Constant Graded Variability ICD (within-subject) 40% 20% ACD (within-subject) 30% 12% ACD (between-subjects) 60% 30% Placebo response ICD 36% 0% ACD 25% 0% Dynamic range Limited Large Entry exclusions 20% 0% Exit exclusions 45% 0% Equipment Basic Complex Patient acceptance Good Good No. of patients per group 25% change 98 22

Estimates of treadmill variability (expressed as the coefficient of variation) and placebo response were taken from published reports of the different treadmill testing methodologies,^{16 20} including a direct comparison between the constant and graded protocols.²² ACD is the absolute (maximal) claudication distance and ICD is the initial claudication distance. The estimated number of entry and exit exclusions (percent of patients no longer limited by claudication within dynamic range of test) were taken from published reports.^{12 23 24} The estimates of sample size were based on a two-tailed test, P<.05, and power of .80 for detecting a 25% difference between treated and control subjects, based on published data.^{10 23}