Cumulative sum failure analysis of a policy change from on-pump to off-pump coronary artery bypass grafting

Background. Use of the sequential probability cumulative sum (CUSUM) techni que may be more sensitive than standard statistical analyses in detecting a cluster of surgical failures. We applied CUSUM methods to evaluate the lea rning curve after a policy change by a single surgeon from routine on-pump (cardiopulmonary bypass [CPB]) to off-pump coronary artery bypass grafting (OPCAB). Methods. Fifty-five consecutive first-time coronary artery bypass patients (CPB group) were compared with the next 55 patients undergoing an attempt a t routine OPCAB using the same coronary stabilizer. The goal in OPCAB patie nts was to obtain complete revascularization, albeit with a low threshold f or conversion to CPB to maximize patient safety during the learning curve. Preoperative patient risk was calculated using previously validated models of the Cardiac Care Network of Ontario. The occurrence of operative mortali ty and nine predefined major complications (myocardial infarction, bleeding , stroke, renal failure, balloon pump use, mediastinitis, respiratory failu re, life-threatening arrhythmia, and sepsis) was compared between the CPB a nd OPCAB groups using Wilcoxon, Fisher exact, and two-tailed t tests, as we ll as CUSUM methodology. An intention to treat analysis was performed. Results. The CPB and OPCAB groups had similar predicted mortality and lengt h of stays (2.2% +/- 2.5%, 8.1 +/- 2.5 days versus 2.4% +/- 3.5%, 8.1 +/- 2 .4 days, respectively). The mean number of grafts per patient was 8.1 +/- 0 .7 in the CPB group versus 3.0 +/- 0.7 in the OPCAB group (p = 0.45). Two o f 55 (3.6%) CPB patients died, as opposed to I of 55 (1.8%) OPCAB patients (p = 0.99). Eight of 55 CPB patients (14.5%) incurred major complications, as opposed to 4 of 55 (7.3%) OPCAB patients (p = 0.36). Median hospital len gth of stay was 6.0 days in the CPB group versus 5.0 days in the OPCAB grou p (p = 0.28). On CUSUM analysis, the failure curve in CPB patients approach ed the upper 80% alert line after eight cases, whereas the curve in OPCAB p atients reached below the lower 80% (reassurance) boundary 28 cases after t he policy change, indicating superior results in the OPCAB group despite th e learning curve. Conclusions. A policy change from coronary artery bypass on CPB to routinel y attempting OPCAB can be accomplished safely despite the learning curve. C USUM analysis was more sensitive than standard statistical methods in detec ting a cluster of surgical failures and successes. (C) 2001 by The Society of Thoracic Surgeons.