LAUSR

In an editorial published in September 2018 in JAMA, Ioannidis discussed the status of nutritional epidemiology and stated that radical reform is needed [1]. Diet is complex, the methods currently available to assess it are inadequate and randomized controlled trials (RCTs) do not support the findings from observational studies. In this issue of the European Journal of Epidemiology, Giovannucci contradicts the arguments of Ioannidis by stating that current methodologies such as food frequency questionnaires and hypothesis-based approaches are sufficient to overcome the complexity of diet and deal sufficiently with confounding, and that considerable RCT data link common dietary factors to risk factors for major diseases. In response, also in this issue of EJE, Ioannidis cautions that diet complexity should not be oversimplified and mentions again the shortcomings of current methods. Ioannidis suggests two options for a radical reform: (1) large pragmatic randomized trials, and (2) giving nutritional epidemiology a chance by adopting a transparent approach of nutrition-wide analyses of single nutrients, foods, and dietary patterns simultaneously that test their robustness to different analytical choices. Antipodal opinions need not be irreconcilable. We believe debates in nutritional epidemiology are just like diet: the key factors are balance and moderation. On one hand, Ioannidis is correct: diet and its assessment are extremely complex for several reasons. First, regardless of whether food is branded or harvested, its composition is extremely varied. Cans of Coke and other top brand fizzy drinks contain widely differing amounts of sugar depending on where they are sold [2]. The median selenium (Se) concentration in Brazil nuts varies from 2.07 mg kg−1 (in Mato Grosso state) to 68.15 mg kg−1 (in Amazonas state); depending on its origin, a single Brazil nut could provide from 11% (in Mato Grosso) up to 288% (in Amazonas) of the daily Se requirement for an adult man; and this is just nuts [3]. Foods can also contain pesticides, mycotoxins, food additives, heavy metals, and environmental contaminants that are rarely measured or taken into account when assessing diet [4, 5]. Second, due to food processing the bioavailability of nutrients may be increased or decreased depending on the process and the food itself [6]. Thermal processing including bleaching, retorting, and freezing can cause loss of lycopene in tomato-based foods [7]. Third, interactions between foods or nutrients are rarely taken into account. Various types of fiber can affect the bioavailability of calcium, iron, and zinc in both positive and negative ways [8–11]. Fourth, relying on biological markers as a solution for the complexity of foods is not a panacea: a single 24-h urine collection cannot predict sodium, potassium, or chloride intake; multiple collections are necessary [12]. Finally, adequate assessment of diet might take much longer than initially planned: a Japanese study reported that the number of days necessary to estimate true average nutrient intake exceeds 50 days for vitamins such as retinol and carotene, and more than a year to adequately estimate intake of seldom consumed foods such as nuts and seeds [13]. A similar conclusion had been reached in a study conducted in the United Kingdom in which the number of days of survey required for an 80%-reliable classification of individuals varied from 2–3 days for sugar or total carbohydrates, to 2–3 weeks for dietary cholesterol or the ratio of polyunsaturated to saturated fatty acids [14]. A single day’s survey classified no nutrients with 80% reliability, whereas one week’s survey classified most nutrients with 80% reliability or better. On the other hand, Giovannucci is correct in stating that studies supplementing specific nutrients are effective when such nutrients are definitely lacking. The treatment of scurvy by oranges and lemons in the eighteenth century by Lind [15] and the iodine supplementation of salt in Switzerland * Sanne Verhoog [email protected]