Known as the Personalized Nutrition Study, participants with a BMI of between 27.0 and 47.5 kg/m2 were recruited to complete a genealogy test (e.g., 23andMe, Ancestry), and were then randomized to either a high-carbohydrate diet (65% calories from carbs, rich in whole grains) or a high-fat diet (40% of calories from fat, rich in unsaturated fats) for a 12-week period. Both diets were designed to have 15% of calories from protein, with an energy intake targeted to be a daily deficit of 750kcal. Although very specific meal plans were provided, including a list of ingredients and amounts for each meal of the day and a food scale, meals themselves were not provided, and relied on participants to comply with the provided instructions.
It’s worth noting that the study did not meet its recruitment goals to have sufficient power to detect weight differences as low as 2kg between genotype concordant vs. discordant groups; they planned to have 32 participants in each group (128 total), but there were only 21 and 16 carbohydrate-responders in the high-fat and high-carbohydrate groups, respectively. Responsiveness to either fat or carbohydrate diets was calculated from a risk score based upon genotyping of 10 SNPs, including variants in the FTO, APOA5, and PPARG genes previously suggested to predict metabolism; for example, rs1801282 is a SNP in the PPARG gene (Peroxisome proliferator-activated receptor γ 2) that was shown to modulate weight loss depending on fat intake, and influence both plasma insulin and HOMA-IR.
Despite the hopes for enhancing weight loss by optimizing diet x gene interactions, this study found no significant differences in weight loss between genotype concordant vs. discordant diets. Additionally, neither fasting insulin nor HOMA-IR was associated with weight loss. While lacking clinically significant findings, they did observe decreased cravings among “carbohydrate-responders” while on a high-fat diet, and lower systolic blood pressure among fat-responders on a high-carbohydrate diet, suggesting at least some possible implications for their predictive algorithm.
It could be argued that this study was underpowered, too short in duration, and by not providing meals to study participants, did not adequately assess for dietary adherence and efficacy. Indeed, they reported having adherence data for only 39% of those on the high-carbohydrate diet; while the data from those participants for whom data was available suggests they were very adherent, missing data for 61% of high-carb participants indicates a large gap in the awareness of what participants were actually eating. While a larger sample size and longer duration would certainly help, the authors indicate that they intentionally wanted to assess how pragmatic these diets were versus how efficacious. This seems like they may have skipped a step in study design; first, they should have determined if the genotype could help to predict weight loss when following a specific diet, and after proving this principle, determined how effective this is in the real world. We are left not feeling confident in their results, but it’s not clear if that’s due to lack of adherence or failure of principle.
It could also be the case that the SNPs (or the related scoring) were not well-chosen, and another set of variants would be more predictive. However, this is not the first study to present disappointing results when attempting to predict optimal weight loss programs by genotype, and a systematic review of trials evaluating the interaction between SNPs and macronutrient intake on weight loss found that the majority of interactions were non-significant (or not replicated after a single finding). For example, the DIETFITS randomized clinical trial published in JAMA found no significant predictive value of genotype with either a low-fat or low-carbohydrate diet (using only 3 SNPs for prediction, however). Yet other studies have found SNPs to have predictive value, including one published in the American Journal of Clinical Nutrition evaluating both a moderately high-protein diet and a low-fat diet. And a study recently published in Clinical Nutrition suggests that personalized diets based solely on the human genome may be missing the target, as models based on the microbiome may have more potential to optimize weight loss. A reasonable conclusion may be that the study published in Nature Communications was insufficiently powered, lacking complexity, and too short; there appears to be a strong genetic component to obesity (and influence of the microbial genome), but it may require more sophisticated algorithms to optimize personalization.