If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Address for correspondence: Xinyun Xu, Department of General Surgery, Changzheng Hospital of the Second Military Medical University, 415 Fengyang Road, Shanghai 200000, China Tel.: +86-21-81885597.
The nomogram in this work is the first 1 of which the target population is LABC patients receiving IBR.
•
Geographical split was used to generate the training group and the test group, which was reported to be preferable to the random split but was rarely used in previous SEER-based studies.
•
10-fold cross validation of the training group was further visualized.
Abstract
Introduction/Background
This study aimed to construct a nomogram to provide prognostic references for patients with locally advanced breast cancer (LABC) to receive immediate breast reconstruction (IBR).
Materials and Methods
All data were obtained from the Surveillance, Epidemiology and End Results (SEER) database. Univariate Cox regression, least absolute shrinkage and selection operator (LASSO) and best subset regression (BSR), separately followed by backward stepwise multivariable Cox, were used to construct the nomogram. Risk stratification was established after validation.
Results
A total of 6,285 patients were enrolled to generate the training group (n = 3,466) and the test group (n = 2,819) by geographical split. Age, marital status, grade, T staging, N staging, radiotherapy, chemotherapy, estrogen receptor status (ER), progesterone receptor status (PR) and human epidermal growth factor receptor type 2 status (HER2) were used to fit the nomogram. The overall Harrell's concordance index (C-index) was 0.772 in the training group and 0.762 in the test group. The area under the receiver operator characteristic curves (AUC) at 3-year and 5-year were respectively 0.824 and 0.720 in the training group, 0.792 and 0.733 in the test group. The calibration curves showed great consistency in both groups. A dynamic nomogram (https://dcpanfromsh.shinyapps.io/NomforLABCafterIBR/) was developed.
Conclusion
A nomogram was developed and validated that predicts prognosis more accurately than the AJCC 7th stage and can be used as a reference for decision-making in LABC patients receiving IBR.
The advantages of IBR over delayed breast reconstruction include better cosmetic outcome, fewer procedures, shorter hospital stays, less financial burden and positive impact on postoperative psychosocial well-being.
However, IBR was usually prudently performed in patients with locally advanced breast cancer (LABC) given the possibility to delay adjuvant therapy by increasing risk of surgical complications.
Long-term oncologic outcomes of immediate breast reconstruction vs conventional mastectomy alone for breast Cancer in the setting of neoadjuvant chemotherapy.
the decision to receive IBR or not is still difficult in consideration of different benefits from discrepant individual prognosis. Therefore, individualized prognostic assessment of these patients is needed to weigh the pros and cons of IBR.
Recent years have seen an increase in the number of studies on nomograms.
Factors affecting pathologic complete response following neoadjuvant chemotherapy in breast Cancer: development and validation of a predictive nomogram.
Long-term oncologic outcomes of immediate breast reconstruction vs conventional mastectomy alone for breast Cancer in the setting of neoadjuvant chemotherapy.
Oncologic safety of immediate breast reconstruction in breast Cancer Patients who underwent neoadjuvant chemotherapy: short-term outcomes of a matched case–control study.
Post-mastectomy immediate breast reconstruction is oncologically safe in well-selected T4 locally advanced breast cancer: a large population-based study and matched case-control analysis.
and no nomogram that focus on the prognosis of these patients. In this study, we retrieved data from the Surveillance, Epidemiology and End Results (SEER) database to construct a nomogram and then established a risk stratification for these patients, aiming to provide a reference to identify high-risk patients prior to surgery.
Materials and Methods
Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) was rigidly referred during the design of this study.
Women with LABC undergoing IBR between January 2010 and November 2017 were identified from the SEER database by using SEER*Stat 8.4.0.1. The inclusion criteria were as follows: (1) age between 20 and 80 years at diagnosis; (2) female; (3) primary breast cancer; (4) American Joint Committee on Cancer (AJCC) stage 7th edition of ⅢA-ⅢC or SEER combined stage of 3A to 3C; (5) underwent IBR (surgery of primary site code 30,43-49, 53-59, 63-69, 73-75). The exclusion criteria were as follows: (1) more than 1 malignant tumor; (2) T0 staging; (3) with unknown marital status at diagnosis, race, grade, T staging, N staging, estrogen receptor status (ER), progesterone receptor status (PR) or human epidermal growth factor receptor type 2 status (HER2) (Figure 1A).
Figure 1Description of the work flow. SEER, the Surveillance, Epidemiology and End Results database; LASSO, least absolute shrinkage and selection operator; BSR, best subset regression.
This cutoff date for this study was November 2017. The following variables were retrieved: age at diagnosis, race (divided into white, black and others), marital status at diagnosis (divorced, separated, single, domestic partner and widowed are defined as unmarried), years of diagnosis, SEER registry (with California and Georgia as whole states), grade (defined as well, moderately, poorly and undifferentiated in ICD-O-2), laterality (the side of the body on which the reportable tumor originated), histologic type (ICD-O-3 8000/3, 8010/3, 8033/3, 8050/3, 8070/3, 8071/3, 8140/3, 8200/3, 8201/3, 8230/3, 8246/3, 8255/3, 8401/3, 8480/3, 8481/3, 8500/3, 8501/3, 8503/3, 8507/3, 8510/3, 8520/3, 8521/3, 8522/3, 8523/3, 8524/3, 8530/3, 8541/3, 8570/3 and 8575/3), derived AJCC stage 7th edition (cases from 2010 to 2015), derived SEER combined stage group (cases from 2016 to 2017), radiation and chemotherapy recode, breast subtype (Luminal A, Luminal B, HER2 enriched and Triple Negative), surgery of primary site, survival time (month) and vital status (until the cut-off date). The optimal cut-point of age was determined by the “surv_cutpoint” function from the package “survminer” in R software (version 4.2.1; http://www.r-project.org/).
Construction and Validation of the Nomogram
The training and testing groups were generated based on the registries obtained from the SEER database. Patients registered in California, Iowa, Louisiana, New Mexico, Seattle (Puget Sound), Utah, Alaska Natives and Hawaii were assigned to the training group. Patients registered in Connecticut, Detroit (Metropolitan), Georgia, Kentucky, New Jersey were assigned to the test group. In the training group, univariate Cox hazard analysis, LASSO regression and BSR were performed to initially identify significant prognostic factors (defined as factors with P value less than .1 or with clinical significance, in univariate Cox hazard analysis), respectively. Then, a backward stepwise multivariate Cox regression was used in each model separately to determine independent significant prognostic factors. Akaike information criterion (AIC) and ROC were used to determine the optimal model (Figure 1B). The nomogram based on the final model was constructed by package “rms” in R. In the meantime, the variance inflation factor (VIF) was calculated to ensure no collinearity among the covariates (collinearity was defined as VIF > 4.0). 10-fold cross-validation was performed in the training group and external validation was performed in the testing group. In both groups, at 3-year and 5-year, the receiver operator characteristic curves (ROC), AUC and C-index were used to evaluate the discrimination of nomogram. Additionally, the calibration curves were plotted by bootstrapped resample with 1,000 iterations to assess the consistency between nomogram-predicted overall survival (OS) and actual OS. Decision curve analysis (DCA) was used to assess the clinical benefit of the nomogram. The nomogram-based point of each patient was calculated by “nomogramFormula” package. The X-tile software was then used to determine the most optimal cut-off. Finally, all patients were divided into low, mid and high-risk groups (Figure 1C).
Statistical Analysis
R software (version 4.2.1; http://www.r-project.org/) was used for all analyzes. A Student's t test was used to analyze normally distributed data as mean ± standard deviation (x ± s). An analysis of non-normally distributed data was conducted using Mann-Whitney U tests. Categorical data were expressed as n (%) and analyzed using the Chi-square test or Fisher's exact test. In order to analyze the overall survival rate, the Kaplan-Meier method along with the log-rank test was used. Statistical significance levels were both two-sided and set at.05. 95% confidence intervals (95%CI) were reported throughout the analysis.
Results
Demographic and Clinical Features
A total of 6,285 eligible patients with a median age of 49 (IQR, 41-57) years were enrolled and divided into training group (n = 3,466) and test group (n = 2,819) according to their geographic information. The median follow-up time of the whole population was 36 (IQR, 18-60) months. The demographic and clinical characteristics of the training group and test group are reported in Table 1. In general, younger than 57 years (74.2%), white (79.8%), married (66.2%), stage-Ⅲ A (69.5%) and Luminal A (66.3%) patients accounted for a larger proportion. Compared with the test group, the training group was associated with more white patients (82.8% vs. 76.1%, P < .001), more patients with invasive ductal carcinoma (69.7% vs. 67.6%, P = .02), more patients with T3 staging (48.5% vs. 41.6%, P < .001) and N1 staging (36.2% vs. 31.6%, P < .001), fewer patients received radiotherapy (67.8% vs. 73.7%, P < .001) and chemotherapy (91.0% vs. 94.6%, P < .001).
Table 1Demographics and Clinical Data of the Training Group and Test Group
The stage was excluded in advance because of its collinearity with T and N staging. Twelve significant prognostic factors (age, race, marital status at diagnosis, grade, histologic type, T staging, N staging, radiotherapy, chemotherapy, ER, PR and HER2) were screened by univariate Cox hazard analysis (Figure 2A). By LASSO regression, we selected the value of lambda.lse (0.01279193) which corresponded to 7 variables (age, grade, T staging, N staging, ER, PR and HER2) (Figure 2B and 2C). In BSR, the maximum value of adjusted R2 was 0.067 and 8 variables (age, grade, T staging, N staging, axillary dissection, ER, PR and HER2) were selected (Figure 2D). After further backward stepwise multivariable Cox regression, 2 models (Uni-cox and LASSO) were established (the model based on LASSO was the same as the 1 based on BSR) (Table 2). Then, additional comparison of the AIC and AUC among models was performed (Figure 3A and 3B). Finally, the Uni-cox (AIC = 5106.73) was selected to construct the nomogram to predict 3-year and 5-year OS (Figure 3C).
Figure 2Univariate Cox regression (A), Least absolute shrinkage and selection operator regression (B and C) and best subset regression (D) for variables screening.
Figure 3Comparison among models and present of the nomogram. (A) Comparison of ROC curves and AUC among Uni-cox, Uni-cox without RT, Uni-cox without CT, Uni-cox with either, LASSO and BSR; (B). Comparison of the value of AIC among models; (C). Nomogram for predicting the 3-year and 5-year overall survival of LABC patients undergoing IBR. Uni-cox, the model based on variables screened by backward stepwise multivariable Cox regression following univariate Cox regression; ROC, receiver operator characteristic curves; AUC, area under the receiver operator characteristic curves; AIC, akaike information criterion; ER, estrogen receptor; PR, progesterone receptor; HER2, human epidermal growth factor receptor type 2.
The overall C-index of the nomogram was respectively 0.772 (95%CI, 0.747-0.796) in the training group and 0.762 (95%CI, 0.735-0.789) in the test group, indicating a favorable discrimination of the nomogram. The mean value of AUC at 3-year and 5-year calculated from 10-fold cross validation were 0.809 (95%CI, 0.807-0.811) and 0.708 (95%CI, 0.706-0.710), respectively. The mean value of C-index at 3-year and 5-year were 0.803 (95%CI, 0.801-0.805) and 0.711 (95%CI, 0.709-0.713), respectively (Figure 4). Meanwhile, the ROC curves at 3-years and 5- years showed the value of AUC reached 0.824 (95%CI, 0.794-0.854) and 0.720 (95%CI, 0.684-0.756) in training group, 0.792 (95%CI, 0.755-0.829) and 0.733 (95%CI, 0.698-0.768) in test group, respectively (Figure 5A and 5B). In addition, calibration curves at 3-year and 5-year of both groups reflected great consistencies between the actual OS and nomogram-predicted OS (Figure 5C and 5D). DCA curves demonstrated the more net clinical benefit provided by nomogram than AJCC 7th stage (Figure 5E and 5F).
Figure 4Visualization of the 10-fold cross validation in the training group. C-index, Harrell's concordance index; CI = confidence interval; AUC = area under the receiver operator characteristic curves.
Figure 5The time-dependent ROC curves of the nomogram predicting (A) 3-year OS and (B) 5-year OS of the training and test groups respectively. The calibration curves of the nomogram for predicting (C) 3-year OS and (D) 5-year OS of the training and test groups respectively. The decision curve analysis of the nomogram and AJCC 7th stage for predicting (E) 3-year OS and (F) 5-year OS of the training and test groups respectively.
Development of Risk Stratification Based on the Nomogram
The risk stratification was developed based on the points of patients calculated by nomogram (points < 191 as low-risk, 191 ≤ points < 241 as mid-risk and points ≥ 241 as high-risk). The Kaplan–Meier OS curves demonstrate the more accurate discrimination of the nomogram than AJCC 7th stage (P < .001) (Figure 6).
Figure 6Kaplan-Meier curves of all patients grouped by (A) nomogram-based risk stratification and (B) AJCC 7th based-TNM stage.
Evaluation of the prognostic stage in the 8th edition of the American Joint Committee on Cancer in locally advanced breast cancer: an analysis based on SEER 18 database.
Consequently, the value of using the classical TNM stage to predict the OS of LABC patients after IBR is limited. As the positive impact of IBR has been gradually acknowledged,
Long-term oncologic outcomes of immediate breast reconstruction vs conventional mastectomy alone for breast Cancer in the setting of neoadjuvant chemotherapy.
it is essential to provide a risk assessment tool to expand the application of IBR in LABC patients. In the current study, we analyzed data from 6,258 patients with LABC who received IBR. Based on ten filtered variables related to prognosis, a nomogram was successfully developed. As distinct from numerous previous researches,
Meanwhile, univariate Cox regression, LASSO and BSR were used to prevent overfitting and underfitting. The performance of the nomogram in predicting 3-year and 5-year OS was great in the test group, despite the significant difference in baseline characteristics between the 2 groups.
Marital status has been shown to affect the prognosis of breast cancer patients. Hinyard L discovered that unmarried status was related with a worse prognosis than married status among women under the age of 65.
According to research by Liu YL, married patients had higher rates of both overall and breast cancer-specific survival than those who were unmarried among inflammatory breast cancer patients.
Marital status is an independent prognostic factor in inflammatory breast cancer patients: an analysis of the surveillance, epidemiology, and end results database.
After being adjusted for via multivariate analysis in this study, marital status was also found to be an independent prognostic factor in the Uni-cox model. This was most likely due to the psychological and social influence of marital status. Married women typically have more social connections and receive more emotional support from their spouses.
Following the IBR, the marriage may provide them with more medical care and material assistance. Furthermore, a study discovered that larger social networks were associated with improved quality of life in breast cancer patients.
As a result, we recommended that prior to surgery, such social supports as marital status for patients who are prepared to receive IBR should be carefully evaluated.
T and N staging are significant prognostic factors underlying the AJCC staging system. Opinions were divided as to the application of IBR in patients with different stages of LABC. Although some studies demonstrate that IBR was a safe option in patients with positive nodes and T4 staging,
Post-mastectomy immediate breast reconstruction is oncologically safe in well-selected T4 locally advanced breast cancer: a large population-based study and matched case-control analysis.
Taqi K et al. demonstrated that patients with T4 breast cancer were less likely to be offered IBR and were found to be associated with higher overall recurrence, overall mortality and disease-specific mortality.
Similarly, a previous study based on the National Cancer Database indicated that IBR was more frequently received in patients with involvement of less than 4 lymph nodes.
Adjusted HR calculated in this study showed that patients with a stage higher than T2 or N1 were associated with a significantly higher risk of receiving IBR. Thus, accurate risk estimates for the T and N stages are required separately. Based on this method, patients in the same TNM stage (eg T4N0M0 and T4N2M0) might have different scores and then to be divided into different risk stratification to receive different surgery. Thus, some high-risk patients based on TNM stage (eg ⅢC) might be assigned to lower-risk population to consider IBR.
Chemotherapy and radiotherapy exert positive influence on the prognosis of LABC patients by improving survival and reducing local recurrence.
Effect of radiotherapy after mastectomy and axillary surgery on 10-year recurrence and 20-year breast cancer mortality: meta-analysis of individual patient data for 8135 women in 22 randomised trials.
Nevertheless, these 2 variables seem to play a less relevant role in our nomogram. One possible explanation was that the benefit of chemotherapy varied depending on the disease.
Given the number of years of research and the wide range of ages, stages, and biological subgroups, the type of adjuvant systemic therapy received would be highly variable. However, limited by the unavailability of details about chemotherapy from the SEER database, we were not able to assess the effect of different treatments on the OS. Another reason was perhaps that the majority of patients in the training group received chemoradiotherapy, which resulted in an inadequate reflection of the risk of not receiving chemoradiotherapy.
For LABC patients with poor prognosis, the benefits of IBR may be limited. As a result, we developed a risk stratification with improved discrimination to identify high-risk LABC patients for IBR, potentially providing a more individualized reference. Meanwhile, we created a web-based dynamic nomogram application (https://dcpanfromsh.shinyapps.io/NomforLABCafterIBR/) to help with scoring. In practice, we believe that the IBR should not hinder adjuvant therapy for any type of breast cancer, even in patients with low-risk scores. Simultaneously, we propose combining patient survival prediction with guidelines such as the 21-gene recurrence-score assay and the 70-gene signature test to develop a more comprehensive treatment plan.
There are several limitations to the current work. The first is about the inevitability of bias in this retrospective study. Aside from that, the nomogram's long-term prediction is hampered by a lack of information about neoadjuvant chemotherapy, Ki-67, and endocrine therapy. We expect to include more variables to improve the utility and applicability of the nomogram. Finally, more external comprehensive data is needed to validate our nomogram.
Conclusion
A survival nomogram for LABC patients undergoing IBR was established by quantifying the associated risks. Surgeons could more adequately assess the risk of these patients to receive IBR through this nomogram and accordingly provide a more appropriate treatment. Moreover, we expect more external validation to improve the generalizability of this nomogram.
Clinical Practice Points
•
In the absence of a tool to quantify risk, decision-making was difficult for patients who were diagnosed with LABC to receive IBR or not.
•
We successfully established a nomogram to predict the 3-year and 5-year OS of LABC patients to receive IBR, which was based on age, marital status at diagnosis, grade, T staging, N staging, radiotherapy, chemotherapy, ER, PR and HER2. In addition, a wed-based dynamic nomogram application was created to facilitate calculating scores (https://dcpanfromsh.shinyapps.io/NomforLABCafterIBR/).
•
The nomogram in this work is the first one of which the target population is LABC patients receiving IBR. Surgeons are accordingly able to estimate long-term outcome of these patients to provide personal treatment.
Disclosure
The authors have stated that they have no conflicts of interest.
References
Fernandez-Delgado J
Lopez-Pedraza MJ
Blasco JA
et al.
Satisfaction with and psychological impact of immediate and deferred breast reconstruction.
Long-term oncologic outcomes of immediate breast reconstruction vs conventional mastectomy alone for breast Cancer in the setting of neoadjuvant chemotherapy.
Factors affecting pathologic complete response following neoadjuvant chemotherapy in breast Cancer: development and validation of a predictive nomogram.
Oncologic safety of immediate breast reconstruction in breast Cancer Patients who underwent neoadjuvant chemotherapy: short-term outcomes of a matched case–control study.
Post-mastectomy immediate breast reconstruction is oncologically safe in well-selected T4 locally advanced breast cancer: a large population-based study and matched case-control analysis.
Evaluation of the prognostic stage in the 8th edition of the American Joint Committee on Cancer in locally advanced breast cancer: an analysis based on SEER 18 database.
Marital status is an independent prognostic factor in inflammatory breast cancer patients: an analysis of the surveillance, epidemiology, and end results database.
Effect of radiotherapy after mastectomy and axillary surgery on 10-year recurrence and 20-year breast cancer mortality: meta-analysis of individual patient data for 8135 women in 22 randomised trials.
00038-1&id=gr1.jpg){kind=link}
00038-1&id=gr2.jpg){kind=link}
00038-1&id=gr3.jpg){kind=link}
00038-1&id=gr4.jpg){kind=link}
00038-1&id=gr5.jpg){kind=link}
00038-1&id=gr6.jpg){kind=link}