A nomogram to predict the high-risk RS in HR+/HER2-breast cancer patients older than 50 years of age

Abstract

Background: The 21-gene recurrence score (RS) testing can predict the prognosis for luminal breast cancer patients. Meanwhile, patients > 50 years with RS > 25 have improved survival with adjuvant chemotherapy. The current study aimed to develop a nomogram with routine parameters to predict RS.

Methods: We included patients diagnosed with hormone receptor (HR)-positive, human epidermal growth factor receptor-2 (HER2)-negative who underwent the 21-gene RS testing and aged > 50 years. The primary outcome was high-risk RS (> 25). Univariate and multivariate analyses were performed to identify significant predictors. A predictive nomogram based on logistic model was developed and evaluated with receiver operating characteristic (ROC) curves. The nomogram was internally validated for discrimination and calibration with bootstrapping method, and externally validated in another cohort. We then assessed the nomogram in different subgroups of patients and compared it with several published models.

Results: A total of 1100 patients were included. Five clinicopathological parameters were used as predictors of a high-risk RS, including tumor grade, histologic subtype, ER expression, PR expression, and Ki-67 index. The area under the curve (AUC) was 0.798 (95% CI 0.772-0.825) and optimism adjusted AUC was 0.794 (95% CI 0.781-0.822). External validation demonstrated an AUC value of 0.746 (95% CI 0.685-0.807), which had no significant difference with the training cohort (P = 0.124). Calibration plots indicated that the nomogram-predicted results were well fitted to the actual outcomes in both internal and external validation. The nomogram had better discriminate ability in patients who had tumors > 2 cm (AUC = 0.847, 95% CI 0.804-0.890). When compared with four other existing models, similar AUC was observed between our nomogram and the model constructed by discriminate Lee et al. CONCLUSIONS: We developed a user-friendly nomogram to predict the high-risk RS in luminal breast cancer patients who were older than 50 years of age, which could guide treatment decision making for those who have no access to the 21-gene RS testing.

Keywords: Breast cancer; Nomogram; Predict; The 21-gene recurrence score.

Fig. 1

Flow chart

Fig. 2

The nomogram predicting the probability of high-risk RS (RS > 25). A nomogram with tumor grade, histological subtype, age, ER expression, PR expression, and Ki-67 index predicting the probability of high-risk recurrence score (RS)

Fig. 3

Discrimination of the nomogram and internal validation of the nomogram. a The discrimination assessed by ROC curve. The AUC is 0.798 (95% CI 0.772–0.825). b The nomogram was internally validated by applying bootstrap sampling for 1000 times. The optimism adjusted AUC is 0.794 (95% CI 0.781–0.822). c Calibration plot of the nomogram. The nomogram was calibrated for the probability of being high-risk RS. (bootstrap 1000 repetitions)

Fig. 4

Subgroup analysis of discrimination of the nomogram. a Comparison of the nomogram in the Luminal B-like and the Luminal A-like cohort. Luminal B-like: The AUC is 0.772 (95% CI 0.739–0.810); Luminal A-like: The AUC is 0.698 (95% CI 0.632–0.764); Delong’s test P = 0.048. b Comparison of the nomogram in the large tumor cohort and the small tumor cohort. Large tumor: The AUC is 0.847 (95% CI 0.804–0.890); Small tumor: The AUC is 0.779 (95% CI 0.746–0.813); Delong’s test P = 0.016

Fig. 5

validation of the model predicting high-risk RS (RS > 25) with clinicopathological characteristics using the current database. The red curve is the ROC curve for the current study, the AUC is 0.798 (95% CI 0.772–0.825); The black curve represents the validation of the model constructed by Hyun-seok Kim et al. using our database, the AUC is 0.766 (95% CI 0.738–0.794), P < 0.001 (delong test); The blue curve represents the validation of the model constructed by Sae Byul Lee et al. using our database, the AUC is 0.757 (95% CI 0.717–0.797), P = 0.090 (delong test); The yellow curve represents the validation of the model constructed by Amila Orucevic et al. using our database, the AUC is 0.695 (95% CI 0.664–0.727), P < 0.001 (delong test); The green curve represents the validation of the model constructed by Shin Hye Yoo et al. using our database, the AUC is 0.699 (95% CI 0.668–0.730), P < 0.001 (delong test)