Reply to: Concerns about the Burden of Proof studies

Health effects associated with exposure to secondhand smoke: a Burden of Proof study

“…Furthermore, the BPRF framework employed in this study addresses many of the limitations of existing meta-analytical approaches 18 . Given the high degree of inconsistency observed across results in the SHS literature, using the BPRF to capture the unexplained sources of variation between studies is particularly relevant for our study.…”

“…In this context, in this Article, we aimed to quantify the exposure–response associations between SHS and nine health outcomes—lung and breast cancer, ischemic heart disease (IHD), stroke, chronic obstructive pulmonary disease (COPD), lower respiratory infections, asthma, type 2 diabetes and otitis media—as well as the strength of the available evidence, using an objective, comprehensive and comparative framework. The Burden of Proof Risk Function (BPRF) derives a conservative estimate of the smallest harmful effects of SHS exposure on given health outcomes that are consistent with the available evidence and to summarize the strength of risk–outcome associations and their underlying evidence into a star-rating measure, ranging from one star (weak evidence of an association) to five stars (consistent evidence of a strong association), to aid the interpretation and comparability of results 18 . The main findings and policy implications of this work are summarized in Table 1 .…”

Health effects associated with chewing tobacco: a Burden of Proof study

“…This heterogeneity resulted in a high degree of sensitivity to model parameters and data point inclusion and more uncertainty in our relative risk estimates than in previous studies. Our analysis advances existing literature by better incorporating between-study variation in our derived estimates, which is particularly important for examining associations such as these where study characteristics diverge substantially and there is a paucity of gold-standard evidence 23 . Similarly, while there is consensus that chewing tobacco is a known carcinogen, existing literature focuses on its association with esophageal cancer and lip and oral cavity cancer, and its relationship to these other head and neck cancers is both less studied and less consistently demonstrated in the case-control studies that are available 8,13 .…”

“…Using the meta-regression-Bayesian, regularized, trimmed (MR-BRT) tool, we estimated a pooled relative risk for each outcome that incorporated significant sources of systematic bias, within-study correlation, and between-study heterogeneity 18 . As presented in previous Burden of Proof studies, this approach better incorporates the uncertainty evident in existing literature than traditional meta-analytic approaches, which is particularly crucial for risk factors like chewing tobacco where the existing literature is quite varied [18][19][20][21][22][23] . We then applied the Burden of Proof Risk Function (BPRF) analytic approach to evaluate the effect size and the strength of the evidence for the relative risk estimates.…”