Response-shift effects in neuromyelitis optica spectrum disorder: estimating response-shift-adjusted scores using...

Area of Science:

• Clinical trial methodology
• Health outcomes research
• Psychometrics

Background:

• Companion paper used random intercept models (RIMs) to investigate response-shift effects in an Eculizumab vs. Placebo trial for neuromyelitis optica spectrum disorder (NMOSD).
• Global Health was predicted using the EQ-5D Visual Analogue Scale (VAS) item, with SF36™v2 mental and physical component scores (MCS and PCS) used to detect response shift.
• This study aimed to back-translate VAS scores into MCS/PCS scores, adjusting for response-shift effects.

Purpose of the Study:

• To back-translate Visual Analogue Scale (VAS) scores into Mental Component Scores (MCS) and Physical Component Scores (PCS) adjusted for response-shift effects.
• To quantify and compare response-shift effects between treatment groups (Eculizumab and Placebo) in NMOSD patients.
• To explain potential null findings in previous clinical trial analyses by accounting for response shifts.

Main Methods:

• Secondary analysis of NMOSD clinical trial data (n=143) assessing Eculizumab's relapse prevention.
• Equated raw scores from VAS, MCS, and PCS to compute scores excluding response-shift effects.
• Employed correlation analysis and descriptive displays to examine response-shift effects comprehensively.

Main Results:

• Crosswalks between MCS/PCS and VAS equated scores with and without response-shift effects.
• Low shared variance between MCS scores (with and without response shift) indicated substantial differences in mental health constructs.
• Physical health shared variance differed significantly only in the Placebo group; larger MCS response-shift effects were observed in the Placebo group at study end, especially at score distribution extremes.

Conclusions:

• Notable differences in MCS response shifts between treatment groups were identified, but not for PCS.
• These group-specific response shifts may explain previously reported null results in NMOSD trials.
• The presented methodology offers a novel approach for analyzing response-shift effects in clinical trial data.