Review of: Pigott HE et al, BMJ Open 2023;13:e063095

Study type: Reanalysis of an open-label, randomized, rater-blinded clinical trial

When it comes to antidepressant efficacy, we often hang our hats on the 67% cumulative remission rate reported in the 2006 landmark Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study. But a new reanalysis has sparked debate over those results.

STAR*D was funded by the National Institute of Mental Health. It included 4,041 patients ages 18–75 who scored >14 on the Hamilton Rating Scale for Depression (HRSD). All patients started on citalopram. If needed, they progressed through four treatment levels involving cognitive behavioral therapy and 14 pharmacologic strategies, with 12 weeks at each level (plus two more weeks if they were close to remission). In the spirit of a real-world study, STAR*D included patients with comorbid medical and psychiatric conditions.

STAR*D’s primary outcome was remission, defined by <8 on the blinded HRSD. The secondary outcome was response (>50% reduction in HRSD). Although nonblinded assessments like the Quick Inventory of Depressive Symptoma­tology–Self Report (QIDS-SR) were not identified as outcomes, investigators used the QIDS-SR as a primary outcome in the published reports. 931 patients who didn’t meet criteria for depression were included in the level 1 (citalopram) analysis. 370 patients dropped out after their first clinic visit and should have been included but weren’t, and 125 patients already in remission were included in analysis for the next level of treatment.

These deviations from the study protocol likely inflated the results, so researchers in this study reanalyzed STAR*D according to the protocol. It didn’t spell out what to do with patients who dropped out without an exit HRSD, so researchers calculated two remission rates: one that assumed none of these patients achieved remission, and another that counted patients with an exit QIDS-SR <6 as in remission.

Under both scenarios, antidepressant efficacy fell below the 67% cumulative remission rate originally reported. The new remission rates were 35% when only patients with an exit HRSD score were included, and 41% when those with a QIDS-SR <6 were added. 

CARLAT TAKE

The remission rate in STAR*D depends on how the data are analyzed. Using the original protocol results in cumulative remission rates that are about half (35%–41%) of what was originally reported (67%). When analyzed according to the original protocol, STAR*D showed modest remission rates of about half (35%–41%) of what was originally reported (67%).