# Predicting lung metastasis after radical colorectal cancer surgery: a two center retrospective cohort study and nomogram development

## TL;DR

This study develops a predictive model to identify colorectal cancer patients at high risk of lung metastasis after surgery, helping guide personalized treatment.

## Contribution

A novel nomogram model using clinical and pathological factors to predict lung metastasis in CRC patients.

## Key findings

- The model incorporated CEA, N stage, perineural invasion, and surgical approach as independent predictors.
- The model showed strong discrimination with AUCs of 0.785 in training and 0.779 in validation cohorts.
- Patients stratified by risk scores had significantly different lung metastasis-free survival (LMFS).

## Abstract

Colorectal cancer (CRC) is a leading cause of cancer-related death, with distant metastasis being the primary driver of mortality. Lung metastases are among the most frequent sites of extra-abdominal metastasis in CRC patients. Early identification of high-risk patients for pulmonary metastasis after radical surgery is critical for timely intervention and potentially improved patient outcomes.

We analyzed a cohort of 399 CRC patients who underwent radical surgery and followed their postoperative imaging outcomes. We developed a predictive model using univariate and multivariate logistic regression analyses based on clinical and pathological characteristics. Patients were stratified into two groups based on the median value of the logistic regression-derived risk score. Subsequently, Kaplan-Meier survival analysis was performed to compare survival curves between the two groups.

The predictive model incorporated four independent factors: carcinoembryonic antigen (CEA), N stage, perineural invasion, and surgical approach. The nomogram demonstrated strong discrimination and calibration in both the training and validation cohorts. The training cohort achieved an area under the curve (AUC) of 0.785 (95% CI: 0.725-0.845), while the validation set demonstrated an AUC of 0.779 (95% CI: 0.696-0.863). The sensitivity and specificity were 73.3% and 71.8%, respectively. LMFS was significantly different (P < 0.001) in both based on the model scores. Decision Curve Analysis (DCA) indicated the potential clinical utility of the model.

Our predictive model can assist clinicians in identifying high-risk patients prone to developing pulmonary metastasis following radical resection of colorectal cancer. This facilitates personalized treatment strategies, potentially leading to improved prognosis for CRC patients.