Prabakaran Viswakumar, Thangaraju Thamizhmathi, Mathew Anil C, Govindan Vimalkumar, Kannan Vignesh, Poulose Tracy Rosalin
Department of General and GI Surgery, PSG Institute of Medical Science and Research, (Affiliated to Tamilnadu Dr. MGR Medical University), Peelamedu, Coimbatore, Tamilnadu 641004 India.
Division of Biostatistics, Department of Community Medicine, PSG Institute of Medical Science and Research, (Affiliated to TamilNadu Dr. MGR Medical University), Coimbatore, India.
Indian J Surg Oncol. 2019 Mar;10(1):174-179. doi: 10.1007/s13193-018-0841-8. Epub 2018 Dec 17.
Preoperative prediction of morbidity in colorectal cancer (CRC) surgery helps to optimize the surgical outcome. In this study, we aim to develop a dedicated equation for predicting operative morbidity using colorectal possum scoring system and also to validate the predictive accuracy of CR-POSSUM scoring system in prognosticating actual complications. We did a retrospective analysis of 322 patients undergoing colorectal cancer surgery from a single centre in South India from 2004 to 2016. Mortality and morbidity risk factors as defined by CR POSSUM were collected from 322 patients who underwent CRC surgery and were used to derive equations to predict morbidity, and the results were compared with the observed morbidity. Logistic regression analysis was used to derive the equation. The model fit and model discrimination were analysed using the Hosmer-Lemeshow statistical test for goodness of fit, the Nagelkerke and area under the receiver operating characteristic (ROC) curve respectively. Out of 322 patients, 103 (32%) patients developed complications and 10 (3%) died due to complications. The regression equation we derived has an overall correct classification of about 70% ( < 0.01) with positive and negative predictive value of 60% and 73% respectively. The Hosmer-Lemeshow goodness of fit was 3.147 ( = 0.829), and the Nagelkerke was 17% and area under ROC as model discrimination was 71.6%. Hence, CR-POSSUM scoring which was originally used for predicting mortality risk can also be extrapolated to predict morbidity.
结直肠癌（CRC）手术术前对发病情况的预测有助于优化手术结果。在本研究中，我们旨在使用结直肠possum评分系统开发一个用于预测手术发病情况的专用方程，并验证CR-POSSUM评分系统在预测实际并发症方面的预测准确性。我们对2004年至2016年期间在印度南部一个中心接受结直肠癌手术的322例患者进行了回顾性分析。从322例行CRC手术的患者中收集CR POSSUM定义的死亡率和发病风险因素，并用于推导预测发病情况的方程，将结果与观察到的发病情况进行比较。使用逻辑回归分析来推导方程。分别使用Hosmer-Lemeshow拟合优度统计检验、Nagelkerke 和受试者工作特征（ROC）曲线下面积来分析模型拟合和模型判别。在322例患者中，103例（32%）出现并发症，10例（3%）因并发症死亡。我们推导的回归方程总体正确分类率约为70%（<0.01），阳性预测值和阴性预测值分别为60%和73%。Hosmer-Lemeshow拟合优度为3.147（=0.829），Nagelkerke 为17%，作为模型判别的ROC曲线下面积为71.6%。因此，最初用于预测死亡风险的CR-POSSUM评分也可外推用于预测发病情况。
