A Nomogram Based on Hematological Parameters and Clinicopathological Characteristics for Predicting Local-Regional Recurrence After Breast-Conserving Therapy

Abstract

Objectives: The aim of this study was to identify the factors for local-regional recurrence (LRR) after breast-conserving therapy (BCT). We established a practical nomogram to predict the likelihood of LRR after BCT based on hematological parameters and clinicopathological features.

Methods: A retrospective analysis was performed on 2,085 consecutive breast cancer patients who received BCT in Shandong Cancer Hospital from 2006 to 2016, including 1,460 patients in the training cohort and 625 patients in the validation cohort. Univariate and multivariate analyses were performed based on hematological parameters (fibrinogen, platelets, mean platelet volume, neutrophils, monocytes, and lymphocytes) and clinicopathological characteristics to identify the independent factors for LRR. Subsequently, a nomogram for predicting LRR was established by logistic regression analysis. The nomogram was validated in 625 patients in the validation cohort.

Results: During the median follow-up period of 66 months, 44 (3.01%) patients in the training cohort and 19 (3.04%) patients in the validation cohort suffered from LRR. Multivariate analysis showed six independent factors related to LRR, including molecular subtype, pathological N stage, re-resection, radiotherapy or not, platelet count*MPV*fibrinogen (PMF), and neutrophil count/lymphocyte count ratio (NLR). Six variables were entered into logistic regression to establish the nomogram for predicting LRR. The nomogram of LRR showed excellent discrimination and prediction accuracy. The area under the receiver operating characteristic curve (AUC) was 0.89 (p < 0.001, 95% CI = 0.83, 0.95) in the training cohort and 0.88 (p < 0.001, 95% CI = 0.8, 0.96) in the validation cohort. Calibration curves for the prediction model in the training and validation cohorts both demonstrated satisfactory consistency between the nomogram-predicted and actual LRR.

Conclusion: The combination of hematological parameters and clinicopathological characteristics can predict LRR after BCT. The predictive nomogram based on preoperative and postoperative indicators of BCT might serve as a practical tool for individualized prognostication. More prospective studies should be performed to verify the model.

Keywords: breast-conserving therapy; clinicopathological characteristics; hematological parameters; local–regional recurrence; nomogram; predicting model.

Figure 1

Consort diagram for the study cohort. BCT, breast cancer therapy; SCH, Shandong cancer hospital; LRR, local–regional recurrence.

Figure 2

Optimal cutoff points for hematologic parameters were on with ROC curves. PF, platelet count*fibrinogen; MF, mean platelet volume*fibrinogen; PMF, platelet count*mean platelet volume*fibrinogen; FMR, fibrinogen-to-mean platelet volume ratio; PMR, platelet count-to-mean platelet volume ratio; NLR, neutrophil count-to-lymphocyte count ratio; MLR, monocyte count-to-lymphocyte count ratio; PLR, platelet count-to-lymphocyte count ratio; ROC, receiver operating characteristic.

Figure 3

Kaplan–Meier curves for local-regional recurrence. Kaplan–Meier curves for local–regional recurrence based on molecular subtype (A), pathological N stage (B), radiotherapy (C), re-resection (D), PMF (E), and NLR (F). TNBC, triple-negative breast cancer; PMF, platelet count*mean platelet volume*fibrinogen; NLR, neutrophil count-to-lymphocyte count ratio.

Figure 4

Nomogram model predicts the probability of local–regional recurrence. Points refers to point for the individual risk factor and add together to the total points. Luminal A, HER-2 (−), ER (+), PR (+) and high expression, Ki67 low expression; Luminal B, HER-2 (−), ER (+), PR (−) or low expression, Ki67 high expression; HER2+, the breast cancer of HER2 positive; TNBC, triple-negative breast cancer; N0, No positive lymph nodes; N1, the number of positive lymph nodes is 1–3; N2+, the number of positive lymph nodes is more than 3; PMF, platelet count*mean platelet volume*fibrinogen; NLR, neutrophil count-to-lymphocyte count ratio.

Figure 5

Evaluation of the LRR nomogram (A–D). Calibration curves for the nomogram in the training cohort (A) and validation cohort (B). The x-axis shows the predicted probability of an LRR event. The y-axis shows the actual LRR outcome. The discrimination assessed by ROC curves for the nomogram in the training cohort (C) and validation cohort (D). The AUCs for LRR prediction were 0.89 (95% CI = 0.83, 0.95) in the training cohort and 0.88 (95% CI = 0.8, 0.96) in the validation cohort. LRR, local–regional recurrence; ROC, receiver operating characteristic; AUC, area under the curve.