One Diet Does Not Fit All

July 18, 2019


By Ashira Lubkin, PhD

Peer Reviewed

For me, one of the most difficult things to do in outpatient medicine is to tell a patient that they need to lose weight. Weight loss is a powerful tool in the prevention and treatment of many common diseases, so it’s very important, if somewhat uncomfortable, counseling to give. But it’s not the awkwardness that bothers me the most. It’s that I don’t know what I should tell the patient to eat.

Despite a lot of research, we are still far away from a consensus on what a healthy diet is. Guidelines on diet have changed over time, leading to confusion among patients and clinicians.1 The aggregate conclusion of many randomized controlled trials of various diets, including low-fat, low-carbohydrate, the Mediterranean diet, etc., is that no one diet is clearly better than the others.2,3 It is perplexing that so many studies on diet contradict each other, and further that the weight loss achieved in them is so minimal, usually under 5%.

We are now on the verge of a possible explanation. A few studies have found that individuals can have vastly different glycemic responses to the same foods.4,5,6 For example, though some people have a larger rise in blood glucose after eating sugar than after eating bread, others have a greater glycemic response to the bread. Similarly, most people experience a greater rise in blood glucose level after eating a banana than a cookie, but some have the opposite response. Further, Zeevi and colleagues used machine learning methods to create a model to predict the rise in blood glucose of individuals after eating specific foods.6 The model is based on clinical parameters including fasting glucose, lipid levels, and waist circumference, as well as stool microbiome. This model was able to predict blood glucose responses to foods far more reliably than by looking at carbohydrate content of meals alone (which is what we advise our diabetic patients to do). Further, it was also able to generate personalized diet recommendations for study participants, which were successful at lowering these individuals’ blood glucose values measured over one week. A recent retrospective analysis showed a somewhat similar finding, that patient characteristics pre-diet can correlate with diet outcomes. Specifically, patients with prediabetes who started out with a high fasting insulin level lost more weight on a low-fat rather than a low-carbohydrate diet, while patients with pre-diabetes and low fasting insulin levels lost more weight on the low-carbohydrate diet.7

What do these findings mean? First, that we all digest and metabolize food slightly differently, and thus have differing abilities to maintain homeostasis in the face of the variability in our diets.8 Therefore, it is unlikely that we will ever find generalizable diet guidelines that we can give to an entire population. Instead, personalized nutrition may become a powerful tool. It may soon be possible to send off some blood tests and a stool sample, and get back an individual diet that is tailored to our patient’s genetics, metabolic state, and microbiome.

We’re not there yet, though. Several nutrigenomic tests already on the market give dietary recommendations based on genetic data. A meta-analysis, however, found that these recommendations are not evidence-based.9 There is a lot more research to do.

It seems likely that, if and when personalized nutrition is ready for prime time, a major barrier will be buy-in by the medical community. In a way, personalized nutrition is the direct opposite of the broadly generalizable guidelines that we are used to relying on from randomized controlled trials. Furthermore, there is a perception that we need to fully understand on a systems level how nutrition impacts health outcomes before we can move towards personalized nutrition.10 As a scientific community, we want to understand the physiology behind the medicine we practice, and as clinicians we want to be able to explain to our patients why we are telling them to eat one food instead of another. However, machine learning-generated models are often not interpretable, and methods that are more open to interpretation are not the most accurate. Post-hoc techniques can be used to interpret machine learning models, but the most common ones often given results that are so oversimplified that they are ultimately inaccurate reflections of what drove the decision-making process.11 Will we be able to tell our patients to eat this diet because the algorithm says so? Will this increase adherence because it is tailored to the patient personally, or decrease adherence because we can’t explain why?

For now, though, our problem is that we don’t know what to say to the patient sitting in front of us who needs to lose weight. Many of our patients can benefit from basic pieces of advice that have good evidence behind them, such as cutting down on sugar-sweetened beverages12 and increasing fruit and vegetable intake.13 For those who need more, hopefully we will soon have more refined dieting tools to offer them.

Commentary by Dr. Sapana Shah

While personalized nutrition may be the way of the future, we are clearly not there yet.  When it comes to who would benefit more from a low-carb versus a low-fat diet, a prospective randomized trial failed to find any significant difference in weight loss between these two diets.14  Basically, reducing calories by any diet resulted in  weight loss. The retrospective study by Hjorth and colleagues7 presented in this review to see if personal metabolic factors such as pre diabetes and fasting insulin levels influenced weight loss response to a low carb vs a low fat diet is flawed. There was a high dropout rate for the low-carb diet and very low numbers in this subgroup analysis to gather any meaningful conclusions.  Moreover, any enthusiasm over a small study showing that a low-carb diet may be associated with weight loss  in the short term in a specific subset of prediabetics with low insulin levels at baseline needs to be tempered by the very real risk of increased mortality  seen in those following a low-carb diet. A  Harvard study following over 100,000 men and women for over 20 years found an increase in overall mortality and cardiovascular mortality in individuals following low-carb diets.15 Interestingly, the increased mortality was only seen in the animal-based low-carb diet followers, but when a plant-based low-carb diet was followed, there was an actual reduction in overall mortality.

Truth be told, there really is no confusion among the experts when it comes to what constitutes an optimal diet. Among the American Heart Association, the American Diabetes Association, the American Institute for Cancer Prevention, and the Academy of Nutrition and Dietetics, there is a consensus. They all agree that to prevent our most common chronic diseases, we must eat more of a whole-foods, plant-based diet rich in vegetables, fruits, whole grains, and legumes while minimizing animal products, processed foods, and added sugars and fats.  In contrast to a low-carb diet, a whole-foods plant-based diet reduces weight without compromising longevity. However, 90% of Americans are not meeting their daily fruit and vegetable intake.16  This highlights that the problem is not confusion over what constitutes a healthy diet but, rather, our reluctance to eat the evidence-based foods that give us the best odds to live our healthiest life.

Ashira Lubkin, PhD is a 4th year medical student at NYU Langone Health

Peer reviewed by Sapana Shah, MD, MPH, Assistant Professor, Department of Medicine, NYU Langone Health

Image courtesy of Wikimedia Commons

References

  1. Gifford KD. Dietary fats, eating guides, and public policy: history, critique, and recommendations. Am J Med. 2002;113 Suppl 9B:89S-106S. https://www.ncbi.nlm.nih.gov/pubmed/12566143
  2. Noakes TD, Windt J. Evidence that supports the prescription of low-carbohydrate high-fat diets: a narrative review. Br J Sports Med. 2017;51(2):133-139. https://bjsm.bmj.com/content/51/2/133
  3. Thom G, Lean M. Is there an optimal diet for weight management and metabolic health? 2017;152(7):1739-1751. https://www.ncbi.nlm.nih.gov/pubmed/28214525
  4. Vega-López S, Ausman LM, Griffith JL, Lichtenstein AH. Interindividual variability and intra-individual reproducibility of glycemic index values for commercial white bread. Diabetes Care. 2007;30(6):1412-1417. https://www.ncbi.nlm.nih.gov/pubmed/17384339
  5. Vrolix R, Mensink RP. Variability of the glycemic response to single food products in healthy subjects. Contemp Clin Trials. 2010;31(1):5-11. https://www.ncbi.nlm.nih.gov/pubmed/19737630
  6. Zeevi D, Korem T, Zmora N, et. al. Personalized nutrition by prediction of glycemic responses. 2015; 163(5):1079-1094. https://www.ncbi.nlm.nih.gov/pubmed/26590418
  7. Hjorth MF, Astrup A, Zohar Y, et. al. Personalized nutrition: pretreatment glucose metabolism determines individual long-term weight loss responsiveness in individuals with obesity on low-carbohydrate versus low-fat diet. Int J Obes (Lond). 2018 Dec 19. [Epub ahead of print]. https://www.ncbi.nlm.nih.gov/pubmed/30568260
  8. van Ommen B, van den Broek T, de Hoogh I, et al. Systems biology of personalized nutrition. Nutr Rev. 2017;75(8):579-599. https://www.ncbi.nlm.nih.gov/pubmed/28969366
  9. Pavlidis C, Lanara Z, Balasopoulou A, Nebel J-C, Katsila T, Patrinos GP. Meta-analysis of genes in commercially available nutrigenomic tests denotes lack of association with dietary intake and nutrient-related pathologies. 2015;19(9):512-520. https://eprints.kingston.ac.uk/32307/2/Nebel-J-C-32307a.pdf
  10. Verma M, Hontecillas R, Tubau-Juni N, Abedi V, Bassaganya-Riera J. Challenges in personalized nutrition and health. Front Nutr. 2018;5:117. https://www.frontiersin.org/articles/10.3389/fnut.2018.00117/full
  11. Lundberg SM, Lee S-I. A unified approach to interpreting model predictions. Advances in Neural Information Processing Systems. 2017;4765-4774. https://arxiv.org/pdf/1705.07874.pdf
  12. Imamura F, O’Connor L, Ye Z, et al. Consumption of sweetened beverages, artificially sweetened beverages, and fruit juice and incidence of type 2 diabetes: systematic review, meta-analysis, and estimation of population attributable fraction. 2015;351:h3576. https://www.ncbi.nlm.nih.gov/pubmed/26199070
  13. Miller V, Mente A, Dehghan M, et al. Fruit, vegetable, and legume intake, and cardiovascular disease and deaths in 18 countries (PURE): a prospective cohort study. 2017;390(10107):2037-2049. https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(17)32253-5/fulltext
  14. Foster GD, Wyatt HR, Hill JO, et al. Weight and metabolic outcomes after 2 years on a low-carbohydrate versus low-fat diet: a randomized trial. Ann Intern Med. 2010;153(3):147–157.
  15. Fung TT, van Dam RM, Hankinson SE, Stampfer M, Willett WC, Hu FB. Low-carbohydrate diets and all-cause and cause-specific mortality: two cohort studies. Ann Intern Med. 2010;153(5):289-298.
  16. Lee-Kwan SH, Moore LV, Blanck HM, Harris DM, Galuska D. Disparities in state-specific adult fruit and vegetable consumption–United States, 2015. MMWR Morb Mortal Wkly Rep. 2017;66(45):1241–1247.