Impact Of Postoperative Radiotherapy On Prognostic Profiles Of Molecular Subtypes Among Breast Cancer Patients

Purpose:To investigate if postoperative radiotherapy brings different predictive profiles of molecular subtypes on prognosis of breast cancer patients.Methods:All 716 women who had the first ever primary surgery of breast cancer in 2008 at the Fourth Hospital of Hebei Medical University were identified for retrospective analyses. Each patient was assigned into one of molecular subtypes of breast cancer luminal A(LA), luminal B-HER2-negative(LB1), luminal B-HER2-positive(LB2), HER2 overexpression(HER2+) and triple-negative(TN) according to 2011 St. Gallen international consensus standards on the basis of ER, PR, HER2 and Ki-67 expression status which were determined through immunohistochemical test of primary tumor. Patients with HER2(++) but without FISH test conducted were grouped as “unassigned” subtype of breast cancer. All patients were followed up to any events of tumor recurrence, metastasis, death and loss of contacts-up or July 1, 2014 whichever occurred earliest since the primary surgery date. For analyses, the main independent variables were breast cancer subtypes assigned and presence postoperative radiotherapy(RT); the analysis covariate pool included age at surgery, menopausal status, surgical approach, tumor size, histological type, grade, number of lymph nodes involved tumor, vascular invasion status, chemotherapy(neoadjuvant or postoperative), and endocrine therapy. Analysis outcome were overall survival and disease-free survival. T-test or ANOVA were used for comparison of continuous variables between or among groups along with statistics of mean and standard deviation. Chi-square test was conducted for percentage comparison for categorical variables. Cox proportional hazard(PH) models were used for survival analyses along with the generation of Kaplan-Meier curves and log rank test. Hazard ratios(HR) and its confidence interval(CI) from models were calculated and reported with p values(P< 0.05 as significant unless stated otherwise). Statistical tools were SAS 9.20 and SPSS 19.0.Results:1 Overall descriptive analysesThe statistics of study population were mean age 51.4 years old(20- 87), 52.7%(377 cases) pre menopause, 8%(63) breast conserving surgery, 95.4%(683) axillary dissection, 69.6 %(498) ductal carcinoma, 15.5%(111) lobular carcinoma, 30.3%(216) postoperative radiotherapy, 71.2%(510) endocrine therapy, 98.6%(706) postoperative chemotherapy. The distributions of breast cancer subtypes were 30.0%(215) LA, 21.6%(155) LB1, 10.5%(75) LB2, 6.8%(49) HER2+, 8%(57) TN, 14.5%(104) “unassigned”, and 8.5%(61) without IHC data at all. The median follow-up was 71.4 months. The overall mortality was 10.5%(75) and treatment failure(defined as either death or relapse or metastasis) was 14.9%(107).2 Comparative analyses between RT and non-RT subgroup patientsThere were statistically significant of age, mastectomy, axillary dissection, pathologic tumor size, vascular invasion, tumor grade, number of lymph node involved between RT(217 cases) and non-radiotherapy subgroup(499 cases). In general, RT patients had more advanced stage than non-RT subgroup patients which reflected the practice pattern of patient selection. No statistical difference was detected for breast cancer subtypes(P = 0.108).3 Comparative analyses among breast cancer subtypesFor all patients, Age has marginal statistical significance among breast cancer subtype patients(P = 0.060). The youngest mean age(48.7 years) was noticed among LB2 patients. There were statistically significant differences of use of endocrine therapy and tumor grade. TN patients had the highest proportion(15.8%) of pathological grade Ⅲ. No differences were noticed for tumor size, number of positive lymph nodes, vascular invasion(all P >0.10).Among the non-RT subgroup patients, the percentages of use of endocrine therapy, pathological grade and others were similar to the overall populations. In addition, the pathological types had statistical significances among patients with different breast cancer subtypes(P = 0.002). Specifically, LA and LB1 patients had relatively higher percentages of lobular carcinoma as 20.5% and 14.7%, respectively.Among the RT subgroup patients, all disease characters listed at method section among subtypes had similar distributions as the non-RT patients. The use of endocrine therapy was obviously significantly related to subtypes(P < 0.05).4 Analyses of OS among breast cancer subtypesFor all patients, the mortality rates for each subtype were LA 7.4%(16/215), LB1 13.6%(21/155), LB2 6.7%(5/75), HER2+ 8.2%(4/49), TN 12.3%(7/57), and “unassigned” 14.4%(15/104). The univariate COX analyses showed that LB1 and “unassigned” subtypes had up to 85% and 101% higher risk compared to LA patients(reference). However, multivariable Cox analyses did not show their significant differences of OS( P > 0.100) although both had high HR 1.632 and 1.591, respectively. That might be related the sample size but appeared to suggest poorer prognosis for both subgroups compared to LA patients.For the non-RT patients, the mortalities for each subtype were LA 6.8%(11/161), LB1 11.6%(11/95), LB2 3.9%(2/51), HER2+ 5.7%(2/35), TN 8.3%(3/36) and “unassigned” 11.6%(8/69). Both univariate and multivariable COX analyses did not show the significant difference of OS among subtypes(P > 0.10). Nonetheless, HRs of LB1 and “unassigned” from multivariable analyses were 2.055 and 1.353, respectively. It appeared to continue indicating their poorer prognosis compared to LA subtype.For the RT patients, neither univariate nor multivariate COX models showed the statistical differences of OS among subtypes(P > 0.10).5 Analyses of DFS among breast cancer subtypesFor all patients, the failure rates were for subtypes LA 10.2%(22/215), LB120.0%(31/155), LB2 10.7%(8/75), HER2+ 14.3%(7/49), TN 19.3%(11/57), and “unassigned” 20.2%(21/104). The univariate Cox analyses indicated that LB1, TN and “unassigned” subtypes had 95%, 96% and 107% higher risk than subtype LA. However, multivariate analyses demonstrated that only LB1 and “unassigned” subtypes were still associated with significant higher risk of failure(P < 0.05).Among the non-RT patients, the failure rates were for subtypes LA 6.8%(11/161), LB1 17.9%(17/95), LB2 5.9%(3/51), HER2+ 8.6%(3/35), TN 13.9%(5/36), and “unassigned” 14.5%(10/69). The univariate analyses showed that LB1 and “unassigned” subtype patients were associated with 172% and 120% higher risk of failure rates than LA patients. However, the multivariable analyses continued showing the significant difference of failure rates for LB1 patients(P = 0.005), but not any more for “unassigned” patients.Among the RT patients, neither univariate nor multivariable COX analyses showed the significant difference in term of DFS subtypes(P > 0.10).6 Effects of radiotherapy on OS and DFSThe mortality rates of non-RT and RT patients were 8.6%(43/499), and 14.8%(32/217), respectively. Univariate Cox analysis showed that radiotherapy was associated with significant higher death risk with HR 1.631(P = 0.036). Clearly that was related to the patient selection for RT at practice. However, the multivariable analysis had lead to the reversed HR 0.662(P = 0.220) which obviously demonstrated the benefit of RT although the p value was still not at 0.05 level.In term of treatment failure, the rates of non-RT and RT patients were 11.0% and 24.0% respectively. Cox analyses of DFS showed HR of radiotherapy was 2.115(P = 0.000) from univariate one and 0.772(P = 0.336) from multivariate one. That seemed to indicate the larger benefit of RT on improvement of DFS could be true.7 Two-dimensional(2D) analysis of OS and DFS for RT*subtypeWhile LA patients with RT were cited as referent subgroup, the 2D univariate analysis of OS did not show any statistically difference(P > 0.10) among 12 subgroups(RT*subtypes). Nonetheless, the multivariate analysis demonstrated that LB1*non-RT subgroup had a marginally statistically higher mortality(HR = 2.706, P = 0.091).In terms of DFS, the univariate analysis showed the statistically significant difference of LA*non-RT(HR = 0.351, P = 0.014) and marginally difference of LB2*non-RT(HR = 0.287, P = 0.055), compared to LA*RT subgroup. However, both differences were no longer noticed from the multivariable analyses, and quite interestingly the LB1*non-RT subgroup was associated with higher failure rates(HR = 2.279, P = 0.057) at statistical marginally level.Additionally, both univariate and multivariate analyses with direct involvement of interaction term RT*subtype at Cox modeling had not indicated their significant roles of associating with OS or DFS(all P > 0.200). This result showed the parallel profiles of breast cancer subtypes between RT and non-RT which also could observed from the 2-D graphics of HRs by RT status*subtypes.Conclusions:1 Breast cancer patients with LB1 subtype associates with low OS and DFS. Their association degree appears to be higher among non-RT patients.2 Postoperative radiotherapy provides the survival benefit for breast cancer patients at some extent.3 Postoperative radiotherapy has similar impact on the predictive profiles of survival among breast cancer subtypes.