INTRODUCTION

Historically, the prognoses of patients diagnosed with schizophrenia were considered variable and pessimistic (Kraeplin, 1919). This is because schizophrenia was believed to be a brain disorder that was expressed in abnormal mental functions, which characteristically appeared in the late second and third decades of life, and usually led to a chronic or recurrent course that produced clinical deterioration in the form of persistent symptoms, impairments in social and work functions, and erosion of the personality (McGlashan, 1988).

Following their introduction in the 1950s, conventional antipsychotic medications, or neuroleptics, were shown to suppress the acute psychotic symptoms of the disorder and prevent their recurrence. However, many patients were found to be unresponsive to neuroleptics and it was generally believed that, despite their ability to suppress symptoms, they did not change the long-term course of the disorder or improve the outcome (Hegarty et al, 1994). In addition, neuroleptics were associated with high rates of neurological side effects (ie acute extrapyramidal signs (EPS) and tardive dyskinesia (TD)), which compromised the therapeutic effects of treatment and impelled many patients to discontinue treatment, which led to higher relapse rates.

In contrast, studies of patients in the first episode of their illness demonstrated high rates of therapeutic response and symptomatic remission in comparison with chronic patients with multiple prior episodes (Lieberman et al, 1993,1998; Sheitman et al, 1997; Lieberman et al, 1998). At the same time, they also experienced high rates of neurological side effects to neuroleptic drugs (Chatterjee et al, 1995; Chakos et al, 1996). In general, first-episode patients appeared to be more sensitive to the pharmacologic effects of antipsychotic drugs than older chronic patients in that they exhibited higher rates of recovery, more frequent extrapyramidal side effects, and required lower drug doses (Lieberman et al, 1998; McEvoy et al, 1991). However, following their recovery first-episode patients experienced a high rate of psychotic relapse particularly if they discontinued antipsychotic medication (Crow et al, 1986; Robinson et al, 1999).

The advent of the second-generation or so-called ‘atypical’ antipsychotic drugs was heralded as a therapeutic advance in the treatment of schizophrenia and related psychotic disorders. It was claimed that atypical antipsychotics had a different mechanism of action and were more effective and safer than conventional antipsychotics (Kinon and Lieberman, 1996). However, these claims are only partially supported by empirical data, and there is a substantial variation in opinion among researchers and clinicians about their comparative effectiveness (Leucht et al, 1999; Geddes et al, 2000; Chakos et al, 2001). Since their introduction into clinical practice, atypical antipsychotic drugs have been examined in comparison with the conventional antipsychotics in heterogeneous patient samples predominantly comprising chronically ill patients whose symptoms may have been (partially) resistant to treatment. The question of the comparative efficacy and safety of these two classes of antipsychotic drugs has not been examined in first-episode patients beyond short-term trials (Emsley, 1999; Sanger et al, 1999). Indeed this question has considerable importance, since first-episode patients are highly responsive to pharmacologic treatment, and the first treatment intervention in drug naive patients is a critical therapeutic opportunity that has the potential to influence the course and outcome of what could be a lifelong illness (Lieberman, 1996). This is particularly the case because longer duration in the first episode of psychosis has been associated with poorer treatment response and outcome (Norman and Malla, 2001). Thus, the rapidity and maintenance of treatment response, as well as tolerability, which is associated with treatment adherence, may be important factors in determining the outcome.

Our hypothesis was that clozapine (CLZ), the prototypical second-generation antipsychotic, had superior efficacy in first-episode patients when compared to a representative conventional antipsychotic. When this study was conceived, CLZ was the only second-generation antipsychotic marketed in any part of the world. By the mid-1990s, CLZ was, depending on the location, the first or second most commonly used antipsychotic in China (Wu and Wu, 1995; Phillips, 2001). Although in most Western countries CLZ is used as a second-line treatment because of its hematological toxicity (Herz et al, 1997), in China it is often used in the first episode of illness. Given this set of circumstances, we felt it was experimentally feasible and clinically justifiable to examine the efficacy of CLZ in first-episode patients in China.

METHODS

Setting

The study was conducted at the Beijing Hui Long Guan Hospital, one of the largest psychiatric hospitals in China, which is located 30 km from central Beijing and serves a catchment area population of 13 million people. Dr Phillips has worked in the PRC since 1985 and is the Director of the Research Center of Clinical Epidemiology at the Hui Long Guan Hospital. The study protocol was designed by the first and second authors (JL and MP). It conformed to international standards for research ethics and was approved by the Research Review Board of the Beijing Bureau of Health prior to its initiation and monitored annually while in progress. Funding for the study and medication supplies were provided by the Novartis Pharmaceutical Company. The authors had unrestricted access to the data; company personnel were not involved in the analysis of the data nor did they have any authority over publication.

Subjects

All hospital admissions were screened from October 1995 to December 1998 for patients who met the following criteria: (1) diagnosis of schizophrenia or schizophreniform disorder; (2) duration of symptoms not longer than 60 months; (3) no prior treatment with antipsychotic medication or, if previously treated, a total lifetime usage of less than 14 days; (4) between 16 and 40 years of age; and (5) current psychotic symptoms of moderate severity or greater measured by one of the five psychotic items in the Brief Psychiatric Rating Scale (BPRS). Patients meeting these criteria, and their families, provided written informed consent to participate.

Procedures

Patients' diagnoses were based on the results of the Structured Clinical Interview for DSM (SCID) using the criteria of the fourth edition of the Diagnostic and Statistical Manual (DSM-IV), physical exams and routine chemistry, and hematology laboratory tests. Attending-level psychiatrists used a Chinese translation of the SCID to interview patients and their family members, and were given freedom to ask additional questions if respondents did not understand the standard probes. There are no major difficulties in employing DSM IV criteria to diagnose schizophrenia in China; current Chinese diagnostic criteria for schizophrenia are similar to ICD-10 criteria, but many locations also use DSM IV criteria. Patients were randomized to treatment with CLZ plus benztropine (BZ) placebo or chlorpromazine (CPZ) and BZ under double-blind conditions. Medication dosages were increased using a standardized regimen during the first 28 days of admission (to a maximum of 400 mg/day of CLZ or 600 mg/day of CPZ) and then adjusted based on the clinical response and side effects. Patients taking CPZ were also given 2 mg BZ twice a day and those on CLZ were given the corresponding placebos. Patients received inpatient care for 12 weeks (a typical length of stay in Chinese psychiatric hospitals), and were then followed as outpatients on a monthly basis for the subsequent 9 months. After the initial 12-month study period, patients were eligible to enter a double-blind extension phase of the study that is still under way.

Patients dropped out of the study because of withdrawal of consent by patients or family members, failure of treatment to relieve symptoms, or severe side effects. Detailed treatment adherence information was recorded; all patients in the study continued the clinical and side effect evaluations as per schedule even if noncompliance occurred, in order to ensure the completeness of data on the total sample.

Assessments

Patients were evaluated using the Chinese version of the BPRS (Phillips et al, 1990), the Chinese version of the Scale for Assessment of Negative Symptoms (SANS), (Phillips et al, 1991), Clinical Global Impression Scale (CGI), Global Assessment of Function Scale (GAF), the Simpson Angus Extrapyramidal Symptoms Scale (SAESS) (with sialorrhea omitted because of its well-known connection with CLZ) and for side effects at study entry, weekly for weeks 1–6, biweekly for weeks 6–12, at hospital discharge, and at 3-monthly intervals following discharge. Side effects were recorded using the Coding Symbol and Thesaurus for Adverse Event Terminology (COSTART) classification terms at each assessment visit. EKGs were performed at study entry, weeks 12 and 52. All patients underwent venipunctures to monitor white blood counts (WBC) weekly while inpatients and monthly after discharge. Serum glucose was assessed under fasting conditions at study entry, week 12, and every 3 months after discharge. Weight was not measured regularly by the protocol or hospital procedure. Consequently, only subgroups of patients in this trial had baseline and follow-up weights. Moreover, most of the follow-up weights were taken after the 52-week period of the formal trial and were not standardized in terms of frequency and time intervals. Since patients who successfully completed the trial were continued on their study medication in a blinded follow-up phase of the study thereafter, we included the first available follow-up weight of each patient in the analysis (n=57 for CPZ and n=58 for CLZ).

Statistical Methods

The primary measure for treatment efficacy is remission, which was defined a priori as a 50% reduction in total BPRS score from baseline with no score greater than mild (3) on the 5 BPRS psychosis items (unusual thought content, suspiciousness, hallucinations, conceptual disorganization, mannerisms, and posturing) and a CGI-severity item of mild (3) or less. Using an intent-to-treatment (ITT) population, we compared the time to first remission between the two treatments using Kaplan–Meier estimates for their survival curves, and refined the model by controlling for baseline BPRS, gender, duration of untreated psychosis (DUP), and age of onset in Cox proportional hazard regression. Owing to the recurrence of remissions and relapses, we identified the week-by-week remission status for each patient during the first 12 weeks of treatment and then monthly thereafter evaluated the overall remission status by taking the proportion of weeks on remission over the total number of weeks followed for each individual, and modeled it using generalized estimating equation (GEE) logistic regression with robust standard error estimates for the possible overdispersion caused by correlated responses within patients.

For secondary efficacy measures, we looked at the BPRS, SANS, CGI, and GAF scores at the end of inpatient treatment (12 weeks) and at the end of the study (52 weeks). We compared the CPZ and CLZ groups using ANCOVA models controlling for baseline, gender, duration, and age of onset of psychotic symptoms. The possible interactions between treatment and covariates were tested and explored in further detail if significant. The comparison was conducted using both the completers at the time point (observed cases (OC)) and the ITT patient population with missing values filled in from the last available observation (last observation carried forward (LOCF)).

The clinical and laboratory parameters (percentage of polymorphonuclear neutrophils and lymphocytes, WBC, glucose, heart rate, and EKG QT interval) and severity of side effects (total SAESS score and parkinsonism subscale) were compared between the two groups using ANCOVA models similar to those for the continuous efficacy measures. For recurrent adverse events (COSTART terms), we recoded all the ranking scales (0–3) higher than 0 for the presence of the side effects. The transformation from ranking to binary scale was conservative; however, by pooling sparse observations of higher rank responses, we were able to get more reliable estimates of odds ratios. The proportion of visits recorded with occurrence was then modeled for each adverse event using the GEE logistic regression similar to that for remission.

The test on primary outcome used a two-sided alpha-level of 0.05. Other tests on numerous clinical measures and side effects were conducted as secondary analysis to supplement our major finding. Our focus was on the efficacy and safety profiles of CLZ and CPZ, instead of on any specific score. p-Values from the tests for the 14 subscale scores considered were reported and no multiple comparison adjustment was applied.

RESULTS

Description of Study Cohort

Of 2708 patients screened for the study, 171 met the criteria for first-episode patients. Of these, 164 patients consented to study participation, 81 were randomized to CLZ, and 83 to CPZ. However, one patient in the CLZ and three in the CPZ group withdrew before the first administration of medication. Thus, the valid number of patients in each treatment group was 80 (Figure 1).

Figure 1
figure 1

The design and patient flow of the study. Of the 2708 patients screened for the study, 171 met the inclusion criteria, 164 patients consented to study participation, 81 were randomized to CLZ, and 83 to CPZ; three patients in the CLZ and one in the CPZ group withdrew before the first administration of medication; so the valid number of patients in each treatment group was 80. Nine CPZ- and 10 CLZ-treated patients withdrew their consent and dropped out of the study. Three patients were withdrawn for lack of efficacy in the CPZ group. Six CPZ-treated and two CLZ-treated patients were withdrawn for adverse effects. The overall attrition rates at week 12 were 10% and 11% for CLZ and CPZ groups, respectively. By 52 weeks, 22.5% for CPZ and 15% for CLZ.

The two treatment groups were comparable in age, gender, diagnosis, baseline symptom severity, and duration and age of onset of psychotic symptoms. Patients ranged in age from 15 to 42 (mean[SD], 28.7[6.9]) years; 48% were female, all were Chinese. (Some patients older than the enrollment criteria were included because their actual age (by their birth certificate) was greater than the age they reported at the time of admission.) Among them 52.5% were diagnosed with paranoid schizophrenia; 23.1% with undifferentiated schizophrenia; and 24.4% with schizophreniform disorder. The first psychotic symptom occurred at the mean age of 27.2[6.5] (median=26.9) in this group of patients. At the time of study entry, patients had had psychotic symptoms for an average of 18.4[17.8] months (median=10.7). At baseline, the average symptom severity score was 43.8[5.1] for total BPRS, 5.6[0.6] for CGI, and 35.8[7.8] for GAF.

The median doses (mg/day) at 12 weeks and at the end of 1-year follow-up were 400 and 300, respectively, for CLZ, and 600 and 400 for CPZ. Nine CPZ- and 10 CLZ-treated patients withdrew their consent and dropped out of the study. Three patients were withdrawn for lack of efficacy in the CPZ group. Six CPZ-treated and two CLZ-treated patients were withdrawn for adverse effects. No severe noncompliance occurred during the acute inpatient phase, but 26 of 72 (36%) CPZ and 23 of 71 (32%) CLZ patients took less than 50% of their prescribed medications, as determined by pill counts, for more than 1 month after discharge. The overall attrition rates were similar at week 12 for CPZ and CLZ groups at 10% and 11%, respectively. By the end of study, more patients had dropped out or been withdrawn in CPZ (22.5%) than in CLZ (15%) group, but this difference was not statistically significant (Figure 1).

Efficacy

Remission

In all, 79% of patients (63/80) on CPZ and 81% (65/80) on CLZ had remitted before leaving or completing 52 weeks in the study. Figure 2 shows the Kaplan–Meier remission survival plots for time to remission for CPZ and CLZ groups. Patients in the CLZ group achieved their remission faster than patients in the CPZ group (χ2(1)=4.10, p=0.04 by Log-Rank test). It took 8 weeks of treatment for half the patients in the CLZ group to achieve their first remission compared to 12 weeks in the CPZ group (χ2(1)=5.56, p=0.02 by Wilcoxon test). The Cox-regression model adjusting for baseline total BPRS, patient sex, DUP, and age of first psychotic symptoms found that the remission rate was almost twice as fast in the CLZ group as in the CPZ group (hazard ratio=1.69, 95% CI (1.18, 2.44), p=0.005).

Figure 2
figure 2

Kaplan–Meier remission survival plots for time to first remission for CPZ (broken line) and CLZ (solid line) groups. The median time to remission in the CLZ group was 8 and 12 weeks in the CPZ group.

The GEE logistic regression model found that patients in the CLZ group spent a significantly larger proportion of time in remission than the CPZ group. Controlling for gender, baseline total BPRS, DUP, and age of onset of psychotic symptoms, the odds for being in remission was almost doubled for the CLZ group than the CPZ group (odds ratio=1.73(1.20, 2.50), p=0.003). The model also identified a significant effect for the DUP on remission status. With a 1-year increase in untreated psychosis, the odds of being in remission were reduced by nearly 15% (odds ratio=0.85 (0.75, 0.95), p=0.006).

BPRS and other clinical measurements at the end of inpatient treatment

In the completer analysis, controlling for baseline, gender, DUP, and age of onset, patients who completed 12 weeks of inpatient treatment in the CLZ group experienced significantly greater reductions in psychopathology than CPZ on multiple rating scales including total BPRS, the BPRS Anergia Factor, total SANS, SANS Affective, Poverty of Speech, Avolition, and Anhedonia subscales, CGI severity and the GAF scales. In the LOCF analysis, CLZ evidenced significantly superior efficacy in the BPRS Anergia Factor, total SANS, SANS Affective and Avolition subscales, and was almost significant on the GAF scale (Table 1).

Table 1 Efficacy Comparison for CLZ and CPZ at Weeks 12 and 52

The analysis also found that patients with longer prior illness duration had higher total BPRS scores at 12 weeks (F(1,138)=6.83, p=0.01, OC) after controlling for their baseline scores, indicating a relation between DUP and therapeutic response. Similar findings supporting worse psychopathology with longer duration of untreated psychotic symptoms were found in other BPRS and SANS subscales (BPRS Anergia Factor (F(1,138)=8.02, p<0.01, OC; F(1,154)=5.93, p=0.02, LOCF), total SANS (F(1,138)=6.95, p=0.01, OC; F(1,154)=8.55, p<0.01, LOCF), SANS Avolition (F(1,154)=4.57, p=0.03, LOCF), Anhedonia (F(1,138)=6.52, p<0.01, OC; F(1,154)=8.46, p<0.01, LOCF), and attention deficit (F(1,138)=17.01, p<0.01, OC; F(1,154)=16.69, p<0.01, LOCF).

Since the major efficacy advantage of CLZ at 12 weeks was on negative symptom scores, to rule out the possibility that they were because of worse EPS in the CPZ group, we repeated the analysis of symptom reduction by treatment controlling for 12-week Parkinsonism severity. The changes in major findings were minimal. A significant efficacy advantage of CLZ over CPZ at week 12 was shown on BPRS total (F(1,137)=3.89; p=0.05), SANS total (F(1,137)=6.17; p=0.01), affective flattening (F(1,137)=10.52; p<0.01), avolition (F(1,137)=4.45; p=0.04), CGI (F(1,137)=5.48; p=0.02), and GAF (F(1,137)=5.28; p=0.02) using OC population; and on BPRS anergy (F(1,153)=4.00; p=0.05) and SANS affective flattening (F(1,153)=10.48; p<0.01) with all ITT population using LOCF.

At 52 weeks there were no differences between the treatment groups on measurements of symptoms. Since the attrition rates were different in the two groups (22.5% in CPZ and 15% in CLZ), an exploratory analysis was conducted that compared BPRS total scores at discharge between the patients who dropped out during follow up vs those remaining on treatment. The dropouts in the CPZ group had significantly higher BPRS scores (26.0[7.5]) at discharge than the other three groups (22.1[5.0] for CPZ on-treatment, 20.6[4.1] for CLZ on-treatment, and 20.3[2.5] for CLZ dropouts). The dropout of sicker patients on CPZ may have contributed to the lack of a persistent difference between the two groups in the LOCF analysis at 52 weeks.

There were 14 rehospitalizations in weeks 13–52. Five patients on CPZ had a total of seven rehospitalizations while six patients on CLZ had seven rehospitalizations. There were no known suicide attempts by any study participants in the 1st year of the study.

Safety

Extrapyramidal symptoms

The means for the total SAESS and Parkinsonism are listed in Table 2. Despite the use of prophylactic BZ, the CPZ group consistently showed more EPS than the CLZ group. At both weeks 12 and 52, patients receiving CPZ had significantly higher total scores on the Simpson Angus Scale and the Parkinsonism subscale with OC analysis. The LOCF analysis corroborated the significant difference at 12 weeks, but the 52 weeks comparison was again not significant. The ANCOVA model also found that females had more severe Parkinsonism than males at week 12 (F(1,138)=5.51, p=0.02, OC; F(1,154)=5.30, p=0.02, LOCF), but the difference was not significant at week 52.

Table 2 Simpson–Angus Extrapyramidal Symptom Scale (SAESS) Total Score and Parkinsonian Score for CLZ and CPZ at Weeks 12 and 52aa

Other side effects

Individual adverse events with significant treatment effects are listed in Table 3. Over the course of study follow up, the CLZ treatment group was at a significantly lower risk for developing side effect symptoms such as dystonia, blurred vision, muscle tension, dry mouth, akathisia, objectively observed restlessness, and decreased urine production, and showed an increased risk for sweating and a marginally increased risk for depressed affect. As stated earlier, eight patients were withdrawn because of adverse effects; six from CPZ (constipation, hyperthermia, ileus, dermatitis, neutropenia, EPS) and two from CLZ (hyperthermia, hypotension). One patient (35-year-old male) died at week 10 of treatment with CPZ while in the hospital after developing hyperthermia (41.5 C) and leukocytosis (16 300 with 78% neutrophils) but no rigidity or marked alteration of consciousness. The cause of death was determined to be hyperthermia possibly because of viral encephalitis. (This event occurred in summer during a heat wave in Beijing.)

Table 3 Significant Treatment Effects (CLZ vs CPZ) on Occurrence of Individual Adverse Events During Follow-Up (up to 52 weeks)aa

At study entry the CLZ and CPZ groups had similar total WBC (7887[2270]), percentages of polymorphonuclear neutrophiles (64%[8%]) and lymphocytes (36%[8%]), fasting glucose levels (4.7[0.96] mmol/l), heart rates (81[17.3]), and QT intervals (0.35[0.04] ms) (Table 4). After treatment there were no significant differences between the two drugs for total WBC or glucose level. OC analysis found a lower percent of neutrophiles and higher lymphocytes at week 12 for patients taking CLZ (F(1,102)=5.75, p=0.02 and F(1,102)=5.12, p=0.03, respectively), but it was not supported by the LOCF analysis, nor were the significant findings maintained at week 52 by OC or LOCF analysis. The CLZ group had a significantly shorter QT interval at both 12 weeks (F(1,126)=16.61, p<0.001, OC; F(1,143)=15.64, p<0.001, LOCF) and 52 weeks (F(1,143)=10.25, p=0.002, LOCF), and a higher heart rate at 52 weeks (F(1,60)=6.88, p=0.01, OC; F(1,149)=6.70, p=0.01, LOCF).

Table 4 Laboratory and other Clinical Parameters at Weeks 12 and 52

Weight

Baseline and follow-up weights were obtained for 72% of subjects (71% for CPZ and 73% for CLZ). The median times for the first weight measurement during follow-up were 131 weeks after starting medication for CPZ vs 137.5 weeks after starting medication for CLZ. For the available data, the mean weights were 60.3 kg (9.7) for CPZ and 61.6 kg (10.1) for CLZ at baseline; and 68.8 kg (12.0) and 71.5 kg (13.3) for CPZ vs CLZ at the first follow-up, which was significant by an ANCOVA model of weight change (controlling for gender, baseline weight, age of first onset and duration of psychotic symptoms and follow-up); weight increases from baseline to follow-up (F(1,108)=36.6, p<0.01). However, the difference in the mean weight gain for CPZ (6.5 kg) and CLZ (9.9 kg) was not statistically significant (F(1,108)=1.10, p=0.30).

DISCUSSION

This study is the first to examine the long-term outcome of treatment-naive first-episode schizophrenia patients in response to atypical and conventional antipsychotic drugs in the context of a randomized double-blind controlled trial. These results are consistent with prior studies in Western countries, which demonstrated that first-episode patients are capable of having good therapeutic responses to antipsychotic medications (Lieberman et al, 1993; Sheitman et al, 1997; Robinson et al, 1999). Moreover, the demographic and clinical characteristics of the sample, including the duration of prior psychotic symptoms, were comparable to those of patients in prior studies from Western countries further indicating the generalizability of these findings (Sanger et al, 1999; Crow et al, 1986; Loebel et al, 1992; Johannesen et al, 2001). The slightly later age of onset reported by patients in this study (mean 27.2 years) relative to that reported in Western countries is an interesting finding that deserves further investigation. The relation between increased duration of psychosis with poorer treatment outcome is consistent with most (Loebel et al, 1992; Wyatt, 1999; Szymanski et al, 1996) but not all (Beng-Choon et al, 2000; Craig et al, 2000; Barnes et al, 2000) prior studies. The study's results also suggest that CLZ, the prototype of atypical drugs, may have superior benefits to conventional medications reflected by faster treatment responses, significantly improved better short-term efficacy, more time spent in remission, reduced EPS and other adverse side effects, and better retention of patients in treatment. The prophylactic use of 4 mg per day of BZ in the CPZ group may have contributed to some of the anticholinergic side effects. At the same time, CLZ was associated with more sweating and increased heart rates. Sialorrhea was not assessed in this study but is known to be common with CLZ use. In spite of the previously reported association of CLZ and other atypical antipsychotic drugs on glucose metabolism (Henderson et al, 2000), there were no significant differences between treatments in this trial. Although weights were not systematically obtained, both CPZ and CLZ were associated with substantial weight gain consistent with previous studies (Allison et al, 1999). Cholesterol and triglyceride elevations, which have also been associated with CLZ treatment, were not assessed (Henderson, 2000). CLZ -treated patients had higher heart rates, but this did not cause any impairment to patients or lead to treatment discontinuations.

Although the rapidity and amount of time in remission was greater for CLZ, the two drugs did not differ on symptom ratings at 52 weeks. The high degree of treatment responsiveness of patients early in the course of schizophrenia may make the demonstration of differences in treatments on the long-term outcomes using traditional measures of psychopathology difficult. Despite the potential ‘ceiling effect’, CLZ produced significantly better symptom reductions than CPZ at 12 weeks. The failure to sustain the significant differences in symptom reduction between CLZ and CPZ beyond 12 weeks could have been because of more posthospitalization dropouts among less responsive patients in the CPZ group than the CLZ group, thereby decreasing the differences in improvement seen at 52 weeks as suggested by the exploratory analysis. Thus, long-term follow-up of all patients who entered the study including those who dropped out would be informative. In addition, it may be necessary to employ additional measures of outcome beyond symptom ratings to characterize more fully the treatment effects of antipsychotic drugs and discern the potential advantages of CLZ as representative of the atypicals. However, it is also debatable to what extent the effects of CLZ in this patient population would generalize to the other atypical agents. If, as these results suggest, atypical drugs produce faster rates of symptom remission, cause fewer side effects, and are associated with better treatment adherence compared to conventional drugs, then follow-up studies longer than 1 year may be necessary to demonstrate putative benefits reflected by reduced relapse rates, illness progression, changes in cognitive functions, and cumulative morbidity. The full extent of the potential superiority in treatment effects of the atypical antipsychotic drugs may only be appreciated in the context of very long-term outcome studies or after extensive clinical experience with these compounds in a new generation of patients who receive them from the beginning of their illness.