Stanford (Diet) Experiment Shows Quality Food Prevails in Weight Loss — While Personalized Nutrition Fails to Show Efficacy

Nothing seems to dominate health and science news lately more than nutrition. It seems like a new study appears every day confirming or retracting the efficacy of a particular way of eating, making it difficult for not only scientists, but also the general public to understand just what we should do with all of this information — with many studies offering contradictory results. Even with this controversy, the discussion on “dieting” has changed. The word “diet” no longer comes with connotations of Atkins, Weight Watchers, or measuring your food on scales. In fact, sometimes dieting comes with no inherent recommendations in terms of how many calories you should consume (more on that later). We have evolved from slimming shakes and 90 calorie snack packs and now focus more on intake of whole foods. With this has come what I’ll call the “macronutrient” debate — specifically the war between low-carbohydrate and low-fat enthusiasts. The former, insisting that our evolutionary roots have metabolically adapted us for a diet high in fat (nuts, seeds, meat) or whatever our ancestors used to eat (probably, whatever they could find). The latter group, into which fewer and fewer health professionals, dietitians, and researchers are falling into, hold to an idea that is becoming largely outdated, that fat may in fact “make you fat” as well as clog your arteries and cause a slew of other health problems, and thus should be avoided. While this theory once “made sense”, more and more research is telling us that restricting fat (at least the good kinds) isn’t the optimal route. Nevertheless, most “low-fatters” still emphasize a quality diet with lots of grains, vegetables, and moderate consumption of some of those “good fats” every now and then.

Low-carb (i.e. high fat) and Low-fat (i.e. high carb) are by no means discrete entities, and this is where “taking sides”and relying epidemiological data fail us. For instance, most of America probably eats both a high-carbohydrate and a high-fat diet and thus receives synergystically,the negative health aspects of each. To say that a high-carbohydrate diet is the reason for the rising obesity levels is to sorely dismiss the fact that the exact people eating too many carbohydrates are the same people who are also eating too much saturated fat, and probably too much sugar (and probably exercising a lot less —can someone say confounding variables?) This is akin to the red meat hypothesis relating high red meat consumption to an increased risk of certain cancers. Who eats the most red meat? Most likely, people avoiding U.S dietary recommendations (which generally recommend limiting red meat consumption), who are also eating red meat along with other generally bad foods.

Finding what is most likely, as Richard Feynman says, is the goal of science. What is most likely, in my opinion, is an interaction of multiple detrimental lifestyle risk factors that predispose our population to disease, disability, obesity, and pathology. Diet definitely plays a role, however, and thus we conduct experimental studies in order to find which diet is “most likely” to produce the most beneficial health effects in the greatest amount of people.

A recent trial by researchers at Stanford University sought to un-muddy the waters when it comes to the debate over which type of diet — one low in fat, or low in carbohydrate — is most beneficial for weight loss. The DIETFITS study (Diet Intervention Examining The Factors Interacting with Treatment Success) enrolled an large amount of participants (609 were enrolled in the trial and 481 completed) to answer multiple questions regarding diet quality, quantity, as well as the interactions with diet and genotype — something that hasn’t been done in many high-quality studies as of yet.

The main question

What is the effect of a healthy low-fat (HLF) diet vs a healthy low-carbohydrate (HLC) diet on weight change at 12 months and are these effects related to genotype pattern or insulin secretion?

Personalized Nutrition

In nutrition (as well as exercise, hell, any intervention studies), there is always substantial variability in the responses between participants; i.e when put in the same training program, one participant may experience a dramatic increase in cardiorespiratory fitness, while one may not even improve at all, and perhaps may get “worse”. While this sometimes may be due to measurement error or a wealth of other factors, the point is that certain individuals will respond differently to an intervention. Muscle phenotype, metabolism, motivation, genetics — all of these factors determine how one responds to an a drug, exercise program or diet. Individual differences have led to the consistent dogma that “no one diet should be recommended universally.” Unfortunately, we don’t understand these individual differences well. We think, however, that genotype may play a role, and this has led to the increasing popularity of “personalized nutrition”, a concept based on the premise that a diet can be particularly tailored to each individual to achieve maximal results.

Thus, one of the novelties of this study included the fact that, in addition to measuring traditional biomarkers related to weight loss in the two diet intervention arms, participants were also categorized by their genotype — either being labeled a “low-fat responsive genotype” (those who would, in theory, be best suited for a low-fat diet) or a “low-carbohydrate responsive genotype” (those who would be best suited for a low-carbohydrate diet). While this method of stratification is in it’s infancy and can get pretty messy (not to mention technical), it essentially involves identifying specific types of gene variants, called SNPs (single nucleotide polymorphisms). (For a pretty cool audio explanation of SNPs, check out this page at the National Human Genome Research Institute). Certain SNPs have been shown to be relevant to carbohydrate and fat metabolism. Researchers in this study analyzed participant’s genotype patterns, with specific focus on genes PPARG, ADRB2, and FABP2, again, those which have been hypothesized to play a role in one’s responsiveness to a low-fat or low-carbohydrate diet.

In addition to testing the interaction of genotype with diet, researchers also wondered whether insulin may help explain weight loss variation among participants. Insulin, as we know, has the primary role of signaling the body to take up glucose into skeletal muscle (in response to hyperglycemia,or high blood glucose). In diabetes, insulin is either no longer secreted, or is not as efficient (termed insulin insensitivity) thus resulting in high blood glucose levels characteristic of the disease. This study utilized a technique of measuring insulin concentration 30 minutes after an oral load of glucose (sugar) to determine participants with higher or lower baseline insulin sensitivity. The rationale? A lower insulin sensitivity (more insulin resistance) may make one more susceptible to weight gain if placed on a high-carbohydrate diet. A lower carbohydrate diet, on the other hand, would be more beneficial for this participant, since it would inherently require less of a demand for insulin to clear the dietary carbohydrate in the circulation.

As we will see, personalized nutrition is a fascinating concept that will remain nothing more than “fascinating”, for now.

Primary Study Objectives

The main outcome of the DIETFITS study was 12 month weight loss. Furthermore, the primary objectives of the trial were two-fold — (1) to test whether a set of 3 single nucleotide polymorphism (SNP) genotype patterns predisposed individuals to differential success in a 12-month weight loss trial and(2) test whether baseline differences in insulin secretion predisposed individuals on either diet to differential weight loss success.

What was the hypothesis?

The researchers hypothesized that first, there is a significant diet x genotype pattern interaction for weight loss (i.e. those best suited for a low-fat or low-carbohydrate diet would preferentially lose weight when placed in the “correct” diet.)The secondary hypothesis was that there is a significant diet x insulin secretion interaction for weight loss (i.e. individuals with high insulin resistance would preferentially lose weight on a low-carbohydrate diet or preferentially gain weight when placed on a low fat, or high-carbohydrate diet).

What was measured?

We get a pretty comprehensive assessment of health biomarkers in this study, including the following. I’ll throw them all in an all-too long list which you can scan at your leisure (or skip if you choose.)

Weight, body mass index, body fat %, waist circumference, lipid levels (HDL,LDL, triglycerides), blood pressure, fasting glucose, fasting insulin, and insulin-30 (a marker of insulin resistance).

A few additional measures are of interest and worth discussion. First, respiratory exchange ratio. This is a resting measure of the amount of C02/02 used and consumed and can be an indicator of what type of fuel your body is primarily using (an RER of .7 indicates using solely fat for fuel, while 1.0 indicates using solely glucose for fuel). Measuring this allows to see whether a particular diet alters substrate utilization in addition to other measured variables.

Resting energy expenditure was also assessed. This measure, although a gross estimate, allows one to see how much “energy” (calories) it takes to support general living (think, metabolism). On many diets, as a result of lowering caloric intake, resting energy expenditure tends to decline, which is thought to be an adaptive process of the body to conserve energy.

The interventions

Before describing the entire diet intervention scheme, one caveat. While it is hard, if not nearly impossible, to completely control dietary intake in a study like this, a major limitation of the current trial was that participants were simply given instructions and guidance on what to eat, and not the actual food. So, even when placed in the low-carbohydrate or low-fat group, participants were still free choose to eat the types and amounts of food they wanted. Furthermore, assessment of dietary “adherence” and determination of each participant’s intake relied on self-report diet records, a process known to be severely flawed — we are, in fact, only human.Regardless, based on the 12 month-data, it appears that participants in each group “adhered” to the advice given, resulting in noticeable differences in macronutrient intake (see tables below).

The entire protocol lasted 12 months. In total, 22 instructional sessions were held for each diet-specific group, led by health educators/dietitians, which were designed to adequately educate participants on how to achieve the goals of their diet (HLC or HLF). Goals of each diet were, per the researchers, to “achieve maximal differentiation in intake of fats and carbohydrates between the 2 diet groups while otherwise maintaining equal treatment intensity.” Simply put, they wanted carbohydrate and fat intake to be the ONLY thing different between the groups’ diets. Emphasis was placed on eating “high quality foods. For the low-fat group, reducing edible oils, fatty meats, whole-fat dairy, and nuts was prioritized, while for the low-carbohydrate group, reducing intake of cereals, grains, rice, starchy vegetables, and legumes was prioritized. BOTH groups were given instructions to maximize vegetable intake, minimize their intake of added sugars, refined flours, and trans fats, and focus on high-quality foods.

One important thing to note. NO advice was given in this trial to reduce caloric intake. This is in complete contrast to many current diets which emphasize reducing caloric intake, often accompanied by an increase in physical activity. The dogma has always been calories in = calories out, and thus calories are the central theme of any weight loss plan. This trial didn’t focus on calories, butinstead on that “whole food” eating approach we discussed earlier.

Additionally, it was instructed that participants follow current physical activity recommendations. They were also given various behavior modification techniques and strategies to help with dietary adherence throughout the trial, details I won’t go into here.


All measures were taken at 0 (baseline), 3, 6, and 12 months during the trial. Below, I present some charts from the article along with a small summary of the significant (or non significant) finding and what it represents.

Total energy intake was not statistically different between the two diet groups. However, despite no recommendation to reduce calories, each group consumed ~500–600kcal less at 12 mo compared to baseline

This finding is significant (in two senses). Since researchers wanted carbohydrate/fat to be the only dietary difference between groups, it helps that one group didn’t lower caloric intake more than another, or else it would be hard to directly compare the diets.

Statistically different intakes of percentage of energy from carbohydrates, fat, protein, saturated fat, fiber, added sugars, and glycemic index and glycemic load.

In the healthy low-fat group (HLF), carbohydrate intake was higher (48% vs 29%), fat intake was lower (28% vs 44%), and protein intake was lower (20% vs 23%) than in the healthy low-carbohydrate group.

It is helpful that the diet advice accomplished what it was meant to. The low-carb group ate more fat, and the lower-fat group at fewer carbohydrates. Success.

In the healthy low-carbohydrate group, fiber intake was higher (14 vs 11g/1000kcal), added sugars were higher (33.1 vs 22.8g), and glycemic index and glycemic load were greater.

What these results indicate (even if they are from self-recall dietary data taken from a random sample of 2 week and 1 weekend days from participants, is that the diets different in macronutrient composition as desired by the researchers. In my opinion, the healthy low-fat group, eating ~30% carbohydrates, wasn’t eating what might technically be considered a low-carbohydrate diet. The diets may not have differed “enough” in composition to realize their benefits. This is just my peanut-gallery commentary, but nevertheless, it is worth noting.

Primary outcome, weight change, significantly decreased in both groups, with no difference between groups

In addition to weight loss, the trial saw changes in secondary outcomes in both groups. At 12 months, both HLC and HLF groups improved lipid profiles, improved blood pressure, insulin, and glucose levels. LDL cholesterol, however, increased in the healthy low-carbohydrate group (thus favoring the healthy low-fat diet).

Respiratory exchange ratio, that number we talked about relating to one’s tendency to utilize fat/glucose at rest, was lower for the healthy low-carb diet than for the healthy low-fat group. This makes empirical sense, since high-fat/low-carb diets have been shown to increase one’s ability to burn fat as fuel.

No interaction effects for genotype or insulin secretion (chart not presented)

Overall, there was no significant difference in weight change among participants matched vs mismatched to their diet assignment based on the single nucleotide polymorphism genotype pattern. In other words, whether or not participants fell into the low-carb, low-bat, or neither genotype, they had similar amounts of weight loss/gain.

Additionally, the test for an interaction effect among insulin secretion (Ins-30 measured at baseline) and diet was not significant. Regardless of whether participants fell into the lower, middle, or upper tertile of insulin secretion, they experienced a similar amount of weight loss/gain on either diet.

Additional individual analysis reveals large variation in weight loss and weight gain among diets

Image courtesy Micheal Hull,

This article was taken from an piece breaking down some data in this study. Data show us that among each group, weight change varied substantially, with individuals in each diet group ranging anywhere from losing 10kg to gaining 30kg. I guess you could say that personalized nutrition worked, just not for the outcomes that were analyzed. The above chart shows that between diet groups, the distribution was essentially equal. However, within each group, much variability was seen in weight change among participants.

This study is generating quite the discussion on social media and in scientific circles. It makes sense, given that low-carb/high-fat diets are increasing in popularity due to the (verb choice my own) demonizing of carbohydrates as of late.

A known low-carb advocate and author Nina Teicholz weighs in on twitter. Not exactly scientific, but a way of rephrasing some of the study conclusions.

Concluding remarks

In the least, the trial can conclude that each diet resulted in a significant amount of weight loss (about 5% total body weight on average), despite varying in their macronutrient composition and validated by changes in certain blood parameters (LDL and HDL cholesterol, RER). The authors cite strengths including large sample size (600+), good retention, substantial weight loss, weight loss variability, good adherence to the diets, and good differentiation between diets. These all increase the validity of the study. In addition, all participants were categorized as obese based on BMI (28–40) and so the applicability of these diets to people who may benefit most is noted.

The fact that no between-group differences were observed and no interactions were seen among genotype or insulin secretion does not downplay the fact that both diets, despite giving no guidance on reducing caloric intake, resulted in weight loss among participants.

The quality vs quantity debate (as well as the composition debate) is far from over, but this trial indicates that emphasizing a whole-food eating approach while simultaneously limiting refined products (sugars, baked goods, etc) may be one of the cornerstones of a healthy weight loss/weight maintenance plan. More trials like this (maybe more adequately controlled, which may mean providing participants with the food they should eat) can help further the literature on what may be the most beneficial diet type. Maybe there isn’t one.

Personalized nutrition, at least in the context of the DIETFITS study genotyping and insulin secretion profiling, doesn’t seem to hold merit quite yet. Further screening for other SNPs that may be involved in diet x weight interactions and a better understanding of how to personalize diet to individuals may, one day, allow us to prescribe eating styles for different people. As of now, that day hasn’t come.

The study was supported by NuSI (Nutrition science Initiative) as well as the NIDDK and the NHLBI

Citation: Gardner CD, Trepanowski JF, Del Gobbo LC, Hauser ME, Rigdon J, Ioannidis JPA, Desai M, King AC. Effect of Low-Fat vs Low-Carbohydrate Diet on 12-Month Weight Loss in Overweight Adults and the Association With Genotype Pattern or Insulin SecretionThe DIETFITS Randomized Clinical Trial. JAMA. 2018;319(7):667–679. doi:10.1001/jama.2018.0245


PhD candidate at the University of Florida — Science writing with a particular focus on exercise and nutrition interventions, aging, health, and disease.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store