Joshua David Wallach
Yale School of Public Health
Co-Authors: Lingzhi Chu, John P.A. Ioannidis, Alex C. Egilman, Vasilis Vasiliou, Joseph S. Ross, Joshua D. Wallach
Vibration of effects in epidemiologic studies on the association between alcohol consumption and the risks of breast cancer
Background/methods: Different analytical approaches can influence the associations estimated in observational studies. For each study included in a recent alcohol-breast cancer meta-analysis, we approximated the variability of effect estimates (vibration of effects [VoE]) by dividing the largest by the smallest effect estimate (relative odds ratio [ROR]) for broadly-defined alcohol consumption and breast cancer associations. Results: Most studies (85/97, 87.6%) reported both harmful and protective relative effect estimates for an alcohol-breast cancer relationship, ranging from 1.1-17.9 and 0.0-1.0, respectively. RORs ranged from 1.0-106.2, with a median of 3.0 (IQR 2.0-5.2); 35 (36.1%) were based on extreme estimates with at least three different analytical characteristics (i.e. adjustment covariates). Conclusions: To provide insight regarding stability and generalizability of findings, observation studies should approximate the VoE across a range of different analytical approaches.