Interpretation of changes in HRQL outcomes: The relationship between distribution- and anchor-based approaches

Approaches to assessment of the impact of treatments for chronic diseases on health-related quality of life (HRQL) in clinical trials are increasing. However, it is difficult to interpret the clinical relevance of changes in HRQL scores. Two basic strategies are: 1) methods that rely on the statistical distribution of changes, such as the effect size (ES); and 2) methods that use some external anchor, such as patient judgement of change, which is then used to compute a minimum clinically important difference (MCID), the proportion benefiting from treatment, p(B), and the number needed to treat. Both approaches involve inherent methodological limitations. There is no evidence that the magnitude of change, identified by ES calculation, is of any clinical relevance. Also, while the simplicity of using an MCID to interpret statistically significant HRQL outcomes may be appealing, this approach relates change in HRQL scores to patients' estimate of their rating of change on a global scale, and the use of retrospective global judgements of change has been criticised. At the same time, the two methods may not be that distinct. Concordance between the ES and MCID methods has been observed, and it has been demonstrated that there is a near-linear relationship between ES and p(B). This relationship is nearly independent of the choice of MCID, which suggests that using the MCID to divide a distribution of changes into important and unimportant changes may be inappropriate. As the science of health-related quality of life develops, it is essential to include patients' and clinicians' perspectives to define clinically meaningful change in health-related quality of life, while concurrently ensuring that methods follow sound statistical principles.