Bupropion-SR, Sertraline, or Venlafaxine-XR after Failure of SSRIs for Depression

Participants

Adult outpatients with a primary clinical diagnosis of nonpsychotic major depressive disorder,²⁴ as confirmed by a checklist completed by the clinical research coordinators, were enrolled at primary and psychiatric public and private practice settings between July 2001 and August 2004. Broad inclusion and minimal exclusion criteria⁴ were used to maximize the generalizability of the findings.^2,4

All study participants provided written informed consent at enrollment into the initial treatment with citalopram (level 1) and into all secondary treatments (level 2). All participants received citalopram as the initial treatment.²³ Participants who were eligible for second-step treatments either had not had a remission or could not tolerate citalopram. A lack of remission was defined as a score of more than 5 on the 16-item Quick Inventory of Depressive Symptomatology — Clinician Rated (QIDS-C-16)^25,26 at the last level 1 visit; scores can range from 0 to 27, with higher scores indicating greater severity of symptoms.

Figure 1. Figure 1. Overview of Study Design.

BUP-SR denotes sustained-release bupropion, SERT sertraline, VEN-XR extended-release venlafaxine, CT cognitive therapy, CIT citalopram, and BUS buspirone.

Figure 1 provides an overview of the study at levels 1 and 2. The study used an equipoise stratified, randomized design⁵ in which patients were strongly encouraged to accept all seven potential second-step treatments — the four switch options (including cognitive therapy) and three augmented treatments. However, to mimic practice, patients could opt to exclude certain level 2 treatment options. They could elect to exclude all switch options or all augmentation options, they could accept or decline cognitive therapy within either the switch or augmentation option, or they could accept only cognitive therapy (both as a switch and an augmentation treatment). In this design, the various acceptable treatments chosen by patients are called acceptability strata. Only the treatments for which patients accepted randomization were compared.⁵

By design, only the treatment groups that included a sufficient number of patients were analyzed to address the primary study aims. For level 2, the primary comparisons were among the three medication switches (sustained-release bupropion, sertraline, and extended-release venlafaxine), which are discussed in this report, and among the two medication augmentations (reported by Trivedi et al. elsewhere in this issue of the Journal ²⁷). Randomization was conducted in a 1:1:1 ratio separately within each regional center and according to which treatments the participants accepted.

The institutional review boards at the national coordinating center, the data coordinating center, and regional centers and at relevant clinical sites and the data safety and monitoring board of the National Institute of Mental Health approved and monitored the protocol.

All authors were involved in study implementation, supervision, data review, and manuscript development. Bristol-Myers Squibb, Forest Pharmaceuticals, GlaxoSmithKline, King Pharmaceuticals, Organon, Pfizer, and Wyeth-Ayerst Laboratories provided medications at no cost for this trial but otherwise had no role in the design, conduct, data analysis, or drafting of the manuscript reporting the results.

Protocol Treatment

To mimic clinical practice, enhance safety, ensure a vigorous dosing regimen, and maximize generalizability, all participants and treating clinicians were aware of treatment assignments and doses. A clinical treatment manual (available at www.star-d.org) with an emphasis on measurement-based care²³ recommended starting doses and dose changes on the basis of scores for the clinician-rated (QIDS-C-16) version of the 16-item Quick Inventory of Depressive Symptomatology (a scale ranging from 0 to 27, with higher scores indicating a greater severity of symptoms)^25,26 and the Frequency, Intensity, and Burden of Side Effects Rating (FIBSER) (a rating scale that collects information on the frequency, intensity, and burden of side effects, each on a 7-point scale with higher ratings indicating greater severity) obtained at each treatment.⁴ (A copy of the manual can be found in the Supplementary Appendix, available with the full text of this article at www.nejm.org.) In addition, didactic instruction, support by the clinical research coordinators, and a centralized monitoring system²⁸ with feedback ensured timely increases in doses when an inadequate reduction in symptoms occurred in the context of acceptable side effects.²³ Treatment was aimed at symptom remission, which was defined as a QIDS-C-16 score of 5 or less at the clinic visit.

At the initiation of the switch in medications, citalopram was discontinued without a tapering or washout period. The recommended daily dose of sustained-release bupropion was 150 mg for seven days, 200 mg from day 8 to 27, 300 mg from day 28 to 41, and 400 mg from day 42 onward. Sertraline was started at a daily dose of 50 mg and increased to 100 mg at day 14, to 150 mg at day 28, and to 200 mg at day 63. For extended-release venlafaxine, the starting daily dose of 37.5 mg for 7 days was increased to 75 mg from day 8 to 14, to 150 mg from day 15 to 27, to 225 mg from day 28 to 41, to 300 mg from day 42 to 62, and to 375 mg from day 63 onward. Dosing recommendations were flexible and were based on clinical judgment as informed by the FIBSER and the QIDS-C-16 scores at each treatment visit.

Concomitant Treatments

Stimulants, anticonvulsants, antipsychotic agents, mood stabilizers, nonprotocol antidepressant medications, and potential antidepressant augmenting agents (e.g., buspirone) were proscribed. Otherwise, any concomitant medication was allowed as necessary to manage concurrent general medical conditions or the side effects of protocol antidepressants (e.g., sexual dysfunction), as were anxiolytic agents (with the exception of alprazolam) and sedative hypnotic agents (including trazodone, at a dose of 200 mg or less at bedtime, for sleep).

Clinical Measurements

Information collected at level 1 baseline to describe the study cohort included scores for the Cumulative Illness Rating Scale (CIRS)²⁹ to assess patients' general medical conditions, the Psychiatric Diagnostic Screening Questionnaire^30,31 to identify coexisting psychiatric disorders, and the 30-item Inventory of Depressive Symptomatology — Clinician Rated^26,32 (IDS-C30) to assess the severity of depression and certain symptom features.³³ Assessments of overall function — including the Short-Form Health Survey (SF-12), Work Productivity and Activity Impairment Questionnaire, Work and Social Adjustment Scale (WSAS) — and satisfaction (Quality of Life Enjoyment and Satisfaction Questionnaire [QLESQ]) were collected by an automated interactive-voice-response telephone system.³⁴

The primary outcome (i.e., symptom remission) was defined as a total score of 7 or less on the 17-item Hamilton Depression Rating Scale³⁵ (HRSD-17), which was obtained in telephone-based, structured interviews (in either English or Spanish) conducted by independent research-outcome assessors who were unaware of treatment-group assignment within five days after entry and exit from the study. The secondary outcomes included results on the Quick Inventory of Depressive Symptomatology — Self-Report (QIDS-SR-16)^2,4,25,26 and the FIBSER⁴ obtained at each treatment visit. QIDS-SR-16 remission was defined as a total score at study exit of 5 or less, and response was defined as a reduction of 50 percent or more (level 2 baseline to exit) on the QIDS-SR-16.

Statistical Analysis

Summary statistics are presented as means (±SD) for continuous variables and percentages for discrete variables. Parametric and nonparametric analysis-of-variance methods and chi-square tests were used to compare the baseline clinical and demographic characteristics, treatment features, and rates of side effects and serious adverse events among treatment groups.

All analyses were conducted according to the intention to treat.³⁶ Logistic-regression models were used to determine whether there was an independent treatment effect on remission and response rates, adjusting for the effect of the regional center and acceptability stratum. Thirteen design variables were included in the logistic-regression models to estimate the effect of differences among the 14 regional centers. For this report, eight treatment-acceptability strata for medication switches were possible. These strata were collapsed into two strata (“medication switch only” and “other”) because of small numbers of patients in several strata. Thus, one design variable was included in the logistic models to control for the effect of the acceptability. Times to first remission (a score of 5 or less on the QIDS-SR-16) and first response (a reduction in the baseline score of 50 percent or more on the QIDS-SR-16) were defined as the first observed point with the use of data from clinic visits. Log-rank tests were used to compare the cumulative proportion with rates of remission and response among the three treatment groups. Exploratory logistic-regression analyses were used to determine whether there was a differential effect of treatment among patients who could not tolerate treatment in level 1.

At the end of this study, patients with missing HRSD-17 scores were assumed not to have had a remission (as originally defined).⁴ To determine whether this assumption affected study results, we conducted sensitivity analyses. The analyses were replicated with the use of two additional methods of imputation, a multiple imputation method³⁷ and imputed values generated from item-response theory, which mapped total scores on the QIDS-SR-16 to corresponding values on the HRSD-17.³⁸ Consistent findings indicated that the results were not affected by this approach to missing data.

Patients

Figure 1 shows how the study groups in this trial (and the study by Trivedi et al.²⁷) were developed on the basis of the various acceptability strata. Only 21 of 1439 patients (1.5 percent) accepted random assignment to any of the seven level 2 treatments. Furthermore, only 369 of 1439 patients (25.6 percent) included cognitive therapy among any of the acceptable treatments.

Of the 727 participants, 239 received sustained-release bupropion, 238 received sertraline, and 250 received extended-release venlafaxine. The majority of these participants were drawn from two groups who were defined by the treatments that they found acceptable (Figure 1), mainly from the 583 patients who accepted only the three medication switch treatments and the 104 patients who accepted only the four switch treatments (i.e., the three medication switches plus cognitive therapy alone).

Demographic and Clinical Characteristics

Table 1. Table 1. Clinical and Demographic Characteristics of Patients.

Participants had index episodes of depression with low function that were defined as recurrent (75.7 percent of the patients), early onset (37.2 percent), or chronic (27.0 percent) (Table 1). The characteristics of the patients were similar to those of the patients who entered level 1 and who could be evaluated.²³ For example, the mean (±SD) scores for the HRSD-17 were 21.8±5.2 at entry into level 1 and 18.9±7.3 at entry into this study. Of the 727 enrollees, 407 patients could not tolerate citalopram (56.0 percent), as defined by the discontinuation of level 1 treatment before four weeks for any reason or after four weeks because of intolerable side effects. The results in the three medication groups were similar. The number of concurrent psychiatric disorders (data not shown) did not differ significantly among the three treatment groups.

Treatment Features

Table 2. Table 2. Characteristics of Treatment.

Table 2 shows the course of treatment with each medication. All three medications were administered in adequate doses for substantial periods, though only 32.8 percent of patients who received extended-release venlafaxine took more than 225 mg per day.

Symptom Outcomes

Table 3. Table 3. Treatment Outcomes, Side Effects, and Serious Adverse Events.

Remission rates, on the basis of the results on the HRSD-17, did not differ significantly among treatment groups (χ²=3.649 with 2 df, P=0.16). Of 239 patients who received sustained-release bupropion, 51 had a remission (21.3 percent), as did 42 of 238 patients who received sertraline (17.6 percent) and 62 of 250 patients who received extended-release venlafaxine (24.8 percent). The treatments did not differ significantly with respect to the QIDS-SR-16 response rates, remission rates, or percent reductions in QIDS-SR-16 scores (Table 3).

Figure 2. Figure 2. Weeks of Treatment until Remission, According to Drug.

The treatments also did not differ significantly with respect to either time to remission (log rank χ²=0.38, P=0.93) (Figure 2) or time to response (log rank χ²=0.65, P=0.72) on the basis of results on the QIDS-SR-16. Among the patients who had a remission, the mean time to remission according to results on the QIDS-SR-16 was 5.4±4.5 weeks (median, 4.0) for those given sustained-release bupropion, 6.2±5.0 weeks (median, 4.9) for those given sertraline, and 5.5±4.7 weeks (median, 4.2) for those given extended-release venlafaxine. Similarly, among patients with a response, according to results on the QIDS-SR-16, the mean time to a response was 5.5±3.5 weeks (median, 4.0) for those given sustained-release bupropion, 6.6±4.3 weeks (median, 5.9) for those given sertraline, and 7.0±4.3 weeks (median, 6.0) for those given extended-release venlafaxine.

Tolerability and Adverse Events

Side effects and serious adverse events were clinically similar among the treatment groups (Table 3). The treatments did not differ significantly in the overall burden of side effects or in the proportion of patients with any serious psychiatric adverse event, though there was a difference in the distribution of the frequency of side effects. Four patients were hospitalized for suicidal ideation or attempted suicide, but none committed suicide during the trial.

In the context of the equipoise stratified, randomized design used in this trial, which gave patients choices in their treatment regimen, most patients opted to have their medication either switched or augmented. Few patients chose to do both, thus preventing a definitive comparison of strategies involving augmentation with those involving a switch medication. Our results suggest that such a choice (i.e., both augmentation and a switch medication) is relatively uncommon in practice when patients are provided options.

With regard to the commonly accepted practice of switching medications, approximately one in four depressed patients had a remission of symptoms with sustained-release bupropion, sertraline, or extended-release venlafaxine after either not having had a remission with or being unable to tolerate citalopram therapy. Remission rates did not differ significantly among the three medication groups nor did QIDS-SR-16 response rates, times to QIDS-SR-16 response, change in QIDS-SR-16 scores from baseline to the end of the study, serious adverse events, or measures of tolerability. Remission rates were slightly higher according to results on the QIDS-SR-16 than to results on the HRSD-17, because patients who did not undergo HRSD-17 evaluation at the end of the study were declared a priori not to have had a remission. Of the 209 patients with missing HRSD-17 scores, 26 had a remission (12.4 percent) on the basis of results on the QIDS-SR-16. On the basis of results on the HRSD-17, no differential treatment effect in regard to remission was found between patients who could not tolerate citalopram and those who could tolerate this agent. These findings are generalizable to most adult outpatients with a nonpsychotic major depressive disorder who are treated in primary or specialty care settings.

Rates of response and remission with each switch medication were lower than rates reported in open-label case series of switch medications after failure of treatment with an SSRI. Such series typically involved patients who were not chronically depressed, had few coexisting medical or psychiatric illnesses, and were treated in research clinics.¹⁰ On the basis of results on the HRSD-17, remission rates were also slightly lower than the overall rate of 30 percent with medication augmentation as the second treatment step (as described by Trivedi et al.²⁷). However, these two trial groups of the STAR*D study involved largely distinct groups of patients who had different outcomes with citalopram treatment. Patients in our study had greater rates of intolerance to and somewhat less benefit from citalopram. Although the remission rates in our trial are clinically meaningful, the relatively low rates were probably not due to inadequate doses of medication or to inadequate durations of treatment, given the mean doses at the end of treatment and durations of treatment for each agent. However, the dose of extended-release venlafaxine was less likely to approach the protocol-recommended maximum of 375 mg per day than was the dose of either of the other two drugs.

These findings have important practical implications. Contrary to the belief that intolerance of one SSRI predicts intolerance of another SSRI, sertraline was tolerated as well as sustained-release bupropion, even though 56.0 percent of patients in this trial could not tolerate citalopram. Thus, intolerance to or the lack of efficacy of one SSRI seems not to imply intolerance or lack of efficacy of another SSRI. These results indicate that both within-class and out-of-class medication switches are reasonable choices.

As for the dual-action agent extended-release venlafaxine, post hoc pooled analyses^39,40 have suggested slightly higher remission rates with venlafaxine than with SSRIs when used as first-step treatment. No studies, to our knowledge, have compared venlafaxine with other potentially active medications at the second treatment step. In this study, higher remission rates were not achieved with extended-release venlafaxine than with the more selective agents among patients who could not tolerate or who did not have a remission with citalopram therapy.

Important limitations to this comparison of three switch medications include the lack of placebo control and unblinded delivery of treatment, though assessment of the primary outcome (according to results on the HRSD-17) was done in a blinded fashion, and the QIDS-SR-16 and HRSD-17 ratings were in agreement. A placebo-controlled study is not needed to discern whether these three switch treatments differ, but without a placebo group, we cannot be certain that any of the treatments was specifically effective (i.e., the results were due to the pharmacologic effects of the medication). On the other hand, switching to a placebo after an initial failed treatment would have aroused concern about ethics,⁴¹ might have limited generalizability (if more severely or chronically ill patients had declined to enroll), or might have led to less vigorous administration of drugs, given the high prevalence of coexisting medical conditions among the patients. In addition, since few patients opted for both augmentation of their medication and switching to another drug, we cannot definitively compare these two strategies. Finally, we cannot determine the basis for the choices exercised by patients, since we did not ask them why they had made such choices.

Among patients who cannot tolerate or who do not have a remission in response to an initial SSRI, approximately one in four patients had a remission on switching to sustained-release bupropion, sertraline, or extended-release venlafaxine; these three drugs had similar efficacy and tolerability. These findings highlight the need for more broadly effective antidepressant treatments.

The following investigators participated in the STAR*D trial: National Coordinating Center — A.J. Rush, M. Trivedi, D. Warden, M.M. Biggs, K. Shores-Wilson, D. Stegman, M. Kashner; Data Coordinating Center — S. Wisniewski, G.K. Balasubramani, J. Luther, H. Eng; Regional Centers: University of Alabama, Birmingham, and Tuscaloosa Veterans Affairs Medical Center, Tuscaloosa, Ala. — L. Davis, K. Rice, A. Berry, P. Johnson, S. Ambrose, M. Jewell, B. Thomas, E. Waldrop, T. Allen, E. St. John, R. Williams; University of California, Los Angeles — A. Leuchter, I. Lesser, I. Cook, M. Epstein, S. Rosenberg, S. Zeim, L. Sulkowski, J. Iribarren, R. Armstrong, A. Rosales, M. Abrams; University of California, San Diego — S. Zisook, K. Harless, C. Gonzalez, M. Smith, C. Lawrence, J. Palica, M. Rohrs, M. Capous-Desyllas, K. Ganadjian; Northwestern University Medical School, Chicago — W. McKinney, W. Gilmer, C. Kelley, C. Cooler, A. Bauer, J. Fleck, C. Endick: Psychiatric Research Institute, University of Kansas, Wichita — S. Preskorn, D. Hilger, A. Klick-Davis, R. Lusk, J. Elmore, D. Soetaert; Massachusetts General Hospital, Boston — J. Alpert, M. Fava, A. Nierenberg, A. Farabaugh, T. Petersen, W. Merens, P. Cassano, N. Craven, H. Yang, M. Candrian, R. Fraguas; University of Michigan, Ann Arbor — E. Young, S. Marcus, J. Greden, H. Briggs, K. Bullard, A. Kennedy, A. Benway, E. Rickard; New York State Psychiatric Institute and Columbia College of Physicians and Surgeons, New York — F.M. Quitkin, P. McGrath, J.W. Stewart, H. Sackeim, K. Tate-Brown, S. Rees, C. Smith, A. Couraud, J. Lavelle, K. Broderick; University of North Carolina, Chapel Hill — R. Golden, B. Gaynes, J. DeVeaugh-Geiss, A. Ford, S. Barnett, B. Pearson; Laureate Healthcare System, Tulsa, Okla. — J. Mitchell, W. Yates, J. Kuehnert, L. Jernigan, B. Williams, J. Hilton; University of Pittsburgh Medical Center, Pittsburgh — M. Thase, R.H. Howland, E. Friedman, J. Callan, S. Berman, L. Shutt, C. Spotts; Vanderbilt University Medical Center, Nashville — S. Hollon, R. Shelton, M. Lovett, T. Crutcher, T. Patton, J. Hart, R.M. Harris-Turner, D. Lilly, B. Sirles, S. Hicks, N. Harris, S. Addington; University of Texas Southwestern Medical Center, Dallas — M. Husain, M. Downing, D. Stegman, E. Shellhorn, B. O'Neal, D. Turner, L. MacLeod, M. Henson, T. Hawley, S. Gardner; Virginia Commonwealth University, Richmond — S. Kornstein, R. Schneider, S. Belyea, B. Perry, K. Schmitt, T. Goff, K. Lamoree, M. Britton, C. Glassman.