Meta-analysis, which refers to the statistical pooling of results drawn from several studies, is prone to dangerous reductionism in terms of adding together studies that it does not make sense to combine.14,15 Thus, before contemplating playing with any statistics, it may seem sensible not to combine studies of childhood atopic dermatitis with those dealing exclusively with adult atopic dermatitis, as they may belong to a different disease group in terms of the aetiology of treatment responsiveness. It may also be sensible not to combine studies of atopic eczema that evaluate one sort of dietary exclusion with another. It may not make any sense to pool a clinically obscure outcome measure such as a "doctor-assessed itch" simply because it was the only outcome that was common to all trials.10 Meta-analyses are only as good as the data from the individual studies that make up such analyses, and great care needs to be taken to avoid adding together things that should not be combined, especially when the statistical output may give rise to a spurious air of precision to those less experienced in assessing the quality of such analyses. It is for this reason that we recommend that meta-analyses are always performed within the context of a systematic review.16 The involvement of "consumers" with the condition being reviewed is also another important aspect that groups such as the Cochrane Skin Group use to protect against reductionism in choice of outcomes.
Was this article helpful?