The value of a risk model combining specific risk factors for predicting postoperative severe morbidity in biliary tract cancer

Abstract

Purpose: Several surgical risk models are widely utilized in general surgery to predict postoperative morbidity. However, no studies have been undertaken to examine the predictive efficacy of these models in biliary tract cancer patients, and other perioperative variables can also influence morbidity. As a result, the study's goal was to examine these models alone, as well as risk models combined with disease-specific factors, in predicting severe complications.

Methods: A retrospective study of 129 patients was carried out. Data on demographics, surgery, and outcomes were gathered. These model equations were used to determine the morbidity risks. Severe morbidity was defined as the complication comprehensive index ≥ 40.

Results: Severe morbidity was observed in 25% (32/129) patients. Multivariate analysis demonstrated that four parameters [comprehensive risk score ≥1, T stage, albumin decrease value, and international normalized ratio (INR)] had a significant influence on the probability of major complications. The area under the curve (AUC) of combining the four parameters was assessed as having strong predictive value and was superior to the Estimation of Physiologic Ability and Surgical Stress System (E-PASS) alone (the AUC value was 0.858 vs. 0.724, p = 0.0375). The AUC for the modified E-PASS (mE-PASS) and Physiological and Operative Severity Score for the Enumeration of Mortality and Morbidity (POSSUM) in patients over the age of 70 was classified as no predictive value (p = 0.217 and p = 0.063, respectively).

Conclusion: The mE-PASS and POSSUM models are ineffective in predicting postoperative morbidity in patients above the age of 70. In biliary tract cancer (BTC) patients undergoing radical operation, a combination of E-PASS and perioperative parameters generates a reasonable prediction value for severe complications.

Keywords: biliary tract cancer; morbidity; risk prediction models; risk score; specific risk factors.

Figure 1

Flowchart of inclusion and exclusion criteria in this study.

Figure 2

ROC curve analysis of severe morbidity prediction according to CCI in BTC patients with age ≥ 70 years. Yellow line indicates ROC plot for predictive value of E-PASS model alone, AUC 0.772, p = 0.005. Red line indicates ROC plot for predictive value of mE-PASS model alone, AUC 0.620, p = 0.217. Green line indicates ROC plot for predictive value of POSSUM model alone, AUC 0.680, p = 0.063. ROC curve, receiver operating characteristic curve; CCI, Comprehensive Complication Index; AUC, the area under the curve; E-PASS, Estimation of Physiologic Ability and Surgical Stress System; mE-PASS, modified Estimation of Physiologic Ability and Surgical Stress System; POSSUM, Physiological and Operative Severity Score for the Enumeration of Mortality and Morbidity.

Figure 3

ROC curve analysis of severe morbidity prediction according to CCI in all patients. Green line indicates ROC plot for predictive value of CRS ≥ 1 combined with disease-specific factors including T stage, Δalbumin, and INR, AUC 0.844. Red line indicates ROC plot for predictive value of E-PASS model alone, AUC 0.724 (p = 0.0375). CCI, complication comprehensive index; CRS, comprehensive risk score; Δalbumin, the reduced values between the first day after surgery and pre-operation; INR, international normalized ratio.

Figure 4

Calibration curve for CRS ≥ 1 + specific factors score for BTC patients with radical operation. Specific factors include T stage, Δalbumin (the reduced values between the first day after surgery and pre-operation), and international normalized ratio. CRS, comprehensive risk score; BTC, biliary tract cancer.