How Do Depression Treatment Success Rates Change with Each Failed Attempt?
A 2023 reanalysis of the STARD study’s patient-level data, published in BMJ Open, found that the trial’s real remission rate was about half of what was reported. STARD was an open-label, semi-randomised study of up to four increasingly aggressive antidepressant therapies in 4,041 adults who screened positive for major depressive disorder, run across 41 North American psychiatry and primary care centres. When the reanalysis applied the study’s own protocol-stipulated outcome measure, the cumulative remission rate after up to four treatment trials was 35.0%, not the 67% STAR*D investigators reported.
Dr. Kumar’s Take
This reanalysis changed how I read the STAR*D numbers that have anchored depression treatment expectations for nearly two decades. The original investigators did not use the blinded rating scale their own protocol required for the headline cumulative figure, and they counted patients who already scored as remitted before treatment began. Those are not small bookkeeping issues. When the protocol is followed, treatment outcomes look considerably worse than the field has assumed, and worse than what comparable open-label trials report. I would rather practise from the honest number than the flattering one, because the honest number is what tells me how much work still needs doing for a patient in front of me.
Study Snapshot
STAR*D enrolled 4,041 adults who screened positive for major depressive disorder. In contrast to most clinical trials, it enrolled patients seeking care rather than recruited volunteers, and included patients with a wide range of common comorbid medical and psychiatric conditions to make the findings generalisable to real-world practice. Patients who failed to gain adequate relief from their level 1 trial on the SSRI citalopram could receive up to three additional treatment trials in levels 2 to 4, and the study evaluated 13 antidepressant therapies across those later levels.
Per the STAR*D protocol, the primary outcome was remission, defined as a score below 8 on the blinded Hamilton Rating Scale for Depression. Response was a secondary outcome, defined as a reduction of 50% or more in HRSD scores. The protocol specifically excluded all non-blinded, clinic-administered assessments from use as research outcome measures.
Results in Real Numbers
The reanalysis found that STAR*D investigators did not use the protocol-stipulated HRSD to report cumulative remission and response rates in their summary article, and instead used a non-blinded clinic-administered assessment. That inflated the reported outcomes, as did the inclusion of 99 patients who scored as remitted on the HRSD at study outset and 125 who scored as remitted when starting their next-level treatment. All of those patients should have been excluded from the analysis.
Against the STAR*D-reported 67% cumulative remission rate after up to four antidepressant treatment trials, the rate was 35.0% using the protocol-stipulated HRSD and the protocol’s inclusion criteria. Combining the HRSD-defined remissions with those from a non-stipulated measure, for patients missing an exit HRSD score, raised the cumulative rate to 41.3%.
For level 1, the treatment remission rate was 25.5% and the response rate 40.5%. Outcomes were worse in treatment levels 2 to 4. Mean change on the HRSD for level 1 patients was 8.4 points, and again worse for patients in the later levels.
Who Benefits Most
The reanalysis compared STARD’s outcomes with a meta-analysis of 7,030 patients enrolled in similar open-label antidepressant comparator trials. Remission and response rates in those comparator trials averaged 48.4% and 65.2%. STARD’s level 1 patients reached 25.5% and 40.5%, and patients in levels 2 to 4 did worse still. Mean HRSD improvement in the comparator trials was 14.8 points versus 8.4 points for STAR*D level 1 patients.
That gap matters because STAR*D deliberately enrolled patients seeking care with comorbid medical and psychiatric conditions, which is closer to the population most clinicians actually treat than a typical recruited trial sample.
Safety, Limits, and Caveats
This is a reanalysis of an existing patient-level data set rather than a new trial, conducted under the guidelines of the Restoring Invisible and Abandoned Trials initiative. Its findings turn on which outcome measure is used and which patients are included, so the headline numbers depend on holding the original protocol as the standard.
The two cumulative figures, 35.0% and 41.3%, differ only in whether remissions from a non-stipulated measure are added for patients missing an exit HRSD score. Both are far below the 67% that has been cited from the original reports.
Practical Takeaways
- Treat the widely cited 67% cumulative remission figure with caution: the protocol-faithful rate after up to four antidepressant trials was 35.0%
- Expect a first-line SSRI trial to produce remission in roughly a quarter of real-world patients, with a 25.5% remission rate and 40.5% response rate at level 1
- Anticipate diminishing returns: remission, response, and symptom improvement were all worse in treatment levels 2 to 4 than in level 1
- Recognise that STAR*D patients did worse than patients in comparable open-label comparator trials on remission, response, and mean HRSD change
What This Means for Depression Treatment
If the honest cumulative remission rate after four sequential antidepressant trials is 35.0%, then sequential monotherapy is a weaker strategy than the field has assumed. That argues for setting realistic expectations with patients at the outset, and for not treating repeated medication trials as an open-ended path that eventually works for most people.
It also argues for scrutiny of how outcomes get reported. STAR*D’s protocol excluded non-blinded clinic assessments from research outcomes for good reason, and substituting one changed the conclusion substantially.
Related Studies and Research
Episode 31: Depression Explained, The Biology Behind the Darkness
Episode 32: Depression Recovery Roadmap: A Step-by-Step, Evidence-Based Plan
FAQs
Why is the reanalysed remission rate so much lower than the published one?
Three reasons: the investigators used a non-blinded clinic-administered assessment instead of the protocol-stipulated blinded HRSD, they included 99 patients who already scored as remitted on the HRSD at study outset, and they included 125 who scored as remitted when starting their next-level treatment.
Should patients continue trying treatments after multiple failures?
Remission, response, and symptom improvement were all worse in levels 2 to 4 than in level 1, so later trials do less work than earlier ones. The decision to continue is an individual one made with a clinician, informed by a cumulative rate of 35.0% rather than 67%.
How did STAR*D compare with other antidepressant trials?
Remission and response averaged 48.4% and 65.2% in a meta-analysis of 7,030 patients in similar open-label comparator trials, versus 25.5% and 40.5% for STARD level 1 patients. Mean HRSD change was 14.8 points in the comparator trials versus 8.4 points for STARD level 1.
Bottom Line
A protocol-faithful reanalysis of STAR*D puts the cumulative remission rate after up to four antidepressant treatment trials at 35.0%, or 41.3% when non-stipulated remission measures are added, roughly half the 67% originally reported. Level 1 remission was 25.5% with a 40.5% response rate, and outcomes were worse in levels 2 to 4.
