Design of comparative clinical studies of percutaneous transluminal coronary angioplasty using estimates from the coronary artery surgery study

The design of a prospective randomized trial of PTCA is discussed. It is suggested that patients with proximal subtotal occlusion of 1 or more vessels may be candidates for such a trial. Medical therapy as a control of 1-vessel CAD, medical or surgical therapy as a control for 2-vessel CAD (with 1 dilatable vessel) cases and surgical therapy as a control for 3-vessel CAD are most congruent with current practice. Possible criteria for comparison of PTCA with a control therapy include vital status, angina with daily activities, angina upon maximal exercise testing, MI, hospitalization, work status, cost, follow-up angiographic assessment, follow-up radionuclide ventriculography, measurement of immediate anatomic or hemodynamic change during initial treatment and quality-of-life indexes. These 10 measures are evaluated with respect to 9 desirable attributes for outcome criteria: the importance of the quantity measured, reproducibility, accuracy, low influence of investigator bias, low influence of subject bias, the measure is inexpensive and easy to obtain, sufficient occurrence to detect important differences, convenient to patients and measuring impairment due to disease. Examination of data from the Coronary Artery Surgery Study suggests that (1) a trial based on mortality alone is not feasible (because of a needed sample size in the thousands); (2) a trial using a combined death and/or MI end point would need approximately 1,000 cases even for patients with multivessel CAD; (3) approximately 950 patients would be needed in a trial to evaluate return to work (within 10%); (4) a trial based on pain relief (within 15% of the control group) would need 500 subjects. The drawback is the potential bias because of lack of blinding. Maximal exercise testing would help to lower the danger of bias; (5) if a trial is performed, it is likely to protect only against very large odds ratio differences in death and MI. Pain relief, hospitalization and cost data would likely provide the criteria used to comparatively evaluate the therapies.