Is everything we eat associated with cancer? A systematic cookbook review
Schoenfeld,J.D. and Ioannidis,J.P Am.J.Clin.Nutr. 10.3945/ajcn.112.047142
Objective: To examine the conclusions, statistical significance and reproducibility in the literature on associations between specific foods and cancer risk.
Design: Review. Authors selected 50 common ingredients from random recipes in a cookbook and identified recent cohort or case-control studies, together with the most recent meta-analysis, which had evaluated the relation to cancer risk.
Setting: n/a
Participants: n/a
Main outcome measures: Authors' conclusion regarding risk, the RR estimate (HR for cohort studies, or OR for case-control studies), the exposure contrast, 95%CI and P value. For the latter, the authors considered P values between 0.05 and 0.001 as weak evidence in epidemiologic studies.
Results: At least one study was identified for 40 of the ingredients. Increased risk, decreased risk, borderline effect and no evidence of effect were concluded in 39%, 33%, 5% and 23% of studies. The authors considered that even in studies which concluded increased risk, 75% of these were weak according to their P value. There was wide variability in exposure contrasts (i.e. tertiles, quartiles, etc. or more arbitrary classifications).
Risk estimates were lower from meta-analyses compared to individual studies.
Conclusions: Associations with cancer risk can be claimed for most food ingredients. Many single studies highlight implausibly large effects, even when evidence is weak, but with lower effect sizes in meta-analyses.
Back to recent research list