Preoperative Albumin to Alkaline Phosphatase Ratio and Inflammatory Burden Index for Rectal Cancer Prognostic Nomogram-Construction: Based on Multiple Machine Learning

Abstract

Purpose: Preoperative albumin to alkaline phosphatase ratio (AAPR) and inflammatory burden index (IBI) are prognostic indicators for a multitude of cancers, and our study focuses on evaluating the prognostic significance of the AAPR and the IBI on rectal cancer (RC) patients to provide a more accurate guideline for patient prognosis.

Patients and methods: This study enrolled patients who underwent laparoscopic rectal cancer surgery from January 2016 to January 2021. We utilized three machine learning approaches to select variables most relevant to prognosis in the training cohort. Finally, based on the screened variables, a nomogram was established to predict RC patients' overall survival (OS). The improvement in predictive ability and clinical benefit was assessed through the concordance index (C-index), receiver operating characteristic (ROC), calibration curve, and decision curve analysis (DCA).

Results: A total of 356 patients were enrolled and they were randomly divided into a training cohort (60%, n=214) and a validation cohort (40%, n=143). Overall survival (OS) was worse for patients in either the low AAPR or the high AAPR group, whereas patients in the low AAPR with both high IBI group had the lowest OS (P<0.001). Finally, five variables were obtained after screening the best variables by three machine learning, and the nomogram was constructed. In both the development and validation cohorts, the C-index values exceeded 0.85, indicating that the predictive model has a strong predictive performance in terms of overall survival. The calibration curves and the decision curve analysis (DCA) showed that the nomogram demonstrated a superior benefit.

Conclusion: Preoperative AAPR and IBI can serve as effective indicators for predicting the OS of RC patients. We have developed a nomogram for predicting the OS of patients who underwent laparoscopic rectal cancer surgery.

Keywords: albumin to alkaline phosphatase ratio; inflammatory burden index; overall survival; prediction nomogram; rectal cancer.

Figure 1

Flow chart of study design.

Figure 2

(A) Kaplan-Meier survival curves for overall survival (OS) of rectal cancer patients classified according to AAPR cut-off values (P < 0.001). (B) Kaplan-Meier survival curves for OS of rectal cancer patients classified according to IBI cut-off values (P < 0.001). (C) Kaplan-Meier survival curves for OS of rectal cancer patients comparing different groups (P < 0.001).

Figure 3

Three important feature selection methods in ML are (A) Lasso regression; (B) Xgboost; and (C) Random Forest.

Figure 4

A Venn diagram showing three important feature selection methods in ML, with blue representing Lasso regression, yellow representing Xgboost, and pink representing Random Forest.

Figure 5

Nomogram model predicting 1-, 3- and 5-year OS in patients with RC.

Figure 6

The ROC curves of the nomogram for predicting 1-year, 3-year, and 5-year overall survival (OS) in rectal cancer patients in the training cohort (A), and the ROC curves of the nomogram for predicting 1-year, 3-year, and 5-year OS in rectal cancer patients in the validation cohort (B).

Figure 7

The calibration curves for 1-year, 3-year, and 5-year overall survival (OS) in the training cohort (A) and the calibration curves for 1-year, 3-year, and 5-year OS in the validation cohort (B). The solid line indicates the performance of the prediction model, and the closer to the diagonal dashed line, the more accurate the prediction.

Figure 8

The DCA curves for 1- (A), 3- (B), and 5-year (C) OS in the training cohort, as well as the DCA curves for 1- (D), 3 (E), and 5-year (F) OS in the validation cohort.