A methodological audit of the gas-stove and childhood asthma meta-analysis

Meta-analysis combines odds ratios or p-values from separate observational studies to answer one question. The paper starts from the point that this combination inherits whatever the individual studies carry: if a claimed association does not reproduce, pooling it does not make it reproduce.

The author counted the number of hypotheses available for testing in 14 of the 27 base papers used for the asthma and wheeze meta-analysis. The median was 15,360, with an interquartile range of 6,336 to 49,152. None of the 14 papers mentioned correcting for multiple testing, and none explained why no correction was used. The 40 odds ratios with confidence limits used in the meta-analysis were converted to p-values and shown as p-value plots; for the gas stove and current asthma association, and for gas stove and current wheeze, the plots are described as showing randomness.

The author's conclusion is stated as a limit on what the evidence supports, not as a finding of safety: the meta-analysis fails to provide reliable evidence for policy making on gas stove harms to children in North America, nitrogen dioxide is not established as a biologically plausible causal explanation for childhood asthma, and multiple testing, p-hacking and selective reporting cannot be ruled out.

This is one author's methodological critique, and it is published as such. The comparison row on this platform describes the literature on gas cooktops as large but mixed and the health causality as debated — a critique of this kind is part of what makes it mixed, not a resolution of it.