Early treatment response affects signal detection in a placebo-controlled depression study

Abstract

We examined the effect of early treatment response on the Quick Inventory of Depressive Symptomatology (QIDS-SR 16) within 2 weeks following randomization on the eventual treatment outcome at 6 weeks in a doubleblind study of subjects with major depressive disorder randomly assigned to a combination treatment (buspirone 15 mg with melatonin SR 3 mg), buspirone 15 mg, or placebo (Clinicaltrials.org: NCT 007005003). The extent of QIDS-SR 16 score improvement between baseline and week 2 was significantly associated with higher treatment response rates at week 6 (≥50% QIDS-SR 16 improvement from baseline) regardless of treatment assignment. Thirty-two of 123 subjects (26.0%) were QIDS-SR 16 treatment responders by week 2 and were excluded in a post-hoc analysis of five clinical metrics: QIDS-SR 16 , the Inventory of Depressive Symptomatology (IDS-c30), clinical global impression of severity and improvement scales, and Hamilton rating scale for anxiety. The effect size favoring the combination-treatment over buspirone and/or placebo increased on each of the 5 clinical metrics in the remaining 91 subjects with < 50% QIDS-SR 16 improvement at week 2. For instance, the effect size favoring the combination treatment over the pooled buspirone and placebo groups improved from 0.33 in the mITT population to 0.64 for the QIDS-SR 16 , and from 0.37 to 0.58 for the IDS-c30. Further, the statistical significance favoring the combination treatment improved from p = .055-.017 for the QIDS-SR 16. This was a post-hoc analysis of a small MDD study, but it is clear that future studies need to explore the mediating factors that affect signal detection and influence individual treatment response.

Baseline reliability and early response in clinical trials of major depressive disorder

“…A post-hoc analysis that excluded the patients who had demonstrated >50% QIDS-SR16 improvement from the baseline score during the first 2 weeks enhanced the effect size favouring the experimental treatment from 0.33 to 0.64 for the QIDS-SR16 despite reducing the sample size (29).…”

“…The symptomatic questionnaires used in MDD trials include many subjective items that require personal perspective that can differ from individual to individual. In fact, the scores recorded by two trained clinician raters about the same patient scored at the same time may be discordant and can often differ from the patient's self-ratings (29,33). It is not possible to affirm the accuracy of any singular measure as a "true" reflection of a patient's current clinical condition.…”

“…However, it is possible to affirm the reliability of paired ratings made by two different raters of the same assessment are reasonably close. The recent availability of remote surveillance methods that offer siteindependent scoring confirmation of recorded site-based interviews has increased confidence about ratings reliability at all study visits (29). Several studies have demonstrated a high inter-rater scoring correlation on depression rating instruments between paired clinician-clinician ratings and between paired clinician-patient self-ratings (29,33).…”

“…Therefore, it is not surprising that many depressed patients show substantial symptomatic improvement immediately following the randomization (baseline) visit regardless of treatment Copyright 2020 KEI Journals. All Rights Reserved http://journals.ke-i.org/index.php/mra assignment (14,16,(21)(22)(23)(24)(25)(26)(27)(28)(29). Baseline score inflation followed by indiscriminate symptomatic improvement regardless of treatment assignment will increase the placebo response and impede signal detection at the end of the study.…”

“…Obviously, a reliable baseline is critical to the reliability of all subsequent assessments of symptomatic change (6,14,20). It is not uncommon that a moderate, early symptomatic improvement will occur within the first week or two following the baseline day regardless of the randomized treatment assignment (14,16,(21)(22)(23)(24)(25)(26)(27)(28)(29). Expectation biases, inflated scores, and other non-specific factors related to study participation may influence this early response (5-7, 11, 28).…”