“…We excluded 14 trials because endothelial function was not evaluated by brachial FMD (n = 9) [43,44,45,46,47,48,49,50,51] or because FMD data were not available at baseline or at the end of treatment (n = 5; fig. 1) [52,53,54,55,56]. Among the 26 trials remaining in the present meta-analysis, 9 had a crossover design [16,17,18,22,24,27,28,32,40] and 17 had a parallel design [19,20,21,23,25,26,29,30,31,33,34,35,36,37,38,39,41]; 11 had a placebo-controlled design [17,18,19,22,24,25,27,28,31,35,38], 13 had an active drug-controlled design [16,20,21,23,26,29,30,33,34,36,39,40,41], and 2 had both of them [32,37].…”