VIRGIL ELECTRONIC CLINICAL OUTCOME ASSESSMENTS (ECOA) PLATFORM: IMPROVING SIGNAL DETECTION IN ALZHEIMER'S DISEASE CLINICAL TRIALS

Abstract

The impact of Alzheimer's disease (AD) globally is astounding; the number of people living with the disease is predicted to quadruple by year 2050 to 106 million. Nonetheless, drug development for AD has had among the lowest success rate for any therapeutic area. The high failure rate in AD trials has, in part, been due to imprecision in endpoint measurements, which introduces noise. Clinical outcome assessments (COAs) that use traditional paper-based administrations are known to be cost- and time-inefficient, but they also are subject to human error, which creates variability in measurement. The Virgil Investigative Study Platform uses electronic source data collection with real-time clinical guidance to standardize outcome measurements and improve data quality. In the present study, we compared paper-based administrations against administration with Virgil to determine the extent to which the use of electronic assessments (eCOA) minimized scoring errors in primary and co-primary endpoints of AD trials. Paper-based assessments from a recent clinical trial of mild cognitive impairment (MCI) were compared against eCOA administrations of the same scales in separate MCI trials. All studies are phase II/III multinational trials. Score discrepancies in CDR, ADAS-Cog, ADCS-ADL-MCI, and MMSE were identified via review of audio recordings and worksheets by the same cohort of expert calibrated reviewers. For each scale, discrepancies were compared between paper-based and Virgil administrations. Percentages of reviews with at least one scoring discrepancy, as well as two or more discrepancies, were significantly and substantially lower in Virgil administrations compared to paper-based for all scales. Paper-based administrations create unnecessary variability around endpoint measurements, which can contribute to inconclusive results. The Virgil eCOA platform with real-time clinical guidance, auto-calculation of scores, and prompts for missing data and out-of-range errors can help standardize scale administration and scoring, thereby substantially reducing error variance and improving signal detection.