To the Editor Henssler et al 1 reported the results of a systematic review and meta-analysis comparing combination antidepressant therapy with antidepressant monotherapy for the treatment of depression. Some aspects of the review should be considered when interpreting the results.The authors concluded that combination therapy was more effective than monotherapy but did not assess the certainty of the evidence, which may be limited for several reasons. First, tests indicated small-study effects, and the authors did not mention searching clinical trial registers or other sources for unpublished trial results. Publication bias should therefore be suspected. We are not aware of empirical evidence suggesting that publication bias, which has been shown to be an issue in antidepressant trials, 2 would not be an issue in trials of combinations of antidepressants. Second, heterogeneity was high for the overall meta-analysis and most subgroup analyses. Third, the authors judged 62% of the included trials to be at unknown or high risk of bias. Information on their assessment of individual domains was not available, nor on how the authors arrived at their overall assessment, but of the 3 largest trials judged by the authors to be at low risk of bias, 2 did not blind assessors or clinicians, and they all had overall dropout rates of 9% to 27%. These issues are likely to result in risk of bias due to deviation from intended treatment, missing outcome data, and bias in the measurement of the outcome. Fourth, many of the trials were of short duration and antidepressant trials have extensive exclusion criteria that may reduce the relevance of trial results for most patients. 3 Taking these issues into consideration, we believe the certainty of the evidence presented by Henssler et al 1 should be rated as low or very low according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach 4 owing to downgrading for publication bias and inconsistency and potentially for risk of bias and indirectness as well.Moreover, Henssler et al 1 did not investigate any harms. Their outcome of dropouts owing to adverse events should not be interpreted as a surrogate for harms because of the potential for bias 5 and does not provide patients with information on the risk of specific adverse effects important for decisionmaking.