The paper, which has 53 references, carefully considers the usual issues of misclassification error, publication bias, and confounding. It carefully and systematically excludes these potential problems and convincingly demonstrates an increase in risk.

It also carefully addresses the fact that the risk of heart disease due to secondhand smoke exposure is large for the dose when compared with the effects of active smoking. Most experts now think that this fact is due to a nonlinear dose-response curve, with the cardiovascular system very sensitive to small doses of toxins in the secondhand smoke.

This evidence comes from studies of platelet and endothelial function in both animal and human studies. There is also a recent epidemiological analysis of this issue (Law M, Morris J, Wald N. Environmental tobacco smoke exposure and ischaemic heart disease: An evaluation of the evidence Br. Med. J. 1997;315:973-980) that supports this view.

Given this strong case, I was surprised to find a commentary in the same issue of NEJM by John Bailar, which concluded that "we still do not know, with accuracy, how much or even whether, exposure to environmental tobacco smoke increases the risk of coronary heart disease." In writing this commentary, Bailer only cited three papers, He's paper, a comparison of meta-analysis with randomized controlled trials, and a commentary he wrote on issues in meta-analysis. He paid no mind to the literature in cardiology.

Bailar dismisses the evidence from experimental and clinical studies which form the backbone of modern cardiology, saying that only epidemiology can answer questions. He says: "Responses in animals may not be like those in humans, laboratory studies involving human subjects must necessarily be of short duration, and reports of clinical series are subject to a range of biases."

In fact, it is just these kind of studies that have solidified the case linking secondhand smoke with heart disease, since it is possible to show rapid and large (compared to the dose) effects on key cardiac and, more important, vascular functions. Rather than engage this literature, Bailer simply dismisses it out of hand.

He also criticises He et al for asserting that they have avoided the issue of publication bias when they failed to detect such bias in a "funnel plot" which people think (including lots of tobacco industry consultants) detect such bias. Bailer says that even though there is no evidence for publication bias, it might still be there.

Bailar even criticises He et al for pointing out that the results of all the studies of secondhand smoke and heart disease are reasonably consistent. I was trained that consistent results are a good thing.

Bailer also says he does not believe the results because they are "too big." Unlike Law et al, who systematically looked at the evidence and ended up concluding that the effects of secondhand smoke were consistent with active smoking, Law just offers a personal opinion, with no evidence.

He then essentially says that the only epidemiology that is useful is a randomized controlled trial. Unfortunately, one cannot do a randomized controlled trial on the effects of secondhand smoke on heart disease. I do not think it would be possible to study ANY environmental toxin in a way that Bailar would find convincing.

I must say that I was amazed at the low quality of this editorial. When an editor invites a commentary on a paper like this, it is customary for the author to present a "mini-review" that cites relevant papers to bolster what ever point he or she is making. Bailar just makes flat statements without citing any evidence to support these statements.