Background: Missing data in Cochrane reviews is a common problem. The problem is often difficult to detect and the impact hard to estimate. There are few methods for assessing missing data from included studies in reviews. Selective outcome reporting may have different effects on different outcomes and time points (e.g. secondary outcomes and follow-up data may be more vulnerable than the primary outcome). The potential impact of missing data should be assessed for each analysis in a review. Objectives: To describe the amount of data potentially missing from studies included in a systematic review, and to assess its potential impact on the results. Methods: We conducted a systematic review and made a table including the number and percentage of (i) included trials and (ii) included participants in each analysis. We then (i) described the relationship between outcomes and the amount of data reported and (i) compared information in the table to tests for bias. Results: Results of analyses with few included studies were sometimes inconsistent with the primary analysis. Effects in analyses with large amounts of missing data were larger than anticipated. Funnel plot asymmetry was large for analyses that included a low percentage of included studies and participants. In large analyses, our table complemented tests for bias and aided assessments of internal and external validity. In analyses of secondary outcomes and time points, these descriptive statistics offered more information than available tests for bias. The table also drew our attention to the different sizes of analyses in the review and differences in their external validity. The table was easy to make. It facilitated our discussion of bias and generalisability. Similar tables might be useful in other reviews.