Clever food hacks from Cornell Food Lab might all be fake

Have you heard that serving your food on smaller plates will make you eat less? I know I have. I even bought smaller plates for our kitchen when I first heard about that study, which was published in 2011.

And did you know that men eat more when other people are watching? Women, though, behave exactly the opposite: they eat about 1/3 less when spectators are present. Perhaps guys should eat alone if they're trying to lose weight.

Or how about this nifty idea: kids will eat more fruits and vegetables at school if the cafeteria labels them with cool-sounding names, like "x-ray vision carrots." Sounds like a great way to get kids to eat healthier foods.

Or this: you'll eat less if you serve food on plates that are different colors from the food. If the plate is the same color, the food blends in and it looks like you've got less on your plate.

And be sure to keep a bowl of fruit on your counter, because people who do that have lower BMIs.

Hang on a minute. All of the tips I just described might be wrong. The studies that support these clever-sounding food hacks all come from Cornell scientist Brian Wansink, whose research has come under withering criticism over the past year.

Wansink is a professor at Cornell University's College of Business, where he runs the Food and Brand Lab. Wansink has become famous for his "kitchen hacks" and healthy-eating tips, which have been featured on numerous media outlets, including the Rachel Ray show, Buzzfeed, USA Today, Mother Jones, and more.

Last week, Stephanie Lee at Buzzfeed wrote a lengthy exposé of Wansink's work, based on published critiques as well as internal emails that Buzzfeed obtained through a FOIA request. She called his work "bogus food science" and pointed out that

"a $22 million federally funded program that pushes healthy-eating strategies in almost 30,000 schools, is partly based on studies that contained flawed — or even missing — data."

Let's look at some of the clever food hacks I described at the top of this article. That study about labeling food with attractive names like "x-ray vision carrots"? Just last week, it was retracted and replaced by JAMA Pediatrics because of multiple serious problems with the data reporting and the statistical analysis.

The replacement supposedly fixes the problems. But wait a second: just a few days after that appeared, scientist Nick Brown went through it and found even more problems, including data that doesn't match what the (revised) methods describe and duplicated data.

How about the studies that showed people eat more food when others are watching? One of them, which found that men ate more pizza when women were watching, came under scrutiny after Wansink himself wrote a blog post describing his methods. Basically, when the data didn't support his initial hypothesis, he told his student to go back and try another idea, and then another, and another–until something comes up positive.

This is a classic example of p-hacking, or HARKing (hypothesizing after results are known), and it's a big no-no. Statistician Andrew Gelman took notice of this, and after looking at four of Wansink's papers, concluded:

"Brian Wansink refuses to let failure be an option. If he has cool data, he keeps going at it until he finds something, then he publishes, publishes, publishes."

Ouch. That is not a compliment.

Soon after Gelman's piece, scientists Jordan Anaya, Tim van der Zee, and Nick Brown examined four of the Wansink's papers and found 150 inconsistencies, which they published in July, in a paper titled "Statistical Heartburn: An attempt to digest four pizza publications from the Cornell Food and Brand Lab." Anaya subsequently found errors in 6 more of Wansink's papers.

It doesn't stop there. In a new preprint called "Statistical infarction," Anaya, van der Zee and Brown say they've now found problems with 45 papers from Wansink's lab. Their preprint gives all the details.

New York Magazine's Jesse Singal, who called Wansink's work "really shoddy research," concluded that

"Until Wansink can explain exactly what happened, no one should trust anything that comes out of his lab."

In response to these and other stories, Cornell University issued a statement in April about Wansink's work, saying they had investigated and concluded this was "not scientific misconduct," but that Cornell had "established a process in which Professor Wansink would engage external statistical experts" to review many of the papers that appeared to have flaws.

And there's more. Retraction Watch lists 14 papers of Wansink's that were either retracted or had other notices of concern. Most scientists spend their entire careers without a single retraction. One retraction can be explained, and maybe two or even three, but 14? That's a huge credibility problem: I wouldn't trust any paper coming out of a lab with a record like that.

But how about those clever-seeming food ideas I listed at the top of this article? They all sound plausible–and they might all be true. The problem is that the science supporting them is deeply flawed, so we just don't know.

Finally, an important note: Brian Wansink is a Professor of Marketing (not science) in Cornell's College of Business. He is not associated with Cornell's outstanding Food Science Department, and I don't think his sloppy methods should reflect upon their work. I can only imagine what the faculty in that department think about all this.

Does white wine give you skin cancer?

Many studies have shown that drinking wine, especially red wine, seems to have modest benefits for heart health, as long as you drink it in moderate amounts. A study published this week, though, offers a more worrisome message: white wine might increase your risk of skin cancer.

The bottom line: a daily glass of white wine carries a 13% increased risk of melanoma, one of the deadliest forms of skin cancer. Surprisingly, though, red wine did not carry the same risk. In fact, when the authors separated out beer, red wine, white wine, and other forms of alcohol, only white wine carried any risk for melanoma. 

Should you cut back on white wine based on this new finding? I read the study to find out more.

The new study, led by Eunyoung Cho from Brown University, appeared in the most recent issue of a journal published by the American Association for Cancer Research. It's a large, carefully-done analysis of survey information from over 210,000 health professionals, about two-thirds of them nurses, followed over an 18-year period. The 210,000 participants came from 3 studies: 2 studies of nurses and one of male health professionals. About three-quarters of the participants were female, and the vast majority were white (which means the findings might not apply to non-white populations).

One feature I always look for in an analysis like this is a dose-response effect. If the effect is genuine, then higher doses (in this case, more alcohol) should increase the risk. The authors here found that people who drank the most alcohol had the greatest increase in risk. The study also controlled for all the usual confounding factors, such as family history of cancer and smoking. 

In some subgroups–those who drank the most white wine–the increased in relative risk of melanoma was 50% or more. Keep in mind, though, that this is "relative risk." I'll get to that in a minute.

When the researchers looked at where the cancer occurred, they found that the risk of melanoma was "far greater" on areas of the body that don't receive regular sun exposure, such as the stomach and back, compared to the arms or neck where people get the most sun. This result suggests that the effect of alcohol is unrelated to sun exposure.

There are several important caveats about this study. Perhaps the biggest is that the mechanism is unclear: how does white wine cause melanoma? The authors suggest that the effect is due to acetaldehyde, a cancer-causing compound that is present in wine. However, both red and white wine contain this compound, so it's not clear why white wine would carry a greater risk. It's also not clear why acetaldehyde should cause skin cancer more than other cancers, and why it should be associated with skin cancer specifically on areas of the body that are not exposed to the sun. 

Before you panic, let's revisit the actual amount of risk here. The 13% increase in risk does not mean that you have a 13% chance of getting melanoma from drinking white wine. I looked at the raw numbers in the study, and the overall rate of melanoma in non-drinkers (about 71,000 people in this study) was 0.61%. About 26,000 people reported drinking 10-20 grams of alcohol per day (the equivalent of 1 or more glasses of wine), but the study didn't report how many of these people drank white wine. Nonetheless, in this group the rate of melanoma was 0.85%. So in raw numbers, the increased risk for melanoma was 0.24%.

That's a rather small number, but it's still worrisome. The National Cancer Institute warns that alcohol is linked to several types of cancer (although not melanoma), especially for those who drink excessively. The new study reinforces that warning, and suggests that for those who drink wine, red wine might carry fewer risks than white.

I'm still a bit skeptical, because we have no good explanation for why white wine–but not red wine, beer, or alcohol–would cause skin cancer.  But still, when you reach for a glass of wine this holiday season, perhaps you should choose red instead of white.

Is being a little bit obese unhealthy?

Everyone knows that being obese is very bad for your health. But how overweight do you have to be before you should worry? A new study covering millions of people attempts to answer this question.

The short answer: being a little bit fat isn't so bad, especially if you're already a senior citizen, but the fatter you are, the shorter your life expectancy. Let's dive into the details.

The new study, published in The Lancet, is a combined evaluation (a meta-analysis) of 239 studies that included over 10 million people from four continents: Asia, Australia, Europe and North America. All the studies followed their subjects for a long time, averaging nearly 14 years of observation. The authors (a large consortium called "The Global BMI Mortality Collaboration") wanted to exclude people who might have already been sick, so their study only looked at people who (a) had never smoked, and (b) who lived at least five years after the study began.

This left them with nearly 4 million people, of whom 385,879 died at some time during the course of the study. From this large data set, the researchers computed the risk of death as a function of body mass index (BMI).

[Aside: BMI is a simple function of your height and weight. For example, someone who stands 5'11" and weights 170 has a BMI of 23.7. A height of 5'6" and weight of 150 gets you a BMI of 24.2. You can calculate your own BMI using this calculator.]

The study divided people into six groups:

• underweight, BMI 15–18.5
• normal, BMI 18.5–24.9
• overweight, BMI 25–29.9
• obesity grade 1, BMI 30–34.9
• obesity grade 2, BMI 35–39.9
• obesity grade 3, BMI 40 or above

The main outcome that they studied was mortality (death) from any cause. Of course, one can argue that this is too simplistic, since if someone dies from, say, an auto accident, it probably wasn't due to their weight. But the results were consistent across all four continents, which argues that the study design was probably good. Here are the main findings for each group:

• BMI 15–18.5: 47% increased risk of death
• BMI 18.5–24.9: no increase (normal)
• BMI 25–29.9: 11% increased risk of death
• BMI 30–34.9: 44% increased risk of death
• BMI 35–39.9: 92% increased risk of death
• BMI 40 or above: 171% increased risk of death

Another way to describe these hazard ratios is this: with a BMI above 40, people are 2.71 times as likely to die during any particular time period as people with a normal BMI. 

If these numbers seem scary, keep in mind that this is relative risk, not absolute risk. So an 11% increase in risk might mean that your chance of dying increases from 1% to 1.11%; it certainly doesn't mean you have an 11% risk of dying. 

To put some real numbers on this risk, consider this comparison: out of 1,075,894 people with a near-optimal BMI between 22.5–25.0, 98,833 died during the course of the study, or 9.2%. (Remember that these data come from 239 different studies, and the average length of followup is 14 years. So fewer than 1% of this group died per year.) Compare this group to people with obesity grade 1, or BMI from 30–35. That group had 330,840 people, of whom 37,318 died, or 11.3%. After various adjustments, this translates into a 44% relative increase, but the actual mortality rate, per year, was about 0.66% versus 0.81% in the two groups.

One mildly positive note: if you're already 70 or older, having a BMI from 25–30 has almost no effect; in other words, it's okay to be a little plump when you're older.

A word of warning to men: the ill effects of obesity are much stronger in men than in women. The study breaks down those hazard ratios by sex, and in each range the risk is higher for men than for women. So for example, if you're a woman with a BMI of 30–35, your hazard ratio is 1.37 (37% higher risk of death), but for men it is 1.70 (70% increase). 

There are many, many more details in the study, including a breakdown of how BMI is associated with four major causes of death (heart disease, cancer, respiratory disease, and stroke), and if you're interested in those, you should read the study. 

Of course, one can think of many caveats to these findings: people die from all sorts of illnesses, and many of them are not caused by being overweight. For any individual case, we might not be able to say whether someone's weight had anything to do with their illness. Nonetheless, this very large study shows clearly that the more obese you are, the greater your risk of dying. That's precisely what we would expect if obesity was causally linked to mortality.

So is it okay to be a little bit fat? The answer is probably yes: people with a BMI of 25 might view themselves as "a bit" fat, even though they are not overweight. But very high BMIs (and very low BMI, below 18.5) are definitely unhealthy. Unfortunately, no one has an easy answer to the problem of losing weight (despite what you might have heard from Dr. Oz), but if you do have a dangerously high BMI, reducing it will likely be good for your health.

Does a 3-day fast reset your immune system?

I’ve been hearing new reports lately that sound an awful lot like pseudoscience: that fasting for an extended time, two days or longer, can reset your immune system and provide other health benefits. I first heard about this when a friend and colleague at Stanford University announced he was going on a 3-day fast. To explain, he pointed to a recent study, which I’ve now read.

Diet advice, including fasting-based diets, can be found all over the Internet, and much of it is nonsense, so I was very skeptical about these latest claims.

This time, though, there might be something to it. The scientific study that my colleague told me about was published back in June by USC’s Valter Longo, who studies aging and longevity. In this paper, Longo and colleagues described remarkable metabolic changes that occurred as a result of prolonged fasting. They found that fasting for 3 days or longer–drinking only water and eating less than 200 calories per day–can truly “reset” some components of your immune system. 

The research looked at both mice and humans. (It’s far easier to run the experiments in mice, of course, but we can’t always trust that the same effects will occur in humans.) In both species, fasting lowered white blood cell counts, which in turn triggered the immune system to start producing new white blood cells. White blood cells (or lymphocytes) are a key component of your body’s immune system.

Longo’s hypothesis is that fasting (or starvation) forces your body to “recycle a lot of the immune cells that are not needed” which explains the drop in the white blood cell count. Two of the key mechanisms are an enzyme called PKA and a hormone called IGF-1, both of which are reduced by fasting. Once you start eating again, your stem cells kick back into high gear to replenish the cells that were recycled.

The human part of the study was much more limited: a group of cancer patients fasted for 1, 2, or 3 days prior to chemotherapy. The idea is that fasting might reduce the harmful side effects of chemotherapy, particularly the immunosuppression caused by some chemotherapeutic drugs. These results are very preliminary: the patients are participating in a phase I clinical trial, which is designed to assess safety, not effectiveness. Nonetheless, the results indicate that a 3-day fast (but not a 1-day fast) was beneficial for these patients.

A key finding in this research is that you have to fast for several days to get any benefit: basically, you have to fully deplete your energy reserves (in the form of glycogen), and it takes your body at least 24 hours, and probably 48 hours or more, to do this. This is much harder than a 1-day fast, which many people do routinely. 

On the other hand, Valter Longo has compared the effects of periodic fasting to long-term caloric restriction, which has been shown to prolong lifespan in mouse and other animals. In a separate review article, Longo wrote: 

“Fasting has the potential to delay aging and help prevent and treat diseases while minimizing the side effects caused by chronic dietary interventions.”

Caloric restriction is extremely difficult to achieve for humans: you have to nearly starve yourself for years. Compared to this, an occasional 3-day fast should be a snap.

Caveats: fasting can be harmful, especially for people who have other health problems. If you’re seriously thinking of trying this, you should consult your doctor first. And this preliminary evidence, though encouraging, is primarily based on mice, not people. We might eventually learn that the benefits of fasting are outweighed by other problems. Fasting for more than two days isn’t easy, either: you’re going to get really hungry. 

Does a 3-day fast truly reset your immune system? Well, maybe not a total reset, but at least a mild refresh. The science suggests that, if you can do it, a prolonged fast for 2-3 days or longer may induce your body to clean out some old immune cells and switch on production of new ones. Stay tuned.

[An interesting aside: this study was lengthy and complex, and the authors apparently went to great lengths to satisfy the peer reviewers: the paper was submitted in October 2012, re-submitted after revision in December 2013, and finally accepted in April 2014. 18 months is a very long time.]

Those fish oil supplements might cause cancer

Eating fish is good for you, especially fish that contain omega-3 fatty acids.  So I was surprised last week to read a new study in the Journal of the National Cancer Institute that found that omega-3 fatty acids increase the risk of prostate cancer.  The risk for both high-grade and low-grade cancer was increased with higher levels of omega-3 fatty acids.  This is a carefully done study, and the results should make anyone who is taking fish oil pills reconsider.

One reason this study caught many people off guard is that there has been much evidence showing that a diet rich in fish that contain omega-3 oils is good for you.  The Mayo Clinic says that  "eating fish helps your heart", especially fish like salmon that contain omega-3 fatty acids.  The American Heart Association (AHA) elaborates:

"Omega-3 fatty acids benefit the heart of healthy people, and those at high risk of — or who have — cardiovascular disease. Research has shown that omega-3 fatty acids decrease risk of arrhythmias (abnormal heartbeats), which can lead to sudden death. Omega-3 fatty acids also decrease triglyceride levels, slow growth rate of atherosclerotic plaque, and lower blood pressure (slightly)."

This all sounds great.  Because of the evidence about the benefits of fatty fish, supplement manufacturers have been marketing and selling fish oil pills for years, with great success.  As I described back in 2010, GlaxoSmithKline even created a high-dose omega-3 fatty acid pill called Lovaza that has FDA approval.

But the evidence for that you can get the same benefit from supplemental omega-3 fatty acids — taking a pill, that is — is much weaker.  In fact, a large review published last year in the Journal of the American Medical Association found no connection at all between supplemental omega-3 and a lower risk of heart attacks, strokes, or death in general.  Other studies have reported similarly negative results.  So it appears that fish oil pills may not have any heart benefits.

And now, with this new study, we learn that supplemental fish oil might increase the risk of prostate cancer.

The bottom line: the AHA recommendations about eating fish are probably still good ones.  The AHA website says:

"We recommend eating fish (particularly fatty fish) at least two times (two servings) a week. Each serving is 3.5 oz. cooked, or about ¾ cup of flaked fish.  Enjoy fish baked or grilled, not fried." 

But popping a fish oil pill is not going to cut it. As we've seen before, supplements often fail to show the benefits that a healthy diet offers.  So save your money and stop buying those fish oil pills — and fire up the grill and throw on a few salmon fillets for this weekend's barbecue.

Eat more chocolate! Lose weight!

It's a holiday weekend, and many of us will be eating chocolate, especially chocolate bunnies and chocolate eggs.  (Where does that come from, anyway?  Bunnies don't lay eggs.) We don't really need an excuse to eat chocolate.  But isn't it wonderful when a scientific study tells us that we should?  Even the skeptics among us really want to believe that eating chocolate is good for you.

Last week, a new study in the Archives of Internal Medicine claimed that eating chocolate more frequently is associated with a lower body-mass index, or BMI.  The study's authors carefully stated their results as a "link," but then they went on to say that other evidence would "suggest the association could be causal."

Of course the media ate this up (pun intended), with headlines like "The Chocolate Diet?" at The New York Times.  The Times story was filled with caveats, pointing out that it was frequency of chocolate consumption (how many times per week you eat it) that was associated with lower body weight,  and that people who simply eat more chocolate tend to be, well, fatter.  Surprise!

Beatrice Golomb, the lead author of the study, explained to the Times that:

“It’s not the case that eating the largest amount of chocolate is beneficial; it’s that eating it more often was favorable.”  

Because the thinner people in the study ate less chocolate overall, they must have eaten much less at each sitting.

Alas, this is just a bad study.  The biggest problem is that it was based entirely on questionnaires, in which 1018 men and women answered questions about how often they excercise and how often they eat chocolate.  Here are some key features of these subjects:

• Their average BMI was 28.  So they are kind of fat. (A BMI of 25-30 is overweight.)
• They claim to exercise 3.6 times per week, on average.
• They report eating chocolate 2 times per week.

We all know how this works. Even though the questionnaire is anonymous, we still want to look like we're doing the right thing.  So we over-state how often we exercise (3.6 times per week? ha!) and we under-state how often you eat unhealthy foods.  The entire result can be explained by a tendency of thinner people to admit eating chocolate more often.  Observational studies like this one are notoriously unreliable.

On this note, I have to give kudos to Gary Taubes at Discover magazine, who isn't buying any of this.  Taubes explains that:

"Both of these studies were classic examples of what is known technically as observational epidemiology, a field of research I discussed at great length back in 2007 in the New York Times Magazine.  I made the argument that this particular pursuit is closer to a pseudoscience than a real science."

Taube looked at multiple observational studies done at Harvard's School of Public Health, by one of the top epidemiology groups in the country.  But when observational studies "discovered" a causal relationship, it never seemed to pan out, says Taube:

"I pointed out that every time that these Harvard researchers had claimed that an association observed in their observational trials was a causal relationship—that food or drug X caused disease or health benefit Y—and that this supposed causal relationship had then been tested in experiment, the experiment had failed to confirm the causal interpretation—i.e., the folks from Harvard got it wrong. Not most times, but every time."

A zero percent validation rate - it doesn't get any worse than that.  Taube makes a very good case that the whole "nutritional epidemiology business" is pseudoscience.  You just can't go around asking people what they eat and expect to get reliable answers.

So nope, it just isn't so, no matter how much we'd like to believe it.  Eating chocolate isn't going to lower anyone's BMI.  But the really important question is, which part of the chocolate Easter bunny do you eat first?  I go for the ears.