Is everything we eat associated with cancer? A systematic cookbook review.
Level 1 - systematic review of randomized trials
Systematic review and meta-epidemiological analysis of nutritional association studies and meta-analyses.
PubMed 23193004 · doi:10.3945/ajcn.112.047142
What was done
Authors randomly selected 50 common food ingredients from cookbook recipes and queried PubMed for recent studies evaluating their associations with cancer risk. They extracted author conclusions and risk estimates from up to 10 of the most recent individual articles per ingredient, as well as relevant meta-analyses, and analyzed effect sizes, statistical significance, and selective reporting in abstracts.
What was found
Studies were found for 40 of the 50 ingredients (80%). Across 264 single-study assessments, 191 (72%) claimed an association with cancer risk (103 increased risk, 88 decreased risk), but 75% had weak (0.05 > P ≥ 0.001) or no statistical significance (P > 0.05). Statistically significant findings were significantly more likely to be reported in study abstracts than in full-text sections only (P < 0.0001). Median relative risks (IQRs) in individual studies claiming increased or decreased risk were 2.20 (1.60 to 3.44) and 0.52 (0.39 to 0.66), respectively. In contrast, 36 evaluated meta-analyses reported a median relative risk of 0.96 (0.85 to 1.10), with only 13 (26%) showing increased (n = 4) or decreased (n = 9) risk.
Why it matters
This review highlights widespread overinterpretation, publication bias, and reporting bias in nutritional epidemiology, where extreme effect sizes in single studies frequently attenuate to null in larger syntheses.
Limits
The sample was restricted to 50 ingredients from cookbook recipes and capped at the 10 most recent studies per food item. It evaluates published literature characteristics rather than conducting primary causal assessments of specific dietary exposures.
Cited by
- supports A meta-research paper analyzing published food studies showed that virtually every food examined had been reported to both cause and prevent cancer multiple times, except bacon, which was only linked to causing cancer.