Corrigendum to 'Meta-analysis of heterogeneous clinical trials: An empirical example' [Contemp. Clin. Trials 32 (2011) 288-298]

摘要

We have updated the application of population weights within meta-analyses to a two step procedure. We had previously advocated multiplying population and bias scores and running the analysis in one step [1]. The latter does not produce expected results especially when there is more than one study per defined population group. Essentially, what should be done is to pool individual studies within population groups using bias adjustment and then pool the single per category estimates using population weights without applying inverse variance weights to the population weight ed pooled estimate (Table 1). These analyses refer specifically to the quality effects model in MetaXL (www.epigear.com). We thus correct our previous recommendation and where standardization for population differences is sought, this two stage application of population weights should be used. Results from this process will then be identical to that commonly used for direct rate standardization. MetaXL does not have the option to exclude inverse variance weights when applying population weights and a separate standardization module will be created for this purpose in due course.