Diet and Health: The Need for New and Reliable Approaches

Andrew Mente; Mahshid Dehghan; Salim Yusuf


Eur Heart J. 2020;41(28):2641-2644. 

In This Article

How can Nutrition Research be Improved to Overcome These Challenges?

To move the field forward, investigations of the impact of diet on health must extend beyond documenting associations between diets and outcomes. Advanced epidemiological and molecular methods (genetics, metabolomics, and proteomics) may be helpful, by discovering intermediate biomarkers that characterize a pathway linking diet to CVD.[23] Further, by linking genetic markers to the same pathways, we can have greater confidence that epidemiological associations observed between diet and outcomes are likely to be causal.[24] Such approaches may be facilitated by measuring a range of biomarkers (through analyses of the metabolome and proteome) and by linking these to the genome; such studies will be likely to identify new pathways characterizing the effects of diet on CVD (or other chronic diseases). This could open up new approaches to prevention which could then be evaluated in randomized trials.

These advanced methods of molecular epidemiology, however, have their own challenges. Development of a comprehensive 'biomarker profile' for CVDs has been limited by several factors, including (i) a disease may be caused through more than one pathway; (ii) a relatively limited number of biomarkers have been examined in CVD and few have been found to be reliably 'causal'; and (iii) the high cost of comprehensive biomarker screening for large studies. These limitations can now be partly overcome through advances in multiplex biomarker assays, which can examine hundreds of biomarkers at relatively modest costs per participant. Nevertheless, even with this approach, the results may not necessarily indicate causality. Combining such approaches with better methods of assessing diet will probably lead to more informative associations, some of which will be more likely to be causal than the results of current epidemiological approaches.

Genetic data can be used to derive individual polygenic risk scores for analysis as instrumental variables, i.e. Mendelian randomization. Mendelian randomization takes advantage of the fact that genetic variants are by nature randomly assigned at birth (and so less confounded or subject to 'reverse causality'), and can be considered to be analogous to the randomization in a randomized trial. However, genetic instruments to study nutritional exposures are uncommon. Such genetic markers have been used to clarify whether the 'protective' effects of alcohol consumption [e.g. using aldehyde dehydrogenase 2 (ALDH2)] are real. Genetic markers for lactose intolerance [in the lactose gene (LCT)] have been used to dissect whether dairy intake is causally related to CVD, but the studies have not been large or provided consistent results.[25] Given that very few genetic markers are currently available to study the effects of diet on health, the extent to which studies using Mendelian randomization will be useful in this context is unclear.

Ultimately, large and long-term clinical trials will be required. These trials do have significant challenges—including maintaining sufficient contrast in differences between the active and control groups in the diets of interest for prolonged durations to detect the moderate effects on outcomes.

At present, no single scientific approach can provide the robust answers on diet and health. Instead, we should encourage diverse methods, such as extensive collaborations between those conducting large studies (to standardize measurements, covariate adjustments, and analytic approaches, and to avoid publication biases), and understand the limits of each type of approach and how to overcome them. The most consistent dietary components (or dietary patterns) should then be evaluated in large and long-term trials. While such trials may be expensive if done using conventional approaches, there are methods to simplify trials and reduce their costs. The global food and agriculture enterprise is far larger than the drug industry (where collaborations with academic groups have been successful in improving health), encouraging collaborations between industry and academia would be useful as long as there are appropriate safeguards to avoid biases. This could provide the funding to improve the quality and size of long-term epidemiological studies and randomized trials.

In the meantime, we suggest eating plenty of fruits and vegetables including legumes, and a moderate amount of whole fat dairy, fish, unprocessed meats, eggs, and whole grains, but limiting overall carbohydrate intake to ~50% of energy. The key to a healthy diet may be one that includes diverse natural foods in moderation, rather than extreme restrictions or high consumption of any single food category.