Second primary breast cancer after unilateral mastectomy alone or with contralateral prophylactic mastectomy

Abstract Background An increasing number of patients undergo contralateral prophylactic mastectomy (CPM) for unilateral breast cancer. However, the benefit of CPM has not been quantified in the setting of contemporary breast cancer therapy. Methods We performed an analysis of 180 068 patients in the Surveillance, Epidemiology, and End Results (SEER) database, diagnosed with unilateral ductal breast carcinoma between 1998 and 2013 and treated with unilateral mastectomy (UM) or CPM. UM was performed in 146 213 patients (81.2%); CPM was performed in 33 855 patients (19.8%). Primary outcome of interest was cumulative incidence of a second primary breast cancer in the ipsilateral or contralateral breast greater than 3 months after initial diagnosis. Cumulative incidence analysis was based on a Cox proportional model to generate curves of second primary breast cancer in any breast, ipsilateral breast only, or contralateral breast only. Results Patients who underwent CPM had a significantly reduced incidence of second primary breast cancer 10 and 15 years after surgery (CPM 0.93% [0.73%, 1.12%] vs UM 4.44% [4.28%, 4.60%]). Patients who underwent CPM had significantly lower adjusted hazard of second primary breast cancer when compared with UM (HR 0.38 vs 1.0, P < .0001). Conclusions CPM offers some protection from a second primary breast cancer, attributable to a reduced incidence in the contralateral breast. These findings provide additional information to providers and patients as they make decisions regarding surgical management. They should also be interpreted in the context of the absolute incidence of second primary breast cancer after UM and previous literature demonstrating no survival benefit.

after initial diagnosis. 1 However, the past 15 years have seen increasing utilization of CPM in low-risk patients with unilateral breast cancer; fear of a second primary breast cancer drives the desire for CPM, even among patients at low risk for CBC. 4,5 To facilitate effective shared decision making with respect to CPM, there is a need for contemporary population-level data on the incidence of any second primary breast cancer or CBC, 6 especially as dissemination of information to the patient is recognized as a contributor to decision making and risk assessment. 7,8 Population-based studies have shown that, in the absence of a predisposition toward developing breast cancer, patients who undergo CPM do not experience improved survival. [9][10][11][12][13] In addition, patients who undergo CPM are at risk for longer operative and hospitalization times, and those who undergo subsequent reconstruction are at a further higher risk for complications related to the reconstructive course. 14 Furthermore, patients who undergo CPM experience increased medical costs with similar quality of life when compared with patients who undergo unilateral mastectomy (UM) and subsequent surveillance. 15 Therefore, it is imperative that information relating the risk of a CBC be conveyed to patients who are considering a CPM.
There are limited data on the incidence of a second primary breast cancer in patients who have had mastectomy for breast cancer 16,17 ; this is even less well understood in patients who have had CPM, especially on a population level. 6,18,19 Small studies with limited cohorts have demonstrated varying impact of CPM on CBC; however, these studies have also demonstrated improved survival after CPM 18,19 -a finding which is now disputed through larger population-level studies. [9][10][11][12][13] Population-level information regarding the risk of a second primary breast cancer with UM or CPM would help patients and physicians make informed decisions regarding the benefit of CPM, especially when taking into consideration that studies have demonstrated no improved survival with CPM. In this study we use population-level data from the Surveillance, Epidemiology, and End Results (SEER) database to determine the incidence and hazard ratio of a second primary breast cancer among patients treated with UM or CPM.

| Data source
Population-level de-identified data were extracted from the SEER cancer database for patients with a primary diagnosis of ductal carcinoma of the breast during the years 1998 through 2013 (Surveillance, Epidemiology,

| Statistical analysis
Patients were divided into two categories-those receiving unilateral mastectomy alone (UM) or unilateral mastectomy with contralateral prophylactic mastectomy (CPM). Patients were determined to have a diagnosis of a second primary breast cancer if they had a subsequent diagnosis of primary breast cancer in the ipsilateral or contralateral breast greater than 3 months after the initial primary breast cancer. Cumulative incidence analysis was based on a Cox proportional model to generate curves of second primary breast cancer in breast, ipsilateral breast only, or contralateral breast only were plotted. Kaplan-Meier analysis was used to generate point estimates for 10-and 15-year cumulative rates. Time-to-event methods were used to analyze the outcome, with time calculated as the months from the date of first breast cancer to the earliest of the following events: second primary breast cancer diagnosis, death, loss of follow-up, or administrative cut-off; for time to second primary breast cancer in the contralateral breast, events were censored at the time of a second ipsilateral primary breast cancer. The proportion with second primary breast cancer diagnosis, accounting for the competing risk of death, was estimated using cumulative incidence estimation ( Figure S1). Fine and Grey competing risk regression was used for all the models. Multivariate model selection compared a base model adjusting only for mastectomy type (UM or CPM), then adding demographic information, and then oncologic information. Models were compared using the maximum likelihood method and comparing Bayesian information criterion (BIC) values. All analyses were performed using R (ver. 3.3.2)[R Core Team, 2013] and packages cmprsk for analysis, Gmisc for plot and table output, and knitr for reproducible research.

| Population characteristics
A total of 180 068 patients (age range 15-105 years, mean 57.8 years (SD 14.3 years)) who underwent mastectomy for a diagnosis of unilateral breast ductal carcinoma were included in this study. Unilateral mastectomy alone (UM) was performed in 146 213 patients (81.2%) while unilateral mastectomy with contralateral prophylactic mastectomy (CPM) was performed in 33 855 patients (19.8%) ( Table 1). CPM patients tended to be younger when compared with UM patients (50 years vs 60 years, P < .0001). In addition, population composition by race was different; a higher proportion of CPM patients were White when compared with UM patients (86% vs 77%, P < .001) ( Table 1).

| Second primary breast cancer incidence
The 10-and 15-year Kaplan-Meier cumulative rate estimates of a second primary breast cancer was substantially lower among CPM patients when compared with UM patients (10- Figure 1). This was attributable to a reduced cumulative incidence of second primary breast cancer in the contralateral breast among CPM patients when compared with UM patients ( Figure 2); a similarly substantial reduction was not observed in the cumulative incidence of second primary breast cancer in the ipsilateral breast ( Figure 3).
Among those with a second primary breast cancer, CPM patients had a significantly reduced proportion of second primary breast cancer in the contralateral breast when compared with UM patients (65% vs 93%, χ 2 = 221, P < .0001).
The mean time to diagnosis of second primary breast cancer was 35 months in CPM patients and 55 months in UM patients (Table 2); the mean size of the second primary breast cancer at time of diagnosis was similar (4 mm) between the two cohorts ( Table 2). Average annual incidence of a second primary breast cancer was higher in the contralateral breast for patients who underwent UM (0.38%) versus CPM (0.04%) ( Table 3).

| Hazard ratio of second primary breast cancer
Multivariate logistic regression was performed to determine the hazard of a second primary breast cancer while accounting for demographic and oncologic factors. CPM patients had a significantly lower hazard of second primary breast cancer when compared with UM (HR 0.38 vs 1.0, P < .0001) ( Table 4). Furthermore, CPM patients had a significantly lower hazard of second primary breast cancer in the contralateral breast when compared with UM patients (HR 0.27 vs 1.0, P < .0001) ( Table 5).

| DISCUSSION
Previous studies have demonstrated that, among patients without a clear predisposition toward developing breast cancer, contralateral prophylactic mastectomy (CPM) does not improve overall or cancer-specific survival when compared with unilateral mastectomy. [9][10][11][12][13] Despite these findings, rates of CPM continue to increase. 12,13,[21][22][23][24][25] Although the reasons for this trend are under investigation, 2,5,7,21,22,26-28 patient anxiety and overestimation of personal contralateral breast cancer risk (CBC) have received substantial attention as a driving force for the increasing utilization of CPM. 4,15,21,24,27,29 Using patient-directed surveys, one previous study has shown that patients overestimate their risk for CBC, with 10-year self-reported risk estimates over 30%. 29 However, these self-reported risk assessments contrast sharply with several population-based studies-Reiner et al reported the 10-year cumulative risk of contralateral breast cancer (CBC) to range from 4.0% to 7.0% among patients with no family history of breast cancer or BRCA1/2 mutation. 17 Even among patients with a family history of bilateral breast cancer, they reported the 10-year cumulative risk of CBC to range from 13% to 24%. 17 However, Reiner et al included only a limited cohort of 1700 patients, and did not directly compare UM with CPM; or did they provide a population breakdown of patients undergoing UM or CPM. Gao et al reported the 10year incidence of CBC to be 6.1% using the SEER database (1973 through 1996) with over 130 000 patients. 16 Similar to Reiner et al, Gao et al did not compare UM with CPM. Using the SEER database, we found the cumulative incidence of CBC among patients with UM to be 4.44% (95% CI [4.28%, 4.60%]). In contrast to the findings reported by Gao et al, our study includes a more contemporary cohort of patients. This is particularly critical, as breast cancer therapy has evolved with improvements in adjuvant radiation and chemotherapy, as well as recommendations for antiestrogen therapy. [30][31][32][33] Our study also includes a broader regional representation due to the expansion of  30,34 We next determined the hazard of CBC among patients who undergo CPM, relative to UM. One prior study by Herrinton et al used a limited cohort of patients diagnosed with breast cancer from 1979 through 1999 (n = 1400) and found the hazard of CBC after CPM to be as low as 0.03, relative to UM. 19 However, that same study found the hazard of death to be significantly lower among patients with CPM, when compared with UM, which is not supported by contemporary literature. [11][12][13]199 In contrast to this study by Herrinton et al, our study includes over 180 000 patients. While we found a significantly reduced hazard of CBC with CPM, our reported reduction was less striking (HR 0.27 vs 1.0). Therefore, CPM does not eliminate the risk of CBC as suggested by Herrinton et al; although this may be obvious to providers, incomplete protection from CBC should be conveyed to patients who desire CPM when considering the already low risk of CBC in patients without predisposing genetic factors. Our findings may be explained by use of a contemporary cohort with a lower incidence of CBC among patients with UM potentially due to improvements in adjuvant radiation therapy, chemotherapy, and antiestrogen therapy. Consistent with this hypothesis, we also found that higher stage or ER-positive first primary breast cancers were associated with reduced hazard of CBC. Several studies have previously reported that receipt of adjuvant chemotherapy or antiestrogen therapy is both associated with reduced risk of CBC. [30][31][32][33] These therapies are most applicable to patients with higher stage tumors or ER-positive tumors, respectively. Regardless, the reduction in hazard of CBC should be considered in the context of the absolute rates of CBC associated with UM, and the risk of morbidity associated with CPM. For example, the odds of wound and overall 30-day complications among patients who undergo CPM have been reported to be two times that of patients who undergo UM. 35 Additionally, patients who undergo breast reconstruction after CPM have increased odds of postoperative complications including infection. 14,36,37 Our study provides contemporary population-level analysis of the incidence and hazard ratio of second primary breast cancer in patients who undergo UM or CPM for unilateral ductal carcinoma; however, it is not without limitations. First, it is subject to the usual biases present in large population-based studies. For example, we are unable to control for tumor properties, such as lymphovascular invasion, and the accuracy of SEER reporting for chemotherapy and radiation therapy is controversial; we did, however, control for tumor stage-a major determinant of therapy. Second, we are unable to account for patient family history or genetic factors such as BRCA1/2 mutations. These patients have a higher risk of CBC and derive a substantial benefit from CPM. 1,24 However, it is likely that a higher proportion of patients with BRCA1/2 mutations comprise the group of CPM patients when compared with the group of UM patients in our analysis; therefore, our analysis may actually overestimate the benefit of CPM to patients who do not have BRCA1/2 mutations. Additionally, we employ a cut-off of 3 months for the diagnosis of a second primary breast cancer; while this time cut-off may be material to the categorization of an ipsilateral second primary breast cancer versus incomplete resection or recurrence, it likely does not impact the diagnosis of a second primary breast cancer in the contralateral breast. Furthermore, we do not perform a direct comparison between CPM and breast conservation therapy (BCT), which is typically unilateral; we did not perform this comparison in order to avoid the potential confounding of radiation therapy which is a component of BCT. 38 Overall, our analysis provides a crucial piece of information for clinicians and patients to consider when deciding upon CPM vs UM. While recent studies have been focused on the impact of CPM on patient survival, patient anxiety may not be derived directly from life or death. Instead, patients may be directed by a fear associated with the uncertainty of diagnosis, further surveillance, and the need for future treatment. A more complete understanding of how CPM may impact a patient's clinical course, which includes surveillance and subsequent treatments, may provide patients and physicians with the information they need to make decisions regarding surgical therapy. In addition, systemic therapy which may be initiated after an initial breast cancer diagnosis (eg, antiestrogen therapy) may provide a protective effect to lower the risk of CBC without requiring CPM. That guidelines now recommend up to 15 years of tamoxifen therapy may also provide a longer-term benefit against CBC. Our findings demonstrate that CPM reduces the hazard of a second primary breast cancer when compared with UM (HR 0.38), but also provides context for this reduction on the basis of cumulative incidence of a second primary breast cancer (10-year estimated K-M rate for UM: 4.44%).