Restoring STAR*D: A RIAT Reanalysis of Medication Augmentation Therapy After Failed SSRI Treatment Using Patient-Level Data with Fidelity to the Original Research Protocol.

Background: The STAR*D trial is the most influential study of sequential antidepressant treatment strategies. However, major STAR*D publications deviated from the protocol-defined analytic plan. Prior re-analyses found lower cumulative remission rates than STAR*D publications reported, sustained remission rates of only 3.1 to 8.4% at 12 months, and high rates of treatment-emergent suicidal ideation (TESI) during medication-switch therapy. A similar reanalysis is warranted for STAR*D's augmentation study in which citalopram was augmented with sustained-release bupropion or buspirone. Methods: We reanalyzed STAR*D's patient-level augmentation dataset with fidelity to the original protocol or relevant STAR*D publications where the protocol did not prespecify an analytic plan. Results: This reanalysis identified 124 patients (21.9% of enrolled subjects) who were inappropriately included in the original STAR*D analysis, including 54 who were in protocol-defined remission before starting augmentation therapy. Remission rates as defined in the protocol were lower than reported in the original publication for bupropion SR (25.0% vs 29.7%) and buspirone (25.8% vs. 30.1%). Using a secondary definition of remission, bupropion SR's rate was significantly lower than reported in original publications (29.2% vs. 39.0%). Sustained remission through 12 months was low (4.9-12.5%). TESI rates were significantly higher for buspirone (13.9%) than bupropion SR (3.6%) augmentation. Conclusion: Compared with the original STAR*D publication, our reanalysis identified inflated remission rates, low sustained remission, and marked differences in TESI risk between augmentation strategies. These findings suggest that both treatments offer lower acute and sustained benefit than is widely understood, with buspirone associated with more TESI.