the associations between demographic and clinical variables and 6MWT performance were assessed by a pearson coefficient of correlation or by analysis of variance with post hoc multiple comparison, depending on whether the variables were continuous or categorical, respectively. Variables found to e significant in univariate analysis were the further analyzed by a multivariate linear regression model in order to identify a set of variable independently associated with 6MWT performance. Besides main effect, we also tested for possible interaction effects between pairs of variable. Reference cross-tabulation tables for 6MWT performance were obtained by grouping the patients for the study according to the value of the variable selected by the multivariate model, and computing relevant 6MWT statistics for each group. As some classification variables were continuous, they were first categorized into two or three levels.