Response-shift effects in neuromyelitis optica spectrum disorder: a secondary analysis of clinical trial data

Abstract

Background: Researchers have long posited that response-shift effects may obfuscate treatment effects. The present work investigated possible response-shift effects in a recent clinical trial testing a new treatment for Neuromyelitis Optica Spectrum Disorder (NMOSD). This pivotal trial provided impressive support for the drug Eculizumab in preventing relapse, but less strong or null results as the indicators became more subjective or evaluative. This pattern of results suggests that response-shift effects are present.

Methods: This secondary analysis utilized data from a randomized, double-blind trial evaluating the impact of Eculizumab in preventing relapses in 143 people with NMOSD. Treatment arm and then relapse status were hypothesized 'catalysts' of response shift in two series of analyses. We devised a "de-constructed" version of Oort structural-equation modeling using random-effects modeling for use in small samples. This method begins by testing an omnibus response-shift hypothesis and then, pending a positive result, implements a series of random-effects models to elucidate specific response-shift effects.

Results: In the omnibus test, the 'standard quality-of-life (QOL) model' captured substantially less well the experience of placebo as compared to Eculizumab group. Recalibration and reconceptualization response-shift effects were detected. Detected relapse-related response shifts included recalibration, reprioritization, and reconceptualization.

Conclusions: Trial patients experienced response shifts related to treatment- and relapse-related experiences. Published trial results likely under-estimated Eculizumab vs. Placebo differences due to recalibration and reconceptualization, and relapse effects due to recalibration, reprioritization, and reconceptualization. This novel random-effects- model application builds on response-shift theory and provides a small-sample method for better estimating treatment effects in clinical trials.

Keywords: Clinical trial; Clinician-assessed outcome; Definitive neuromyelitis optica; Neurologic; Neuromyelitis optica spectrum disorder; Patient-reported outcome; Response shift.

Fig. 1

Timing of PRO measurement collection in the clinical trial design. This study schema provides the timing of clinician- and patient-reported outcome collection over the course of the trial

Fig. 2

Mean change from baseline to study end in SF-36™ domain scores and EQ-5D VAS by treatment arm. The Eculizumab group evidenced bigger Changes in the SF-36™ physical domains compared to the Placebo group, which showed larger changes in the mental domains

Fig. 3

Discrepancy Score Distributions for the whole sample. The distribution was centered around zero, and slightly left-skewed

Fig. 4

Discrepancy Score Distributions by Treatment Arm. The distributions of discrepancy scores are different by treatment arm, with the largest and more-negative discrepancies found in the Placebo group, as compared to the Eculizumab group. One low-scoring outlier was not shown in the graph but was retained in calculations

Fig. 5

Discrepancy Score Distributions by Relapse Group. The distributions of discrepancy scores are different by relapse group, with the largest and more-negative discrepancies found in the Adjudicated-Relapse as compared to the Clinician-Assessed and No-Relapse Groups. One low-scoring outlier was not shown in the graph but was retained in calculations