ESTIMATES OF MINIMAL CLINICALLY IMPORTANT IMPROVMENTS VARY WITH THE RESPONSIVENESS OF THE SAMPLE

ABSTRACT Objective. Minimal clinically important improvements (MCII) are known to vary with the baseline level in the sample. We examined if MCIIs are also larger in samples with higher responsiveness. Study design and setting. In a prospective longitudinal study of patients with active rheumatoid arthritis, we assessed arthritis activity before and after new treatments. We estimated anchor-based MCIIs for three outcomes (pain severity, physical functioning by the Health Assessment Questionnaire, and Simplified Disease Activity Index, a composite measure) using receiver operating characteristic curves. We compared MCIIs among patients treated with three interventions of different impact (dose escalation, new disease-modifying medication, or prednisone). Separately, we used simulations to estimate MCIIs in five groups of responsiveness. Results. Among 250 patients, standardized response means (SRMs) increased across the dose escalation, disease-modifying treatment, and prednisone treatment groups (-0.74, -1.00, and -1.53, respectively). MCIIs were also highest in the prednisone group. For example, corresponding MCIIs were -5.5, -8.9, and -13.8 for the composite measure. In the simulations, MCIIs (range -4.6 to -11.9) varied directly with SRMs (range -0.40 to -1.33). Results were similar for pain and the Health Assessment Questionnaire. Conclusion. The MCII is not an intrinsic measurement property but varies directly with sample responsiveness.