Introduction: Whilst it has long been recognised that the presence of publication bias may produce misleading results, the effect in a meta-analysis of within study selective reporting has received little attention to date. We consider selective reporting in relation to outcomes and subgroups. It may be that particular outcomes or subgroups with more striking results are more likely to be reported.

Discussion: Methods based on the effect size allow results from studies with different outcomes to be combined. However, the possibility of selective reporting of outcomes must be considered. The effect of selective reporting on estimates of effect size and significance levels is presented, and sensitivity analyses suggested. A meta-analysis of anthelminth therapy is re-assessed allowing for within study selection, as it is clear that more outcomes had been measured than were reported. The sensitivity analyses show that the robustness of the results is dependent on what assumption is made about the reporting strategy for the largest trial. Subgroup analysis is a commonly used method for investigating heterogeneity. We examine a meta-analysis on the use of malaria chemoprophylaxis in pregnancy which drew the conclusion of a subgroup effect. We propose a method of sensitivity analysis, involving data imputation for the missing subgroup information, to assess the robustness of results to this problem. Our investigation suggests that the odds ratio for primigravidae reported in the review has potentially been underestimated, drawing the possibly false conclusion of a beneficial effect of chemoprophylaxis limited to primigravidae. It is useful to be able to investigate how sensitive meta-analysis conclusions are to various kinds of selectivity. The simple procedures suggested here can provide information on the magnitude of the possible change in estimates which might result from analysis of individual patient data.