Background
Schizophrenia is often a severe and persistent mental illness, characterized by cognitive deficits, thought disorganization, mood abnormalities, and multiple functional deficits. Expert consensus guidelines for the treatment of schizophrenia [
1] have identified psychosocial interventions and continuous antipsychotic medications as core treatment modalities. Interruptions in antipsychotic therapy have been shown to diminish treatment effectiveness [
2‐
4] and increase the risk of hospitalization [
4‐
6], even when the medication is interrupted for as little as 10 days [
3]. Despite consistent treatment recommendations and recognition of the importance of continuous medication treatment, only about 50% of schizophrenia patients are adherent to antipsychotic medication regimens [
7].
Consensus on the importance of continuous treatment with antipsychotics has been accompanied, however, by an ongoing debate over whether the newer "atypical" antipsychotics are superior to the older "typical" antipsychotics in the treatment of schizophrenia [
8‐
13]. Some have reported comparable efficacy [
12] and questioned the claims of superior efficacy of atypicals relative to typicals [
8,
12,
13] because clinical trials have often established efficacy of the atypicals in comparison to haloperidol, a high-potency typical antipsychotic known to have pronounced treatment-emergent extrapyramidal symptoms (EPS) that may contribute to poorer treatment outcomes. Some researchers have claimed [
8] that atypicals are of comparable efficacy to typical antipsychotics of medium or low potency, because lower-potency antipsychotics have a more favorable EPS profile, and thus provide greater tolerability and better outcomes. Others also reported [
12] that olanzapine is not more efficacious than haloperidol when haloperidol is provided with prophylactic anticholinergic treatment to help ameliorate EPS.
Time to all-cause medication discontinuation has been recognized as an important global index of antipsychotic effectiveness, because it reflects the judgment of both patients and clinicians on the medication's effectiveness, safety, and tolerability [
14]. This global proxy measure of medication effectiveness is the primary outcome measure in the National Institute of Mental Health Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) project [
14], an important 18-month, randomized, double-blind study comparing 1 typical antipsychotic (perphenazine) and 4 atypical antipsychotics (olanzapine, risperidone, quetiapine, and ziprasidone) in the treatment of schizophrenia in the United States. The CATIE trial concluded that olanzapine was the most effective in terms of discontinuation rates and time to discontinuation [
14]. The median time to discontinuation of treatment for any cause was significantly longer in the olanzapine group (9.2 months) than in the quetiapine (4.6 months, p < .001) or risperidone group (4.8 months, p = .002). The median time to all-cause treatment discontinuation was also longer for olanzapine compared to perphenazine (5.6 months, p = .021) and ziprasidone (3.5 months, p = .028), but these differences lost their statistical significance following adjustments for multiple comparisons. It is important to note that for various reasons [
14], the specific comparisons involving perphenazine and ziprasidone had a lower statistical power to identify group differences in treatment discontinuation (statistical power of 76% and 58% for perphenazine and ziprasidone, respectively, rather than 85%). It is unclear whether findings from CATIE, a randomized, double blind study, may generalize to open-label, naturalistic treatment in usual care settings. The objective of this study is to compare atypical and typical antipsychotics on time to all-cause medication discontinuation in the usual care of patients with schizophrenia spectrum disorders. We used data from a large, prospective, non-randomized, non-interventional, multisite, naturalistic study of schizophrenia patients in the United States. Using time to all-cause medication discontinuation for 1 year following medication initiation, we first compared 5 atypical antipsychotics (singly and combined) with typical antipsychotics of high-, medium-, or low-potency level. Next, we compared the atypicals (singly and combined) with 2 specific typical antipsychotics: (a) perphenazine, a medium-potency typical antipsychotic, and (b) haloperidol with prophylactic anticholinergic agents. Consistent with meta-analytical findings from randomized clinical trials (RCTs) [
9‐
11], we hypothesized that during treatment of schizophrenia patients in "real-world settings," time to medication discontinuation for any cause is longer for patients treated with atypical than typical antipsychotics, regardless of potency level, and that time to all-cause medication discontinuation will also differ among atypical antipsychotics compared to typical agents.
Discussion
Consistent with findings from meta-analyses of randomized, double-blind clinical trials [
9‐
11], this large naturalistic observational study found significant differences between atypical and typical antipsychotics in the treatment of schizophrenia. The present findings may help extend and complement previous observations on treatment duration in RCTs to treatment in usual clinical practice, using a clinically meaningful proxy measure of medication acceptability that appears to capture the patient and clinician judgments of the medication, including symptomatic response, safety, and tolerability [
14]. Longer treatment duration was previously shown to be associated with a lower risk of psychotic relapse and psychiatric hospitalization [
3‐
5,
14,
27], thus likely to decrease patients' personal burden and lower the economic costs for acute care in the mental health system [
6]. Longer treatment duration was also found to be associated with greater symptom improvement [
28] and better functional outcomes in the treatment of patients with schizophrenia [
29].
More notably, this naturalistic study presents novel and clinically important information about differences in treatment duration between specific atypicals and specific typicals (of high-, medium-, and low-potency levels; perphenazine, and haloperidol with prophylactic anticholinergic agents). Findings demonstrate that compared to typicals, the atypicals are not a homogeneous group in usual clinical practice. Remarkably, not all the 5 studied atypicals differentiated significantly and consistently from typicals of various potency levels, as findings were primarily driven by clozapine and olanzapine, and to a lesser extent by risperidone. Furthermore, only clozapine and olanzapine therapy showed consistently and significantly longer treatment duration compared to perphenazine, a medium-potency typical antipsychotic medication.
This study has also shown that some but not all atypical antipsychotics are associated with significantly longer treatment duration compared to haloperidol, when haloperidol is provided with prophylactic anticholinergics. Specifically, only clozapine, olanzapine, and risperidone were found to be associated with significantly longer treatment duration compared to haloperidol with prophylactic anticholinergic agents.
Current findings are consistent with several core findings from the CATIE trial [
14]. Despite substantial differences in design and methodology between CATIE and US-SCAP (e.g., CATIE is a randomized, double blind trial and US-SCAP is non-randomized and non-interventional), both studies found atypical antipsychotics to be a heterogeneous group on time to all-cause medication discontinuation. When excluding clozapine, which was not included in Phase 1 of CATIE, both studies found olanzapine-treated patients to have the longest time to medication discontinuation, followed in descending order by risperidone, quetiapine, and ziprasidone. In both studies, treatment with olanzapine, but not with risperidone, quetiapine, or ziprasidone, was associated with significantly longer time to all-cause medication discontinuation when compared with perphenazine. The statistically significant difference versus perphenazine found in CATIE (p = .021) lost, however, its significance following corrections for multiple comparisons (required p value ≤ .017). However, when similar stepwise corrections for multiple comparisons were performed in the US-SCAP analyses, the findings were essentially unchanged, still favoring olanzapine over perphenazine. Current findings suggest that if time to all-cause medication discontinuation in usual care were to reflect medication effectiveness as it does in CATIE [
14], than the effectiveness of clozapine and olanzapine relative to perphenazine, and the effectiveness of clozapine, olanzapine, and risperidone relative to haloperidol, may not be overestimated in usual care settings.
Our findings of differential treatment duration in usual care between typicals and atypicals are also consistent with previous reports of a lower rate of treatment failure and longer time to drug discontinuation [
11,
30‐
33] among schizophrenia patients treated with atypicals compared with typicals. Generally, there appears to be a 2-fold increased risk of drug discontinuation with typicals compared with atypicals [
34]. Current findings are, however, inconsistent with a meta-analysis by Wahlbeck and colleagues [
35], who reported lack of significant differences in dropout rates between typicals and atypicals other than clozapine. That meta-analysis included 163 randomized clinical trials, most of which were short-term, placebo-controlled trials. Importantly, the authors did not perform direct head-to-head comparisons of typicals and atypicals. When direct comparisons between typicals and atypicals were performed in another meta-analysis [
36], the findings supported the superiority of atypicals over typicals on "treatment failure" that captured dropout rates and symptomatic relapse.
Consistent with prior research in usual care setting [
37], we also found clozapine to be associated with the longest time to medication discontinuation, a finding that is congruent with clozapine's greater efficacy in meta-analytical reviews [
8,
9,
11,
35]. It is unclear, however, whether certain confounds are responsible for this finding. Unlike treatment with other antipsychotics, treatment with clozapine requires periodic blood monitoring to assess the risk of developing agranulocytosis. In addition to the likely initial selection of more adherent patients for treatment with clozapine in usual care, the frequent monitoring may help increase treatment duration with clozapine. Support for this hypothesis comes from randomized double-blind trials of clozapine versus olanzapine [
38,
39], in which the treatment groups did not significantly differ on dropout rates, possibly because the double-blind design required both treatment groups to undergo regular monitoring with blood tests.
There is growing interest in naturalistic studies that compare, among other variables, the differences in treatment duration between specific atypical and typical antipsychotics. Two very large, naturalistic, prospective, 3-year studies have been recently completed, reporting findings consistent with US-SCAP. One is the European Schizophrenia Outpatient Health Outcomes (EU-SOHO, N = 10,972), which was conducted in 10 European countries [
30], and the other is the Inter-Continental SOHO (IC-SOHO, N = 7658), conducted in 27 other countries [
31,
32]. In addition, a recent naturalistic study of 2230 first-episode schizophrenia patients hospitalized in Finland [
40] compared monotherapy treatment with various typical and atypical antipsychotics. The authors found the first- and second-generation antipsychotics to be highly heterogeneous groups with regard to effectiveness in a real-world setting, and reported that patients treated with clozapine or olanzapine had substantially lower discontinuation rates of their initial treatment than patients treated with haloperidol. In another recent study [
41], our group conducted a post hoc analysis of time to all-cause medication discontinuation using data from a 1-year, randomized, open-label study of schizophrenia patients treated with olanzapine, risperidone, or typical antipsychotics. Findings from that post hoc analysis were almost identical to the present study. Namely, the 1-year survival rate for patients randomized to olanzapine was significantly higher than it was for patients randomized to typical antipsychotics of high-, medium-, or low-potency levels, and compared to perphenazine. The survival rate for patients randomized to risperidone was significantly higher than typicals of high- or medium-potency levels but did not significantly differ from olanzapine, from typicals of low potency, or from perphenazine.
Current findings need to be interpreted in light of their limitations. First, the non-randomized, naturalistic design of this study introduced potential selection bias, which cannot be completely eliminated despite our use of statistical adjustments for a set of socio-demographic and treatment characteristics. The treatment groups may have differed on unobserved (thus unadjusted), pre-existing measures, such as symptom severity at the time of medication initiation. As noted earlier, US-SCAP did not assess patients' symptomatology at the time of medication initiation, only at predetermined intervals. Other confounds may have also affected the findings. These include the potential period bias, episode bias, and the potential for preference of the sponsor's medication (olanzapine) over other atypicals by participating physicians. One cannot completely adjust for these biases using statistical models, especially not the potential "sponsor bias." It should be noted, however, that the treating clinicians in the US-SCAP study were not connected to the study and were generally unaware that particular patients were participating in the study. Furthermore, if treating clinicians were biased in favor of the sponsor and its drug, one would have expected clinicians to quickly switch patients to olanzapine from antipsychotics they were receiving at enrollment. This phenomenon was not found when assessed with a sensitivity analysis. It is also important to note that the 3 statistical approaches used in this study and the 5 additional sensitivity analyses did provide consistent findings. Furthermore, our core findings on differential effectiveness between typicals and atypicals are highly consistent with those found in prior RCTs [
9‐
11,
15,
42], in retrospective claims database studies [
4,
21,
43,
44], and in naturalistic prospective studies [
4,
30‐
33,
40].
A second study limitation is the small sample size of the ziprasidone treatment group, which limited the statistical power of the analyses. It should be noted, however, that the current findings are consistent with CATIE [
14], in which the ziprasidone-treated patients had the shortest median time to all-cause discontinuation (3.5 months in the 18-month trial) among the 5 treatment comparators, and the highest rate of discontinuation rate for any cause (79%). These findings were not statistically significant, likely due to much-reduced statistical power (change from 85% power for comparisons versus other atypicals to 58% power versus ziprasidone).
Third, our study was not designed to assess reasons for initiation or discontinuation of the antipsychotic medications. Findings from recent research suggest, however, that in addition to patient preferences, antipsychotic medication efficacy [
14,
28], rather than safety or poor tolerability, is the primary reason for medication discontinuation in the treatment of schizophrenia.
And lastly, this study did not require patients (or treatment episodes) to be confined to antipsychotic monotherapy. Although our approach did not distinguish between the augmenting and augmented antipsychotics, it is important to note that augmentation roles change over time due to prevalent switching of antipsychotics, a phenomenon we captured in analyses with multiple treatment episodes. Furthermore, antipsychotic switching processes are often characterized by medication overlap with cross-titration, which amounts to short-term (and at times prolonged) antipsychotic polypharmacy, thus making true monotherapy treatment episodes relatively rare and possibly not representative of the complex treatment regimens in real-world practice. Most importantly, the current findings are highly consistent with RCTs [
14,
42] and 3 large naturalistic studies conducted outside of the U.S. that included only monotherapy-treated patients [
30‐
32,
40].
The strengths of this study lie primarily in its prospective, naturalistic, real-world, long-term perspective; its large sample size; the use of an observational approach without any planned treatment intervention; the ability to provide comparative data on a number of commonly used antipsychotics; the ability to generalize the findings to patients treated in large public systems of care across the United States; and the availability of medication information during hospitalizations, a type of data that is typically absent in claims databases. Further, this study offers comprehensive medication information because patients were also queried about use of psychiatric medications outside of their regular treatment site. When this occurred, systematic efforts were made to abstract this out-of-site information. Additional strengths of this study lie in its use of multiple statistical approaches and sensitivity analyses, to help check the robustness of the findings, and the ability to proactively address several analytical challenges that are inherent in analyses of naturalistic, longitudinal data.
Acknowledgements
The US-SCAP study was supported by Eli Lilly and Company and administered by the Medstat Group. We wish to thank the site investigators and others who collaborated in the research. By site, they include Maryland: A.F. Lehman M.D., M.S.P.H., University of Maryland School of Medicine and G. Gallucci, M.D., M.H.S., Johns Hopkins Bayview Medical Center; Colorado: C. Harding Ph.D., University of Colorado (previously); Florida: D. Shern, Ph.D., Florida Mental Health Institute, University of South Florida, and T Saunders, M.S., (previously Florida Mental Health Institute); North Carolina: J. Swanson, Ph.D., L.A. Dunn, M.D., and M. Swartz, M.D., Duke University Medical School; California: R.L. Hough, Ph.D., and C. Barrio, Ph.D, Child and Adolescent Services Research Center and San Diego State University; Connecticut: R.A. Rosenheck, M.D., and R. Desai, Ph.D., VA Connecticut Health Care System; Medstat Group: P. Russo, Ph.D. M.S.W., R.N., (previously), L. Palmer, Ph.D., L. Torres, MBA, and B. Cuffel, Ph.D. (previously); Eli Lilly and Company: D. Buesching, Ph.D., Bryan M. Johnstone, PhD., and T. Croghan, M.D. (previously); Consultants: D. Salkever, Ph.D., Johns Hopkins University, E. Slade, Ph.D. (previously Johns Hopkins University), W. Hargreaves, Ph.D. and M. Shumway, Ph.D., University of California, San Francisco.
Competing interests
Drs. Ascher-Svanum, Zhu, Faries, and Landbloom are employees of Eli Lilly and Company, Indianapolis, Indiana, and minor stockholders in that company.
Dr. Swartz has received research funding from Eli Lilly and Company, and consulting and educational fees from AstraZeneca Pharmaceuticals LP, Bristol-Myers Squibb, Eli Lilly and Company, and Pfizer Inc.
Dr. Swanson has received funding support from Eli Lilly and Company as a SCAP site investigator.
Authors' contributions
• HAS conceived of the study, participated in its design, the analytical plan, the interpretation of the results, and drafted the manuscript.
• BZ and DF participated in the design of the study, the analytical plan, the interpretation of the results, performed the statistical analyses, and assisted drafting the manuscript.
• RL participated in the design of the study, the interpretation of the results, and assisted drafting of the manuscript.
• MS and JS were involved in data collection of US-SCAP as co-PIs at a study site, participated in the design of the study, the analytical plan, the interpretation of the results, and assisted drafting the manuscript.