[Establishment of nomogram model to predict peritoneal metastasis in colon cancer patients without distant metastasis by preoperative imaging examination].

OBJECTIVE

To establish a nomogram model to predict the peritoneal metastasis in colon cancer patients without distant metastasis by preoperative imaging examination.

METHODS

Clinicopathological data of colon cancer patients without distant metastasis by preoperative imaging examination who underwent surgery in our department between January 2000 and December 2014 were retrospectively analyzed. Predictors of peritoneal carcinomatosis were analyzed by univariate and Logistic multivariate analyses. Base on the independent predictors by multivariable analysis results, a nomogram model was formulated with further use of R software. The total score was calculated by the addition of each predictor score, indicating the corresponding risk of peritoneal metastasis. The score was greater in the nomogram, and the risk was higher in peritoneal implantation metastasis. A receiver operating characteristic(ROC) curve was then constructed to evaluate the predictive abilities of the various preoperative factors and nomogram.

RESULTS

A total of 1 417 patients were defined as above and enrolled in the study. The median age was (60.5±13.3) years, 835 cases (58.9%) were male, and 132 cases (9.3%, 132/1417) were diagnosed with synchronous peritoneal carcinomatosis during operation. Univariate analysis showed that peritoneal metastasis was associated with age, incidence of abdominal pain, incidence of mucous bloody stool, CEA level, traversible rate, tumor diameter, ratio of infiltrating type cancer, differentiation, histological type, cT staging and cN staging (all P<0.05). Logistic multivariate analysis revealed that younger age (OR:0.974, 95%CI: 0.958 to 0.990, P=0.001), later clinical T stage (OR: 2.949, 95%CI: 1.588 to 5.476, P=0.001), lesion not traversible(OR: 0.519, 95%CI: 0.314 to 0.858, P=0.011), infiltrative gross type (OR: 1.812, 95%CI: 1.099 to 2.987, P=0.020), larger tumor (OR: 1.044, 95%CI: 0.998 to 1.093, P=0.061), higher preoperative serum CEA level(OR:1.004,95%CI: 1.001 to 1.007, P=0.007) and histopathologic type of mucinous or signet ring cell adenocarcinoma (OR:1.642, 95%CI: 1.009 to 2.673, P=0.046) were independent risk factors. The nomogram model was further established based on above 7 independent risk factors, whose total score was 350 and area under the ROC curve was 0.753(P=0.000).

CONCLUSION

The nomogram model can be helpful to screen the colon cancer patients with high risk of peritoneal metastasis and to avoid unnecessary laparotomy for colon cancer patients without distant metastasis by preoperative imaging examination.