New Measures of Neurological Disability Derived from NARCOMS Registry Data

Background:  A challenge in MS clinical research has been to account for the multiform manifestations of the disease.

Objectives: 1) To estimate linear scores of global neurological disability and uncorrelated scores of residual variability for specific disability domains using North American Research Committee on Multiple Sclerosis (NARCOMS) data and Item Response Theory (IRT) methods; 2) to assess relations between disability scores and patient characteristics.

Methods:  Study data included recruitment surveys collected in 1998-2011. Disability measures consisted of the Patient-determined Disease Steps (PDDS), which attempts to approximate steps of the Expanded Disability Status Scale (EDSS), and 11 single-item performance scales (mobility, hand, vision, fatigue, cognition, bladder/bowel, sensory, spasticity, pain, depression, and tremor). Analyses were restricted to 7,851 patients with nearly complete data. Factor and IRT analyses were used to evaluate the dimensional structure of disability in study patients and to generate IRT scores of global and domain-specific disability expressed on the standard deviation (z-score) metric. We used linear regression to identify patient’s characteristics associated with each disability score and logistic regression to test the hypothesis of independent association of the scores with unemployment.

Results:  The measurement structure of disability was best represented by a bifactor model composed of a general scale of global disability and two uncorrelated scales of extra-variability in physical (mobility, hand, bladder/bowel, spasticity and tremor) and mental disability (cognition, fatigue, pain, depression, vision). In multivariable analysis, disease duration was the only characteristic positively associated with all three disability z-scores. Age at first symptoms was related in complex ways with disability scores. Disease modification therapy was associated with lower scores of general and physical disability, but not of mental disability. Native Americans had higher scores of general disability, and men and African American higher scores of physical disability. All three scores of disability were independent predictors of unemployment.

Conclusions:  Controversies about risk factors for MS progression might be attributable, at least in part, to limitations of the composite outcome measures most frequently used in MS clinical research. IRT and bifactor modeling appear to be promising approaches to improve these measures.