Revisiting the impact of age and molecular subtype on overall survival after radiotherapy in breast cancer patients

Adjuvant radiotherapy (RT) in breast cancer (BC) is often used to eradicate remaining tumor cells following surgery with the goal of maximizing local control and increasing overall survival. The current study investigated the impact of age and BC molecular subtype on overall survival after RT using a meta-analysis of the METABRIC and TCGA BC patient cohorts. We found that RT significantly prolonged survival across the whole BC patient population. The survival benefit of RT was predominantly observed in stage II BC patients treated with breast conserving surgery. Patients were then stratified by age and molecular subtype to investigate survival rate associated with RT. An increase in survival for the luminal-A and basal BC molecular subtypes was observed after RT. Stratifying patients based on age revealed that increased survival was restricted to younger patients (≤60 years of age at diagnosis). There was a significant survival benefit of radiotherapy for younger patients with tumors of the luminal A and basal molecular subtypes. Patients with other breast tumor subtypes or older breast cancer patients did not seem to benefit effects of RT. Therefore, alternate local treatment strategies should be considered for older, luminal B, and HER2 driven BC patients.

two large BC patient cohorts (METABRIC and TCGA), which contain clinical data including radiotherapy, molecular subtype, age, overall survival (OS) and other patient characteristics. We first investigated the demographic differences between patients that received radiotherapy versus those that did not using the METABRIC and TCGA data (Table 1). A higher proportion of young patients, patients receiving breast conserving therapy and patients with high grade and late stage tumors received radiotherapy (Table 1). Overall we found that patients who receive radiotherapy survive significantly longer compared to those who did not receive radiotherapy in both datasets ( Figure S1; METABRIC: p = 0.007; TCGA: p = 1.12E-04).

Effect of radiotherapy on patient survival is dependent on clinical stage and surgery.
To study how clinical stage and surgery influence the effect of RT on patient survival, we first combined the METABRIC and TCGA cohorts and confirmed that radiotherapy significantly increased patient survival in the combined patient cohort ( Fig. 1A; p = 2.14E-04). We then stratified the patient cohort by clinical stage and assessed whether RT offered a survival benefit in different subgroups. We found that radiotherapy increased patient survival in patients with stage II disease ( Fig. 1B; p = 1.35E-05), whereas a tendency for survival benefit was observed in patients with stage III/IV disease ( Fig. 1B; p = 0.059). No benefit was observed in patients with stage I disease ( Fig. 1B; p = 0.78). We further stratified the patient cohort by surgery type (lumpectomy vs. mastectomy). Our data showed that in the cohort of patient that underwent lumpectomy, RT significantly increased survival ( Fig. 1C; p = 0.012), especially in patients with stage II disease ( Figure 1D; 0.014). In the cohort of patients that underwent mastectomy, RT had a tendency to increase survival in stage II patients ( Fig. 1C; p = 0.07).

Impact of radiotherapy on patient survival is independent of clinical factors. To determine
if the impact of radiotherapy on patient survival was independent of age at diagnosis, tumor size, estrogen-and progesterone-receptor status, tumor grade and molecular subtype (as determined by Pam50) we used multivariate Cox regression with these factors including radiotherapy as covariates. In multivariate analysis the difference of OS attributable to radiotherapy remained significant (HR = 0.873: 95% CI: 0.771-0.988; p = 3.16E-02). We also found that age, tumor size, ER status, tumor grade and PAM50 subtype were significantly associated with OS (Table 2).  [3][4][5][6] . Many studies have demonstrated an association between molecular subtype and patient prognosis 14,15 . The basal and HER2 subtypes are generally more aggressive and associated with poorer survival compared to normal-like and luminal breast tumors 16 . To investigate whether radiotherapy benefits patients equally among different molecular subtypes, we stratified our patient cohorts into different molecular subtypes based on the PAM50 molecular score 17 . Surprisingly, in the METABRIC cohort we found that radiotherapy increased patient survival only in the luminal A subtype (p = 3.66E-04), whereas a tendency for increased survival was observed for the basal subtypes (p = 0.13) ( Figure S2). No survival benefit was observed for the other subtypes (luminal B: p = 0.78; HER2: p = 0.57) ( Figure S2). In the TCGA cohort, radiotherapy significantly increased overall survival only in the basal subtype (p = 2.25E-04) ( Figure S3). A tendency for survival benefit associated with radiotherapy was observed for the luminal A (p = 0.053) subtype ( Figure S3). Again, no survival benefit was observed for the luminal B (p = 0.26) and HER2 (p = 0.32) subtypes ( Figure S3). A meta-analysis combining the METABRIC and TCGA cohorts confirmed that radiotherapy significantly increased patient survival in the luminal A (p = 7.68E-05) and basal subtypes (p = 7.13E-03) (Fig. 2).
Effect of age at diagnosis on patient survival after radiotherapy. Age at diagnosis is another well-known prognostic factor in BC 18,19 . We asked whether age at diagnosis in combination with molecular subtype could further clarify the survival benefit from radiotherapy. We split our patient cohorts into two age groups: "young" (age ≤60 years) and "old" (age >60 years). In the "young" group, radiotherapy significantly increased

Discussion
In this study, we revisited the beneficial effects of radiotherapy on OS in BC patients by a population study using the METABRIC (N = 1980 patients) and TCGA (N = 1100 patients) databases. We first addressed the association between radiotherapy and overall survival time of breast cancer patients stratified by clinical stage, and found a significant increase in survival time in stage II breast cancer patients that received RT, but not in stage I and II-IV patients. Further stratification of patients by surgery type showed increased survival time in RT treated stage II breast cancer patients after lumpectomy compared to stage II patients after lumpectomy that did not receive RT treatment. RT only offered a marginal survival benefit in stage II breast cancer patients that underwent mastectomy. It has to be realized here that mastectomy patients do not routinely undergo radiation therapy, so this much smaller subgroup was probably selected for risk factors that we do not know and cannot analyze. Several studies have demonstrated improved survival and reduced local recurrence in patients receiving a combination of breast conserving therapy and RT compared to breast conserving therapy alone [20][21][22][23] . In addition, a number of studies have demonstrated in randomized controlled trials that breast conserving therapy in combination with RT is at least equivalent to mastectomy alone 24,25 . A retrospective cohort study including 5,335 women concluded that breast conserving surgery plus radiation resulted in better overall survival than mastectomy alone 26 . In addition, a Dutch population based study of 37,207 early breast cancer patients that showed that breast-conserving surgery plus radiotherapy showed improved survival compared to patients that underwent mastectomy 27 . However, the latter study could have been biased by "confounding by indication". We consistently found a significant survival benefit of radiotherapy for patients with luminal A or basal molecular subtypes who were diagnosed younger than 60 years of age. To translate our discoveries into clinical practice, future prospective clinical trials are warranted to validate our findings. Wang et al. found that adjuvant radiotherapy reduces the risk of relapse in breast tumors of the luminal A subtype, but not luminal B 28 . Consistent with this discovery, we also showed that tumors of the luminal A subtype showed a significant survival benefit after radiotherapy. However, such survival benefit was driven predominantly by younger patients (age at diagnosis ≤60 years), since no survival benefit was observed in the older age group (age at diagnosis >60 years). Regardless of age at diagnosis, patients with tumors of the luminal B subtype did not show any survival benefit from radiotherapy.
Similar to the luminal B subtype, patients with the HER2 subtype did not show survival benefit after radiotherapy. This data is consistent with previous observations 29 . In contrast, several groups did report an increased risk of recurrence for breast tumors overexpressing HER2 following radiotherapy 30,31 . This discrepancy could be due to additional heterogeneity within tumors of the HER2 subtype. Additional biomarkers are required to separate HER2 positive patients into radiosensitive and resistant cohorts.
Radiotherapy is a major treatment modality for patients with tumors of the basal molecular subtype, since their triple negativity offers no options for hormonal-or HER2 therapy. A few studies have shown increased mortality after radiotherapy in triple-negative breast tumors 29,30 . Surprisingly, we found a significant survival benefit associated with radiotherapy in patients with tumors of the basal subtype, especially those diagnoses in younger patients (age at diagnosis ≤60 years). Our result is consistent with a report that showed that adjuvant radiation was associated with improved overall survival in triple-negative breast cancer 32 . While not synonymous, basal-like breast cancers are predominantly triple-negative. Interestingly, in the present study improved survival was only observed in the lumpectomy group, not the mastectomy group. The main limitations of our study are (1) incomplete clinical information in the TCGA dataset; for example, radiotherapy data was missing in ~20% of the data, (2) the clinical follow-up was different between METABRIC (approximately 30 years) and TCGA (majority less than 10 years), (3) confounding effects of systemic therapy were not taken into account in our study since this information was not present in the TCGA dataset and (4) our analysis is focused on OS due to incomplete clinical information for disease free survival. Nevertheless, from our studies we think we can conclude that there is a significant survival benefit of radiotherapy for especially younger patients with tumors of the luminal A and basal molecular subtypes, but not luminal B and HER2 driven subtypes. Future studies will need to be conducted to identify predictive biomarker panels that can accurately identify older, luminal A and basal subtype patients most likely to benefit from radiotherapy treatment, which can then be validated in prospective cohorts to ultimately improve clinical practice. Currently all breast cancer patients are treated with the same RT regimen with no prior knowledge as to which tumors are likely to respond and which are not. Our study takes an initial step towards more personalized treatment with the ultimate goal of reducing overtreatment with radiotherapy while retaining low breast cancer mortality.  35,36 . TCGA specimens were collected from newly diagnosed patients with invasive breast adenocarcinoma at different US based tissue source sites. METABRIC primary fresh-frozen breast cancer Figure 4. Effect of radiotherapy on overall survival in older breast cancer patients across different molecular subtypes. Kaplan-Meier overall survival curves comparing survival for breast cancer patients diagnosed at age > 60 years ("old") who did or did not receive radiotherapy across different molecular subtypes: luminal-A (A), luminal-B (B), HER2 (C), basal (D) (METABRIC cohort). P-values were obtained using the log rank (Mantel-Cox) test. specimens were obtained from tumor banks in the UK and Canada. Additional clinical data including PAM50, clinical stage, surgery, patient age at diagnosis and radiotherapy for the TCGA cohort was obtained from the UCSC Genome Browser and UCSC Cancer Browser (http://genome-cancer.ucsc.edu). Summary details of the two patient cohorts are presented in Table 1. With regard to PAM50 data, the small normal-like group was excluded because of doubt as to the relevance of this group (material analyzed may not have contained cancer) and the generally perceived lack of clinical relevance of this group. The claudin-low group was also excluded for being only available in METABRIC.

Methods
Survival and statistical analysis. All survival and statistical analyses were performed using SPSS. Patients were stratified based on clinical stage, surgery type, age and molecular subtype (based on PAM50). Kaplan-Meier survival curves were generated to show differences in overall survival (p-values were generated using log rank (Mantel-Cox) test) between patients with and without radiotherapy. Multivariate analyses were carried out to examine whether radiotherapy has an independent benefit for survival when adjusting for other covariates (age, ER, PR, grade, tumor size) or the molecular subtypes using Cox proportional-hazard regression. Significance level was set at p < 0.05.