Theauthors of the paper decided a priori that the data were clinicallyheterogeneous, and used a random effects calculation for statisticalsignificance.