Audio-digital recordings for surveillance in clinical trials of major depressive disorder

Abstract

Ratings surveillance is used in clinical trials to assure ratings reliability of site-based scores. One surveillance method employs audio-digital recordings of site-based clinician interviews to obtain remote, site-independent scores for assessment of paired scoring concordance and interview quality. We examined the utility of this surveillance strategy using paired site-independent scores derived from recorded site-based Montgomery-Asberg depression rating scale (MADRS) interviews obtained from patients with major depressive disorder (MDD) participating in 5 clinical trials. High correlations were noted between the 3736 paired site-based and site-independent scores across all visits. Some rater "outliers" were identified whose ratings performance improved following remediation. In 3 studies with available outcome data, the blinded remote ratings yielded a high predictive value (91.2%) for replicating treatment response rates. The magnitude of the total MADRS scores affected the directionality of paired scoring deviations in each of the 5 studies. Across all visits, site-based raters scored the more severe MADRS scores (≥30) higher than site-independent raters and the less severe MADRS scores (<20) lower than site-independent raters. Individual MADRS items were similarly affected by the directionality of symptom severity. This analysis affirms the utility of audio-digital recording of site-based interviews as a surveillance strategy for quality assurance (monitoring and remediation). In addition, the high predictive value of blinded remote ratings to replicate site-based treatment outcomes may be useful to affirm primary site-based results when there is a potential of functional unblinding. The use of remote ratings as a primary measure beyond its utility for quality assurance needs further exploration.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 12 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.