The Spectrum of Functional Rating Scales in Neurology Clinical Trials

Abstract

The selection of an appropriate outcome measure is crucial to the success of a clinical trial, in order to obtain accurate results, which, in turn, influence patient care and future research. Outcomes that can be directly measured are mortality/survival. More frequently, neurology clinical trials evaluate outcomes that cannot be directly measured, such as disability, cognitive function, or change in symptoms of the condition under study. These complex outcomes are abstract ideas or latent constructs and are measured using rating scales. Functional rating scales typically assess the ability of patients to perform tasks and roles for everyday life. Rating scales should be valid (measure what they are supposed to measure), reliable (provide similar results if administered under the same conditions), and responsive (able to detect clinically important changes over time). The clinical relevance of rating scales depends on their ability to detect a minimal clinically important difference, and should be distinguished from statistical significance. Most rating scales are ordinal scales and have limitations. Modern psychometric methods of Rasch analysis and item response theory, termed latent trait theory, are increasingly being utilized to convert ordinal data to interval measurements, both to validate existing scales and to develop new scales. Patient-reported outcomes are being increasingly used in clinical trials and have a role in clinical quality assessment. The PROMIS and NeuroQoL databases are excellent resources for rigorously developed and validated patient-reported outcomes.

Keywords: Functional rating scales; Rasch analysis; classical test theory; clinimetrics; item response theory; outcome measures.

Fig. 1

Random error and systematic error. The black center represents the “truth”. The individual dots represent the results of studies. (A) The dots do not fall on the black center is because of a small degree of random error. (B) The dots are more widely dispersed because of a larger degree of random error. (C, D) The dots or results of individual studies are off target because of systematic error. In (C) there is a small degree of random error, while (D) shows a larger degree of random error (from Gary Gronseth, MD, with permission)

Fig. 2

Rasch model interval “ruler”. The Rasch model conceptualizes a measurement scale where subjects are ranked by their ability from low to high, and items on a rating scale are ordered by the level of difficulty from easiest to most difficult. The thick lines on the ruler indicate the location of items of increasing difficulty from left to right. (A), (B), and (C) represent 3 subjects with increasing levels of physical ability