With CATIE, context is key,
when sorting out Phases 1, 1A, 1B, 2E, and 2T
by Leslie Citrome, MD, MPH
Dr. Citrome is from the New York University School of Medicine, Department of Psychiatry, and the Nathan S. Kline Institute for Psychiatric Research, Orangeburg, New York.
DISCLOSURE: Dr. Citrome is a consultant for, has received honoraria from, or has conducted clinical research supported by the following: Abbott Laboratories, AstraZeneca Pharmaceuticals, Avanir Pharmaceuticals, Barr Laboratories, Bristol-Myers Squibb, Eli Lilly and Company, Forest Research Institute, GlaxoSmithKline, Janssen Pharmaceuticals, Jazz Pharmaceuticals, Pfizer Inc., and Vanda Pharmaceuticals.
Abstract
The NIMH-funded Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) for schizophrenia enrolled close to 1500 patients and is the largest randomized clinical trial to date that compares several of the second-generation antipsychotics on overall effectiveness. This article reviews the use of the evidence-based medicine tool “number needed to treat” to interpret what the CATIE results mean when choosing among the different antipsychotics available. Depending on the phase of CATIE, different antipsychotics had different rankings for overall effectiveness. CATIE can be viewed as a switch study, and depending on the circumstances or reason for the switch and the medication the patient was switched from, different outcomes were seen for the antipsychotics tested. Olanzapine had advantages in terms of all-cause discontinuation and efficacy, particularly in Phase 1. Quetiapine (and olanzapine) had advantages in terms of all-cause discontinuation in Phase 1B where patients had failed perphenazine. Clozapine was superior to risperidone and quetiapine for patients who discontinued a second-generation antipsychotic in Phase 1 (or 1B) because of poor efficacy. Risperidone had advantages in terms of overall tolerability in Phase 1, 2E, and 2T. Ziprasidone had the most benign metabolic profile, and in phase 2T was associated with a higher likelihood of weight loss for patients who gained greater than seven percent of their initial body weight in Phase 1. Treating clinicians need access to all of these medications in order to optimize treatment for the individual patient.
Key Words: schizophrenia, antipsychotics, effectiveness, evidence-based medicine, number needed to treat
Introduction
A current therapeutic controversy in the treatment of schizophrenia is the relative merit of using different antipsychotic medications. Recently reported are the results of all the randomized phases of the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) for schizophrenia, a NIMH-funded project that cost over 40 million dollars.[1–4] A very large amount of clinical trial data is now available for the practitioner to sort out and apply to his or her own patients. This is not easy to do, given the sophistication of the study and the resultant publications. However, by using the tools of evidence-based medicine (EBM),[5,6] the front-line clinician can begin to interpret the CATIE results. The concept of number needed to treat (NNT) in interpreting and applying data can be very helpful in this instance.[7] Defined as “the number of patients who must receive an intervention of therapy during a specific period of time to prevent one adverse outcome or produce one positive outcome,”[5] NNT is one of the essential tools of EBM that helps us gauge effect size or clinical significance.[8,9] It is different from knowing whether or not something is statistically significant. NNT allows us to place a number on how often we can expect to see a difference between two interventions. If a difference in therapeutic outcome is seen once in every five patients being treated with one intervention versus another (an NNT of 5), it will likely influence our day-to-day practice of medicine. However, if we see a therapeutic difference once every 100 patients (an NNT of 100), the difference between two treatments is not of great concern under most circumstances (there are exceptions, such as in the assessment of immunization against a rare but very dangerous illness). To appropriately interpret a NNT, statistical significance can be examined by inspecting its confidence interval.10 This article will review the different phases of CATIE (Table 1) and see how the context of each phase impacts on results, and more importantly, informs us how to apply this information to the treatment of our own individual patients. See Sidebar for instructions on how to calculate NNT.
Where you end depends on where you start
In Phase 1 of CATIE, 23 percent of the patients randomly assigned to olanzapine already were receiving olanzapine monotherapy at the time of study entry and 18 percent of patients randomly assigned to risperidone were already receiving risperidone monotherapy (Table 2). This was not unexpected as both olanzapine and risperidone were the two most commonly used first-line, second-generation antipsychotics when CATIE was conducted. A post hoc analysis11 reported that patients randomly assigned to olanzapine and risperidone who were continuing with their baseline medication had significantly longer times until discontinuation than did those assigned to switch antipsychotics. This information tells us that switching antipsychotics is not necessarily a benign process. However, we do not know what exactly motivated the patients to participate in CATIE. We do not know with any precision how efficacious or tolerable their medication regimens were prior to being randomized.
Before applying Phase 1 results to your patients, additional questions to ask are whether or not the demographics of the patients enrolled (Table 2) are similar to that of the ambulatory patients you treat. Mean age was approximately 40 years (CATIE was not a study of patients in their adolescent or early adult years; first-episode patients were explicitly excluded) and 40 percent were married or were previously married.[1] Although concomitant medication and substance abuse were not exclusions, history of non-response to olanzapine, quetiapine, risperidone, ziprasidone, or perphenazine, or history of receiving clozapine excluded patients from participating in CATIE.1 Moreover, patients with schizoaffective disorder were not able to enroll in the study. The issue of substance abuse is especially important. Approximately 60 percent of the participants in CATIE were found to use substances, including 37 percent with current substance use disorders.12 Users, compared with nonusers, were more likely to be male, African-American, have lower educational attainment, have a recent period of homelessness, report more childhood conduct problems, have a history of major depression, have lower negative symptom and higher positive symptom scores on the Positive and Negative Syndrome Scale (PANSS), and have a recent illness exacerbation. Although the primary outcomes for substance users who participated in CATIE have not yet been published, it would seem that their prognosis would not be as good as for the non-users.
Overall, NNT for all-cause discontinuation in Phase 1 for the five tested antipsychotics can be presented in a 5 by 5 table (Table 3), together with the respective 95-percent confidence interval. A negative number indicates a disadvantage for the primary antipsychotic versus the comparator. When we compare an antipsychotic versus itself, the NNT to see a difference is infinite (there is no difference). These NNT tables are similar in concept to mileage charts found on maps where distances between different cities are efficiently displayed. For example, the NNT for all-cause discontinuation for risperidone versus quetiapine is 13, with a 95-percent confidence interval of 7 to 54. This means for every 13 patients randomized to risperidone instead of quetiapine in Phase 1, one additional patient will have completed the entire 18 months on his or her initially randomized medication. The 95-percent confidence interval is considered “wide;” with 95-percent certainty we can say that the actual difference between risperidone and quetiapine on all-cause discontinuation is between a low of every seven patients to a high of every 54 patients. When comparing quetiapine with perphenazine, the NNT is -15, meaning a disadvantage for quetiapine seen every 15 patients randomized to quetiapine versus perphenazine in terms of all-cause discontinuation. Here too the confidence interval is wide, from -8 to -336. The NNT tables are highly dependent on the phase of CATIE being scrutinized and the results can differ in a remarkable way. This will be demonstrated next.
What about patients with tardive dyskinesia?
By design, patients with pre-existing tardive dyskinesia could not be randomized to perphenazine, a first-generation antipsychotic, and were enrolled in Phase 1A where they were randomized to olanzapine, risperidone, quetiapine, or ziprasidone. Descriptions of baseline differences between those with tardive dyskinesia and those without have been published.[13] Probable tardive dyskinesia was identified in 212 patients, whereas 1,098 patients had no history or current evidence of tardive dyskinesia. Patients with tardive dyskinesia were older (mean age 47.2 vs. 38.9 years), had more years since first prescribed an antipsychotic (21.5 vs. 12.8 years), and were more likely to have been receiving a first-generation antipsychotic (28% vs. 14%). Patients with tardive dyskinesia had higher baseline PANSS ratings (78.2 vs. 75.1). Differences in outcome among the different second-generation antipsychotics may be different for those with tardive dyskinesia than those without, but have not yet been published.
CATIE is a switch study!
Relatively underappreciated is that CATIE provides information as to what happens when a switch occurs from one randomized medication to another. The design allowed for the evaluation of effectiveness (all-cause discontinuation) for patients who prospectively failed their initially randomized antipsychotic for reasons such as poor tolerability or lack of adequate efficacy before the 18-month trial period was over. Three papers have been published from these switches out of Phase 1: Failure of the prior second-generation antipsychotic presumably because of poor tolerability (Phase 2T),[2] failure of the prior second-generation antipsychotic presumably because of poor efficacy (Phase 2E),[3] and failure of perphenazine for any reason (Phase 1B).[4] Phase 2T involved randomized to double-blind ziprasidone, olanzapine, risperidone, or quetiapine (but not to the same antipsychotic they may have received in Phase 1). Phase 2E involved randomization to open-label clozapine or to double-blind olanzapine, risperidone, or quetiapine (but again, not to the same antipsychotic they may have received in Phase 1). Phase 1B was for the special circumstance that the patient was randomized to perphenazine in Phase 1 but discontinued before 18 months. Before these patients could enter Phase 2E or 2T, they were required to participate in Phase 1B where they were randomized to double-blind olanzapine, quetiapine, or risperidone. Patients who prematurely exited Phase 1B were then offered the opportunity to enroll in Phase 2E or 2T. Table 1 lists the randomized phases of CATIE, all of which involved assignment to double-blinded medication, except for clozapine in Phase 2E.
Failures Because of Tolerability
Phase 2T was designed to test what advantages might there be for switching to ziprasidone for patients who discontinued their Phase 1 second-generation antipsychotic because of poor tolerability.[2] In reality, because the clinicians and their patients could elect with Phase 2 pathway to enter (2T or 2E), approximately equal numbers of patients who discontinued Phase 1 because of poor efficacy entered Phase 2T as those who discontinued Phase 1 because of poor tolerability. This makes Phase 2T more difficult to interpret. For the subgroup of patients who entered Phase 2T after discontinuing their Phase 1 medication because of poor efficacy, the overall pattern of results resembles that for Phase 1. For the subgroup that entered Phase 2T because of poor tolerability with their Phase 1 second-generation antipsychotic, differences between the antipsychotics were difficult to ascertain; however, risperidone had the lowest numerical all-cause discontinuation rate.
Although patients randomized to ziprasidone did not have an advantage over the other groups in terms of all-cause discontinuation (NNT comparing ziprasidone with the other individual antipsychotics were not statistically significant), the benign metabolic profile for ziprasidone was clearly evident. There were fewer discontinuations due to weight or metabolic concerns (NNT for ziprasidone vs. olanzapine, risperidone and quetiapine were 12, 21, and 11, respectively, on this measure). Of particular interest is the scenario where patients gain greater than seven percent of their initial body weight. Of the 61 patients who did so in Phase 1, 42 percent of those randomized to ziprasidone in Phase 2T lost over seven percent of their body weight, compared with 20 percent for risperidone, seven percent for quetiapine, and zero percent for olanzapine. This translates to NNTs comparing ziprasidone to risperidone, quetiapine, and olanzapine of 5, 3, and 3, respectively.
Overall, NNT for all-cause discontinuation in Phase 2T for the four tested antipsychotics can be presented in a 4 by 4 table (Table 3). Of interest is that the only statistically significant NNT results were with comparing quetiapine with olanzapine or risperidone, showing a disadvantage for quetiapine every six patients (vs. olanzapine) or every five patients (vs. risperidone). This comparison of quetiapine vs. risperidone in Phase 2T is the opposite of what was observed in another phase of CATIE (1B).
Failures Because of Efficacy
Phase 2E was designed to test what advantages might there be for switching to clozapine for patients who discontinued their Phase 1 second-generation antipsychotic because of poor efficacy.[3] Patients enrolled in Phase 2E had baseline PANSS total scores that were higher than patients who enrolled in phase 2T (87.6 vs. 77.0) and were more likely to be male (81% vs. 69%). Fewer patients randomized to olanzapine in Phase 1 entered Phase 2E compared to 2T (19% vs. 26%), but more patients who were randomized to quetiapine in Phase 1 entered Phase 2E compared to 2T (37% vs. 31%). Of importance is the fact that patients who entered phase 2E knew they could be randomized to clozapine, and once randomized knew if they were actually getting open-label clozapine versus double-blinded olanzapine, quetiapine, or risperidone. In contrast to Phase 2T, 86 percent of enrollees in Phase 2E had discontinued Phase 1 because of inadequate therapeutic effect, and only five percent had discontinued Phase 1 because of unacceptable side effects. This makes the Phase 2E group of patients somewhat more homogeneous in terms of clinical and demographic characteristics than the Phase 2T group, and the results are easier to interpret.
Overall, NNT for all-cause discontinuation in Phase 2E for the four tested antipsychotics can be presented in a 4 by 4 table (Table 3). Statistically significant differences are evident for the comparisons of clozapine with risperidone or quetiapine, with NNTs showing an advantage for clozapine every four or five patients, respectively, in terms of all-cause discontinuation. Although not statistically significant, the comparison of clozapine versus olanzapine shows an advantage every seven patients for patients receiving clozapine.
Failures on Perphenazine
Phase 1B was an interim phase for those patients who had been randomized to the first-generation antipsychotic perphenazine in Phase 1 but who discontinued for any reason.[4] Note that this did not include patients with tardive dyskinesia as they were excluded from possible randomization to perphenazine (these patients entered Phase 1A). Similar to the pattern in Phase 2T, 48 percent of enrollees in Phase 1B had discontinued Phase 1 because of inadequate therapeutic effect and 32 percent had discontinued Phase 1 because of unacceptable side effects. Thus the enrollees in Phase 1B represent patients who did not do well enough because of either efficacy or tolerability issues on an older agent with relatively “tight” dopamine D2 receptor binding (perphenazine). In this phase, olanzapine and quetiapine performed relatively well, whereas risperidone, a “tighter” D2 blocking agent than olanzapine or quetiapine, did not.
Overall, NNT for all-cause discontinuation in Phase 1B for the three tested antipsychotics can be presented in a 3 by 3 table (Table 3). In contrast to the Phase 1 and Phase 2T NNT results, in Phase 1B quetiapine demonstrated an advantage over risperidone in terms of all-cause discontinuation evident every four patients, with a relatively narrow confidence interval of 3 to 15.
What’s Next: Phase 3
Phase 3 results have not yet been published but will provide information on what happens after a third switch (or fourth switch for those who went through the intermediate Phase 1B). Phase 3 medications were all provided open-label and consisted of one or two antipsychotics chosen among the following: Any of the medications tested in Phase 1 or 2 (olanzapine, quetiapine, risperidone, ziprasidone, perphenazine, or clozapine), fluphenazine decanoate, or aripiprazole. The numbers assigned to each medication choice will likely be small and statistical power limited in terms of resolving differences between the antipsychotics. However, patterns may emerge in terms of what “pathways” might be helpful to consider, depending on which antipsychotic the patient has failed and the reason for that failure.
Conclusions
The calculation of NNT can help quantify differences between two interventions. How much importance should be placed on a NNT for your particular patient will depend on the context from which the NNT was calculated. The patients being compared in the different phases of CATIE can be contrasted on a number of clinical and demographic measures, including what they had been receiving in terms of prior antipsychotic medication. In the randomized phases of CATIE, advantages were evident for each of the second-generation antipsychotics tested, but all in a different context. Olanzapine had advantages in terms of all-cause discontinuation and efficacy, particularly in Phase 1. Quetiapine (and olanzapine) had advantages in terms of all-cause discontinuation in Phase 1B where patients had failed perphenazine. Clozapine was superior to risperidone and quetiapine for patients who discontinued a second-generation antipsychotic in Phase 1 (or 1B) because of poor efficacy. Risperidone had advantages in terms of overall tolerability in Phases 1, 2E, and 2T.
Ziprasidone had the most benign metabolic profile and in Phase 2T was associated with a higher likelihood of weight loss for patients who gained greater than seven percent of their initial body weight in Phase 1. Treating clinicians need access to all of these medications in order to optimize treatment for the individual patient.
References
1. Lieberman JA, Stroup TS, McEvoy JP, et al. Effectiveness of antipsychotic drugs in patients with chronic schizophrenia. N Engl J Med 2005;353(12):1209–23.
2. Stroup TS, Lieberman JA, McEvoy JP, et al. Effectiveness of olanzapine, quetiapine, risperidone, and ziprasidone in patients with chronic schizophrenia following discontinuation of a previous atypical antipsychotic. Am J Psychiatry 2006;163(4):611–22.
3. McEvoy JP, Lieberman JA, Stroup TS, et al. Effectiveness of clozapine versus olanzapine, quetiapine, and risperidone in patients with chronic schizophrenia who did not respond to prior atypical antipsychotic treatment. Am J Psychiatry 2006;163(4):600–10.
4. Stroup TS, Lieberman JA, McEvoy JP, et al. Effectiveness of olanzapine, quetiapine, and risperidone in patients with chronic schizophrenia after discontinuing perphenazine: A CATIE study. Am J Psychiatry 2007;164(3):415–27.
5. Guyatt G, Rennie D. Users Guides to the Medical Literature: Essentials of Evidence-Based Clinical Practice. Chicago, IL: AMA Press, 2002.
6. Straus SE, Richardson WS, Glasziou P, et al. Evidence-based Medicine: How to Practice and Teach EBM, Third Edition. Edinburgh, UK: Elsevier/Churchill Livingstone, 2005.
7. Citrome L, Stroup TS. Schizophrenia, Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) and number needed to treat: How can CATIE inform clinicians? Int J Clin Pract 2006;60(8):933–40.
8. Citrome L. Dissecting clinical trials with ‘number needed to treat.’ Curr Psychiatry 2007;6(3):66–71.
9. Citrome L. Show me the evidence: Using number needed to treat. South Med J 2007;100(9):881–4.
10. Citrome L. Can you interpret confidence intervals? It’s not that difficult. Curr Psychiatry 2007;6(8):77–82.
11. Essock SM, Covell NH, Davis SM, et al. Effectiveness of switching antipsychotic medications. Am J Psychiatry 2006;163(12):2032–3.
12. Swartz MS, Wagner HR, Swanson JW, et al. Substance use in persons with schizophrenia. Baseline prevalence and correlates from the NIMH CATIE study. J Nerv Ment Dis 2006;194(3):164–72.
13. Miller DD, McEvoy JP, Davis SM, et al. Clinical correlates of tardive dyskinesia in schizophrenia: Baseline data from the CATIE schizophrenia trial. Schizophr Res 2005;80(1):33–43.
