To our knowledge, this is the first study that has assessed and compared clinical outcomes (effectiveness/tolerability) of switching from olanzapine to risperidone and vice versa in the usual course of schizophrenia care. Our results showed that patients who were treated with risperidone and subsequently switched to olanzapine for clinical reasons experienced a more favorable treatment course than patients who were treated with olanzapine and subsequently switched to risperidone. Patients who switched to olanzapine experienced significant improvements in symptom control (as assessed by the CGI-SCH overall score) and in all tolerability outcomes (EPS, loss of libido, impotence/sexual dysfunction, and amenorrhea/galactorrhea) except for weight gain. On the other hand, patients who initiated with olanzapine and switched to risperidone did not experience improvements in any of the tolerability outcomes, although they experienced significant improvements in symptom control. The differential effect between the two groups remained even when the patient characteristics before switching were taken into account in multivariate analyses. Patients who switched to olanzapine were more likely to improve in symptom control and less likely to experience relapse, EPS and amenorrhea/galactorrhea, compared with those who switched to risperidone.
Consistent with previous research [5, 15], patients experienced symptom improvement after medication switch in both the RIS-OLZ and OLZ-RIS groups. However, symptom improvement was greater in those who switched from risperidone to olanzapine. This finding was partially in agreement with recent results from the Clinical Antipsychotic Trials for Intervention Effectiveness study (CATIE) . In phase 2T of the CATIE study (conducted in patients who primarily experienced intolerability to their previous antipsychotic), the olanzapine group achieved greater improvement than the risperidone group in positive symptom control as assessed by the Positive and Negative Syndrome Scale (PANSS). However, there were no differences between groups in negative and overall symptoms. Data from other head-to-head trials are available , but they do not suggest that one drug is clearly more effective than the other in symptom control. However, consistent with our findings, olanzapine appears to be superior to risperidone in relapse prevention, EPS and reproductive adverse events .
Relapse prevention is a primary goal in the long-term treatment of schizophrenia. Surprisingly, however, there is only limited information on the comparison between risperidone and olanzapine. Nevertheless, this small body of literature also reports a lower risk of relapse in patients treated with olanzapine compared with those treated with risperidone [14, 18, 19].
EPS are common adverse events during treatment with antipsychotics, especially for conventional agents. They are the most common reason for non-adherence with antipsychotic medication, leading to treatment failures, relapses, poorer long-term outcomes and preventable economic costs [20–22]. Atypical antipsychotics have a lower risk of EPS than conventional antipsychotics, and this is a considerable asset from a tolerability and safety point of view . However, the risk of EPS varies among atypical antipsychotics and dose-dependent increases in EPS have been observed with risperidone . Several studies have provided the evidence that patients treated with olanzapine may have a lower risk of EPS than patients treated with risperidone . Previous SOHO publications also showed a lower risk of EPS with olanzapine compared with risperidone . Moreover, a UK population-based study reported reduced use of antiparkinsonian drugs (treatment for EPS) in patients who switched from conventional agents to olanzapine, but not in those who switched to risperidone . A reduced use of antiparkinsonian drugs was also observed in patients who switched from risperidone to olanzapine, but not in those switched from olanzapine from risperidone . Lowering the risk of EPS is crucial in the management of schizophrenia, given their clinical and economic implications.
Consistent with previous studies [27, 28], our study demonstrated a lower risk of amenorrhea/galactorrhea with olanzapine compared with risperidone. One possible reason for this may be the potent effect of risperidone on prolactin elevation, although plasma prolactin concentrations were not measured in the SOHO study. Prolactin elevation is a well-known adverse event of conventional antipsychotics and of some atypical antipsychotics such as risperidone , although treatment with olanzapine may also elevate prolactin level and the elevation may persist during chronic administration . An increase in prolactin levels can cause amenorrhea, galactorrhea and other sexual disturbances [29, 31]. A number of studies have shown that the risk of prolactin elevation with risperidone is similar to or greater than that of conventional antipsychotics [29, 31, 32]. The switching study from olanzapine to risperidone by Takahashi and colleagues also reported a significant increase in plasma prolactin concentrations after the medication switch . However, the findings of a recent study in patients with first episode psychosis indicated that such reproductive adverse events can occur even when prolactin levels are normal . While the mechanisms underlying such reproductive adverse events in antipsychotic-treated patients are not yet clearly understood, it is well known that hyperprolactinemia can cause amenorrhea and galactorrhea.
In addition, weight gain has been commonly reported during treatment with most atypical antipsychotics, including olanzapine and risperidone . However, our study did not find any significant differences in weight change before and after medication switch between the two groups (i.e. RIS-OLZ and OLZ-RIS groups). Notably, however, patients who were on olanzapine at baseline gained an average of 3.14 kg before switching to risperidone. In comparison, those who were initially on risperidone gained an average of 1.43 kg before switching to olanzapine.
The results of this study also confirmed that medication switch is a common practice in the management of schizophrenia . More than one in three outpatients with schizophrenia who initiated treatment with either olanzapine or risperidone switched antipsychotic medication at least once during the 3-year follow-up in the normal course of care. In addition, more than half of the switchers also made subsequent switches (52.4% for patients who switched to olanzapine and 61.1% for patients who switched to risperidone). While medication switch constituted a logical and common treatment strategy for patients who did not respond adequately to the prescribed antipsychotic treatment or could not tolerate treatment-emergent adverse events, it may not always result in improvement, as indicated in our study. Although the results of our study confirmed that switching to olanzapine would be of benefit to patients who failed treatment with risperidone, the converse was not confirmed.
Our results need to be interpreted in the context of the following study limitations. Firstly, the W-SOHO study was originally designed and powered to compare clinical and economic outcomes between olanzapine and other antipsychotics. Comparison of the impact of switching antipsychotics between olanzapine and risperidone was a post hoc analysis and, consequently, the relevant subgroup sample size was small. This, in turn, might have affected the statistical results. Secondly, our study sample was initially drawn from 11,078 patients, who completed the study and had no more than one missing visits during follow-up (i.e., 64% of the baseline sample). Although the retention rate of 64% was relatively high given the study duration of three years and thereby the same size of study completers was still very large, the clinical prognosis could differ between study completers and those who were not. It is, however, unclear whether the inclusion of study completers only disproportionally influenced the outcomes of the OLZ-RIS and RIS-OLS groups. Thirdly, as the outcome assessments were not performed blind, we cannot exclude the possibility that the treating psychiatrists were likely to evaluate patients more favorably after the switch. Nevertheless, this bias is unlikely to differ between medications. In addition, as the clinical outcomes were assessed by the treating psychiatrists, the assessment could have been subjective and influenced by their prior knowledge and expectations (i.e., observer bias). This could be of particular concern in the absence of blinding if more favourable assessments were made towards their preferred treatment. Notably however, a previous study  explicitly investigated the observer bias in SOHO by comparing patient- and investigator-reported outcomes, and found no such bias. Fourthly, although our findings were adjusted for pre-switch clinical and demographic characteristics of patients when comparing the impact of switching on clinical outcomes, there could be unobserved differences between the treatment groups, which may confound our results. Finally, for patients who switched medication between visits, we assumed that clinical status was similar to that at the visit before the switch.