Development and validation of a collagen signature-based nomogram for preoperatively predicting lymph node metastasis and prognosis in colorectal cancer

Background: Current preoperative evaluation approaches cannot provide adequate information for the prediction of lymph node (LN) metastasis in colorectal cancer (CRC). Collagen alterations in the tumor microenvironment affect the progression of tumor cells. To more accurately assess the LN status of CRC preoperatively, we developed and validated a collagen signature-based nomogram for predicting LN metastasis in CRC. Methods: In total, 342 consecutive CRC patients were assigned to the training and validation cohorts. A total of 148 fully quantitative collagen features were extracted based on preoperative biopsies using multiphoton imaging, and the least absolute shrinkage and selection operator method was utilized to construct the collagen signature. A collagen signature-based nomogram was developed by multivariable logistic regression in the training cohort. Nomogram performance was evaluated for its discrimination, calibration, and clinical usefulness and then validated in the validation cohort. The prognostic values of the nomogram were also evaluated. Results: A seven-feature-based collagen signature was built. We found that the collagen signature showed a significant association with LN metastasis in CRC. Additionally, a nomogram incorporating preoperative tumor differentiation, computed tomography-reported T stage and LN status, carcinoembryonic antigen level, carbohydrate antigen 19-9 level and collagen signature was developed. This nomogram had good discrimination and calibration, with AUROCs of 0.826 and 0.846 in the training and validation cohorts, respectively, and had a sensitivity of 86.5%, a specificity of 68.2%, an accuracy of 76.9%, a negative predictive value of 84.9%, and a positive predictive value of 71.2% for all patients. Compared to the clinicopathological model, which consisted of the clinicopathological risk factors for LN metastasis, the collagen signature-based nomogram demonstrated a significantly improved ability to discriminate LN status. Moreover, a nomogram-predicted high-risk subgroup had remarkably reduced survival compared with that of the low-risk subgroup. Conclusions: The collagen signature in the tumor microenvironment of preoperative biopsies is an independent predictor for LN metastasis in CRC, and the collagen signature-based nomogram is helpful for tailored treatment and prognostic predictions in CRC preoperatively.