A new review of depression treatments suggests that the scientific evidence for many common strategies used when a first antidepressant fails is not as strong as widely believed. The findings, which reexamine the influential Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial, indicate that the benefits observed in that study may stem more from factors like patient expectations than the specific pharmacological action of the medications. The analysis was published in the Journal of Clinical Psychopharmacology.
The new review was conducted by a team of researchers led by Kevin P. Kennedy at the Corporal Michael J. Crescenzo VA Medical Center. Their work was prompted by the widespread influence of the STAR*D trial.
Before STAR*D, most clinical trials for antidepressants studied medications in carefully selected patient groups. These trials often excluded individuals with other medical conditions, co-occurring psychiatric disorders, or chronic depression, meaning the results were not always applicable to the more complex patients seen in everyday clinical practice.
Published in the early 2000s, STAR*D was a large and ambitious study designed to fill this knowledge gap. As a pragmatic trial, it was conducted in real-world primary care and psychiatric clinics and enrolled a diverse group of over 4,000 patients, making it the largest study of its kind.
All participants began treatment with the antidepressant citalopram. Those who did not achieve remission after this first step could proceed through a sequence of up to three additional treatment levels, where they were offered different strategies, such as switching to another antidepressant or augmenting their current medication with a second one.
The findings from STAR*D have shaped depression treatment for years. The study reported that while only about a third of patients recovered after the first treatment, sequential treatment steps offered continued hope. The most widely cited conclusion was that by trying up to four different strategies, a cumulative remission rate of nearly 70% could be achieved among patients who remained in the study.
However, the original STAR*D study was open-label, meaning both patients and their doctors knew which medication was being prescribed, and there was no placebo group for comparison. This design makes it difficult to separate the true effect of a drug from other factors like a patient’s expectations or the natural course of the illness.
Since STAR*D’s publication, many of its treatment steps have been tested in double-blind, placebo-controlled randomized trials, which are considered a more rigorous way to measure effectiveness. Kennedy and his colleagues set out to compare the results from STAR*D with this newer body of evidence.
The researchers performed a detailed review of the scientific literature, searching for meta-analyses and high-quality randomized controlled trials that investigated the specific treatment strategies used in the different stages, or “levels,” of the STAR*D study.
They then systematically compared the findings from these blinded, controlled studies to the outcomes reported in the original STAR*D trial for each corresponding strategy. These strategies included increasing the dose of an initial antidepressant, switching to a different one, and augmenting an antidepressant with a second medication.
The first strategy examined was the practice of increasing an antidepressant dose if a patient does not respond to the initial starting dose. In STAR*D, patients who did not achieve remission on the antidepressant citalopram had their dose systematically increased.
The review by Kennedy and colleagues found that this common practice is not well supported by subsequent controlled trials. Multiple meta-analyses that pooled data from thousands of patients provided evidence that increasing the dose of a selective serotonin reuptake inhibitor (SSRI), the class of drug that includes citalopram, offers no significant benefit over simply continuing the original, standard dose.
These analyses also suggested that higher doses tend to be associated with a greater likelihood of side effects, potentially making the treatment less tolerable for patients without providing additional antidepressant effects. While a couple of analyses identified a very modest benefit at certain higher doses, the overall body of evidence points toward a flat dose-response relationship for SSRIs, meaning that once a standard therapeutic dose is reached, higher doses do not appear to provide a clinically meaningful improvement.
The next strategy evaluated was switching to a different antidepressant after the first one proved ineffective. This was a core component of Levels 2 and 3 in the STAR*D trial.
The review found a similar lack of supporting evidence from blinded trials for this approach. A meta-analysis of studies in which patients were randomly assigned to either switch to a new antidepressant or continue their original one found no advantage for the switching strategy. In these controlled settings, patients who switched medications did not experience greater symptom reduction than those who stayed on their initial medication.
The researchers also examined the strategy of augmentation, which involves adding a second medication to the first antidepressant. In STAR*D’s Level 2, patients could have their citalopram augmented with either bupropion or buspirone. For buspirone, the review found consistent evidence from blinded trials that it performs no better than a placebo when added to an SSRI. This finding stands in contrast to STAR*D, where buspirone augmentation was associated with remission rates nearly identical to bupropion augmentation.
The evidence for bupropion augmentation was more complex but generally did not replicate STAR*D’s positive results. A comprehensive meta-analysis found that when all trials were considered, adding bupropion was not superior to antidepressant monotherapy. While a small subset of trials involving patients who had previously not responded to treatment showed a marginal benefit, these studies had limitations. The larger, higher-quality trials failed to show a clear advantage for the combination treatment.
The review then moved to the augmentation strategies used in STAR*D’s Level 3, which were reserved for patients who had not responded to two previous treatment attempts. These strategies involved adding either T3 thyroid hormone or lithium. For T3, the available evidence from controlled trials is limited, but existing meta-analyses do not suggest that it outperforms a placebo. Studies looking at both T3 augmentation and co-prescribing it with an antidepressant from the start have not found a significant benefit in remission or response rates.
Lithium augmentation, on the other hand, appeared to be one of the few STAR*D strategies with some support from controlled trials. Meta-analyses of placebo-controlled studies have consistently found that adding lithium to an antidepressant is an effective strategy for treatment-resistant depression. However, the researchers noted an important limitation. The evidence base is surprisingly small, and very few of these trials have specifically studied lithium in combination with the modern SSRIs that are most commonly prescribed today.
Finally, the researchers looked at the Level 4 strategy of combining the antidepressants venlafaxine and mirtazapine for highly treatment-resistant patients. A large meta-analysis provides evidence of a benefit for this type of combination therapy compared to monotherapy. This finding seems to support the strategy used in STAR*D.
Yet, the review authors point to significant limitations within that meta-analysis. They note that the positive result appears to be heavily influenced by many small studies, while the five largest and highest-quality trials on the topic were all negative. This suggests the possibility of publication bias, where smaller studies with positive results are more likely to be published than larger studies with negative results. After accounting for this potential bias, the benefit of the combination was reduced to a level that may not be clinically meaningful.
The authors of the review acknowledge several limitations in their own analysis. The patients included in randomized controlled trials are often different from the “real-world” patients in STAR*D, who had more co-occurring medical and psychiatric conditions. It is possible that treatments that fail in controlled trials could still have an effect in a more diverse population.
Additionally, the specific treatment protocols in the controlled trials did not always perfectly match the steps taken in STAR*D, and the review itself was not a formal systematic one, meaning some relevant studies may have been missed.
The findings from this review have several important implications. They suggest that many treatment guidelines, which were shaped by STAR*D, may be based on strategies whose effectiveness is not confirmed by blinded, placebo-controlled evidence. The discrepancy between STAR*D’s outcomes and the results of controlled trials highlights the powerful role of non-pharmacological factors in treating depression. These factors, such as patient expectancy and the therapeutic relationship, may account for much of the improvement seen in open-label settings.
Future research should focus on conducting high-quality, blinded trials for second- and third-step depression treatments to provide clinicians and patients with clearer guidance. The review also suggests that findings from pragmatic trials should be interpreted with caution until they are validated by more rigorous studies.
The study, “What if STAR*D Had Been Placebo-Controlled? A Critical Reexamination of a Foundational Study in Depression Treatment,” was authored by Kevin P. Kennedy, Jonathan P. Heldt, and David W. Oslin.
