Establishment and application of Nomogram model for prediction of recurrence after breast cancer surgery based on multi-dimensional indicators

Objective To establish a Nomogram model for prediction of recurrence after breast cancer surgery based on multi-dimensional indicators.

Methods A total of 313 female patients with surgical treatment for unilateral breast cancer from March 2014 to April 2020 in authors' hospital were retrospectively selected as research objects, and they were divided into modeling set with 219 cases and validation set with 94 cases by the random number table method at a ratio of 7 to 3. In the modeling set, the patients were divided into recurrence group and non-recurrence group according to recurrence condition during follow-up after operation. Single factor and multi-factor Logistic regression models were used to analyze the risk factors of postoperative recurrence of breast cancer. Based on the screening results, the R software was used to draw the Nomogram prediction model. The performance of the model was evaluated by the receiver operating characteristic (ROC) curve and deviation test of goodness of fit. The calibration map was verified and drawn by computer-simulated repeated sampling method (Bootstrap); the decision curve was used to evaluate the clinical benefit rate of the model.

Results Among the 219 case in the modeling set, 63 cases (28.77%) had recurrence of breast cancer after surgery. Irregular tumor margin (OR=1.692, 95%CI, 1.154 to 3.794), high maximum elastic modulus of tissues around 3 mm from focus measured by Shell function and tracing method (Shell3 E_max) (OR=2.869, 95%CI, 1.795 to 5.392), lymph node metastasis (OR=2.071, 95%CI, 1.486 to 4.578), high expression of D-dimer (OR=2.264, 95%CI, 1.574 to 5.307) and high ratio of fibrinogen to albumin (OR=3.089, 95%CI, 2.053 to 6.156) were risk factors for postoperative recurrence of patients with breast cancer (P < 0.05). Nomogram model for risk prediction was established based on the five factors mentioned above, the area under the curve of the ROC curve of the model was 0.872 (95%CI, 0.829 to 0.917), the best cut-off value (threshold probability) was 0.32 (32%), and the corresponding sensitivity and specificity were 0.871 and 0.837 respectively; the test of goodness of fit showed that there was no over-fitting phenomenon in the prediction model (χ²=4.204, P=0.826); the Bootstrap method based on 1 000 sampling verifications found that the average absolute error of the calibration curve was 0.019, indicating that the prediction model had a good consistency. The area under the curve of the ROC curve of the Nomogram model for prediction of validation set was 0.864, the sensitivity was 0.862, and the specificity was 0.815; the correction curve was close to the ideal curve. When the threshold probability value in the decision curve was set to 32%, the clinical benefit rates of the population in the modeling set and the validation set were 56% and 62% respectively.

Conclusion Nomogram model based on tumor margin, Shell3 E_max value, condition of lymph node metastasis, serum D-dimer and the ratio of fibrinogen to albumin of breast cancer patients has a certain value in predicting the risk of postoperative recurrence.