Proposer of the vote of thanks to Cork et al. and contribution to the Discussion of ‘Methods for estimating the exposure–response curve to inform the new safety standards for fine particulate matter’

My concerns about the paper do not relate to the work itself (which is generally excellent), but rather they relate to the causal inference framework in which the work has been placed. Causal inference: what's in a name?Since the time of John Snow, or earlier, epidemiologists have always done causal inference.Epidemiologists have successfully established causality -not with certainty, because nothing is certain in science, but "beyond reasonable doubt" -for a long list of hazards including tobacco smoking and lung cancer, asbestos and lung cancer, and LDL-cholesterol and cardiovascular disease.All of this was done without RCTs, although some already established causal associations were subsequently confirmed by RCTs.Overall, this is an excellent paper, which explores the effects of different analytic approaches.However, it provides little or no evidence that "traditional regression" approaches are in any way inferior to weighting approaches.Moreover, it does not provide a comparison of these traditional approaches with causal inference methods.