TO THE EDITOR: We thank Drs. Capehart and Holsinger, Dr. Barbui, and Dr. Mattes for sharing their opinions. Capehart and Holsinger suggest that we combined patients from distinct diagnostic groups. However, only subjects with DSM-III-R schizophrenia or schizophreniform or schizoaffective disorder were randomly assigned to treatment. We do not feel that these diagnoses represent distinct groups. All three are classified by the DSM nomenclature under "schizophrenia and other psychotic disorders." Furthermore, ICD-10 includes schizophreniform disorder under schizophrenia, which underscores the diagnostic similarities. Moreover, 83% of the 1,996 patients had the specific diagnosis of schizophrenia. In addition, there was no statistically significant difference among diagnostic groups in efficacy or safety conclusions. Capehart and Holsinger's second point alludes to different severity levels. We assume that they are referring to baseline severity as defined by total score on the Positive and Negative Syndrome Scale. In our opinion, mean scores of 90.1 (olanzapine group) and 92.1 (haloperidol group) do not suggest a clinically relevant diversity in symptom severity. Furthermore, these baseline differences did not contribute to analytical differences in the statistical methods, e.g., analyses of covariance that used patients' baseline score as the covariate. Under those analyses, all measures of efficacy were statistically significant in favor of superior efficacy within the olanzapine group.

Drs. Capehart and Holsinger also raise the issue of the relatively high fixed-dose haloperidol regimen. We are perplexed by this interpretation, since this was not a fixed-dose study. We clearly indicated that patients were assigned to treatment with 5 mg/day of either olanzapine or haloperidol and that either drug could be subsequently increased by 5 mg/week to 20 mg/day or decreased to a minimum of 5 mg/day as clinically warranted. There was no requirement to treat patients with a dose of haloperidol in excess of 5 mg/day. Doses lower than 5 mg/day or higher than 20 mg/day would have been subject to greater critique. In a review (1), King noted that "the majority of studies with haloperidol have found that there is no advantage in exceeding standard doses (10–20 mg/day) even in treatment-resistant cases." Thus, our study employed an optimal dose range under relatively naturalistic conditions.

The next question raised is the possible impact of prior suboptimal responses to oral or depot haloperidol. We addressed the issue by excluding patients who had not responded to conventional dopamine (D₂) antagonists (i.e., had failed to achieve during the previous 2 years a significant clinical response to 6-week trials of three neuroleptics from three different chemical classes given at a minimum of 800 mg chlorpromazine equivalents).

Capehart and Holsinger mistakenly suggest that the trial had a single-blind design. As we stated in the summary, the design of this international multicenter trial was double-blind. Thus, clinicians were not "nonblind." As for the likelihood of identifying familiar (haloperidol) or unfamiliar (olanzapine) adverse events, it is our suspicion that investigators would be more likely to identify and note those events with which they were unfamiliar.

The last criticism is the one that is most disturbing. Capehart and Holsinger suggest that a responsibility to promote olanzapine sales for our employer subtly influenced our decision as to what data should be included in the article. This comment is inappropriate and unsubstantiated. Biased selective reporting of data, in any form, has no place in a source review of scientific data. We can report with confidence from personal experience that pharmaceutical-based studies are conducted with the same rigor as psychopharmacology research in academic settings. To imply that support for one's salary biases the integrity of submitted work calls into question any academic or private researchers who receive funding from federal, foundation, or other sources that are vital to their programs. This study and its results have been scrutinized by regulatory agencies from more than 30 countries. As part of the peer review process of the Journal, our report was further reviewed for accuracy and scientific validity by independent experts in the field before being accepted for publication.

Dr. Barbui requests more information on several aspects of the international collaborative trial. Space limitations did not permit citing all 174 sites in the manuscript. Analyses for a possible treatment-by-geographic region interaction were conducted. These analyses revealed no statistically significant treatment-by-geographic region interactions. Of those patients assigned to olanzapine (N=1,336), 83% were diagnosed with schizophrenia (paranoid subtype: 50%, undifferentiated: 21%, disorganized: 6%, residual: 5%, and catatonic: 1%). Approximately 15% of the total group had schizoaffective disorder, and 2% had schizophreniform disorder. The demographic and clinical distribution of the 660 patients assigned to haloperidol did not significantly differ from that of the olanzapine group.

We stated that the olanzapine and haloperidol groups were similar with regard to patient and illness characteristics, which included their course of illness. Approximately 60% of subjects had a chronic course, 27% had a chronic course that was in acute exacerbation, and 6% had a subchronic course that was in acute exacerbation. The remainder had a subchronic, unspecified, or remitted course. There were no statistically significant differences between the two treatment groups.

Dr. Barbui states that around 77% of the subjects were intolerant of current antipsychotic therapy. Correct reading of the results shows that these patients either had to discontinue or were otherwise judged unresponsive to their last course of antipsychotic therapy according to the overall clinical impression of the investigator at the time of screening. Not all had been intolerant. We characterized the group as moderately ill on the basis of the mean BPRS total scores for the olanzapine and haloperidol groups (33.1 and 34.1, respectively). The a priori choice of a minimum BPRS score of 18 was selected because of the inclusion of both inpatients and outpatients. A slightly reduced minimum severity from the conventional standard of 24 used in other olanzapine trials was felt to be more concordant with an outpatient population, while still maintaining a symptomatic study group. Overall, we believe that the description of the population provided a reasonable approximation of the type of patient in the trial.

It is further suggested that the comparison group had more severe symptoms than the experimental group at baseline. While group differences may be statistically significant, one must consider that these differences are evaluated not only by p values but also by group size. In this case, there was only a 1-point difference in mean BPRS total scores, with comparable standard deviations, between the two treatment groups. The Clinical Global Impression (CGI) severity scores for the two groups were identical. Thus, there is little clinical evidence that the haloperidol group was more symptomatic. We are somewhat perplexed by the comment that the haloperidol group received lower drug doses than the olanzapine group. The study employed a flexible-dose design. The mean modal dose for patients treated with haloperidol (11.8 mg/day) was within a standard dose range (1). Furthermore, there is no reason to believe that two pharmacologically distinct agents would end up with identical mean modal doses. We do not define this as an "unbalanced situation."

Dr. Barbui hypothesizes that a temporal relationship between negative symptoms and extrapyramidal side effects may have existed. The protocol was structured to minimize the effect of extrapyramidal symptoms by permitting the use of benztropine, which, incidentally, was used significantly more often within the haloperidol arm. We believe that a very reasonable way to address secondary negative symptom efficacy with olanzapine relative to haloperidol is to use fixed-dose, placebo-controlled trials. This, in fact, was the subject of our companion article in the Journal. We trust that the author read this article, which employed path analytic techniques and demonstrated that olanzapine had negative symptom treatment advantages over both haloperidol and placebo beyond those substantially accounted for by extrapyramidal symptoms.

We agree with Dr. Mattes that the majority of patients who participated in the international collaborative trial had not enjoyed a robust clinical response to one or more conventional D₂ antagonists. One of the inherent challenges in conducting a large registration trial with a novel psychopharmacological agent is that interested subjects are typically not those who have had an optimal or satisfactory response to conventional therapies. What incentive would exist for their participation? Investigational drug trials are most likely to include subjects with a suboptimal past treatment experience.

In his critique, Dr. Mattes suggests that haloperidol-treated patients demonstrated a poor response. We would like to point out that 34% of the haloperidol-treated patients achieved the a priori-defined criterion of response (a 40% or greater improvement from baseline on the BPRS), which is consistent with response rates reported in the literature for haloperidol at similar doses (2). Among the patients assigned to haloperidol in our study, the mean improvement in CGI severity score was 0.7 (SD=1.1) points. Specifically, 50.6% demonstrated some degree of CGI improvement. Within the haloperidol treatment arm, change score analyses revealed statistically significant improvements from baseline on the BPRS, the Positive and Negative Syndrome Scale, and the CGI. Dr. Mattes' equation of improvement with terms from Hegarty et al.'s review is exceedingly optimistic. These terms regrettably characterize only a minority of chronic schizophrenic patients who receive conventional neuroleptic agents. Additional instruments used in this study further spoke to the improvement enjoyed by haloperidol-treated patients. For example, these patients had a mean improvement of 3.1 points on the Heinrich-Carpenter Quality of Life scale and averaged 8.7 BPRS-defined disability-free days at the end of the acute phase of the clinical study.

The importance of looking at specific symptom subclasses is acknowledged. We reported statistically significantly superior effects of olanzapine for negative symptoms and a trend for improvement in positive symptoms that also favored olanzapine. Given that 72% of the group was schizophrenic and chronic, these data are felt to be representative. In addition, there was no statistically significant difference in efficacy among diagnostic subgroups.

Dr. Mattes calls for studies that include neuroleptic-responsive patients. He is correct in noting that such patients are not commonly available for current studies, but this does not have so much to do with length of hospitalization as the reality that physicians would likely recommend for their patients an initial trial with an already approved agent. If there is a satisfactory response, it is unlikely that these patients would be candidates for a randomized trial with placebo or an investigational agent.

As to whether or not a novel antipsychotic agent is as effective for positive symptoms as standard neuroleptics in neuroleptic-responsive patients, we would refer Dr. Mattes and interested readers to the Beasley et al. report on olanzapine versus placebo and haloperidol (3). This study demonstrated that patients assigned to haloperidol had a statistically superior improvement in positive symptoms over placebo. Olanzapine (10 mg [SD=2.5] and 15 mg [SD=2.5]) also achieved significantly superior results in BPRS symptom improvement over placebo, and both olanzapine doses were numerically comparable to the haloperidol dose (10–20 mg).

Finally, there is a lack of consensus on the definition of what is a typical neuroleptic-responsive schizophrenic patient. The specific safety and efficacy profile of olanzapine in such a population would depend on the definition itself. However, it is noteworthy that among the 1,996 patients were 83 first-episode schizophrenic patients (nearly one-third were neuroleptic naive). As previously reported (4), significantly more olanzapine-treated than haloperidol-treated first-episode patients completed the trial (73% versus 38%). Significantly more haloperidol patients discontinued treatment because of adverse events (17% versus 2%). Olanzapine-treated first-episode patients also had significantly greater improvement in total and negative symptom scores on the BPRS and total and positive symptom scores on the Positive and Negative Syndrome Scale. They were also comparable but numerically superior to haloperidol-treated patients on the BPRS positive symptom subscale. Overall, olanzapine-treated first-episode patients had a significantly higher response rate (67%) than their haloperidol-treated counterparts (29%). In contrast to Dr. Mattes' supposition, this first-episode group of olanzapine-treated patients actually had a more robust treatment response than their chronic, multiepisode counterparts. We look to additional studies in such populations as described by Dr. Mattes to further corroborate these observations.