How diet-related follow-up period affects risk of adverse health outcomes

Regression dilution bias (RDB) occurs when baseline exposure measurements such as 25-hydroxyvitamin D or dietary intake fail to account for changes over time, potentially weakening observed associations in prospective cohort studies (PCS). In this narrative review, we examine how follow-up duration after food-frequency questionnaire (FFQ) completion influences relative risk (RR) estimates in meta-analyses of diet-related chronic disease outcomes.

A narrative review and graphical reanalysis of studies in six meta-analyses involving diet-related exposures and outcomes was conducted. RR values from PCS in populations with Western dietary patterns were included and plotted against mean follow-up periods. Linear regression analyses were performed to assess strength and direction of association between follow-up time and RR estimates.

Plots were made of RR versus mean follow-up period after completion of FFQs for incidence of type 2 diabetes mellitus from coffee, processed meat, and whole grains; for risk of coronary heart disease and cardiovascular disease from nuts; and for risk of ischemic heart disease from processed and red meat. Linear regression fits to the data were made with and without inclusion of Harvard University's nurses and health-professionals' studies. Those studies last for up to 26 years but have participants complete FFQs every four years.

Longer follow-up periods after baseline dietary assessments may introduce RDB, weakening observed associations in nutritional epidemiology. Future studies should consider repeat FFQs every few years and adjust for RDB. Meta-analyses should account for temporal misclassification to avoid underestimating dietary risk estimates.