Long Follow-Up Times Weaken Observational Diet–Cancer Study Outcomes: Evidence from Studies of Meat and Cancer Risk

Abstract

For years, prospective cohort studies of diet and cancer incidence have reported smaller effects than do retrospective case–control (CC) studies. The differences have been attributed to problems with CC studies, including dietary recall bias, poor matching of cases and controls, and confounding. The hypothesis evaluated here is that long follow-up periods between ascertainment of diet and cancer incidence weaken the findings. Prospective studies of cancer incidence with respect to serum 25-hydroxyvitamin D concentration have already shown reduced benefit of higher concentrations for longer follow-up periods. Evaluating that hypothesis for dietary factors involved searching the journal literature for meta-analyses of red meat and processed meat and cancer incidence. I used findings from observational studies for bladder, breast, colorectal, and gastric cancers. To evaluate the effect of duration of follow-up time, I used two approaches. First, I plotted the relative risks for CC studies for gastric cancer with respect to consumption of 100 g/day of red meat and for bladder cancer for 50 g/day of processed meat against the interval between the dietary data and cancer incidence. Second, I compared nested CC studies of meat and cancer incidence for five breast cancer studies and one colorectal cancer study. Both approaches yielded an inverse correlation between interval or follow-up time and relative risk. My findings strongly suggest that diet near time of cancer diagnosis is more important than for longer intervals, that results from meta-analyses should be revised when possible with appropriate adjustments for duration of follow-up, and that dietary guidelines be revised accordingly.