To the Editor:

Polkey and colleagues attempt to address the important topic of determining the minimal clinically important difference (MCID) for 6-minute-walk distance in patients with chronic obstructive pulmonary disease (COPD) (1). Although their observations are important, this interesting study brings up a few issues that deserve comment.

Although the association of change in walk distance with mortality is important, it is difficult to accept death as the defining feature of “MCID.” When the magnitude of change in walk distance due to an intervention crosses a threshold value, it becomes large enough to be perceptible by the patient. This threshold or minimal change is the MCID. It is valuable for health outcomes researchers and clinical trialists, who are interested in identifying therapies for COPD that lead to a noticeable change in their patients’ exercise capacity. Because there is no outcome worse than death, a change in walk distance that predicts death is a maximal rather than minimal clinically important difference.

In the authors’ analysis, what were the hazard ratios for death, with 95% confidence intervals, for 20 and 25 m changes in walk distance? If these shorter/minimal distances were significant predictors of death, then they should have been the “MCID” as defined by the authors. If they were not significant, it is likely they had a hazard ratio well above 1, and their P value would have been below 0.05 if the sample size had been larger. Therefore, the authors’ methodology of estimating MCID will yield estimates that are a function of statistical power, rather than a property of the 6-minute-walk test. In contrast, anchor-based methods estimate MCID by calculating the change in walk distance that is statistically equivalent to change in a quality-of-life measure. These estimates will not decline with increasing sample size, and will yield the same result in any study that has requisite statistical power.

Also, the author’s analysis is inherently time dependent, unlike anchor- and distribution-based methods, which have been used to estimate MCID for COPD successfully in the past (2). If a decline in walk distance of 30 m over 1 year is considered the MCID, what would the MCID be for 3- and 6-month time intervals: 7.5 and 15 m, respectively, or 30 m?

In conclusion, we applaud the authors for demonstrating that prior estimates of MCID around 30 m by investigators in pulmonary hypertension (3), primary fibrosis (4), and COPD (2), additionally predict death. However, we are hesitant to conclude that their methodology represents a novel definition or estimate of the minimal clinically important difference; rather, we think it represents a time- and sample size–dependent maximal important difference.