Study population
In this cohort study, data were retrieved from the Taiwan Cancer Registry Database (TCRD). We enrolled women with HF with reduced ejection fraction (LVEF ≤ 40%; HFrEF)[7-9] who had received a diagnosis of left-side breast IDC between January 1, 2008, and December 31, 2018. The index date was the date of BCS, and the follow-up duration was from the index date to December 31, 2019. The TCRD of the Collaboration Center of Health Information Application contains detailed cancer-related information of patients, including their clinical stage, pathologic stages, chemotherapy regimens, dose of chemotherapy, molecular status, drug use, hormone receptor status, radiation modalities and doses, and surgical procedures.[10-13] The study protocols were reviewed and approved by the Institutional Review Board of Tzu-Chi Medical Foundation (IRB109-015-B).
Inclusion and exclusion criteria
The diagnoses of the enrolled patients with HFrEF were confirmed after their pathological data were reviewed, and the women with newly diagnosed left-side IDC were confirmed to have no other cancers or distant metastases. The women with HFrEF were included if they had received a left-side IDC diagnosis, were 20 years old or older, and had clinical stage IA–IIIC (American Joint Committee on Cancer [AJCC], 8th edition) without metastasis. Patients with HFrEF were excluded if they had a history of cancer before the IDC diagnosis date, unknown pathologic types, missing sex data, unclear staging, or non-IDC histology. In addition, patients undergoing neoadjuvant chemotherapy or with unclear differentiation of tumor grade, missing HR status, missing data on trastuzumab or anthracycline use, or unclear staging were excluded. Other adjuvant treatments such as adjuvant chemotherapy, hormone therapy, or the human epidermal growth factor receptor 2 inhibitors did not constitute exclusion criteria based on the National Comprehensive Cancer Network (NCCN) guidelines.[14] We also excluded patients with HFrEF with unclear data on surgical procedures such as BCS or TM, ill-defined nodal surgery, unclear Charlson comorbidity index (CCI), or unclear differentiation from our cohort. Hormone receptor positivity was defined as ≥1% of tumor cells demonstrating positive nuclear staining through immunohistochemistry.[15]
After applying the inclusion and exclusion criteria, we enrolled 294 women with HFrEF and AJCC clinical stage IA–IIIC and left-side IDC who had received a BCS followed by sentinel lymph node biopsy (SLNB) or axillary lymph node dissection (ALND) and divided them into two groups based on their adjuvant WBRT status to compare all-cause mortality: Group 1 (women with left-side IDC and HFrEF who received BCS followed by adjuvant WBRT) and Group 2 (women with left-side IDC and HFrEF who received BCS and no adjuvant WBRT). We also excluded women in Group 1 receiving nonstandard adjuvant WBRT (contrast with standard adjuvant radiotherapy consisting of irradiation to the whole breast with a minimum of 50 Gy). Contemporary RT techniques were included in our study and the conventional two-dimensional RT technique was excluded. The included contemporary RT techniques were three-dimensional RT and intensity-modulated radiation therapy. The incidence of comorbidities was scored using the CCI.[16, 17] Ischemic heart disease, heart valvular disease, cardiomyopathy, hypertension, diabetes, and arrhythmias and conduction disorders were excluded from the CCI scores to avoid repetitive adjustment in multivariate analysis. Only comorbidities observed within 6 months before the index date were included; they were coded and classified according to the International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) codes at the first admission or based on more than two repetitions of a code issued at outpatient department visits.
Study covariates and statistical analysis
Significant independent predictors, namely age, diagnosis year, CCI score, differentiation, pT, pN, hypertension, ischemic heart disease, heart valvular disease, cardiomyopathy, arrhythmias and conduction disorders, diabetes, adjuvant chemotherapy with anthracycline-based regimen, hormone receptor status, trastuzumab use, nodal surgery, hospital level (academic or nonacademic), and income, were analyzed using a multivariate analysis of the propensity score–weighted population to determine hazard ratios (HRs). We calculated the propensity score and applied inverse probability of treatment weighting (IPTW) to create a pseudo-study cohort; the weighted cohort avoids covariate bias and mimics randomized adjuvant WBRT or no adjuvant WBRT assignment: IPTW for patients with WBRT = 1/p(WBRT); IPTW for patients without WBRT = 1/(1 − p[WBRT]).[18, 19] The independent predictors were examined in multivariable analyses after IPTW adjustment. Moreover, they were controlled for and were stratified in the analysis. The endpoint was all-cause death in the women with left-side IDC and HFrEF who received BCS followed by adjuvant WBRT (Group 1, case group) and in the women with left IDC and HFrEF who received BCS and had no adjuvant WBRT (Group 2, control group).
The cumulative incidence of death was estimated using the Kaplan–Meier method, and differences in the overall survival (OS), locoregional recurrence (LRR)–free survival, and distant metastasis (DM)–free survival between women with left IDC and HFrEF receiving BCS followed by adjuvant WBRT versus no adjuvant WBRT were determined using a log-rank test. After confounders were adjusted for, IPTW-adjusted models were used to determine the time from the index date to all-cause mortality in the women with left IDC and HFrEF who received BCS followed by adjuvant WBRT or no adjuvant WBRT. Subsequently, in a multivariate analysis, HRs were adjusted for age, diagnosis year, CCI scores, differentiation, pT, pN, hypertension, ischemic heart disease, heart valvular disease, cardiomyopathy, arrhythmias and conduction disorders, diabetes, adjuvant chemotherapy with anthracycline, hormone receptor status, trastuzumab use, nodal surgery, hospital levels, and incomes. All analyses were conducted using SAS (Version 9.3; SAS, Cary, NC, USA), and a two-tailed P value <0.05 was considered statistically significant.