Redefining mortality after pancreatic cancer resection

Abstract

INTRODUCTION: Distinct outcome measures such as in-hospital and 30-day mortality have been used to evaluate pancreatectomy results. We posited that these measures could be compared using national data, providing more precision for evaluating published outcomes after pancreatectomy. METHODS: Patients undergoing resection for pancreatic cancer were identified from the linked SEER-Medicare databases (1991-2002). Mortality was analyzed and trend tests were utilized to evaluate risk of death within ≤60 days of resection and from 60 days to 2 years post-resection. Univariate analysis assessed patient characteristics such as race, gender, marital status, socioeconomic status, hospital teaching status, and complications. RESULTS: One thousand eight hundred forty-seven resected patients were identified: 7.7% (n = 142) died within the first 30 days, 83.6% of whom died during the same hospitalization. Postoperative in-hospital mortality was 8.1% (n = 150), 79% of which was within 30 days, greater than 90% of which was within 60 days. Risk of death decreased significantly over the first 60 days (P < 0.0001). After 60 days, the risk did not decrease through 2 years (P = 0.8533). Univariate analysis showed no difference between the two groups in terms of race, gender, marital status, and socioeconomic status, but patients dying within 60 days were more likely to have experienced a complication (41.1% vs. 17.0%, P < 0.0001). CONCLUSIONS: In-hospital and 30-day mortality after resection for cancer are similar nationally; thus, comparing mortality utilizing these measures is acceptable. After a 60-day post-resection window of increased mortality, mortality risk then continues at a constant rate over 2 years, suggesting that mortality after pancreatectomy is not limited to early ("complication") and late ("cancer") phases. Determining ways to decrease perioperative mortality in the 60-day interval will be critical to improving overall survival.