Vol. 22, No. 4 / November 2010
Switch or augment? Lessons from STAR*D
Associate Director, Depression Clinical and Research Program, and Co-Director, Bipolar Clinic and Research Program, Massachusetts General Hospital, Boston, Massachusetts; Professor of Psychiatry, Harvard Medical School, Boston, Massachusetts
Dr Nierenberg has disclosed that he has received grants or research support from Bristol-Myers Squibb Company; Cederroth AB; Cyberonics, Inc; Eli Lilly and Company; Forest Laboratories, Inc; GlaxoSmithKline; Janssen Pharmaceutica; Lichtwer Pharma; NARSAD; NIMH; Pfizer Inc; Shire; The Stanley Foundation; and Wyeth-Ayerst Laboratories; and honoraria from Bristol-Myers Squibb Company; Cyberonics, Inc; Eli Lilly and Company; Forest Laboratories, Inc; GlaxoSmithKline; MGH Psychiatry Academy; Physicians Postgraduate Press, Inc; Shire; and Wyeth-Ayerst Laboratories. He is a retained consultant for Abbott Laboratories; AstraZeneca; Basilea Pharmaceutica, Ltd; BrainCells Inc; Bristol-Myers Squibb Company; Eli Lilly and Company; EPI-Q, Inc; Genaissance Pharmaceuticals, Inc; GlaxoSmithKline; Innapharma Inc; Janssen Pharmaceutica; Jazz Pharmaceuticals; Merck & Co, Inc; Novartis; Pfizer Inc; PGxHealth; Schering-Plough; Sepracor Inc; Shire; Somerset Pharmaceuticals, Inc; Takeda Pharmaceuticals North America, Inc; and Targacept, Inc; and a stock shareholder of Appliance Computing II, Inc (Mindsite) and BrainCells Inc.
According to the World Health Organization (WHO), depression affects approximately 121 million people worldwide. The WHO predicts that by 2020, depression will be a primary cause of disability globally, second only to ischemic heart disease.1,2 In the United States, the lifetime prevalence of depression is 15.7%.3
The treatment goal in major depressive disorder (MDD) is remission, defined as the virtual absence of depressive symptoms. Remission is associated with a better prognosis than treatment response, which is defined as a clinically meaningful reduction in symptoms.4 Despite the importance of achieving remission in patients with depression, few clinical studies of treatment-resistant MDD have used remission as a primary outcome of treatment efficacy. Studies that have used remission as a primary outcome often have lacked validity, have not been generalizable to patients most frequently seen in clinical practice, and have failed to compare management strategies.4,5 Therefore, treatment guidelines for management of patients with MDD following medication failure have been based on clinical consensus or open uncontrolled trials rather than on clear evidence.6 The Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial was designed to address the question of what is the most effective “treatment of next choice” in patients with MDD who fail to achieve remission on initial antidepressant therapy.7
The STAR*D study, which enrolled 4041 outpatients from 41 sites (18 primary care and 23 psychiatric care settings), was designed to allow for generalization to “real-world” patients, such as those routinely seen in clinical practice. The study included adults aged 18 to 75 years from primary and specialty psychiatric practices who had at least moderate depression and who had failed to attain remission with ≥1 adequate trial of treatment. Unlike participants in previous efficacy trials, more than one-fourth of patients in STAR*D had chronic depression (index episode >2 years) and three-fourths had recurrent depression.5 Benefits to patients enrolled in STAR*D included free medication, nominal reimbursement, and careful monitoring by clinicians. Approximately 38% of study participants were from primary care clinics, reflective of real-world patients with multiple comorbidities. Prior studies of MDD treatment strategies excluded such patients, making generalization difficult. However, STAR*D had broad inclusion and exclusion criteria, which allowed for a wider, more varied patient population, including those with substance abuse disorders and chronic or recurrent depression, as long as their clinicians considered antidepressant therapy appropriate.5,8 No advertising was used to enroll patients, and the evaluating and treating clinicians were practicing physicians rather than clinical trial researchers. Clinical research coordinators were trained and certified in the protocol and data collection to support the clinicians.8
STAR*D was designed to preserve the role that patients typically play in treatment decisions in the clinical setting. The trial design utilized equipoise-stratified randomization to maintain the significant influence of patient preference in selection of treatment after medication failure.8,9 Patients were asked 4 questions about the next step in their treatment: 1) whether or not they wished to discontinue citalopram and switch to another therapy; 2) whether or not they were willing to continue citalopram and have another treatment added; 3) if they decided to switch to another medication, would they be willing to have cognitive-behavioral therapy (CBT) added; and 4) if they decided to have another medication added, would they be willing to have CBT added as well. Participants could not select among the specific treatment options, and they were encouraged to accept both switch and augmentation possibilities at each stage of decision-making. This design structure allowed patient acceptability to be incorporated into the decision-making process, while preserving the power of randomization. Neither patients nor study clinicians were masked to treatments.8
The primary outcome of STAR*D was depressive symptom severity, as measured with the 17-item Hamilton Rating Scale for Depression (HAM-D) by independent telephone interviewers at entry and exit of each treatment level and at 3-month intervals during the 12-month follow-up. Treatment remission was defined as a HAM-D total score <8 at treatment level exit. The secondary outcome was based on remission defined by the Quick Inventory of Depressive Symptomatology—Self Report (QIDS-SR); a QIDS-SR score ≤5 was considered symptom remission.4,8,10
TREATMENT AND LEVELS
Participants enrolled in STAR*D (FIGURE 1) received treatment with the selective serotonin reuptake inhibitor (SSRI) citalopram (CIT) (Level 1). All STAR*D clinicians used a brief clinical rating system provided by the trial group to define adequate response to therapy. At any treatment level, patients who demonstrated adequate response, as defined by the clinician, entered a 12-month follow-up phase. Patients without adequate response to CIT at Level 1 were eligible to enter a series of randomized clinical trials. The first trial (Level 2) compared the effectiveness of 7 treatments, including 4 switch options (extended-release venlafaxine [VEN-XR], sertraline [SERT], sustained-release bupropion [BUP-SR], and CBT) and 3 augmentation options (CBT, BUP-SR, and buspirone [BUS]). STAR*D sought to evaluate both the comparative effectiveness of switching vs augmentation as well as the effectiveness of treatment options within and across these 2 strategies.
Patients who did not experience an adequate response to both CIT at Level 1 and CBT at Level 2—regardless of whether they chose switching or augmentation—entered Level 2A, which compared the effectiveness of 2 medication switch options, BUP-SR and VEN-XR. This design ensured that all participants who entered Level 3 had failed to respond to 2 antidepressant medications, which is one of the guideline-proposed criteria of treatment-resistant MDD. Level 3 compared the relative effectiveness of 2 switch options, mirtazapine (MRT) and nortriptyline (NTP), and 2 augmentation options, lithium (Li) and the thyroid hormone triiodothyronine (T3). Patients who did not achieve adequate response at Level 3 entered Level 4, which compared the effectiveness of 2 switch options, tranylcypromine (TCP) and the combination of MRT and VEN XR. Participants who failed to achieve adequate response at Level 4 exited the study and were managed as clinically indicated at the discretion of their treating physicians.10
A sample of 2876 patients was available for analysis.4 Researchers expected that multiple patients would accept randomization to all 7 treatment options; however, only 1% of patients did so. It could be inferred that patients were willing to accept a partial response to treatment rather than randomization to another treatment as long as they were not experiencing significant adverse effects. Approximately 50% of patients accepted medication augmentation and 57% accepted a medication switch, although, surprisingly, most patients agreed to switch or augment, but not both. The study thus was not adequately powered to compare outcomes for switching vs augmentation. Patients who accepted a switch after CIT monotherapy had experienced greater intolerance and less improvement with CIT than patients who chose augmentation, indicating that patients have preferences about switching vs augmentation as a second step in therapy.5
Reprinted with permission from Warden et al. Curr Psychiatry Rep. 2007;9:449-459.
BUP-SR, bupropion sustained-release; BUS, buspirone; CIT, citalopram; Li, lithium; MIRT, mirtazapine; NTP, nortriptyline; SERT, sertraline; T3, triiodothyronine; TCP, tranylcypromine; VEN-XR, venlafaxine extended-release.
a Remission rates at treatment exit based on the 16-item Quick Inventory of Depressive Symptomatology–Self Report (QIDS-SR16).
At Level 4, TCP monotherapy and VEN+MRT combination therapy demonstrated similar effectiveness, although TCP was associated with higher discontinuation rates due to intolerance.17
For all medical treatments in STAR*D, the remission rate was approximately 33%, based on QIDS-SR scores, and the overall response rate, including those who remitted, was 47%. Diminishing remission rates were seen at increased levels of treatment resistance (FIGURE 2)5: the greater the number of acute treatments that patients tried and failed, the less likely they were to achieve remission. By the time patients reached Level 4, the remission rate was 16% with VEN+MRT and 14% with TCP.5
Relapse rates also increased at greater levels of acute treatment. For example, overall relapse rates rose from 40% at Level 1 to 55% at Level 2, and to 65% and 71% at Levels 3 and 4, respectively. Also, patients who achieved remission at time of entry into follow-up demonstrated lower relapse rates (34% to 50%) than patients who did not achieve remission by entry into follow-up (59% to 83%) (FIGURE 3).5
Reprinted with permission from Warden et al. Curr Psychiatry Rep. 2007;9:449-459.
a Relapse rate was calculated based on results for patients with ≥1 post-baseline interactive voice response call.
Patients who required more levels of acute treatment had greater general illness burdens, longer illness durations, longer index episodes of depression, more chronic and recurrent depression, and more anxiety features.5 A study by Fava et al found that time to remission was longer and the probability of achieving remission was lower in depressed patients who were anxious compared with nonanxious depressed patients (FIGURE 4).18
Reprinted with permission from Fava et al. Am J Psychiatry. 2008;165:342-351.
Log-rank statistic=41.7 (P<.0001).
Because of its broad inclusion-exclusion criteria and study enrollment method, most patients in STAR*D would be considered “difficult to treat.” In fact, the majority of patients enrolled in STAR*D would not have been eligible to participate in traditional efficacy trials. In a study by Wisniewski et al, data from STAR*D were used to evaluate whether Phase III clinical trials recruit outpatients with depression who would be representative of patients seen in clinical practice. The study found that those STAR*D patients who would have met the stringent inclusion and exclusion criteria of a traditional efficacy trial (efficacy sample) had better rates of remission and response than did the STAR*D patients who would not have met the trial criteria (nonefficacy sample) (TABLE). Only 25% of patients in the nonefficacy sample—the group that was more representative of real-world patients—achieved remission compared with 34% of patients in the efficacy sample. Response rates also were different for the 2 samples: 52% of patients in the efficacy sample vs 39% of patients in the nonefficacy group achieved response to therapy.19
QID S-SR16 remission
QID S-SR16 response
a Patients in the efficacy sample met typical phase III trial inclusion and exclusion criteria.
b Patients in the nonefficacy sample did not meet these criteria.
All STAR*D patients who achieved adequate response to any treatment were entered into the 12-month naturalistic follow-up study.8 Rush et al looked at relapse rates in patients who achieved remission prior to entry into follow-up and compared them with relapse rates of patients who did not achieve remission before follow-up. Investigators found that, for all treatment levels, the proportion of patients relapsing increased over 12 months of follow-up. Additionally, patients who did not achieve remission by entry into follow-up had a higher likelihood of relapse during the 12-month follow-up than did patients who achieved remission at the time of follow-up entry (FIGURE 5).20 These long-term results highlight the high rate of relapse in treatment-resistant MDD and underscore the need to aggressively treat MDD to achieve remission as early as possible. The results also emphasize the need to carefully monitor patients for relapse once remission or response has been achieved.5
Adapted from rush et al. Am J Psychiatry. 2006;163:1905-1917.
a Relapse defined as Quick Inventory of Depressive Symptomatology–Self Report (QIDS-IVR16) score ≥11.
Valuable lessons about MDD can be learned from STAR*D. With relatively low rates of remission and high rates of relapse, MDD is difficult to treat in the real-world setting. The coexistence of medical and psychiatric conditions with treatment-resistant MDD can complicate treatment decisions. STAR*D demonstrated that if a first treatment intervention does not result in remission or response, many effective treatment options are available for second-step treatment and beyond.
STAR*D emphasized that it takes time and persistence on the part of both the patient and clinician to achieve response or remission in MDD. STAR*D outcomes also emphasized that close monitoring of patients is crucial, even after remission or response is achieved. Although some patients with MDD may respond to antidepressant therapy in as few as 3 to 6 weeks, many patients can take up to 12 weeks to experience the full benefits of treatment. In the STAR*D trial, one-half of those who ultimately remitted did so after 6 weeks. This suggests that treatment with an antidepressant for only 6 weeks may be inadequate for many patients with MDD. Patients who demonstrate a modest improvement after 6 weeks of treatment might achieve remission by 12 weeks.21 Thus it is important for patients to work with their clinicians to adjust dosing and treatment strategy as needed.7 Additionally, practicing clinicians often use patient and physician subjective judgments to determine efficacy and tolerability. STAR*D demonstrated that an objective measurement-based approach might help ease difficult decisions about dosing, time frame, adverse-effect reduction, and symptom reduction.7
Surprisingly, STAR*D did not demonstrate clinically significant differences among medication choices, despite general perceptions and individual experiences of practicing clinicians that some treatments are more effective than others. STAR*D was designed with the hope that patients could be matched with treatments, but this was not the case. For example, it was expected that patients with depression and anxiety on CIT would do worse if switched to BUP-SR than if switched to a different SSRI, but the study did not demonstrate this. The question of how to select the most effective treatments or treatment sequences for individual patients remains unanswered. Future studies can help determine whether there are clinical characteristics or biomarkers that could help clinicians choose the most effective treatment strategies for their patients with treatment-resistant MDD.
New medications for treatment of depression continue to be developed and studied. STAR*D was designed prior to publication of data on the efficacy of antidepressant treatment augmentation using atypical antipsychotic agents. Future studies on treatment strategies and effectiveness in treatment-resistant MDD should include atypical antipsychotic options, as well as examination of how these drugs fit into the treatment paradigms described in STAR*D.7 Another important finding of STAR*D was that CBT, both alone and in combination with CIT, was as effective as the various second-step pharmacologic strategies, although medication augmentation produced remission more rapidly.15 Future studies that incorporate comparisons of psychotherapy would be useful. STAR*D has provided important insights into the management of treatment-resistant MDD, and its results will continue to be utilized to formulate future studies in this difficult-to-treat patient population.
Annals of Clinical Psychiatry ©2010 Quadrant HealthCom Inc.