What is new?
- •
This is the first study to test if patients who comprise the subgroup with a high minimal clinically important improvements (MCIIs) on one measure of disease activity also constitute subgroups with high MCIIs for other measures of disease activity.
- •
Comparing the concordance of patient responses across measures, MCII in the high baseline subgroups segregate with the measure rather than with particular patients.
- •
Variation in the MCII with the baseline value is attributable to differences in maximum possible changes and opportunities for misjudgments at different baseline values, and was similar for three measures of rheumatoid arthritis activity.
- •
A single MCII can be estimated for all patients provided they meet a minimum level of disease activity or severity.
Patient-reported outcomes have become recognized as central components in the assessment of health, and are now routinely included as endpoints in clinical trials and observational studies. Although the comparison of responses between treatment groups provides an estimate of the effects of treatment, this comparison does not provide information on whether the improvement was substantial or trivial. Full interpretation requires knowing what degree of change in a measure represents an important or clinically meaningful change, and whether a higher proportion of patients in one group met this threshold [1]. In addition to facilitating the interpretation of trial results, the minimal clinically important improvement (MCII) of a study's primary outcome is important in study design as a guide to sample size estimation. Although the MCII has most often been assessed for patient-reported outcomes, similar issues pertain to measures that are not patient reported.
Of several approaches used to estimate the MCII, anchor-based methods are the most direct and frequently use the patient's explicit judgment of improvement as an external standard [2]. Most often, investigators intend to determine a single MCII for a given measure. However, the MCII may vary with the analytic approach or the nature of patients assessed [3], [4]. Several studies have examined potential sources of variation in the MCII, including for example whether the MCII was similar for men and women, as an indication of whether group-specific MCIIs were needed [5]. A notable observation has been that when patients are stratified by their value on the measure at study baseline, estimates of the MCII are substantially larger for subgroups of patients with high baseline values (or values indicating more severe disease) than for subgroups with midrange or low values (or values indicating less severe disease). Dependence of the MCII on the baseline value was observed in each of the 27 studies we identified that examined the baseline value as a source of variation in the MCII [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31]. This dependence was irrespective of the nature of the outcome, which ranged from pain scales and functional indices to urinary symptom scales, and irrespective of the format of the measure, suggesting that it may be axiomatic.
This association has commonly been interpreted to indicate that patients with more severe symptoms require a larger improvement to appreciate that they are better than those with less severe symptoms. Although this interpretation is logical, the universality of this association across studies, conditions, and measures suggests that the dependency of the MCII on the baseline value may be a consequence of the measurement process, rather than a truism of how patients perceive health changes. Most measures are bounded, and improvements, by definition, are unidirectional. Floor and ceiling effects have been invoked as possibly contributing to this observation, but this possibility has not been explored in detail [8], [19], [32]. In this study, we examined whether floor and ceiling effects might account for the baseline dependency of the MCII in a study of patients with active rheumatoid arthritis (RA). We examined three different measures of RA activity, namely the patient global assessment, a widely used patient-reported measure of overall arthritis activity; the swollen joint count, a physician-derived measure; and walking time, a performance measure. In addition to testing if the MCII varied with the baseline value of each measure, we examined if the same subset of patients was identified as having a high MCII for each RA activity measure. We hypothesized that if baseline dependency of the MCII was owing to the “requirements” or judgments of a particular subgroup of patients, the same subgroup should be identified by each measure of RA activity. In contrast, if each measure identified different sets of patients as having a high MCII, the baseline dependency of the MCII would relate to the measure rather than to the requirements or expectations of particular patients.