Background:
Network meta-analysis is used increasingly more frequently if various treatments are available for the same disorder. But sometimes the results of network meta-analysis do not seem to match those of conventional pairwise meta-analysis.

Objectives:
To present such a case in a network meta-analysis of the effects of antipsychotic drugs for treatment-resistant schizophrenia.

Methods:
We calculated both a network meta-analysis and a series of conventional pairwise meta-analyses. We initially compared the results qualitatively.

Results:
To date, only preliminary results are available. However, it appears that the results are not entirely inconsistent. The discrepancy only emerges if hierarchies produced by network meta-analysis are interpreted without distinction of significance of differences, i.e. if numerical differences are overestimated.

Conclusions:
The results of network meta-analyses, especially the hierarchies of treatments derived from them, need to be carefully interpreted.