Purpose In a companion paper, we found no statistical reason to favor percentage syllables stuttered (%SS) over parent-reported stuttering severity as a primary outcome measure for clinical trials of early stuttering. Hence, considering the logistical advantages of the latter measure, we recommended parent-reported stuttering severity for use as an outcome measure. The present report extends the prior analysis to a comparison of %SS with self-reported stuttering severity (SRSS) for use as an outcome measure in clinical trials of stuttering treatments for adults. Method We analyzed data from four randomized clinical trials for adults that incorporated %SS and SRSS data at prerandomization and at 6 months post randomization. We analyzed the distributions associated with the two measures, their agreement, and their estimates of effect sizes. Results The positively skewed distribution of %SS warrants much reservation about its value as a clinical trial outcome measure. This skew causes inherent instability because of spurious data associated with low scores, which occur commonly at the low end of such a distribution. This inherent instability is compounded by inherent problems with absolute reliability of %SS measures. These problems are reduced with the much more normal distribution of SRSS. Conclusions The logistical arguments in favor of SRSS apply similarly to adults as they do when parents report the stuttering severity of their children. However, there are statistical reasons to favor SRSS over %SS measures as a primary outcome of clinical trials with adult participants: SRSS has acceptable discriminant validity and a normal distribution, and it is less error prone than %SS. We recommend SRSS as a primary outcome for clinical trials of adults with stuttering.
Read full abstract