OBJECTIVES

: We evaluated the consistency of effectiveness results from model-based cost-effectiveness analyses (CEA) for antipsychotics in schizophrenia.

METHODS

: Model-based CEAs comparing antipsychotics and reporting quality-adjusted life-years (QALY) as health outcome were identified via Medline and Embase. The review included studies published from 2000 to 2016, in English language. QALYs were extracted for each antipsychotic, and incremental QALYs were derived for each possible comparison. For each possible 1:1 comparison between antipsychotics, estimates of incremental QALYs were considered as consistent if they were either all negative or all positive.

RESULTS

: Among 34 studies included in the review, 142 antipsychotic comparisons were reported, and 506 incremental differences were available overall, as some studies reported results for several subgroups or scenarios. Among these comparisons, 73 (51%) were reported at least twice: 40 (55%) lead to inconsistent estimates. When considering comparisons reported at least 3 (n=65), 4 (n=63), 5 (n=41) or 6 (n=25) times, the inconsistency rate reached 57%, 57%, 78% and 92% respectively. The most frequent comparator was oral olanzapine, compared with 25 other antipsychotics in total, accounting for 142 estimates. At least 2 estimates were available for 16 of these comparisons (64%), with 12 (80%) comparisons leading to inconsistent estimates. The most frequent comparison was oral olanzapine vs. oral risperidone with 13 studies and 23 estimated differences in QALYs: 7 studies with 7 estimates were in favor of oral olanzapine; 4 studies with 13 estimates, in different treatment lines or scenarios, were in favor of oral risperidone; 1 study reported 2 opposite estimates for different timeframes; 1 study reported similar QALY for oral risperidone and oral olanzapine.

CONCLUSIONS

: Though these could be partially explained by drivers such as timeframe, treatment line or population definition, the high rates of inconsistencies between results raise concern about the credibility of effectiveness results of pharmacoeconomic models for schizophrenia.