Tuesday, March 1, 2011

Impact of EXPRESS Pagoclone Trial Group Inhomogeneities

In the published results of the EXPRESS (EXamining Pagoclone for peRsistent dEvelopmental Stuttering Study) pagoclone trials*, the pagoclone group and the placebo group were inhomogeneous with respect to at least one characteristic. Specifically, the mean social anxiety of the groups as measured by the Liebowitz Social Anxiety Scale (LSAS) differed--the pagoclone group measured 46, while the placebo group scored 56. So the placebo group tended to have a higher degree of social anxiety.

Given that stuttering is both a function of the brain as well as the mind (anxiety being a mind state), we can attempt on a priori grounds to determine the direction of potential bias of the LSAS inhomogeneity with respect to trial outcomes.

One question that arises concerns a potential relationship between social anxiety and suggestibility. Might individuals with high levels of social anxiety be more suggestible?  For example, a high anxiety individual might be more strongly motivated to hope or believe in a placebo's positive outcome if only as a relief from the discomfort of an anxious mind.  If this were the case, a placebo might have a stronger effect on the higher anxiety placebo group and the difference of the fluency outcomes between the pagoclone and placebo groups would be narrowed. Thus, the efficacy of pagoclone, as measured by the extent that fluency is improved beyond the placebo response, might be biased downward (i.e., underestimated). In addition, the significance level associated with the outcome might also be reduced.

Even if there were no significant correlation between social anxiety and suggestibility, the placebo group, having the higher average LSAS score, might be more responsive to a placebo. In the several previous posts on "The Anxious Mind Affects Stuttering," we emphasized that stuttering is not only a physiological problem associated with the brain, but also a mind problem. A placebo tends to be particularly efficacious for the mind portion of a mind/body problem such as stuttering. So, if this were the case, we might again expect pagoclone efficacy to be underestimated.

The problem of group inhomogeneities is endemic to ANOVA types of analyses, such as was performed in the pagoclone EXPRESS study. Granted that the group numbers in this study were relatively small, but even for larger trials various inhomogeneities may still be a problem.

A possible resolution to this problem might be the use of regression analysis approaches of the type suggested in two of the recent posts discussing a proposal for the analysis of pagoclone trials. The units of observation become the individuals participating in the trial rather than the groups, and the LSAS can be used as an additional explanatory variable in any of the models suggested in those posts. For example, a logarithmic model might be (ignoring the suggestibility variable):

ln(DFA) = a + b*ln(DFB) +c*T + d*ln(LSAS) + e

where the variables DFA, DFB, and T were defined in the post, "A Proposal for Analyzing Pagoclone Trials." In addition, non-linear effects of the LSAS can be explored by entering quadratic terms as well as cross products of LSAS with the other explanatory variables in the model.

*  G. Maguire, et al, Journal of Clinical Psychopharmacology (Feb. 2010), pp. 48-56, Vol. 30, Number 1.

(copyright 2011)

No comments: