Sequenced Treatment Alternatives to Relieve Depression: Recent Results

Associate Director, Depression Clinical and Research Program, Massachusetts General Hospital, Associate Professor of Psychiatry, Harvard Medical School

Dr. Nierenberg is associate director of the Depression Clinical and Research Program at Massachusetts General Hospital and associate professor of psychiatry at Harvard Medical School in Boston. He is also director of the National Institute of Mental Health Bipolar Trials Network, a federally-funded infrastructure for the next generation of clinical trials to help patients with bipolar disorder. He was the site principal investigator for the Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD) and was involved in the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study. His research interests focus on bipolar disorder, treatment-resistant depression, and the longitudinal course of affective disorders. Dr. Nierenberg has been listed in The Best Doctors in America for the treatment of mood and anxiety disorders in every edition since 1994.

What were the objectives and design of the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study?

STAR*D was a large, hybrid study sponsored by the National Institute of Mental Health created to instruct clinicians on what to do next if one treatment does not bring patients to remission. There was never a study that compared competing alternatives, such as augmentations or switches.

The study was unique in that it involved patients who were presenting for care in real-world settings. It was designed so that the results of the study would have maximum generalizability. Given the nature of the study and how people were invited to participate, it is probably the most representative clinical trial of depression that has ever been conducted. The patients who presented to the study are just like patients who will present to any clinician’s office, and the sites where this occurred were in the community. Many of these sites had never conducted any sort of clinical trials before.

Advertising was rejected as a means to solicit patients. Patients who presented to primary care or psychiatric practices with depressive symptoms were directed by their clinicians to consider participating in the STAR*D study.

How did STAR*D differ from other studies?

There were 4,041 patients who initially entered into the study. Some of them were not eligible for parts of the study. Nevertheless, many patients continued treatment.

There were very few exclusion criteria and very few targeted inclusion criteria. Patients needed to have met the criteria for Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition,¹ major depressive disorder (MDD). They had to be willing to at least have pharmacotherapy. In some places, they could speak English or Spanish, depending on if there was a Spanish-speaking clinician available. They could have other problems that would usually exclude them from other studies. The STAR*D study is in sharp contrast to the randomized clinical trials that are used to register new medications because such trials have very restrictive exclusion criteria. They disallow for many comorbid conditions and the patients who enter those studies are not like the patients who present for clinical care.

Another unique aspect of STAR*D was that psychotherapy was offered. If someone did not react well to medication and they were willing to accept psychotherapy, they could have it added or be switched to it.

What made the STAR*D study feasible was that all the treatments were open but the assessments were blind. This allowed for great flexibility, depending on how a patient was doing at any particular site. The STAR*D study was conducted throughout the United States in many settings. Because it was conducted in 41 clinical settings, it was very important to be able to make this an open study as opposed to one that was so restricted that it could not be done at all.

What were the major findings of the STAR*D study?

We found that it was feasible to use something called “measurement-based care,” guided by measuring the patient’s progress rigorously and by measuring their side effects. The clinicians would use guidelines to make the clinical decisions about, for example, how quickly a dose of medication could be escalated. Using this enhanced measurement-based care, it was discovered that very vigorous treatment with the selective serotonin reuptake inhibitor (SSRI) citalopram brought approximately 33% of these patients into full remission.

These results match what would be expected in regular, randomized, clinical trials. Patients in the STAR*D study had other burdens of both medical and psychiatric comorbid illnesses, which would lead one to expect a lower remission rate. Remarkably, almost regardless of the next step taken (mostly switching or augmenting patients), another 20% to 30% of STAR*D patients improved.

Another unique and important aspect of the STAR*D study is how patients were randomized to the next treatment. Using equipoise randomization, patients were randomized only to the groups of treatments that they would find acceptable. This made a lot of clinical sense. The goal of each level of treatment was to help patients achieve remission. If a patient was substantially better but did not meet the criteria for remission it would make no sense clinically to then randomize the patient and either add or switch to another drug.

Patients could be treated for up to 12 weeks. The study was designed so that if a patient felt that he or she could not stay on the drug any longer, he or she could elect to go to the next level. The mean duration of treatment was 9 weeks, which is also much longer than any other study that has been conducted for treatment-resistant depression, in terms of the first phase.

Most patients decided to have another drug added if they were tolerating the current medication and doing somewhat better. If they could not tolerate the drug and they were not doing better, most elected not to have another drug added, but rather elected to switch. In that second level, approximately 20% to 30% of the patients who were willing to be re-randomized went into remission.

Thereafter, the story is not as positive. Perhaps 10% to 15% of patients got better in the next level and about the same amount got better in the level after that. Thus, patients had two shots at finding the treatment that would have the best effect; thereafter, it became more difficult to get to the point of remission.

Were researchers surprised by any of the study findings?

One of the surprises was that augmentation options in the second level helped including adding either buspirone or bupropion. We were uncertain whether these would work or not, specifically regarding buspirone. However, as many other researchers were adding these medications, it was important for us to do so as well. It turned out that patients reacted fairly well to these drugs. The remission rates with buspirone or bupropion augmentation were about the same. Patients who elected to augment improved somewhat more than patients who elected to switch, just as would be expected. However, switches cannot be compared to augmentation, as they are really different groups.

The other surprise was in the switching arm. Patients could either be switched to sertraline or bupropion, which work completely differently, or they could be switched to the dual-action agent venlafaxine. We found that there was no substantial difference when people were switched from the initial SSRI to any of those three switches. One would have thought that either switching to a drug with a different or dual mechanism of action would have been substantially better, but it actually was not.

Is it true that an unexpected portion of patients did not show a meaningful response until after 8 weeks?

There were early responders and late responders. Some of the previous literature would support this. Part of the reason for this is that approximately 80% of the patients who entered into STAR*D study had either chronic or recurrent MDD as opposed to having had just one episode.

As there was no placebo group in the STAR*D trial, is there any way of knowing to what extent patients in remission were simply reflecting placebo or spontaneous remission?

We do not know. In all placebo-controlled trials there is a very interesting bias that is never really talked about. That is, placebo-controlled trials only include those patients who are willing to take placebo. The other limitation of registry-type placebo-controlled trials is that they exclude people who are severely ill. For example, if patients have suicidal ideation and there is deemed to be some suicidal risk, these patients are immediately excluded from the trials. In STAR*D, we included people who were suicidal as long as they could be managed as an outpatient.

One does not quite know exactly what the external generalizability is of the randomized controlled trial that includes placebo, though we included different people. We do not know what would have happened if we did have placebo. There most likely would have been a different population willing to go into STAR*D and a different population that clinicians were willing to put into the trial.

In view of the fact that a higher response or remission rate was anticipated when designing the trial, is it possible that, despite its large size, the study was underpowered?

There are certain comparisons that were not powered enough to show small differences, but, in fact, we were less concerned about that. We wanted to be able to find differences that would change practice, and we were probably powered enough to show that. We also had many sites that consisted mostly of patients with low socioeconomic status. There were also patients who used alcohol but did not meet criteria for abuse. Patients who left acute treatment prematurely had substantial socioeconomic disadvantages, more general medical conditions, more comorbid Axis I disorders, a greater illness burden, and did not reach criteria for remission.

How has your involvement with the STAR*D study changed the way you practice?

It has become routine for me to have every patient conduct a self-rated measurement. This is one of the other unique aspects of this trial. The measurements that were used in the trial are the measurements that can be used in one’s office. Specifically, there is the Quick Inventory of Depressive Symptomatology-Self-Rated, which is downloadable for free.² One can actually use the instruments that were used in this trial, as opposed to trials that use the Hamilton Rating Scale for Depression (HAM-D), which is hardly used by clinicians. The other nice thing about using the self-rated score, which we found was as good as the HAM-D, is that the patient completes it, the report takes 3 minutes to fill out, and it is very easy to score. I think that in order to get patients to remission it is absolutely essential that clinicians measure what they treat. I do this in my practice all the time.

The STAR*D study has also showed the effectiveness of buspirone. I have started to use it and find that some chronically depressed patients do quite well on it. I was surprised about these results.

In light of the antidepressant effects of some mood stabilizers as well as the debate about diagnostic boundaries of unipolar versus bipolar depression, would you have included an arm for lamotrigine or another mood stabilizer as monotherapy?

Inclusion of mood stabilizers may have been interesting. However, when we designed the study in 1998, we were choosing to study from the treatments being used at the time. Lamotrigine was not on the radar screen for unipolar depression at that time.

Can you speculate on whether or not a mood stabilizer should be started for a patient who fails on two different antidepressant regimens?

I do not think we really have the data to say that. What is remarkable is that the rate of people presenting with mania—which might be a surrogate marker for the propensity to have bipolar disorder—was unbelievably low in STAR*D. We do not know exactly why patients left the study, and it is difficult to track them down, so we do not know what happened to them. However, I was the safety officer for the study and saw every single serious adverse event. Mania was very infrequent. I do not think that many of these patients were likely to have some sort of bipolar or bipolar-spectrum disorder.

What additional data are being analyzed?

The group is analyzing much more of the longitudinal data in detail. We are also doing moderator and mediator analyses to try to get a finer understanding of who has responded to what and when. We have a long list of other sorts of analyses that we are in the middle of. Plenty should be coming out in the next year or so. Results of the cognitive therapy arm are also in the press now.

Is there anything else of interest regarding STAR*D?

I think that remission is achievable for most people with depression, but even with this enhanced care many patients did not get or stay well. Fundamentally, we do need better treatments. Rather than the conventional use of one monotherapy, we need to start thinking about the possibility of using combined therapy, which may help patients achieve remission sooner and stay well. I think that is a hypothesis worth testing.

References

1.  Diagnostic and Statistical Manual of Mental Disorders. 4th ed, text rev. Washington, DC: American Psychiatric Association; 2000.

2.  Quick Inventory of Depressive Symptomatology Self-Rated. Available at: www.ids-qids.org. Accessed January 4, 2007.