The primary aim of the present study was to investigate whether differential anti-depressive effectiveness is present among olanzapine, quetiapine, risperidone and ziprasidone in a clinically relevant sample of patients acutely admitted to hospital for symptoms of psychosis. The study was funded independently of the pharmaceutical industry and the patients were followed for up to 2 years during everyday clinical circumstances. This strengthens the applicability of the results to acutely admitted patients with psychosis in general.
There were no substantial differences among the SGAs on the primary outcome measure. The results are in line with those of the recently published EUFEST and CATIE effectiveness studies that included schizophrenia patients in first episode and chronic phase, respectively [15
]. The collected evidence from naturalistic studies in psychosis thus indicates that if differential anti-depressive effectiveness exists among the drugs, this is likely to be of only marginal magnitude in clinical practice. As the sample was diagnostically heterogeneous the primary outcome was measured by two different inventories: the PANSS-D and the CDSS. The correlation between the sum scores of the inventories was good, but left substantial variance unexplained, thus underlining the rationale for using both inventories in the present study. The majority of prior studies indicating anti-depressive differences among antipsychotics are short term [17
]. If the anti-depressive effects of the drugs occur mainly within the first weeks to months of treatment, differential effectiveness may, in theory, be blurred in a longer time frame. Based on the sensitivity analyses restricted to the first 90 days of follow-up our data do not support this hypothesis. Consistent with our findings there was a significant overall decrease of the CDSS score in the CATIE study. Neither study had a placebo arm, which makes interpretation of this result difficult with regards to assessing the anti-depressive effectiveness of the drugs.
Depression is highly prevalent in first-episode and acute phase psychosis [44
]. The mean CDSS sum score in the BPP at baseline was 6.5. The CDSS sum score was > 6 for 42.7% of the sample. DeNayer et al. [21
], using the same CDSS-score cut-off as in our study, found significant reductions both in the groups with CDSS sum scores > 6 and ≤6 points, respectively. Our analyses demonstrated no statistically significant differences among the SGAs in either group based on this subdivision. The equal anti-depressive effectiveness found also in the group with CDSS sum score > 6, supports our main findings. Caution should be given to the fact that subgroup analyses increase the risk of statistical type II errors. In the CATIE study quetiapine was found to be superior to risperidone in patients with a CDSS score ≥6 [16
]. However, the more depressed group became relatively larger in the CATIE-trial as a consequence of the lower cut-off (CDSS ≥ 6) for depression.
The sample was diagnostically heterogeneous, though with equal diagnostic distribution among the RGs. Hypothetically, different diagnostic groups could differ in anti-depressive susceptibility from the SGAs, which could blur the overall picture. We therefore conducted sensitivity analyses, excluding the affective psychoses and substance-induced psychoses which did not skew the results. The proportion of primary affective disorders was rather low.
Some limitations apply to the study. In a moderately sized clinical trial like the Bergen Psychosis Project, the possibility of a type II statistical error exists. The BPP has, however, proven statistically powerful enough to disclose differences among the SGAs on several outcomes [24
]. Furthermore, power analyses indicate that the study should have a sufficient number of subjects to detect clinically significant differences in anti-depressive effectiveness among the drugs, if present. The pragmatic design was chosen to address issues relevant to everyday clinical practice. The resulting heterogeneous sample does not have enough power to conclude statistically inside particular diagnostic subgroups. The randomization procedure allowing the patient or clinician to choose a different drug than the first one could potentially introduce bias if there were differences among the groups in the proportions accepting the first SGA on the list. No such differences were unveiled. Furthermore, the primary analyses were intention to treat analyses based on the randomization groups. It could be argued given the naturalistic design of the study, with assessments not restricted to the time frame of actual use of the first SGA, that the outcomes may not be related to that particular SGA, but to subsequent medications. We have, however, demonstrated that about three-quarters of the patients did not change their original SGA. Moreover, there were no differences among groups in the rate of antipsychotic medication changes or the choice of a new antipsychotic agent for those who did change. Furthermore, time until discontinuation was generally the same for all SGAs. Finally, the analyses restricted to the period of actual use of first chosen drug revealed generally the same results as the primary analyses. Inherent to the pragmatic design which permits the use of concomitant psychotropics, the net effects of the SGAs under investigation may be somewhat blurred by effects of the concomitant psychotropics. The randomization was open to the patient and the treating clinician in order to imitate a clinically realistic setting. This could have introduced bias if some of the SGAs were more popular among the clinicians or patients. There was a high attrition rate, although not significantly different between the randomization groups. To our best knowledge this is a major problem in all clinical antipsychotic drug trials. Leucht and collaborators [46
] state in their methodology paper that even in short-term trials of only 4 to 10 weeks more than 40% of participants discontinue prematurely. Given the long follow-up of our study we expected a high drop-out rate. This was the main reason for the choice of the mixed-effects statistical method applied, as this method is one of the preferred ones in such a situation. Reasons for drop-out were not recorded and the possibility that the more depressed patients dropped out cannot be ruled out. Still, comparisons on baseline characteristics between those with long term follow-ups and the ones leaving the study early, do not point to substantial clinical differences among the groups. Of those assessed for eligibility, only 30% were included in the trial. Theoretically, including only a fraction of eligible participants could limit the applicability of the results to the whole population. On the other hand, other clinical trials studying antipsychotics included only between 7-14% [47
]. The diagnoses were determined by psychiatrists or specialists in clinical psychology, and structured clinical interviews were not systematically used, which may decrease the validity and reliability of the diagnoses. The ITT-analyses may lead to an underestimation of treatment effect. However, the substantially equal results of the ITT- and FCG-analyses indicate that this was not the case in this trial. Analyses involving single items in the CDSS should be interpreted with caution as the data are unlikely to be normally distributed. Finally, the CDSS-instrument is designed to measure depressive symptoms in schizophrenia specifically. Our sample was diagnostically heterogeneous. Somewhat surprisingly, considering that few trials indicate a superior antipsychotic effectiveness of quetiapine [15
], the recently published results of the Bergen Psychosis Project demonstrated a significant superiority of quetiapine compared with olanzapine and risperidone on several psychometric scales. The present study shows that the superiority of quetiapine could not be explained by a stronger anti-depressive effect.