When National Public Radio’s Science Friday devotes most of a program to a psychiatric study, it must be newsworthy. The STAR*D study got this special treatment recently, with Ira Flato interviewing John Rush about the ins and outs of the NIMH-funded project.
What is the STAR*D study? Are the initial results as earthshaking as promised? And should we be changing our practices as a result?
In this issue of The Carlat Psychiatry Report, we’ll try to answer these questions to your satisfaction.
STAR*D stands for “Sequenced Treatment Alternatives to Relieve Depression.” Funded by NIMH, with almost no pharmaceutical company involvement (they supplied the meds and may have consulted on dosing), the idea behind STAR*D was to conduct a large research trial that would mimic what happens in the real world of psychiatric practice.
Most large clinical trials of antidepressants are funded by pharmaceutical companies, and the companies have one overarching goal: to make their antidepressant look really good, so that the FDA will approve it and clinicians will prescribe it.
How do you make a drug look good? First, do not compare it with another drug; instead, compare it with a placebo. Chances are pretty good that your active drug will outperform placebo. Second, choose your study patients carefully, in order to maximize the chances that they will get better. For most trials, this means excluding patients who are suicidal, have medical problems, are substance abusers, have depression that is either too severe or too mild, or have other psychiatric problems. Such exclusion criteria have been shown to exclude more than 80% of typical clinic patients (Am J Psychiatry 2002;159:469- 473). These design strategies have led to plenty of FDA approvals but poor generalizability to the real world of real patients.
The STAR*D trial, on the other hand, allowed just about any nonpsychotic depressed patient to enroll, except those with a primary diagnosis of bipolar disorder, eating disorders, or OCD. Secondary bulimia and OCD were allowed, however, as were the full gamut of other Axis I disorders, including GAD, PTSD, panic disorder, and others.
If anything, the researchers may have gone too far in the direction of including significant pathology, since many of the study patients are lower functioning than those commonly seen in private practices; the average STAR*D patient had
• Six prior episodes of depression
• Been continuously depressed for two years
• Recurrent depressions for over 15 years
• At least three medical conditions
• Education below college level
In addition, 38% of these subjects were unemployed, and 18% had attempted suicide in the past.
Thus, as the results of STAR*D come trickling in over the next couple of years, you will have to consider whether you feel these patients are generalizable to those you see in your practice.
Now, onto some words about the STAR*D methodology. Think about how you manage your own patients, and you’re most of the way toward understanding STAR*D. Chances are that you a) do not give your patients placebo pills, and b) do not insist that you and your patient be blinded to the name of the medication. In order to mimic these natural situations, STAR*D is not a typical placebo-controlled, double-blind study.
Patients were enrolled because they actively sought treatment through regular medical channels rather than via newspaper or radio advertisements, the typical method in industry trials, which often results in the inclusion of unusually treatment-responsive volunteers. They were told by their doctors what drug and what dose they would be on–neither patients nor doctors were blinded to this information. Patients returned for followup visits every two to four weeks (probably more often than is the standard in the community), during which doctors assessed the response and asked about side effects. Then, based on their scores on the depression scales used, clinicians made dosing adjustments according to a dosing algorithm created for the study.
STAR*D’s research protocol was designed to answer a series of practical questions that we all have as we treat depressed patients. The initial assumption is that most psychiatrists and primary care doctors start their depressed patients on one of the SSRIs. Accordingly, all 4,041 patients were started on Celexa (citalopram). Why was Celexa chosen? Largely because of its minimal drug-drug interactions and its relative lack of discontinuation side effects.
Patients for whom Celexa did not work out (either because of side effects or lack of remission) went on to Step Two and were randomized to an array of possible treatments within two broad categories: switching to another treatment, or continuing on Celexa but augmenting with an additional treatment. The switch group was randomized to Wellbutrin SR (bupropion), Effexor XR (venlafaxine), Zoloft (sertraline), or cognitive therapy; the augmentation group was randomized to treatment combining Celexa with Wellbutrin SR, BuSpar (buspirone), or cognitive therapy.
Thus far, only results from the first two steps have been published; the third and fourth phases of the study are currently underway.
Celexa as Monotherapy
Let’s move onto the results. First, how well did Celexa monotherapy do for these patients’ depression?
All patients were started on 20 mg/day, were raised to 40 mg/day by week four, and then to 60 mg/day by week 6 if necessary. This relatively aggressive dosing schedule resulted in a final average Celexa dose of 41.7 mg/day.
The rate of remission, defined as a score of seven or less on the 17-item Hamilton Depression Rating Scale (HDRS-17), was 30%. The rate of response, defined as a 50% reduction of symptoms on a different rating scale (the QIDS), was 47%.
On average, it took patients 5.7 weeks to achieve response, and 6.7 weeks to achieve remission. The average length of treatment on Celexa (including all patients, whether remitted or not) was 10 weeks.
In practical terms, the implication of the Celexa monotherapy results is that you need to push the dose up to at least 40 mg/day and that your patients will need six to seven weeks to achieve remission, if they are destined to achieve remission at all. Because remember, 70% of these patients did not achieve remission, and were entered into other treatment arms (see below).
Is there any way to tell which of your patients will end up in the lucky 30%? In order to answer this question, researchers did some secondary analyses to see what patient characteristics predicted remission. Positive prognostic qualities included
• College education or higher
• Fewer comorbid psychiatric or medical conditions than average
Having any one of these characteristics increased odds of remission by a factor of about 1.3. (However, you can’t add them all up to to achieve a stratospheric remission rate–the statistics don’t allow that tempting calculation!)
So did Celexa do well or poorly? If you were to compare these numbers to the numbers we typically see in placebo-controlled trials, they are a little worse. Response rates of 50-60% and remission rates of 30-40% are standard in these trials, but remember that the patients enrolled in these studies are a healthier population at baseline.
These results imply that you can look your depressed patients squarely in the face and say, “If you take this SSRI at a good dose for at least two months, you have a 30% chance of having your depression disappear completely.” Does this sound impressive to you? Will it sound impressive to your patients? You will have to be the judge, based on your own clinical experiences and practices.
Antidepressant Augmentation: BuSpar versus Wellbutrin SR
While 30% of STAR*D patients remitted on Celexa alone, 70% did not. This remaining large group of patients was randomized in a complicated way to different options. 565 of these patients were assigned to augmentation treatment with either Wellbutrin SR (279 pts, mean dose 267.5 mg/day) or BuSpar (286 pts, mean dose 40.9 mg/day). Remission took an average of 6.3 weeks for the Wellbutrin SR patients and 5.4 weeks for those assigned to BuSpar (not a statistically significant difference). The remission and response numbers are in the table below.
Although both BuSpar and Wellbutrin SR augmentation produced about 30% remission, various aspects of effectiveness and tolerability, reproduced in the table below, put Wellbutrin SR ahead.
How significant are these differences? The study authors were certainly impressed: “Overall, these findings reveal a consistently more favorable outcome with sustained-release bupropion than with buspirone augmentation.”
But if you look at the raw numbers, the clinical significance of the difference in depression scores is debatable. Wellbutrin SR nudged patients’ QIDS score about one point lower than did BuSpar. Since the QIDS scale is a 27- point scale, a one-point difference is awfully small. For example, on the QIDS weight loss question, one point is the difference between a patient who reports no weight loss and one who “feels as if some slight weight loss has occurred.”
Thus, it appears that there was little meaningful clinical difference between Wellbutrin SR and BuSpar augmentation, which is troubling. Why? Because prior double-blind studies of BuSpar augmentation of SSRIs have shown it to work no better than placebo, implying that BuSpar essentially operated as a placebo in the STAR*D trial. This, in turn, would mean that Wellbutrin augmentation works no better than placebo either.
This is a controversial point, and one that STAR*D’s principal investigator, John Rush, rebuts in our interview with him in this issue.
Nonetheless, the fact that Wellbutrin was better tolerated than BuSpar provides some research support for the common practice of using Wellbutrin augmentation more readily than BuSpar. Of course, in clinical practice, we tend to choose BuSpar augmentation for our more anxious patients, and STAR*D doesn’t give any further guidance here, since patients were assigned randomly to either option (based on their preference as explained below) rather than assigned according to a specific symptom profile.
Switching strategies: Effexor XR, Wellbutrin SR, or Zoloft?
Of the Celexa nonremitters, 727 patients were randomly assigned to a switch from Celexa to one of the following: Effexor XR (average dose 193.6 mg/day), Wellbutrin SR (282.7 mg/day) or Zoloft (135.5 mg/day). Patients stayed on these meds for an average of eight to nine weeks, depending on the treatment.
Here are the results:
Granted, it looks like Effexor XR edges out its competitors here, but the differences weren’t statistically significant. Nonetheless, Effexor has such a numerical edge over Zoloft that we wonder if this difference, though not statistically significant, might represent a clinically significant effect. This is a crucial point, because the legendary Thase meta-analysis showed that Effexor is slightly better at bringing about remission than SSRIs (Brit J Psychiatry 2001;178:234-241).
The remission rates of these switch patients are lower than the remission rates of the augmentation patients, tempting clinicians to conclude that augmentation works better than switching. Unfortunately, we can’t affirm that conclusion, because patients weren’t randomly assigned to switching versus augmentation. In fact, patients were allowed to choose whether they wanted to switch or augment, so that patients who ended up in the switch group were different from those in the augmentation group. Predictably, many patients who chose to switch were intolerant of Celexa (56%), and they were somewhat more depressed as well. So the lower remission rates of the switch patients may simply mean that they were a sicker and more medication-intolerant population.
Summing Up the STAR*D Results
Based on this first set of STAR*D results, the take-home clinical points are as follows:
• SSRI monotherapy can bring chronically depressed patients into remission, but the dose needs to be high and the length of treatment at least seven weeks.
• Augmentation of SSRIs with either Wellbutrin SR or BuSpar can increase remission rates to some extent, but it’s not clear whether this effect is any greater than a placebo effect.
• Switching to antidepressants with vastly different mechanisms of action leads to identical results, a discouraging result for those psychiatrists who believe that a strong command of neurotransmitter knowledge leads to better treatment decisions.
On the brighter side, many more results are forthcoming, including data on cognitive therapy and secondary analyses of the medication data. Let’s hope that this information provides even more data to enhance our clinical toolboxes.
TCR VERDICT: STAR*D: A promising first round, but we hope for more.