Data points indicate observed data.
A, Estimated mortality among all patients if surgical procedure were randomly assigned (analysis of average treatment effect). B-D, Estimated mortality from another surgical procedure among patients who had a specific surgical procedure (analysis of average treatment effect for those treated). B, For patients receiving breast-conserving surgery with radiation, bilateral mastectomy would have resulted in marginally higher mortality, on average, and unilateral mastectomy in higher mortality. C, For patients receiving unilateral mastectomy, bilateral mastectomy would have resulted in unchanged mortality and breast-conserving surgery with radiation in lower mortality. D, For patients receiving bilateral mastectomy, breast-conserving surgery with radiation would have resulted in unchanged mortality and unilateral mastectomy in higher mortality.
eTable. Patient Characteristics According to Surgery Type: Bilateral Mastectomy, Breast-Conserving Surgery With Radiation, and Unilateral Mastectomy, Stages 0-III Breast Cancer, 1998-2011, California
Customize your JAMA Network experience by selecting one or more topics from the list below.
Kurian AW, Lichtensztajn DY, Keegan THM, Nelson DO, Clarke CA, Gomez SL. Use of and Mortality After Bilateral Mastectomy Compared With Other Surgical Treatments for Breast Cancer in California, 1998-2011. JAMA. 2014;312(9):902–914. doi:10.1001/jama.2014.10707
Copyright 2014 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
Bilateral mastectomy is increasingly used to treat unilateral breast cancer. Because it may have medical and psychosocial complications, a better understanding of its use and outcomes is essential to optimizing cancer care.
To compare use of and mortality after bilateral mastectomy, breast-conserving therapy with radiation, and unilateral mastectomy.
Design, Setting, and Participants
Observational cohort study within the population-based California Cancer Registry; participants were women diagnosed with stages 0-III unilateral breast cancer in California from 1998 through 2011, with median follow-up of 89.1 months.
Main Outcomes and Measures
Factors associated with surgery use (from polytomous logistic regression); overall and breast cancer–specific mortality (from propensity score weighting and Cox proportional hazards analysis).
Among 189 734 patients, the rate of bilateral mastectomy increased from 2.0% (95% CI, 1.7%-2.2%) in 1998 to 12.3% (95% CI, 11.8%-12.9%) in 2011, an annual increase of 14.3% (95% CI, 13.1%-15.5%); among women younger than 40 years, the rate increased from 3.6% (95% CI, 2.3%-5.0%) in 1998 to 33% (95% CI, 29.8%-36.5%) in 2011. Bilateral mastectomy was more often used by non-Hispanic white women, those with private insurance, and those who received care at a National Cancer Institute (NCI)–designated cancer center (8.6% [95% CI, 8.1%-9.2%] among NCI cancer center patients vs 6.0% [95% CI, 5.9%-6.1%] among non-NCI cancer center patients; odds ratio [OR], 1.13 [95% CI, 1.04-1.22]); in contrast, unilateral mastectomy was more often used by racial/ethnic minorities (Filipina, 52.8% [95% CI, 51.6%-54.0%]; OR, 2.00 [95% CI, 1.90-2.11] and Hispanic, 45.6% [95% CI, 45.0%-46.2%]; OR, 1.16 [95% CI, 1.13-1.20] vs non-Hispanic white, 35.2% [95% CI, 34.9%-35.5%]) and those with public/Medicaid insurance (48.4% [95% CI, 47.8%-48.9%]; OR, 1.08 [95% CI, 1.05-1.11] vs private insurance, 36.6% [95% CI, 36.3%-36.8%]). Compared with breast-conserving surgery with radiation (10-year mortality, 16.8% [95% CI, 16.6%-17.1%]), unilateral mastectomy was associated with higher all-cause mortality (hazard ratio [HR], 1.35 [95% CI, 1.32-1.39]; 10-year mortality, 20.1% [95% CI, 19.9%-20.4%]). There was no significant mortality difference compared with bilateral mastectomy (HR, 1.02 [95% CI, 0.94-1.11]; 10-year mortality, 18.8% [95% CI, 18.6%-19.0%]). Propensity analysis showed similar results.
Conclusions and Relevance
Use of bilateral mastectomy increased significantly throughout California from 1998 through 2011 and was not associated with lower mortality than that achieved with breast-conserving surgery plus radiation. Unilateral mastectomy was associated with higher mortality than were the other 2 surgical options.
Randomized trials have demonstrated similar survival for patients with early-stage breast cancer treated with breast-conserving surgery and radiation or with mastectomy.1,2 However, older data show increasing use of mastectomy, and particularly bilateral mastectomy, among US patients with breast cancer.3-5 Bilateral mastectomy represents both treatment (for the affected breast) and prevention (for the contralateral breast), with the uncommon exception of patients having bilateral tumors. The causes of the increasing use of bilateral mastectomy are unknown; one possibility is the dissemination of sensitive diagnostic tests such as breast magnetic resonance imaging and genetic testing of BRCA1 (unigene cluster number Hs.194143) and BRCA2 (unigene cluster number Hs.34012).4,6 Although it may be cited as a justification for bilateral mastectomy, evidence for a survival benefit appears limited to rare patient subgroups, including women with BRCA1/2 mutations or strong family history of cancer.7-9
Because bilateral mastectomy is an elective procedure for unilateral breast cancer and may have detrimental effects in terms of complications and associated costs10,11 as well as body image and sexual function,12,13 a better understanding of its use and outcomes is crucial to improving cancer care. Because patients’ preferences drive its use, patients are unlikely to accept randomization to a less extensive surgical procedure in a clinical trial; thus, observational studies offer a feasible alternative to address an important clinical question. To minimize selection bias, we designed a population-based study of the use and outcomes of bilateral mastectomy compared with other surgical treatments, using the California Cancer Registry (CCR, part of the National Cancer Institute [NCI] Surveillance, Epidemiology and End Results [SEER] program), which comprises about 99% of all breast cancer cases statewide.
The study population consisted of all female California residents newly diagnosed with a first primary breast cancer (International Classification of Diseases–Oncology, 3rd edition, morphology codes C50.0-50.9), of American Joint Commission on Cancer stages 0-III, from January 1, 1998, through December 31, 2011. Approval for human subjects research was obtained from the Cancer Prevention Institute of California institutional review board. We obtained CCR data routinely abstracted from medical records on age at diagnosis, race/ethnicity (from patients’ medical records and registry categorization; assessed because prior research indicates that the use of and survival after surgical procedures vary by race/ethnicity, and because we aimed to evaluate these associations in a population-based context), marital status, stage, tumor grade, tumor size, histology, lymph node involvement, metastasis, and biomarkers.14 Tumors with histologic morphology codes 8500-8508 and 8521-8523 were coded as ductal and those with codes 8520 and 8524-8525 as lobular. We also obtained CCR information on initial treatment (surgery, chemotherapy, and radiation therapy), primary health insurance, census block group of residence at diagnosis, and vital status (determined by CCR through hospital follow-up and database linkages, including the Social Security Administration) as of December 31, 2010, and, for the deceased, the underlying cause of death.
Estrogen-receptor and progesterone-receptor status were each categorized as positive (≥5% nuclear staining), negative, borderline, not tested, not recorded, or unknown. Tumors were considered estrogen receptor–/progesterone receptor–positive if they were estrogen receptor–positive, progesterone receptor–positive, or both, and as estrogen receptor–/progesterone receptor–negative if both were negative. Given that CCR did not systematically collect v-erb-b2 avian erythroblastic leukemia viral oncogene homologue 2 (ERBB2, also known as HER-2/neu, unigene cluster number Hs.446352) testing results before 2006, ERBB2 data are not included.
For each case, we assigned a previously developed measure of neighborhood socioeconomic status (SES). For cases diagnosed in 1998-2005, we used a measure of neighborhood-level SES quintiles based on distribution across California, incorporating block group-level data from the 2000 Census on income, education, housing costs, and occupation.15 For cases diagnosed in 2006-2010, we used data from the American Community Survey of the US Census to derive a similar index.
The CCR records the facility reporting each case. Using the aforementioned index, we determined the SES distribution of all cases for each facility and identified facilities that were NCI–designated cancer centers.
We used polytomous logistic regression to model surgery use. Survival time was measured in days from diagnosis to death. Women who died from other causes were censored at time of death for the analysis of breast cancer–specific mortality. Women alive at the time of last follow-up or December 31, 2010, were censored then. We used Cox proportional hazards to model the association of various factors with overall and breast cancer–specific mortality. The proportional hazards assumption was confirmed by testing the correlation of Schoenfeld residuals with time. For both models (surgery use and mortality), covariates included age, race/ethnicity, tumor size, grade, histology, nodal and estrogen receptor/progesterone receptor status, receipt of adjuvant chemotherapy and radiation, neighborhood SES quintile, marital and insurance status, the SES composition of patients at the reporting hospital, care at an NCI-designated cancer center, and diagnosis year. Stage was included as a stratifying variable in the Cox regression, allowing baseline hazards to vary by stage. Multicollinearity in the models was assessed using the variance inflation factor. We did not test for a priori interactions but did conduct stratified analyses by age and stage. Missing data were coded as unknown and retained as a separate category for analyses.
We used SAS version 9.3 for all analyses except those of surgical use trends, for which we used Joinpoint (Joinpoint Regression Program version 4.0.4 [Statistical Research and Applications Branch, NCI]). This program uses Monte Carlo Permutation tests to model data and identify up to 3 points (“joinpoints”) at which there was a statistically significant change in linear trend.16 Results of joinpoint analysis were used to inform grouping of diagnosis years in logistic regression analysis.
Propensity score analyses defined surgery type as the patient attribute for which scores were calculated.17 We used generalized boosting models, a nonparametric machine-learning classifier, in the R package twang, setting the search limit to 15 000 trees.18 All independent variables in Table 1 and Table 2 were used to calculate per-patient scores, except 3 variables highly correlated with others (radiation therapy with surgery type; chemotherapy and adjuvant treatment with administration of chemotherapy before or after the surgical procedure).
We used graphical analysis to assess the postbalance maximum standardized effect difference for each variable17 and calculated weights for the average treatment effect (average outcome for the whole population after one surgery vs another); and average treatment effect for those treated (average outcome for those treated after one surgery vs another). The svykm and svylogrank functions from the survey package19 were used to calculate weighted Kaplan-Meier curves and P values; the svycoxph function was used for weighted Cox proportional hazard models, with outcome regressed on treatment and stratified by stage. Weighted CIs for mortality rates were calculated by the survfit function in the R survival package.
A total of 291 117 stages 0-III breast cancer cases were diagnosed and reported to CCR from January 1, 1998, through December 31, 2011. Cases were excluded if missing essential data for categorization or if ineligible for breast-conserving surgery with radiation according to practice guidelines,20 as follows: diagnosed by death certificate or autopsy only (n = 33); tumor larger than 5 cm or unknown, microscopic or diffuse tumor, Paget disease of breast or mammographic diagnosis only, or inflammatory carcinoma (n = 41 853); no pathology report confirmation (n = 283); unknown lymph node involvement (n = 1771); surgery other than bilateral mastectomy, breast-conserving surgery with radiation, or unilateral mastectomy (n = 52 343); and diagnosis of bilateral tumors or a second primary breast tumor within 60 days (n = 5100), resulting in 189 734 women included in analyses of surgery use. Mortality analyses excluded women diagnosed after 2010 because of incomplete mortality data for 2011 (n = 14 331), those having zero or invalid survival time (n = 11), and those having unknown cause of death (n = 475). Mortality analyses included 174 917 women; median follow-up time was 89.1 months (interquartile range, 54.8-129.9 months).
The proportions of all patients who underwent each surgery were 6.2% (95% CI, 6.1%-6.3%) for bilateral mastectomy, 55.0% (95% , 54.8%-55.3%) for breast-conserving surgery with radiation; and 38.8% (95% CI, 38.6%-39.0%) for unilateral mastectomy (Table 1 and eTable in the Supplement). Among all patients, the rate of bilateral mastectomy increased from 2.0% (95% CI, 1.7%-2.2%) in 1998 to 12.3% (95% CI, 11.8%-12.9%) in 2011, an annual increase of 14.3% (95% CI, 13.1%-15.5%) (Table 2 and eTable). The increase in bilateral mastectomy rate was greatest among women younger than 40 years: the rate increased from 3.6% (95% CI, 2.3%-5.0%) in 1998 to 33.0% (95% CI, 29.8%-36.5%) in 2011, increasing by 17.6% (95% CI, 14.9%-20.4%) annually. Use of unilateral mastectomy declined in all age groups (Figure 1).
Factors associated with having undergone bilateral mastectomy (vs breast-conserving surgery with radiation) included age younger than 50 years, non-Hispanic white race/ethnicity, larger tumor size, nodal involvement, lobular histology, higher grade or estrogen receptor–/progesterone receptor–negative status, care at a hospital predominantly serving patients with lower SES or at an NCI-designated cancer center, having higher neighborhood SES, and recent diagnosis. Factors inversely associated with having undergone bilateral mastectomy (vs breast-conserving surgery with radiation) included age 65 years or older, minority race/ethnicity, receipt of adjuvant therapy, married status, and insurance type other than private (Table 3).
Characteristics associated with having undergone unilateral mastectomy (vs breast-conserving surgery plus radiation) included diagnosis at age other than 50 to 64 years, Asian, Hispanic, and American Indian race/ethnicity (with notable associations for Filipina and Hispanic women vs non-Hispanic white women), larger tumor size, nodal involvement, lobular histology, higher grade, estrogen receptor–/progesterone receptor–negative status, married status, public/Medicaid insurance, or care at a hospital predominantly serving patients of lower SES (Table 3). Factors inversely associated with having unilateral mastectomy (vs breast-conserving surgery with radiation) included black race, receipt of adjuvant therapy, care at an NCI-designated cancer center, higher neighborhood SES, and recent diagnosis.
Compared with breast-conserving surgery with radiation, bilateral mastectomy was not associated with a mortality difference (hazard ratio [HR], 1.02 [95% CI, 0.94-1.11]), whereas unilateral mastectomy was associated with higher mortality (HR, 1.35 [95% CI, 1.32-1.39]) (Table 4). Other factors associated with overall mortality included age 65 years or older or younger than 40 years, black race, larger tumor size, nodal involvement, higher grade, estrogen receptor–/progesterone receptor–negative status, lower neighborhood SES, unmarried status, having Medicare or public/Medicaid insurance, and receiving care at a hospital predominantly serving patients of lower SES. Higher mortality was associated with unilateral mastectomy in all age groups. Similar mortality between bilateral mastectomy and breast-conserving surgery with radiation was observed in all age groups except women 65 years or older, whose survival was slightly better after breast-conserving surgery with radiation. Findings were similar for breast cancer–specific mortality (Table 5). Compared with unilateral mastectomy, bilateral mastectomy was associated with lower overall mortality (HR, 0.75 [95% CI, 0.70-0.82]) and breast cancer–specific mortality (HR, 0.85 [95% CI, 0.76-0.94]).
Figure 2A shows estimated mortality among all patients if surgical procedure were randomly assigned (analysis of average treatment effect). The estimated 10-year mortality rates were 18.8% (95% CI, 18.6%-19.0%) for bilateral mastectomy, 16.8% (95% CI, 16.6%-17.1%) for breast-conserving surgery with radiation, and 20.1% (95% CI, 19.9%-20.4%) for unilateral mastectomy. Figure 2B-D shows estimated mortality from another surgical procedure among patients who had a specific surgical procedure (analysis of average treatment effect for those treated). For patients receiving breast-conserving surgery with radiation, bilateral mastectomy would have resulted in marginally higher mortality, on average, and unilateral mastectomy in higher mortality. For patients receiving unilateral mastectomy, bilateral mastectomy would have resulted in unchanged mortality and breast-conserving surgery with radiation in lower mortality. For patients receiving bilateral mastectomy, breast-conserving surgery with radiation would have resulted in unchanged mortality and unilateral mastectomy in higher mortality. Proportional hazards regression models showed similar results (Table 6).
This observational study comprising 189 734 women with unilateral early-stage breast cancer compared 3 surgical treatments and found a substantial increase in the rate of bilateral mastectomy throughout California from 1998 through 2011. To our knowledge, this is the first side-by-side comparison of all 3 common surgical treatments for early-stage breast cancer. Previous SEER studies have compared 2 treatments at a time: some reported a survival advantage with bilateral vs unilateral mastectomy21,22 and others reported improved survival after breast-conserving surgery with radiation compared with unilateral mastectomy.23,24 By comparing all 3 surgical options for a patient with early-stage breast cancer, we found no mortality benefit associated with bilateral mastectomy compared with breast-conserving surgery, and higher mortality associated uniquely with unilateral mastectomy.
For the surgical treatment of early-stage breast cancer, available randomized trial data are limited to those showing no survival difference between unilateral mastectomy and breast-conserving surgery.1,2 There is no randomized trial evidence to inform whether bilateral mastectomy improves survival, and it is unlikely that such a trial will ever be performed. Thus, conclusions about surgical treatments must rely on observational studies that compare the effectiveness of different procedures in practice21,22,25,26; however, a recent meta-analysis judged the existing data inadequate to enable conclusions about the effect of bilateral mastectomy on survival.27 Patient selection attributable to unmeasured factors probably explains much of the higher mortality that we observed with unilateral mastectomy relative to the other 2 surgical procedures. In prior SEER-based studies, both we24 and Agarwal et al23 reported worse survival associated with unilateral mastectomy vs breast-conserving surgery with radiation, results that persisted after propensity analysis. We agree with previous suggestions that patients with tumor features suggesting poor prognosis, such as lymphovascular invasion or extranodal extension, which SEER does not record and for which we cannot control, are more likely to undergo unilateral mastectomy than breast conservation and also to experience worse survival.23,24 The current study offers another potential explanation, namely confounding related to sociodemographic differences between women who underwent bilateral mastectomy and women who underwent unilateral mastectomy.
Women who underwent bilateral mastectomy were more likely to be non-Hispanic white and privately insured, to live in high SES neighborhoods, and to be treated in NCI-designated cancer centers. By contrast, women who underwent unilateral mastectomy were more likely to be Asian, Hispanic, or non-Hispanic American Indian/other/unknown; to have public/Medicaid insurance, and to be treated in hospitals serving patients of lower SES; they were less likely to live in high SES neighborhoods or to be treated in NCI-designated cancer centers. Cancer registry data lack details about comorbidities and specific regimens of endocrine, radiation, and chemotherapy. However, prior studies enriched for clinical data, including our own within the Kaiser Permanente Northern California health care system, reported treatment-limiting comorbidities (for example, diabetes and myocardial infarction) and reduced treatment intensity among the same racial/ethnic minority, low SES patients who most frequently underwent unilateral mastectomy in our current study.28-30 In addition to signifying unmeasured poor prognostic factors,21,22 unilateral mastectomy might correlate with subtle disparities in effective access (for example, diabetic neuropathy that limits chemotherapy dosing; lack of transportation to the postsurgical radiation treatments required for breast conservation) that we could not identify using registry data and that may mediate higher mortality. By contrast, patterns of bilateral mastectomy use suggest that affluent non-Hispanic white women, women of high SES, or both seek more aggressive preventive care, consistent with reported associations between greater use of expensive diagnostic tests (such as breast MRI and genetic testing) and bilateral mastectomy within this patient subgroup.4,31
The increase in bilateral mastectomy use despite the absence of supporting evidence has puzzled clinicians and health policy makers. Proposed explanations include the increasing use of highly sensitive breast magnetic resonance imaging, with increases in anxiety-producing recall and biopsy rates that may drive patients to undergo preventive surgery,6,31,32 and the dissemination of genetic testing, which facilitates identification of high-risk patients who benefit from bilateral mastectomy.7,8,33 Although fear of cancer recurrence may prompt the decision for bilateral mastectomy, such fear usually exceeds the estimated risk.34,35 Other studies found recurrence fears less influential than aesthetic considerations, notably those that arise with new reconstruction approaches that achieve cosmetic symmetry through bilateral tissue flap placement.6,36 Because cosmesis may be inferior if both breasts are not reconstructed simultaneously, these new approaches encourage use of immediate bilateral mastectomy. We found that bilateral mastectomy use over time increased most among patients younger than 40 years at diagnosis, which may be attributable to their relatively high probability of carrying genetic mutations (an evidence-based indication for bilateral mastectomy)37 or to the greater likelihood that they have young children and may therefore seek maximal intervention in hope of extending their lives (an emotional rather than evidence-based decision).34,35,38 Although some studies reported patient satisfaction after bilateral mastectomy,39 others observed deleterious effects on body image, sexual function, and quality of life12; moreover, repeat operations and complications (including flap failure, necrosis, and infection) are substantially more common with bilateral mastectomy than with other surgical procedures.10,11
In a time of increasing concern about overtreatment,40 the risk-benefit ratio of bilateral mastectomy warrants careful consideration and raises the larger question of how physicians and society should respond to a patient’s preference for a morbid, costly intervention of dubious effectiveness.
Our study used a population-based statewide data set, multiple regression analysis, and propensity scores. However, given its observational design, it cannot prove causation and may be subject to selection bias and uncontrolled confounding. As discussed above, unmeasured patient selection factors related to cancer prognosis and access to care may explain the higher mortality observed with unilateral mastectomy. Other limitations include the lack of SEER data on diagnostic testing (eg, magnetic resonance imaging, genetic testing for BRCA1/2 and other inherited mutations, tumor analysis for ERBB2 amplification, and broader genomic profiling), details of systemic treatments, family cancer history, and comorbidities. Additional information gaps include patient preferences and physician recommendations, which influence surgical decisions.38 Future research with more comprehensive data sets that integrate detailed clinical, treatment, and patient-reported information will be essential to advance understanding of breast surgery use and to enhance the quality of cancer care.
Among all women diagnosed with early-stage breast cancer in California, the percentage undergoing bilateral mastectomy increased substantially between 1998 and 2011, despite a lack of evidence supporting this approach. Bilateral mastectomy was not associated with lower mortality than breast-conserving surgery plus radiation, but unilateral mastectomy was associated with higher mortality than the other options. These results may inform decision-making about the surgical treatment of breast cancer.
Corresponding Author: Scarlett L. Gomez, PhD, Cancer Prevention Institute of California, 2201 Walnut Ave, Ste 300, Fremont, CA 94538 (firstname.lastname@example.org).
Author Contributions: Drs Kurian and Gomez had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Kurian, Clarke, Gomez.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Kurian, Nelson, Clarke, Gomez.
Critical revision of the manuscript for important intellectual content: Kurian, Lichtensztajn, Keegan, Nelson, Clarke, Gomez.
Statistical analysis: Lichtensztajn, Nelson.
Obtained funding: Kurian, Gomez.
Administrative, technical, or material support: Clarke, Gomez.
Study supervision: Kurian, Clarke, Gomez.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Clarke and Dr Gomez reported receiving grants from Genentech outside the submitted work. No other authors reported disclosures.
Funding/Support: This study was supported by the Jan Weimer Junior Faculty Chair in Breast Oncology, the Suzanne Pride Bryan Fund for Breast Cancer Research at Stanford Cancer Institute, and the NCI SEER program under contract HHSN261201000140C awarded to the Cancer Prevention Institute of California (CPIC). The collection of cancer incidence data was supported by the California Department of Health Services as part of the statewide cancer reporting program mandated by California Health and Safety Code Section 103885; the NCI SEER program under contracts HHSN261201000140C awarded to CPIC, HHSN261201000035C to the University of Southern California, and HHSN261201000034C to the Public Health Institute; and the CDC National Program of Cancer Registries, under agreement 1U58 DP000807-01 awarded to the Public Health Institute.
Role of the Funders/Sponsors: None of the funders/sponsors had any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.
Disclaimer: The ideas and opinions expressed herein are those of the authors, and endorsement by the University or State of California, the California Department of Health Services, the NCI, or the CDC or their contractors and subcontractors is not intended and should not be inferred.
Create a personal account or sign in to: