The review question was clear and it was supported by well-defined inclusion criteria. Relevant databases were searched, with no language restrictions. No search for grey or unpublished literature was reported, so it is possible that relevant trials were missed. It was unclear how many reviewers were involved in the review processes, and whether these processes were performed independently, thus reviewer error and bias was possible. Quality assessment was reportedly performed, but no results were presented, leaving the possible bias within trials unclear.
Trial details were presented and the methods of synthesis seem to have been appropriate. Extensive effort was made to explore the heterogeneity between trials, but it remained unexplained. Many variables were assessed in a meta-regression with just a few trials (nine), which limits the reliability of the results per variable. Only two trials investigated the same combination of diagnosis, D-cycloserine treatment regimen, and outcome, limiting the generalisability of the overall pooled result, and the reliability of the subgroup analyses. The authors acknowledged that only a limited number of trials could be included in the review.
Given the unknown risk of bias and the lack of evidence, the authors' conclusions seem overly strong.