Misleading meta-analysis

BMJ 1995;310:1603-1604 (17 June)

Letters

Misleading meta-analysis

Public policy is based on results of epidemiological meta-analysesthat contradict common sense

EDITOR,--Matthias Egger and George Davey Smith's scepticismabout the validity of meta-analyses whose conclusions are contradictedby those of single megatrials¹ can be extended to meta-analysesof epidemiological studies whose findings contradict biologicalcommon sense. The alleged causative contribution of passivesmoking to lung cancer is based, among other things, on a meta-analysisof 10 case-control studies and three prospective studies thatshowed a highly significant 35% increased risk among non-smokersliving with smokers compared with non-smokers living with non-smokers.²This risk is especially pronounced for adenocarcinoma in theperiphery of the lung (a histological category not associatedwith smoking in Doll and Hill's studies of doctors) rather thanfor the central squamous cell cancers seen most commonly inactive smokers.³ If the conclusion of this meta-analysis isto be believed it is necessary to suppose that . . . [Full text of this article]