Background: Meta-analyses can be useful in deciding whether or not an intervention is feasible. However, it is subject to discussion which data should be included. An inclination to pool as much data as possible to get a sturdy estimate must be balanced against a possible increase in heterogeneity and difficulties in interpretation of the results within the clinical context. We highlight different choices which have to be made when comparing rivaroxaban to enoxaparin, drugs which are used as prevention of deep venous thromboembolism (DVT) in patients undergoing elective total hip or knee replacement surgery.

Objective: The objective is to raise awareness regarding which choices have to be considered when pooling data and assessing what are their possible impact on the results and interpretation.

Methods: A systematic literature search was performed. Data were extracted and pooled in different ways to give risk ratios (RR) and 95% confidence intervals (CI).

Results: An overview of the data for prevention of DVT is presented in Figure 1. Pooled data for hip replacement surgery, regardless of dose and treatment duration gained RR 0.46 (0.26-0.84). However, the approved dose of rivaroxaban is 10 mg and limiting the analysis to this gave RR 0.48 (0.23-0.98). Finally, the pooled effect estimate based on the doses and treatment durations representative of clinical use within EU countries is limited to one study with an RR 0.28 (0.16-0.49). On the other hand, instead of narrowing the focus one might want to expand to pool data regardless of hip or knee replacement therapy.

Conclusions: Always adjust the meta-analysis according to the research question. Use of subgroups may facilitate getting an overview of the data, interpretation and extrapolation to similar research questions.