Interferon alfa therapy for malignant melanoma: a systematic review of randomized controlled trials

CRD COMMENTARY:

The research question was clearly defined in this review, although some of the inclusion and exclusion criteria could have been specified in more detail. For instance, the authors did not fully rationalise why only IFN-alpha monotherapy was eligible. This may limit the clinical applicability of the findings.
The search strategy was reasonable, although the authors did not search Cancerlit. The authors attempted to identify unpublished material and there were no language restrictions. This suggests that most of the relevant studies are likely to have been identified. However, publication bias was not assessed. In general, the inclusion and exclusion criteria and methods for assessing validity were described adequately, although the authors did not provide details about how the studies were selected or how many of the reviewers performed the study selection and quality assessment processes. This makes it difficult to assess the quality of the review.
The studies included may have been of limited quality. One of the nine studies identified was excluded without clear reason. None of the studies included were placebo-controlled. The authors state that details of the participants, disease characteristics, study designs and outcomes were not always clearly reported in the primary studies.
The authors did not describe potential biases that could affect the findings of the review. A narrative synthesis was provided, supplemented by summary statistics in a table. This is likely to be appropriate given the (stated) heterogeneity of the data. The authors calculated summary statistics for individual studies, which could have been better presented in the narrative summary.
There are also some difficulties with the outcomes used. The authors did not report the statistical significance of median survival for the individual studies, or state how disease-free survival was analysed. Using outcomes that count the number of participants is problematic in cancer studies, as this does not account for confounding and differences in the time period over which the participants were recruited. A better outcome measure may have been to compare time to events using the Kaplan-Meier method (life tables). It is also possible that non significant findings are due to the sample size of the included studies. Care should be taken when interpreting the findings of small and heterogeneous studies.
Overall, the data appear to support the authors' general conclusions, although the lack of detail about the included studies makes it difficult to assess the generalisability of the findings.