In an era when 11 meta-analyses are published every day, there are sometimes 2 on the same topic which do not agree. As such a situation can be very confusing, systematic reviewers should discuss their findings in the light of existing reviews to make the differences understandable to readers. ^{Reference Helfer, Prosser, Samara, Geddes, Cipriani and Davis1} In a pairwise meta-analysis by Siskind et al, ^{Reference Siskind, McCartney, Goldschlager and Kisely2} clozapine was shown to be superior to other first- and second-generation antipsychotics in treatment-resistant schizophrenia, which is in sharp contrast to our recently published network meta-analysis on the same topic. ^{Reference Samara, Dold, Gianatsi, Nikolakopoulou, Helfer and Salanti3} As the publication of the two studies overlapped, the authors could not discuss the results of the other. Thus, readers might find themselves confused and unable to understand where the discrepancies lie.

First of all, there are differences in included trials. Siskind et al included studies in children and adolescents (Kumra 1996, 2008), whereas we did not because we considered that this population requires different pharmacological treatment from adult patients. They included Chinese studies (Cao 2003, Shaw 2006, Wang 2002), whereas we did not because it has been reported that studies from mainland China are often not reliable. ^{Reference Woodhead4} Moreover, Siskind et al included the study by McEvoy et al (from CATIE phase II) assuming that it was a blind trial, but this did not hold true for the crucial clozapine arm, which was open label. Last but not least, five studies (Breier 1999, Conley 2003, Daniel 1996, Honigfeld 1984 and McGurk 2005) were missed by our colleagues. There are also differences in the use of end-point or change data and the handling of short-term and long-term studies.

Finally, a major difference lies in the statistics applied in the two meta-analyses. Siskind et al conducted a pairwise meta-analysis, whereas we conducted both a network and a pairwise meta-analysis. But even in the pairwise meta-analysis alone, results differed. Siskind et al combined all comparator drugs versus clozapine, ignoring possible efficacy differences between the various comparators, which may explain the significant heterogeneity in many outcomes, limiting the robustness of the results. In our pairwise meta-analysis of all clozapine trials, even lumping the other antipsychotics and comparing them with clozapine revealed no significant difference in overall symptoms. We found clozapine to be better than the first-generation antipsychotics chlorpromazine and haloperidol, but not better than the second-generation antipsychotics olanzapine, risperidone and ziprasidone.

Whether the superiority of clozapine has been sufficiently proven by blinded trials is essential for clinical practice. Therefore, we and our Australian colleagues plan a joint re-analysis of these meta-analytic data.