We’ve all seen it: one day, you read “Study Says: Saturated Fat Is Killing You”. A year later, a new article in the same publication claims, “Saturated Fat Is Good For You.” Or we find sensational claims that seem too good to be true: “Eating Blueberries Can Add 6 Years To Your Life.”
This visualization showing certain foods both cause and protect against cancer (according to research) from Vox tells the story:
Is this good or bad research? Good or bad reporting? These answers matter, but few of us have the time to for the deep dives necessary to make smart personal health decisions based on new headlines.
Here at The Unwinder, we’ve always focused on evidence-based health and wellness, citing studies and explaining the science behind our recommendations.
Our team goes beyond the quoted evidence, diving deep into research context and study design to point to where the truth lies, and an appropriate level of confidence in that hypothesis..
This month we’re starting a new monthly column– one in which we critique science articles in the media, as well as the underlying studies. We’ll explain what the media gets right and wrong about the studies they write about, what the studies actually do or do not prove, how they could have been better, and what needs to be studied next. And of course, what practical takeaways you can actually get from the research.
For our inaugural column, here are two recent studies reported in popular media outlets, about health topics including diet, metabolism and aging.
Business Insider: “Fast carbs don’t necessarily make you fat”
The glycemic index (GI) is a measure of how much and how fast a food causes a rise in blood glucose after consumption. Glucose (pure sugar) has a glycemic index of 100; everything else has a lower glycemic index. The lower a food’s GI, the less it causes a spike in blood sugar, and, so the theory goes, the healthier and less fattening it is.
White or processed carbohydrates, like white bread, white rice, flour and breakfast cereal, are usually high-GI. Darker carbohydrates like whole grains and most fruits are usually low- to medium-GI.
However, as reported by a recent article in Business Insider, a new study suggests that low-glycemic foods don’t aid weight loss, nor inhibit weight gain, any better than high-glycemic foods. This will surprise many dieters– and hell, even some doctors and dieticians– who have hung their hats on low-GI diets.
That said, this isn’t actually new. The study Business Insider cites, led by Dr. Glenn Gaesser of Arizona State University, is actually a systematic review that looked at 27 observational (i.e. non-experimental) studies as well as 30 meta-analyses of randomized controlled trials. In other words, this finding reflects the sum total of decades of research.
This may sound like it’s saying that the conventional wisdom that advocates eating a high-fiber, low-sugar diet is incorrect. In fact, that’s not at all what the study is saying.
The truth is quite a bit more complicated. First, let’s look into what the findings actually were.
Out of 27 cohort studies, 12 found no difference between high- and low-GI diets, 8 found low-GI diets to be better, and 7 actually found high-GI diets to be better. Since these were observational rather than experimental studies, that’s open to a few possible interpretations.
Unlike experiments, observational studies don’t make subjects behave a certain way– they merely observe how subjects behave and what results they get. Now, observational studies can control for variables that can be measured, by correcting for them during statistical analysis. However, by randomly assigning subjects to different groups, experimental studies can control for variables that the researchers either can’t effectively measure or don’t even think of.
As such, there are plenty of possibilities that cohort studies can’t eliminate, such as that the low- and high-GI groups are genetically different, or that the high-GI groups eat more fast carbs to fuel higher levels of exercise. It would take experiments to figure that out– but we have those too.
As for the experimental evidence, GI does still seem to matter a little bit. Here’s what the study abstract says: Results of 30 meta-analyses of RCTs [Randomized Controlled Trials] from 8 publications demonstrated that low-GI diets were generally no better than high-GI diets for reducing body weight or body fat. One notable exception is that low-GI diets with a dietary GI at least 20 units lower than the comparison diet resulted in greater weight loss in adults with normal glucose tolerance but not in adults with impaired glucose tolerance.
In other words, a large reduction in dietary glycemic index can provide a small but significant advantage, though this is relevant mostly to the people least in need of losing weight. That is, impaired glucose tolerance (or pre-diabetes) generally goes hand in hand with obesity. There are however some people who have impaired glucose tolerance without being overweight– and sometimes impaired glucose tolerance is a precursor to dangerous levels of weight gain. Those are the people most likely to benefit from a low-GI diet.
Why Glycemic Index (GI) Is Not The Best Measure
Logically, it seems like low-GI carbs should be healthier; they would allow for more stable blood sugar levels and keep you full for longer. To understand why that may not always be the case, we should first look at how the glycemic index is measured.
An individual food is consumed in isolation on an empty stomach. Usually this is done in the morning, before breakfast to ensure stomach emptiness, meaning that the subject hasn’t eaten for 12 hours or so. Blood sugar is measured right before ingestion and then 2 hours after, and the difference between the two measurements is taken is where the food’s glycemic index is taken from.
Blood sugar may also be measured at more frequent intervals– often every fifteen minutes– to gather further data, but this doesn’t factor into the GI calculation.
There are a few problems with measuring a food’s impact on blood sugar this way:
First, this isn’t how food is generally eaten. You usually eat several foods together in a meal, rather than one in isolation. And it turns out that mixed meals consistently have a lower glycemic index– not just lower than one food, but lower than a weighted average of all the foods in the meal.
Second, most food is also not consumed on a completely empty stomach, other than at breakfast, so in practice foods usually digest more slowly than they do when GI measurements are taken.
Third, the two-hour benchmark is arbitrarily chosen for standardization purposes, but the largest peak in blood sugar probably doesn’t occur at two hours after a meal. It may occur after, or more commonly before. In fact, in many cases a meal causes a quick spike in blood sugar, then a crash, then a second smaller rise in blood sugar, and that second rise is what gets measured two hours post-meal.
Fourth, GI mainly measures the impact of carbohydrates, but ultimately caloric intake is what matters. Protein can be converted into carbohydrates via gluconeogenesis, but that takes longer than 2 hours. Fat, for its part, can be deposited directly into fat cells for later use without first being turned into blood sugar. Thus, GI measurements aren’t very appropriate for food that contains a mix of macronutrients– which again includes most mixed meals.
Fifth, some carbohydrates are just processed differently by the body. This mainly applies to fructose, but it’s an issue with most sugar alcohols (used in low-sugar foods and many protein bars) as well.
Sixth, a food’s glycemic index can be affected by preparation factors such as ripeness, cooking methods, processing, temperature and freshness. White rice has a lower GI if it’s allowed to cool for 24 hours after cooking and before eating.
Note that many of these limitations also apply to the thermic effect of food, another supposedly-important measurement which ends up not mattering very much (or rather, being uniformly good) as long as you eat mixed meals composed of healthy foods.
As a result, many foods have GI measurements that would surprise you. White rice and potatoes, while maybe not the best carbs, are high-GI despite being healthy in moderation. Pure fructose, and many fatty, sugary junk foods like cake and ice cream, are low-GI.
In the case of fructose, this is because it mostly has to be processed into glucose by the liver, which may take more than 2 hours– but it does happen and eventually impact blood sugar.
In the case of cake and ice cream, this is because they have so much fat, which is fattening even if it doesn’t affect the GI the way carbohydrates do. Again, these calories can go straight to fat without first becoming blood sugar.
The glycemic index also doesn’t measure other factors that are important to health, such as vitamin and mineral content, although it isn’t meant to. Low-GI foods generally have better nutritional value, though as you’ve seen this can’t be taken as a hard and fast rule. Thus, apples are still healthier than white bread, just not because of their glycemic index.
On the other hand, some low-GI foods, especially whole grains, also have more anti-nutrients such as phytic acid, which impair absorption of nutrients and can damage the gut lining.
This study does not in any way imply that low-GI foods are unhealthy– only that their low glycemic index, in and of itself, doesn’t make them healthy.
You should probably forget about the glycemic index (and Business Insider’s article), or at least treat it as only a minor factor in your food choices. Instead, eat mixed meals with high protein and some fat, low caloric density (as measured in calories per gram) and choose carb-heavy foods which are minimally processed and have a lot of vitamins, minerals and fiber. These foods usually will be on the low-GI side, but that shouldn’t actually be your goal.
A good listener can help stave off Alzheimer’s disease, new study shows
That first study was basically correct, and the media articles about it mostly described it well.
Now, though, we get to a study that in my opinion, has been badly misreported, and whose authors have failed to consider a few likely confounders.
Alzheimer’s disease is a neurodegenerative disorder that is responsible for 60-70% of all dementia cases. It affects roughly 30 million people a year, and kills 2 million, worldwide. After diagnosis, Alzheimer’s patients typically have a life expectancy of 3-9 years.
It would be amazing if we could do something to slow or prevent the progression of Alzheimer’s. So, is that what this study found?
According to iNews, a new study shows that having a good listener available can help stave off Alzheimer’s.
However, the actual study is observational– a cross-sectional cohort study– not experimental, so it doesn’t prove causation. The language used in the study is far more conservative: social support in the form of supportive listening is associated with greater cognitive resilience. That sort of “is associated with” phrasing is common in observational studies to avoid implying that causation has been demonstrated.
Here’s why they used that method:
It would take an experiment to actually prove causation, however it’s hard to see how this question could ethically be investigated experimentally. That would require deliberately depriving the control group of social support– obviously unethical.
You could maybe pick out subjects who don’t have social support to begin with and just not provide any to the control group, but even that is ethically fraught, since we do know that, Alzheimer’s aside, social support definitely improves quality of life and protects against depression. Not helping people– especially nursing home patients– when you know you could do so is still unethical.
And that’s an important lesson to take home: some topics are nearly impossible to ethically study via experimental methods.
Now let’s look at what the study actually found. Five forms of social support were measured. Only listener availability showed a negative correlation with Alzheimer’s progression; advice, love-affection, emotional support and sufficient contact did not.
The study also found that social support did not prevent the brain from shrinking; rather, it was associated with greater cognitive resilience despite continued brain shrinkage. In other words, it broke or weakened the usual correlation between brain shrinkage and cognitive decline.
Because most social support came from family members, genetic factors should have been considered, but don’t seem to have been. That is, people with genetic resilience against Alzheimer’s may also have relatives who are better listeners, or more available to listen.
It would have been better if the study had distinguished between “listener availability” that comes from family vs friends vs nursing home staff. That said, the lack of correlations with love-affection and emotional support does suggest that the listener doesn’t need to be family; nursing home staff often listen to residents, despite not loving them the way their family does.
Now, this may have limited practical importance since social support is well-known to be important for other reasons like mood and quality of life. However, it would be good to know just how much and what kinds of social support, if any, help against Alzheimer’s, and also how that interacts with other Alzheimer’s treatments.
It would also be great to see how the effect sizes compare against other treatments; potentially this could lead to very cost-effective treatments in the form of having volunteers or low-skilled staff available for the sole purpose of talking with residents.
It would also be good to know if other activities that use the same brain capabilities as speaking (or being listened to) have a similar effect. Do word games such as Balderdash or CodeNames help?
All in all, it is highly plausible that the implied finding here is correct, and having someone to talk to helps stave off Alzheimer’s. After all, the brain benefits from exercise just like the body does. The study just wasn’t well-designed to reach that conclusion, and even given its observational nature, could have controlled for more factors like who was providing social support.
Ultimately, social support probably does protect against Alzheimer’s to some degree, but the details of what kind of social support, what confounders (especially genetic and who the social support comes from) you control for, how big the effect size is and how you define your outcome measures matter a lot.