Defining cut-off values for disease activity states and improvement scores for patient-reported outcomes: the example of the Rheumatoid Arthritis Impact of Disease (RAID)

INTRODUCTION: The Rheumatoid Arthritis Impact of Disease (RAID) is a patient-reported outcome measure evaluating the impact of rheumatoid arthritis (RA) on patient quality of life. It comprises 7 domains that are evaluated as continuous variables from 0 (best) to 10 (worst). The objective was to define and identify cut-off values for disease activity states as well as improvement scores in order to present results at the individual level (for example, patient in acceptable state, improved patient). METHODS: Patients with definite active RA requiring anti-tumour necrosis factor (anti-TNF) therapy were seen at screening, baseline and after 4 and 12 weeks of etanercept therapy. Answers to "Gold standard" questions on improvement (MCII: Minimum Clinically Important Improvement) and an acceptable status (PASS: Patient Acceptable Symptom State) were collected as well as the RAID score and Disease Activity Score 28- erythrocyte sedimentation rate (DAS28-ESR). Cut-offs were defined by different techniques including empirical, measurement error and gold standard anchors. The external validity of these cut-offs was evaluated using the positive likelihood ratio (LR) based on the patient's perspective (for example, patient's global) and on low disease activity status (such as DAS28-ESR). RESULTS: Ninety-seven (97) of the 108 recruited patients (age: 54 ± 13 years old, female gender: 75%, rheumatoid factor positive: 81%, disease duration: 8 ± 7 years, CRP: 18 ± 30 mg/l, DAS28-ESR: 5.4 ± 0.8) completed the 12 weeks of the study. The different techniques suggested thresholds ranging from 0.2 to 3 (absolute change) and from 6 to 50% (relative change) for defining MCII and thresholds from less than 1 to less than 4.2 for defining PASS. The evaluation of external validity (LR+) showed the highest LR+ was obtained with thresholds of 3 for absolute change; 50% for relative change and less than 2 for an acceptable status. CONCLUSIONS: This study showed that thresholds defined for continuous variables are closely related to the methodological technique, justifying a systematic evaluation of their validity. Our results suggested that a change of at least 3 points (absolute) or 50% (relative) in the RAID score should be used to define a MCII and that a maximal value of 2 defines an acceptable status. TRIAL REGISTRATION: Clinicaltrial.gov: NCT004768053