the health of the entire human race. #prop37	139	?
RT @AmberLyon: California voters: “The Video Monsanto Does Not Want You to See” #Prop37 t.co/uRYhyhz3	131	Journalist
RT @vanderjames: Hey CA – #Prop37 would require that food with GMO’s be labeled. Vote Yes on 37 to know what’s in your food. #Simple htt …	125	Celebrity
RT @SophiaBush: CA Voters – regardless of who you’re voting for in the POTUS race, PLEASE Vote YES on Prop37. We have a right to know wh …	124	Celebrity
RT @mercola: The more you can avoid genetically modified foods, the healthier you and your family will be. Vote “Yes” on #Prop37! http:/ …	122	Celebrity
RT @BarbraStreisand: On Election Day I support the right to know with a #YesOn37 vote to #LabelGMOs. Get informed &amp amp; rock the vote: h …	119	Celebrity
RT @mercola: Today, #California will be voting on the most important health decision in history: #Prop37. t.co/oiZH5JbH #yeson37 …	117	Quack & majority funder of Prop 37
RT @AbbyMartin: It’s sad how much a disinfo $ campaign can influence people. 90% in US WANT to label #GMOs, yet CA polls show #Prop37 NO …	116	?
RT @vanderjames: Huge food corps have spent $40 million to scare you into thinking #Yeson37 will raise your grocery bills. It won’t. ht …	113	Celebrity
RT @kennyflorian: RT &amp amp; don’t support these “All Natural Companies” who helped defeat #Prop37 #NoToGMO #SellOuts t.co/mES42inH	112	MMA Fighter
RT @mercola: Your child is not a lab rat. And that’s what eating GM foods are turning them into. t.co/0tSA0uYK #gmos #prop37	107	Quack & major funder of Prop 37
RT @CARightToKnow: #YesOn37 is a true Peoples Movement. 90% of consumers want to #LabelGMOs so they can make informed choices: t …	104	Official campaign
RT @CARightToKnow: Food labels are important. Information is power and that power belongs in the hands of consumers. #LabelGMOs #YesOn37 …	102	Official campaign
RT @d_seaman: Why would you NOT want to know if GMOs are in your food? The corporate smear campaign against #Prop37 should be illegal. H …	99	?
RT @OrganicLiveFood: Syngenta’s #GMO cow food killed many cows &amp amp; protein used 4 cow food maize is also used in our food #YesOn37 htt …	98	Organic restaurant owner

Inflating the Campaign

Due to evidence suggesting that the official campaign promoting prop 37 purchased followers, I was on alert for other abnormalities. It quickly became obvious that there were a large number of fake accounts (with zero followers and following counts) tweeting the same thing over and over on the #yeson37 hashtag- and linking to the campaign’s website. So after expanding the links of all tweets, I counted how many contained #yeson37, contained a link to the campaign website (carighttoknow.org), and had 0 followers: 10,209 from 965 different accounts! It appears that someone (the campaign denies it)- paid for a huge number of fake accounts to tweet the website. If you look through the tweets, they almost look normal, but the screennames are all names with random numbers at the end, are highly repetitive, and include various keywords or hashtags that were trending at the time. This is likely for 2 reasons: 1) to try to get their website out on various popular hashtags to increase awareness, 2) twitter doesn’t allow you to tweet the same thing multiple times within a short time period, so using trending hashtags would slightly change the tweets. Very shady stuff, and all of this is grounds for suspending of accounts. Annoyingly, it made more work for me. So I removed all tweets with follower and following counts of 0 (10,527), leaving 22,578. I used these to explore what links were being tweeted and for associations. Here is additional proof: by plotting the time of each tweet for each screenname, we see that most only tweet a few times and are then suspended. Below that plot is a plot of non-fake accounts.

Sources/Links

I wrote some code to expand all links and running a frequency on them revealed pretty poor top information sources. Here are the top 50. Many suggest health risks from GMO, ignoring consensus, or propagate erroneous stories. For the top 15, I added some notes.

Associations

I ran some word associations to see what terms appeared together most frequently. Because I don’t have much RAM and R was struggling, I took a random sample of 5,000 tweets of the 44,561 (full dataset minus fake accounts) for this.

First I explored words that appeared at least 150 times:

[1] “california” “companies” “democracy” “food” “foods” “gmo” “gmos” “health” “label” “labelgmos” “labeling”
[12] “labels” “monsanto” “movement” “organic” “people” “please” “prop” “prop37″ “support” “video” “vote”
[23] “voters” “yeson37″

Here are some associations (values are how often the words occur together, x100 for %):

“prop37″

“gmo”

So, some evidence that people who wrote GMO in their tweets tend to think they are harmful to health, and not surprisingly Monsanto was a popular word. We can dig a bit further:

“health”

“Terrorism”- really??

“monsanto”

Sentiment

I ran each tweet through the sentiment analysis method described here. It scores each tweet by the net number of positive and negative words that it matches from a list here. The average was just above neutral, because half the tweets were neutral. Interestingly, as I analyzed the ratios of increasing sentiment (positivity and negativity), the tweets became more negative.

This is easier to visualize:

Conclusions

Though it is difficult to draw overarching conclusions without more manual classifying of tweets, tweeters, and sources, and those that I identified (the top ranking of each) may just represent a small portion of the overall activism of Prop 37, they paint a picture of misinformation. Although some in favor of Prop 37 just don’t want corporations controlling the food supply, the top tweeters and sources of information clearly think there are health risks to GM foods. The official “Yes on 37″ campaign did nothing as far as I could tell to correct this thinking and in fact promoted it at times.

So that is just the tip of the iceberg of what could be done with this data I’m sure, and if I come across new ways of digging around I will update. Let me know if you have any ideas. I will also note that Becca Harrison is working on some qualitative analysis of prop 37 tweets, so those results will be more interesting!

Methods

Unfortunately, I didn’t have the foresight to write a script to grab tweets directly from the #FNCE hashtag at the start of the conference, and because there is a limit to how many tweets you can backtrack on, I had to go about it from a very inefficient way. First, I wrote a python script to gather the twitter handles of those posting to the #FNCE hashtag starting at 0300 CST on Saturday, October 7th every 10 minutes through the end of the conference until October 10th (the conference took place October 6-9). This should represent the vast majority of tweeters, only missing those who may have tweeted or retweeted only before 0300 on Saturday and did not tweet for the rest of the conference. However, 40% of people posting to the #FNCE hashtag within the official conference start and end time did so only one time. I removed those that had 0 followers and accounts that I could not access (suspected spammers – there were 132(!)) using another python script. Finally, another script then read and searched each user’s last 400 tweets in their timeline for instances of the #FNCE hashtag. This took about 37 total hours to complete (had to scan over a half million tweets using 3 accounts to get additional API access, sorry twitter!). Tweet information and text containing the hashtag along with tweet ID, datetime, if it was a retweet, links, user bio, website, following and follower count, listed count, total tweet count, and location were added to a CSV file by this script, which I then performed some analysis with & made some graphs with in R.

Results

After restricting tweets to those posted after the official start of the conference up until the official end of the conference, there were 14512 total tweets (including old and new style retweets) from 1570 different people.

Here are the top 20 people who posted to the #FNCE hashtag (includes all retweets by the person):

Sore thumbs?

The mean, median, and standard deviation show that most actually did not tweet a lot:

Similarly, here are the top retweets, retweeted at least 10 times (an imperfect measure as retweets that add a message before the “RT” aren’t counted- this could probably be more precisely quantified by checking if a certain percentage of the tweet is the same as another). The top retweets are dominated by a small group of people:

There were more occurrences of “New-Style Retweets” (retweeting without adding the RT in front of the message) than “Old-Style”:

Because we have all tweets, if each tweet was an “old-style” or “new-style” retweet, and the dates for each, along with information like follower and friend count for each user, we could really do some complex analysis to visualize interactions. But that will have to wait for another time.

But since I collected the home locations of each person who tweeted with the hashtag (if they had one listed in their profile), I wrote a python script to geocode these into coordinates using the google maps API and plotted them on a map with R. Now we can visualize where people are from:

Here is only the US map:

Next I ran a check of frequent terms appearing at least 150 times. I did this for tweets that I stripped of usernames, hashtags, and links. These included all retweets. In alphabetical order:

[[1] “academy” “amazing” “awesome” “blog” “blogging” “booth” “breakfast” “calories” “check”
[10] “conference” “dairy” “day” “diet” “dietitians” “eat” “eating” “education” “energy”
[19] “excited” “exercise” “expo” “fat” “fnce” “food” “foods” “forward” “free”
[28] “fun” “getting” “grains” “health” “healthy” “help” “hope” “industry” “info”
[37] “kids” “krieger” “learn” “look” “looking” “loss” “love” “loved” “media”
[46] “meet” “meeting” “milk” “morning” “nutrition” “obesity” “people” “philly” “protein”
[55] “rdchat” “rds” “recipe” “reduce” “research” “session” “social” “stop” “sugar”
[64] “talk” “thank” “thanks” “time” “tips” “tweets” “twitter” “usa” “veggies”
[73] “weight” “world”

We can visualize associations between some of these terms. Lots of positivity sprinkled in there.

A word cloud isn’t that helpful in this case as most of the terms except for fnce are mentioned are similar frequencies:

We can dig deeper and examine correlations between specified terms. As I was following the tweet stream myself I noticed it was often very conflicted over the sponsors of the conference. I queried to see what terms were most associated with some relevant terms. The numbers under the terms represent the correlation with the queried term.

“sponsors”

“industry”

(Margo was a speaker at a relevant session)

“corporate”

The high association with “elated” is deceiving; in context it was a popular quote “elated” at critical tweets about corporate sponsors. The 3 digit numbers come from tweets/retweets of a session that included times and room numbers.

“healthy”

This only had 1 term that was correlated more than 20%:

And it appears to be from a popular tweet/retweet:

@Fooducate: “Healthy eating: education cannot by itself counteract convenience, pricing, culture, & marketing” #FNCE”

“GMO”

I was disappointed to read that there was an anti-GMO session (edit: “non GMO event” per tweets) (edit 2: this event was off-site and not part of the conference), so I was curious if this was reflected in the tweets. Indeed:

So it appears that there is a trend of dislike toward sponsorship, but this could be my added bias in selecting the terms. We can try a sentiment analysis of all tweets and see how positive or negative they are to add some objectivity. This scores each tweet by the number of positive or negative words it matches (positive or negative 1 per word) (match words from the opinion lexicon here, method based on this great example).

Finally we can plot the tweets by everyone over time to see the consistency of people joining in with their 1st tweet over time. As you can see, the people who tweeted earliest tended to tweet most frequently.

R is powerful but if you are new to it like me, it can be difficult to get it to work. Without the numerous examples shared by bloggers I would not have been able to complete these so quickly (or most likely at all). Special thanks to the following examples: wordclouds, associations, visualizing frequency, geocoding, sentiment.

Please let me know if the comments if you have any ideas or requests for additional things to look for (or tips!).

A new paper was published recently by Elise Reitshamer, Margaret Steffey Schrier, Nancie Herbold, and Elizabeth Metallinos-Katsaras who surveyed the attitudes of AND members about corporate sponsors. The results are worth exploring, so here is a summary.

The authors designed the survey with questions about demographics, membership, and sponsorship opinions. Oddly, the Academy of Nutrition and Dietetics declined to offer them a list of members to make the survey representative of the membership. Elise told me in an email instead they would have had to pay for it but did not have the budget. That the Academy could not assist in research on a question that is beneficial for them to know is unfortunate. So instead they had to contact state or district affiliates, most who agreed to send out emails to members. 2,968 surveys were completed out of a possible 29,500. They note in the discussion of course that self-selection based on interest in the topic may skew the results somewhat, but the demographic results are close to those of the 2009 ADA Compensation and Benefits Survey which used probability sampling.

Results

• Interestingly, only 13% disapproved of all corporate sponsorship, but 68% replied that it depends who the sponsor is
• 83% say they should have a say in who sponsors the AND
• 97% said yes when asked if they AND should verify that the sponsor’s corporate mission is in line to the AND’s before accepting them
• 80% think that sponsorship implies endorsement by the AND
• 47% of members are not willing to pay more in membership costs to reduce sponsorship. But 34% are willing to pay 1-5% more, and 14% are willing to pay 6-10% more. To completely make up the difference in corporate sponsors, I estimated that each member would need to pay an extra $41/year (at least in 2009), which is about a 27% increase (on average). According to these results, only 2% of members are willing to pay 21% or more so this doesn’t seem to be an option.

It is interesting that most of the sponsors are actually heavily rated acceptable with the exception of Mars, Incorporated, Coca-Cola, and PepsiCo which are rated mostly unacceptable.

The authors also found a inverse associations between length of Academy membership and years of practice with whether members should have a say in the decisions of corporate sponsorship, with younger members more likely to want a say.

Respondents were also allowed to provide comments which were coded into themes by the authors, and the top response was that strict policies and guidelines should be in place but sponsors are ok.

The full paper is worth a read for a nice overview on these issues, especially of the policies and sponsorship of other organizations. Notably, the American Public Health Association does not accept donations, and the American Medical Association recently eliminated sponsorship but make it up in advertising.

The Academy states that internal survey results support corporate sponsorship approval by members, but clearly not all are approved and members want input. Will this force some reconsiderations?

Reference

Elise Reitshamera, Margaret Steffey Schriera, Nancie Herbolda, & Elizabeth Metallinos-Katsarasa (2012). Members’ Attitudes Toward Corporate Sponsorship of the Academy of Nutrition and Dietetics. Journal of Hunger & Environmental Nutrition : 10.1080/19320248.2012.704748

If you use Mendeley, thanks to their great example client, I was able to add the ability to push papers from links to your Mendeley library. This can make for a nice to-read list in Mendeley, or enable you add things if you don’t have Mendeley installed. To do so, first you must authorize @ScholarlyBot to access your library. Follow these steps (you only need to do this once):

1. Follow @ScholarlyBot on twitter.
2. Wait for it to follow you back (may take up to 1 hour, or longer if my computer loses connection). This is necessary because it needs to be able to send you a Direct Message and you need to be able to send one back. You can unfollow it after you are authorized if you want.
3. Send it this message (not in a Direct Message): ” class=”autohyperlink” title=”https://twitter.com/nutsci/status/246354640776601600]” target=”_blank”>twitter.com/nutsci/status/246354640776601600]
4. It will then Direct Message you a link; follow the instructions and Direct Message back the verification code.
5. It will Direct Message a notification that you are authorized and you are finished; you won’t have to authorize it again.

Now you can push papers to your library. At the moment, you can send it links to pubmed abstracts like so:

” class=”autohyperlink” title=”https://twitter.com/nutsci/status/245566349332582401]” target=”_blank”>twitter.com/nutsci/status/245566349332582401]

Or if you have a link to a PDF file that is freely available, you can push that as well. Example:

” class=”autohyperlink” title=”https://twitter.com/nutsci/status/246287856681746433]” target=”_blank”>twitter.com/nutsci/status/246287856681746433]

I will add support for links to non-pubmed journal abstracts very soon.

@ScholarlyBot will send you a message if the upload was successful.

You can view the document in your “Recently Added” folder in your Mendeley Library. @ScholarlyBot attempts to find some of the document info for you, such as the title, authors, year, and DOI or PMID. You can then have Mendeley lookup the rest of the details for you and automatically fill them in within the desktop client. However, it is likely that it will not be able to find details for many PDFs- especially if they are new- this is something I will need to continue improving. Please send me reports if it fails or gets the wrong information!

• “No yolk: eating the whole egg as dangerous as smoking?” writes the LA Times
• “Eggs found to be as dangerous for the heart as smoking, new study suggests” says the Global Post
• “Eggs move back to the naughty list” notes The Globe and Mail
• Along with many more

These were prompted by a press release from The University of Western Ontario titled “Research finds egg yolks almost as bad as smoking.” The press release is based on a new study by Drs. David Spence, David Jenkins, and Jean Davignon. Maybe you recognize the university and authors from a similarly scary press release from 2010 titled “KFC’s Double Down or an egg – what’s worse?” In my opinion, both of these press releases are fear-mongering and don’t accurately portray the research. The new one, for example, implies causality in the first sentence, downplays the big limitations until the final sentence, and doesn’t put the study in context of existing research, as press releases often neglect to do. Those things aside, the study is worth looking at with existing research.

Two years ago I explored much of the egg literature and wrote a long post describing it. I concluded that epidemiological studies and short term interventional trials seem to show that eggs are safe for healthy populations, and given that they may be good for weight maintenance for breakfast (because of the protein- eggs have been studied specifically for this), they are inexpensive, and nutrient rich, it seems like a no-brainer to eat them if you enjoy them. Yet, some well designed observational studies showed an increased risk for cardiovascular disease and mortality in diabetics with higher egg intake, and I noted that caveat in the conclusion. I didn’t go into detail about these studies though.

About 6 months later, the review paper that inspired the KFC Double Down to egg comparison was published. The authors blame the egg industry for downplaying the risks of dietary cholesterol and emphasizing that postprandial effects of cholesterol and LDL oxidation rather than fasting cholesterol be considered, and explore dietary cholesterol and cardiovascular research. It is worth a read to gain their perspective. Incidentally, they hint at their new study:

“In our opinion, stopping egg consumption after a myocardial infarction or stroke would be like quitting smoking after lung cancer is diagnosed: a necessary act, but late.”

Onto their new paper, titled “Egg yolk consumption and carotid plaque.” To summarize, the study consisted of 1231 patients who had ischemic attacks or strokes who were attending vascular prevention clinics. At referral, the patients filled out a questionnaire that included smoking and egg yolk consumption (eggs per week and number of years consumed). Given that someone with vascular disease is probably aware of eggs and their cholesterol content and government recommendations, and our memories are bad as demonstrated on case-control studies, it seems like this retrospective recall is probably very biased. Unfortunately, data on alcohol intake, other dietary factors, and exercise were not accurately quantifiable so they weren’t included, another major limitation as I will discuss later. The patients had their carotid total plaque measured at referral as well. The authors split the data into quintiles of egg consumption. With each increasing quintile (and smoking group), plaque area increased after adjustment for age. Eggs remained associated with plaque area after adjusting for sex, total serum cholesterol, systolic blood pressure, diabetes, BMI, and smoking. They found no correlation between egg consumption and smoking. Shockingly, the effect from the upper quintile of egg consumers was equivalent to 2/3 of the upper quintile of smoking.

So does this provide evidence that eggs are causing plaque buildup and heart disease?

The authors casually mention that a weakness of their study is that they don’t have exercise or other dietary data. But they don’t comment on other research that has found tendencies of egg eaters to have different lifestyle habits. For example, Hu et al (1999) looked at eggs and cardiovascular disease in the Health Professionals Follow-up Study and Nurses’ Health Study cohorts. There is a consistent decrease in physical activity across quintiles in the HPFS as subjects consume more eggs (not in WHS though), and energy intake consistently increases across quintiles in both cohorts (the difference between highest and lowest is 500+ kcal). Red meat intake increases consistently in the HPFS, which has a small relationship with cardiovascular disease. Data from the Physicians’ Health Study I (Luc Djousse et al, 2008) shows a similar pattern for red meat, alcohol consumption, and (strikingly/oddly) whole milk. They were also less likely to exercise more than 1 time per week, albeit only slightly. I could continue, but the point is we know there are differences in the lifestyle habits of people who eat more eggs from other research. I understand that the data simply isn’t there, but I certainly wouldn’t make such strong conclusions knowing this as Dr. Spence did in the press release as:

“What we have shown is that with aging, plaque builds up gradually in the arteries of Canadians, and egg yolks make it build up faster — about two-thirds as much as smoking. In the long haul, egg yolks are not okay for most Canadians.”

By the way, there are two nice tables in this paper on page 405 & 406 with relative risks of various lifestyle habits that influence coronary heart disease risk. Obviously, diet quality (which can be estimated with the Healthy Eating Index or other tools), exercise, alcohol have strong influences. The new study measures none of these or the numerous other differences that could be apparent.

Also not evident in the new paper is a discussion on mechanisms in which eggs are doing harm. Between quintiles of egg consumption, HDL, LDL, triglycerides, and total cholesterol are not different. This isn’t surprising based on other research I’ve read. As I noted before in their previous review paper the authors lay out mechanisms independent of these measures that could influence risk; it is too bad none of these were quantified for this study.

Even though I don’t think this paper is strong enough to demonize eggs, there are still other well designed prospective studies that suggest risk for diabetics, though there are limitations here as well. I will quote from the 5 papers that Spence, Jenkins, and Davignon cite as evidence for harm:

Hu et al (1999): “In subgroup analyses, egg consumption appeared to be associated with increased risk of CHD among individuals with diabetes. This result should be interpreted cautiously due to numerous subgroup analyses, but the consistency of the association in the 2 cohorts argues against a chance finding. The increased risk may be related to abnormal cholesterol transport due to decreased levels of apolipoprotein E³⁷ and increased levels of apolipoprotein C-III³⁸ among patients with diabetes.”

That study found no increased risk for healthy people at 1 egg per day.

Djousse and Gaziano (2008):“At present, we can only speculate that perhaps among diabetic subjects, dietary cholesterol might lead to a less favorable lipoprotein profile in terms of serum concentration and particle size, with a shift to smaller and dense LDL particle size, leading to accelerated atherosclerosis and its complications. Testing of such hypothesis in an experimental design among diabetics is warranted.

Our study has additional limitations. We cannot exclude unmeasured confounding or residual confounding as possible explanation of the observed positive association among diabetic subjects. In particular, we were not able to examine the effects of saturated fat, markers of insulin resistance, lipids, and other nutrients or relevant biomarkers on the observed association.”

This was not as strong as a design as other studies and the authors acknowledge it. They have a good discussion on the diabetic issue that I recommend reading. Also, in healthy male physicians, up to 6 eggs per week was not related to cardiovascular disease and mortality but 7+ was associated with a small increased risk of total mortality.

Eckel (2008): I discussed this editorial in my previous egg blog post- it points out the dramatic differences in consumption patterns in these physicians compared to the general population. Also of note:

“It is of interest that the trend to higher risk in subjects with diabetes begins at lower egg (yolk cholesterol) intakes. The expected effect on plasma LDL cholesterol would be expected to be small. It is noteworthy that this relation between egg consumption and CVD risk in patients with diabetes is not a novel finding (7, 8). Cholesterol absorption has been shown to be higher in patients with type 1 diabetes (9) but not in patients with type 2 diabetes (10). It is not stated—but may be presumed—that most of the subjects with diabetes in the study by Djoussé and Gaziano had type 2, not type 1, diabetes.”

Trichopoulou et al (2006): This seems like a well designed study that showed exercise reduces mortality in diabetics and saturated fat and eggs increase it- but they only had 1013 subjects, 4579 person-years, and only 80 total deaths!

Djousse et al (2009): This one suggests eggs increase the risk for diabetes, using large cohorts and lots of data. But it is a novel finding and there is a lack of good evidence for biological mechanisms.

On the other hand, Spence, Jenkins, and Davignon don’t cite the following prospective study that found no problem in diabetics:

Scrafford et al (2011): Using NHANES data, this study included a subgroup analysis on diabetics which found no association with eggs and coronary heart disease or stroke mortality, though there were very few deaths in the diabetic group so statistical power is limited.

I’d love to completely absolve eggs from any wrong-doing, but the data in diabetic populations can’t be ignored, and there are some plausible mechanisms. I certainly do not think it is strong enough right now to recommend that these populations stay away from eggs, but it deserves further study, and randomized trials aren’t practical in this case so we need more well designed prospective studies.  But I do still think in healthy people the evidence that they are safe greatly outweighs that suggesting they are of risk.

It does 2 3 4 5 things:

1. You can @ it a Google Scholar query and it will check if the first result has an indexed full text available for free. It sends back the result of the check. Because it only grabs the first result, it helps to include information like an author, year, or journal if the title is generic. You don’t have to be following the account to do this. Note: put the ‘@ScholarlyBot’ before the query.
1. (Added 8.29) Now you can @ it pubmed or journal article links and it will check for the full text. Pubmed links will be more reliable until I am able to test further. Be sure to begin the tweet with the @ symbol.
2. If you follow the account, it will follow you back and monitor your tweets for links to the pubmed database, get the abstract title, then query Google Scholar for the full text. Only if it finds a full text does it send you a link. Edit 8.17: I changed this to reply if it doesn’t find a link as well. Just unfollow it if you want to stop receiving messages. Note that there is a delay in Google Scholar indexing new abstracts and full texts (I don’t know if it is variable or not and I don’t think they publish this information), so keep that in mind when it searches papers from this year.
3. (Added 8.27) If you are following the account, it will now scan for non-pubmed journal links and search Google Scholar for the full text. At this point it seems fairly reliable in my testing, but since there doesn’t seem to be a standardization in how journals encode their data, I’ll continue monitoring and improving.
4. (Added 9.14) You can push abstracts and PDFs to your Mendeley library. See this post for details.

There are a couple limitations: twitter restricts the number of API calls that can be made per hour, so there will be a small delay after sending it a query or when it is checking for pubmed links (at most ~20 seconds). It also currently runs from my laptop, so I can’t guarantee it will be working all the time. I hope to move it to a platform in the cloud soon. There are numbers in parentheses after the messages- twitter shoots back an error message if a duplicate tweet has been sent too soon after the last, so this is just a way to get around that. I tested it pretty thoroughly, but can’t promise that everything works perfect. If you see some strange behavior please let me know.

It isn’t the most amazing thing, but now that I know how to grab and manipulate twitter data it will help for future projects. I’ll continue to develop it as I think of how it could be more useful, but if anyone has any ideas let me know in the comments!