Drs. Nelson and Papakostas Reply

To the Editor: Dr. Carroll has raised several points about our meta-analysis of atypical antipsychotic agents when used as adjunctive agents in the treatment of depression (1). His primary point is that our meta-analysis did not adequately account for the risks associated with these drugs; hence, our evaluation of overall effectiveness is flawed.

His first point is that the risk of tardive dyskinesia was minimized. He cites a recent article (2), using olanzapine in the treatment of psychotic depression, in which the rate of probable tardive dyskinesia was 8.5%. Indeed, this rate would be reason for concern. In that study, there was no mention of exclusion of patients with pre-existing tardive dyskinesia. In addition, the authors indicated that because of the seriousness of the illness, the intended washout of prior antipsychotic agents was sometimes abandoned in order to proceed quickly with treatment. As a result, it is possible that there was an unmasking effect of tardive dyskinesia with the rapid withdrawal of the previous agent. Tarsy and Baldessarini (3) noted the potential influence of prior neuroleptic exposure on the vulnerability to tardive dyskinesia, which Dr. Carroll does not mention. The tardive dyskinesia rate in that study was based on an Abnormal Involuntary Movement Scale score of ≥2 on any item, without requiring persistence of the score 1 week later. (The Schooler–Kane criteria for tardive dyskinesia require a score of ≥3 on one Abnormal Involuntary Movement Scale item or ≥2 on two items. A score of 2 on one item would require the presence of symptoms 1 week later. [4]). However, Barnett Meyers, M.D., indicated that treatment was seldom discontinued in patients with these Abnormal Involuntary Movement Scale ratings, raising some questions about whether clinicians considered the rating clinically meaningful (personal communication with Barnett Meyers, M.D., October 7, 2009).

Compared with other studies, the aforementioned tardive dyskinesia rate appears to be an outlier. Prior reviews of tardive dyskinesia with atypical antipsychotic agents report rates of tardive dyskinesia averaging 0.8%, which is approximately sixfold less than those associated with typical antipsychotic agents (5). Published analyses of pooled tardive dyskinesia rates during randomized trials of patients receiving olanza­pine (N=1,192) (6) and risperidone (N=3,298) (7) yielded rates of 0.5% and 0.6%, respectively. These rates should be considered in relation to reported spontaneous rates of tardive dyskinesia in schizophrenia of 0.5% (8). It should be noted that large samples are required to provide reasonable estimates of rarely occurring event rates. A rate of 5 cases per 1,000 individuals in a sample of 1,000 has a fairly broad 95% confidence interval (CI) (95% CI=1.6–11.6 per 1,000). A sample of 10,000 individuals is needed to provide a more reasonable estimate (95% CI=3.7–6.6 per 1,000).

These tardive dyskinesia rates cited were observed in studies of schizophrenia and related psychotic disorders. It is possible that rates of drug-induced tardive dyskinesia are different in patients with nonpsychotic major depression. Yet, in a 12-month open-label study of aripiprazole in the treatment of nonpsychotic major depressive disorder (9), four cases of tardive dyskinesia were observed out of 1,002 case subjects. Adjusted for exposure, the annualized risk was 0.7%. For comparison, an extension study of schizophrenia patients, comparing patients receiving aripiprazole (N=786) and haloperidol (N=391) over a 12-month period, found annualized tardive dyskinesia rates of 0.45% for aripiprazole and 9.09% for haloperidol (10).

It should be emphasized that comparison of tardive dyskinesia rates among different studies is limited by the use of different tardive dyskinesia criteria that vary in sensitivity. In addition, long-term exposure in large numbers of patients is required to accurately estimate annualized rates that may be less than one case per 100 patients. Often, prior neuroleptic exposure, which may influence the development of tardive dyskinesia, is not taken into account (3). Nevertheless, to date, the data regarding tardive dyskinesia risk with atypical agents in the treatment of schizophrenia indicate lower rates than those for tardive dyskinesia risk with haloperidol. We have not found published rates of tardive dyskinesia with risperidone, olanzapine, or quetiapine in the treatment of major depressive disorder. In the aforementioned 12-month major depressive disorder extension trial, the rate for aripi­prazole was relatively similar to the rate for schizophrenia. As we suggested, it appears that the rate cited for olanzapine in the psychotic depression study is an outlier and may reflect other factors. Dr. Carroll calls attention to an important issue in mentioning that neuroleptic-naïve major depressive disorder patients are a high-risk population. We agree that the risk of tardive dyskinesia in neuroleptic-naïve major depressive disorder patients is not well studied. As part of standard clinical practice, clinicians should disclose the tardive dyskinesia risk (as well as other risks), when presenting the relative risks and benefits of adjunctive treatment, as well as provide alternative strategies for antidepressant nonresponders with major depressive disorder to enable patients to make informed decisions.

Dr. Carroll also questions the efficacy of atypical antipsychotic agents in the treatment of major depressive disorder. He states that "the trials of aripiprazole were unblinded by akathisia and other side effects that can introduce rater bias." While this is a possibility, it is not clearly demonstrated. He also notes that efficacy was lacking in men in the first study of aripiprazole cited in our meta-analysis (11). In the second study, there was no gender interaction with outcome (12). In the third study, the drug was efficacious in both men and women, and the magnitude of change was slightly greater in men (13). Perhaps more important, however, is that in our meta-analysis, our primary focus was on the overall efficacy of the 16 studies of the drug cited rather than subanalyses of individual studies.

Dr. Carroll suggests that atypical antipsychotic agents are not as effective as lithium augmentation. He cites a number needed to treat of four to five for lithium. As we noted, lithium augmentation has been studied in 10 placebo-controlled trials that have been previously reviewed (14). Nine of these studies were small, with less than a total of 36 patients. Perhaps the most important difference in the atypical antipsychotic trials is that only one of the lithium studies required that patients had to have failed at least 6 weeks of prior treatment (15). In that study, lithium was not more effective than placebo. In the largest, albeit uncontrolled, study of lithium augmentation (performed in the Sequenced Treatment Alternatives to Relieve Depression [STAR*D] study), lithium and thyroid hormone triiodothyronine (T3) augmentation were compared after two prior failed trials in 127 patients (16). While remission rates did not differ significantly in lithium- and T3-treated patients (15.9% versus 24.7%, respectively), significantly more patients taking lithium discontinued treatment because of intolerance (23.2% versus 9.6%, p=0.03). If the benefits are weighed against the risks, 11 of 69 patients receiving lithium remitted, while 16 were intolerant. This is not a favorable balance. In the pooled data from our meta-analysis, the remission rate for the atypical agents was 30.7%, and the discontinuation rate for adverse events was 9.1% (range: 4.4%–12.2%). Of course, these data are not from direct comparisons, and the patients in the atypical trials may have been less treatment resistant than those in the STAR*D study. Ultimately, only studies involving a direct comparison between atypical antipsychotic treatment and treatment with lithium can definitively inform clinicians and patients whether lithium augmentation is more effective, better tolerated, and safer than augmentation with an atypical antipsychotic.

Dr. Carroll suggests that the manufacturers of the atypical agents have "studiously avoided testing their products against lithium." It is not possible for us to know whether this was intentional. We do wonder if lithium is the current "treatment to beat." There are several augmentation and combination strategies that might offer advantages in terms of their safety and tolerability (17, 18). It is true that there are few studies comparing augmentation or combination strategies in resistant depression funded by pharmaceutical companies. There are also few such studies that are federally funded.

Finally, Dr. Carroll asserts that there is no evidence that atypical augmentation is effective in the long-term. He cites a failed relapse prevention trial of risperidone (19). That study was problematic given the very high initial nonresponse rate (89%) with a selective serotonin reuptake inhibitor, followed by a high remission rate (63%) with risperidone augmentation during open-label treatment. We agree that currently there are no controlled long-term trials for adjunctive atypical use in major depressive disorder that confirm efficacy. But then there are no controlled trials of any augmentation strategy that demonstrate efficacy in relapse prevention. This is an uncharted area, and we cannot assume that usual relapse prevention designs for monotherapy will necessarily be appropriate for patients who respond to augmentation or combination treatments.

In our consideration of the risks and benefits of the atypical agents, we stated the following in our article: "The atypical agents also are associated with a variety of relatively serious adverse effects, such as metabolic syndrome, extrapyramidal symptoms, and rare but serious symptoms such as tardive dyskinesia and neuroleptic malignant syndrome. As a consequence the risk benefit ratio appears to be different from that of several alternative treatments for major depressive disorder " (p. 989). Because of these risks, atypical agents are likely to be placed later in an algorithm, even though their efficacy appears well established. We stand by this statement.