*Geek Box: Prospective Cohort Studies
In the biomedical model, a randomised controlled trial is considered the ‘gold standard’ of scientific evidence, particularly where an RCT tests an effect of an intervention on hard endpoints. However, in nutrition, doing long-term RCTs on hard endpoints is a practical and ethical impossibility; trying to maintain behaviours over the long-term is a practical issue, while it would be unethical to place subjects on a diet deficient in a given nutrient, or excessively high in a potentially harmful dietary constituent, over the long-term. Nutritional epidemiology is therefore where the effects of diet on long-term health outcomes is primarily examined. Amongst the different observational designs available, prospective cohort studies provide the best option for assessing the impact of a long-term exposure (diet) on diseases with long latency periods (cardiovascular disease, cancer, etc.). The prospective nature of cohort studies makes them more robust as they are able to mitigate a number of biases, for example recall bias, reverse causation, and selection bias, that often arise in other observational designs (16). A prospective design allows researchers to update potential confounders to be adjusted for, as additional or new information on potential confounders may be included prior to analysis (16). If all relevant confounders are accounted for, a well-conducted prospective cohort study can mimic an RCT; and ultimately any truly long-term RCT in nutrition becomes about as ‘controlled’ as a prospective cohort study over time. In this respect, prospective cohort studies are the best observational design to examine the relationship between diet and health outcomes over the true long-term. They are not perfect, no design is: when coming to conclusions, we consider the totality of evidence.