Clinically meaningful changes in QL scores

Although most methods of analysis can provide us with a p-value and 95 per cent confidence intervals to indicate the statistical significance of a comparison between two treatments, this inevitably needs to be translated into clinical significance to be useful. For example, given a large enough sample of patients even a 1 per cent difference will be statistically significant, but such a difference would rarely influence practice.

Given that most of the standard questionnaires require individual items to be summed into subscales and their score converted to a 0-100 scale, this poses problems in interpretation, as this is a scale without clinical meaning. For instance what does a score of thirty on emotional functioning mean, and would a patient notice the difference between thirty on one treatment and thirty-five on another treatment? Osoba et al. [27] asked patients to complete not only the EORTC QLQ-C30 on repeated occasions but also to rate their perception of change since the previous assessment. Associating the replies, they found that when the functional scale scores changed by 5-10 points patients described their change as a little better (or worse). A change of 10-20 points was associated with a moderate change, and 20+ points with very much better (or worse). How widely such general definitions can be applied is debatable, as King [28] collating data from fourteen studies, concluded that definitions of large changes differed for different scales. As a corollary, it should be noted that such definitions of change can have major impact on the sample size calculations, as relatively small numbers of patients are required if large differences are expected.

0 0

Post a comment