The impact of risk-reducing hysterectomy and bilateral salpingo-oophorectomy on survival in patients with a history of breast cancer—A population-based data linkage study


  • Andreas Obermair,

    Corresponding author
    1. Queensland Centre for Gynaecological Cancer School of Medicine, The University of Queensland, Royal Brisbane and Women's Hospital, Brisbane, QLD, Australia
    • Correspondence to: Prof. Andreas Obermair, School of Medicine, The University of Queensland, Brisbane Royal Brisbane and Women's Hospital, 6th Floor, Ned Hanlon Building, Butterfield Street, Herston, QLD 4029, Australia, Tel.: +61-7-3636-8501, Fax: +617-3636-5289, E-mail:

    Search for more papers by this author
  • Danny R. Youlden,

    1. Cancer Council Queensland, Brisbane, QLD, Australia
    Search for more papers by this author
  • Peter D. Baade,

    1. Cancer Council Queensland, Brisbane, QLD, Australia
    2. School of Public Health, Queensland University of Technology, Brisbane, QLD, Australia
    Search for more papers by this author
  • Monika Janda

    1. School of Public Health, Queensland University of Technology, Brisbane, QLD, Australia
    Search for more papers by this author

  • Conflict of interest: A.O. is the managing director of He has a private practice in gynaecologic oncology. He has received travel support from Gate healthcare, and honoraria for speaking by Johnson and Johnson and Bayer


Prophylactic surgery including hysterectomy and bilateral salpingo-oophorectomy (BSO) is recommended in breast cancer susceptibility gene (BRCA)-positive women, whereas in women from the general population, hysterectomy plus BSO may increase the risk of overall mortality. The effect of hysterectomy plus BSO on women previously diagnosed with breast cancer is unknown. We used data from a population-base data linkage study of all women diagnosed with primary breast cancer in Queensland, Australia between 1997 and 2008 (n = 21,067). We fitted flexible parametric breast cancer-specific and overall survival models with 95% confidence intervals (also known as Royston–Parmar models) to assess the impact of risk-reducing surgery (removal of uterus, one or both ovaries). We also stratified analyses by age 20–49 and 50–79 years, respectively. Overall, 1,426 women (7%) underwent risk-reducing surgery (13% of premenopausal women and 3% of postmenopausal women). No women who had risk-reducing surgery compared to 171 who did not have risk-reducing surgery developed a gynaecological cancer. Overall, 3,165 (15%) women died, including 2,195 (10%) from breast cancer. Hysterectomy plus BSO was associated with significantly reduced risk of death overall [adjusted hazard ration (HR), 0.69; 95% confidence interval (CI), 0.53–0.89; p = 0.005]. Risk reduction was greater among premenopausal women, whose risk of death halved (HR, 0.45; 95% CI, 0.25–0.79; p < 0.006). This was largely driven by reduction in breast cancer-specific mortality (HR, 0.43; 95% CI, 0.24–0.79; p < 0.006). This population-based study found that risk-reducing surgery halved the mortality risk for premenopausal breast cancer patients. Replication of our results in independent cohorts and subsequently randomised trials are needed to confirm these findings.

Breast cancer is the most frequently diagnosed cancer and the leading cause of cancer death among females globally, accounting for 23% of total cancer cases and 14% of cancer deaths in 2008.[1] Breast cancer incidence has been rising in Asia and Africa,[2, 3] whereas rates have largely stabilised in North America, Europe and Australia,[4, 5] although in young women (25–39 years) an increase in breast cancer with distant involvement has been observed (United States SEER data 1996–2009[6]).

Risk factors for breast and uterine cancers are well described and include prolonged exposure to and higher concentrations of endogenous oestrogen.[7, 8] Women in Queensland (QLD), Australia (including mutation carriers) who were diagnosed with breast cancer subsequently have a more than 150% increased risk of developing uterine cancer and also a higher than 40% increased risk of developing ovarian cancer compared to the general population.[9] Risk-reducing hysterectomy and bilateral salpingo-oophorectomy (BSO) could reduce the risk of subsequent gynaecological and breast cancers in these patients.

In breast cancer susceptibility gene (BRCA) carriers, risk-reducing BSO significantly reduces ovarian cancer risk[10] and incidence of new breast cancers in premenopausal women.[11-13] Consequently, in BRCA carriers BSO decreases all-cause, breast cancer-specific and ovarian cancer-specific mortality.[11-13] However, only 5–6% of all breast cancers are directly attributable to inheritance, and the cumulative risk of developing breast cancer by age 70 for a mutation carrier in Australia is ~40%,[14] less than had been estimated from studies in other countries.[15] Furthermore, the cumulative risk of ovarian cancer by age 70 is estimated at 40–50% for BRCA1 mutation carriers and 10–25% for BRCA2 carriers.[16-19]

In women at average cancer risk without a previous diagnosis of breast cancer, two large prospective studies and one retrospective population-based cohort study found that hysterectomy plus BSO reduced risk of ovarian cancer by more than 96% and the risk of breast cancer in women 45 years or younger by 40%.[20-27] However, these benefits were counteracted by a significantly increased risk of death from other causes (e.g., cardiovascular disease) compared with women who preserved their ovaries, particularly among premenopausal women. In a meta-analysis of 12 case–control studies and a recent case–control study, hysterectomy alone without BSO was reported to reduce the risk of ovarian cancer by 34%[28] and breast cancer risk by 16%,[29] respectively. The exact mechanism of this is unknown, but it is suspected to be induced by reduced follicle-stimulating hormone levels.

Risk-reducing surgery could potentially form one of the many options in the breast cancer treatment armamentarium already complex to a degree that it requires decision-making algorithms. Currently, for patients diagnosed with breast cancer, the benefits and risks of BSO are unknown, especially for the majority (>90%) of patients who are BRCA1/2 negative. Therefore, we used a population-based data linkage approach to examine if patients with a personal history of breast cancer who had risk-reducing BSO with or without hysterectomy experienced different overall and breast cancer-specific survival compared to women with breast cancer who did not have prophylactic gynaecological surgery.



All cases of invasive breast cancer (ICD-O-3 code C50) diagnosed among women 20–79 years in Queensland (QLD) between 1997 and 2008 were selected from the population-based QLD Cancer Registry (QCR). Cases based on autopsy or death certificate only were excluded. Other data items available from the QCR included breast cancer cell type (morphology), indigenous status (self-identified), laterality and size of the tumour, number of lymph nodes surgically excised, number of lymph nodes positive as well as information regarding second primary cancers. Cause of death was ascertained through routine matching with the Australian National Death Index, with follow-up to December 31, 2009.

The QCR also holds a record of the most recent admission to every public and private hospital within QLD for each cancer patient. This facilitated a deterministic linkage between the QCR data and the QLD hospital admitted patient data collection for all admissions on or after the date of diagnosis of breast cancer until the end of 2009 as well as any gynaecological surgery that occurred between 1995 and 2009. Matching was performed using a unique hospital record number that was stored in both datasets. Once this link was in place we could then identify all admitted episodes of care for each woman during the study period. In particular, we were able to obtain details of breast cancer-related surgical treatment as well as any gynaecological surgery (BSO +/− hysterectomy). Data on selected comorbidities [atherosclerosis, cerebrovascular disease, cholesterol (hypercholesterolaemia), dementia, deep vein thrombosis, diabetes, heart disease, osteoporosis/bone fractures and pulmonary embolism] that were documented during admission were also obtained (see Table 1 for definitions). After the data linkage was completed, deidentified data were extracted by the data custodians for analysis.

Table 1. ICD-9-CM and ICD-10-AM procedure and disease codes
Type of surgery/diseaseICD-9-CM codesICD-10-AM codes
  1. a

    These procedures may also involve a USO or BSO. They were assigned to the category “hysterectomy only” for the main analysis, and sensitivity analyses were subsequently performed to determine what effect assigning these procedures to the categories “hysterectomy and USO” or “hysterectomy and BSO” would have on the results.

  2. Abbreviations: BSO: bilateral salpingo-oophorectomy; USO: unilateral salpingo-oophorectomy.

Gynaecological procedures
Hysterectomy and BSO 35653-03, 35673-01, 35753-01, 35756-02
Hysterectomy and USO 35653-02, 35673-00, 35753-00, 35756-01
Hysterectomy only68.3–68.935653-00, 35653-01, 35653-04a, 35657-00, 35661-00a, 35664-00a, 35664-01a, 35667-00a, 35667-01a, 35670-00a, 35673-02a, 35750-00, 35753-02a, 35756-00, 35756-03a, 90443-00, 90448-00, 90448-01, 90448-02a
BSO only65.51, 65.52, 65.61, 65.6235638-03, 35638-12, 35717-01, 35717-04
USO only65.3, 65.435638-02, 35638-11, 35638-13, 35713-07, 35713-11, 35717-05
Breast cancer-related procedures
Breast-conserving85.20–85.2330342-00, 30342-01, 30346-00, 30346-01, 30347-00, 30348-00, 30350-00, 30350-01, 31500-00, 31515-00
Mastectomy85.41–85.4830338-00, 30338-01, 30338-02, 30351-00, 30351-01, 30353-00, 30353-01, 30353-02, 30354-00, 30354-01, 30356-00, 30356-01, 31518-00, 31518-01, 31524-00, 31524-01, 30359-00, 30359-01, 30359-02, 30359-03, 30359-04, 30359-05, 30359-06, 30359-07
Cerebrovascular disease430–438G45–46, I60–I69
Deep vein thrombosis451.1–451.9I80.2–I80.9
Heart disease401–404, 410–411, 413–414, 428I10–I12, I20–I22, I24–I25, I50
Osteoporosis/bone fractures733M80–81
Pulmonary embolism415.1I26

Surgical procedures

Relevant gynaecological and breast cancer-related surgical codes are shown in Table 1, classified by type of procedure. For the aim of this study we defined “risk-reducing” gynaecological surgery as those surgical procedures that were performed electively at least 30 days before a diagnosis of gynaecological cancer. Four procedure groups were formed: (i) hysterectomy only [including hysterectomy plus unilateral salpingo-oophorectomy (USO)]; (ii) BSO only (including two separate USOs); (iii) both hysterectomy and BSO and (iv) neither (including single USO only). Procedures were conservatively classified as “hysterectomy only” in situations where it was unclear whether a hysterectomy also involved a USO or BSO (see Table 1).

Some women with breast cancer did not have a matching hospital treatment record. The reasons for this are unclear, but may include those who received treatment either interstate or overseas. As we could not be sure that these cases did not undergo any risk-reducing gynaecological surgery, they were excluded from the study to ensure that they were not incorrectly included in the group who did not have surgery.

Statistical analyses

Survival time was calculated as the number of days between diagnosis and either death or December 31, 2009, whichever came first. The follow-up period for each patient was divided between the four risk-reducing gynaecological surgical procedure groups, depending on type and timing of procedures. For instance, if a patient who survived for 8 years had a risk-reducing hysterectomy without BSO 2 years after her breast cancer diagnosis, then the first 2 years of her follow-up were assigned to the group with no surgery, whereas the remaining 6 years were assigned to the group of “hysterectomy only.”

Flexible parametric survival models (also known as Royston–Parmar models) were used for this analysis.[30, 31] The baseline survival distribution is represented as a restricted cubic spline function in Royston–Parmar models. This leads to several advantages over the traditional Cox proportional hazard models, particularly the ease with which nonproportional effects can be handled.

Royston–Parmar models may be fitted using various scales for the restricted cubic spline function, including hazards (Weibull models), odds (log-logistic models) and normal (probit models). A differing number of internal “knot points” (where the pieces of the spline function join) can also be defined. The aim is to choose the scale and number of knot points which result in the best proportionality assumption for the covariates, and is determined by the combination that minimises the Bayes information criterion statistic. Significant covariates are selected via backward elimination using a multivariable fractional polynomial approach.[31]

We conducted modelling for all-cause survival, breast cancer-specific survival and survival due to causes other than breast cancer. The analysis of breast cancer-specific survival was further stratified for “premenopausal” and “postmenopausal” women (20–49 and 50–79 years). For the all-cause and breast cancer-specific survival models, the normal scale with 3 degrees of freedom (two internal knot points) provided the best fit, whereas for nonbreast cancer survival the optimum model was on the odds scale also with 3 degrees of freedom.

The main variable of interest was risk-reducing gynaecological procedure group. Delayed entry survival models were used to account for the fact that the risk-reducing gynaecological procedure group for an individual could alter during their time at risk. Other covariates that were considered included age group at diagnosis of first primary breast cancer, indigenous status, area-based socioeconomic status, locality of residence, morphology, tumour size, lymph node ratio, laterality, type of breast cancer surgery, hospital type, diagnosis of second primary cancer (breast, gynaecological and other) and the comorbidities listed above. In addition, significant covariates (p ≤ 0.20), including risk-reducing gynaecological procedure group, were tested for time dependency within each model, by fitting interactions between the covariates and time using additional spline functions.

Unadjusted and adjusted estimates of 10-year survival with 95% confidence intervals were calculated. Differences in survival by risk-reducing gynaecological procedure group were determined using the model coefficients (β), with the reference group being “neither hysterectomy nor BSO.” The significance of the overall effect for risk-reducing gynaecological surgery group was also assessed using the Wald test and expressed in terms of a χ[2] statistic. Individual estimates were considered significant only if p ≤ 0.05 for the overall effect. Adjusted survival curves were produced by averaging the predicted survival curve for each subject in a particular stratum.

Propensity score analysis was retrospectively applied to breast cancer-specific survival among younger women, in an attempt to minimise selection bias that could have explained survival differences by risk-reducing gynaecological procedure group.[32, 33] The propensity score is defined as the probability of treatment assignment conditional on observed baseline covariates. Covariates recorded at the time of breast cancer diagnosis and known to influence survival were age, tumour size and positive lymph node ratio expressed as continuous variables along with categorical groupings for indigenous status, locality of residence and cell type/morphology (as shown in Table 2). Observations were randomly sorted before matching. Propensity scores for treatment ranged from 0.056 to 0.319 for breast cancer patients aged 20–49 years. Those who had some form of risk-reducing gynaecological surgery were matched with three others who did not have surgery using nearest neighbour matching without replacement, with a maximum absolute difference of 0.01 allowed in the propensity score for each matched pair (one woman was excluded as there were only two suitable matches). Paired t-tests were used to ensure that there were no biases in the distribution of the matching variables between the treated and untreated subjects.

Table 2. Cohort of first primary breast cancer patients diagnosed in Queensland by type of prophylactic gynaecological surgerya, 1997–2008
 Total cohortNeitherHyst. onlyBSO onlyBoth
  1. Abbreviations: Hyst. only: hysterectomy only; df: degrees of freedom; BSO: bilateral salpingo-oophorectomy; **: data withheld—cell count < 5.

  2. a

    The cohort was categorised depending on the type of prophylactic gynaecological surgery performed as at the end of the follow-up period for each woman.

  3. b

    Includes hospital/s where the patient was treated for breast cancer and/or where prophylactic gynaecological surgery was performed.

  4. c

    χ2 test excludes the category “unknown/not applicable.”

Number of eligible women21,06719,650634286497
Row %
Total years at risk119,340110,8103,5591,6223,349
Median years at risk5.
 nRow %Row %Row %Row %
Age group at diagnosis     
 χ2 = 491.2; df = 12; p < 0.001
Indigenous status     
Not stated2,05594.
 χ2 = 13.85; df = 6; p = 0.031
Area-based socioeconomic status (SES)     
Most disadvantaged2,58693.
Middle SES14,88793.
Most advantaged3,52593.
 χ2 = 12.33; df = 9; p = 0.195
Locality of residence     
Major city12,68393.
Inner regional4,59792.
 χ2 = 23.31; df = 9; p = 0.006
Infiltrating duct carcinoma (8500-3)15,81393.
Lobular carcinoma (8520-3)2,48894.
Infiltrating duct and lobular carcinoma (8522-3)76391.
 χ2 = 18.12; df = 9; p = 0.034
Tumour size     
≤20 mm12,37393.
21–50 mm6,74192.
>50 mm1,20893.
Not recorded74595.
 χ2 = 14.43; df = 9; p = 0.108
Lymph node status     
No lymph nodes excised2,48595.
No positive lymph nodes11,23093.
At least one positive lymph node6,85092.
Not recorded50294.
 χ2 = 34.33; df = 12; p < 0.001
Not stated8894.3**0.0**
 Fisher's exact test; p = 0.563
Type of breast cancer surgery     
Breast-conserving surgery13,39593.
No curative surgery recorded1,49391.
 χ2 = 11.12; df = 6; p = 0.085
Hospital typeb     
Unknown/Not applicable471100.
 χ2 = 216.61; df = 6; p < 0.001c
Multiple primary cancers     
Breast cancer—Yes1,00691.
Breast cancer—No20,06193.
 χ2 = 6.40; df = 3; p = 0.094
Gynaecological cancer—Yes171100.
Gynaecological cancer—No20,89693.
 Fisher's exact test; p < 0.001
Other cancer—Yes86893.
Other cancer—No20,19993.
 χ2 = 5.78; df = 3; p = 0.123
Other reported diseases and conditions     
 Fisher's exact test; p = 0.104
Cerebrovascular disease—Yes46197.2****1.5
Cerebrovascular disease—No20,60693.
 χ2 = 12.63; df = 3; p = 0.006
 χ2 = 6.62; df = 3; p = 0.085
 Fisher's exact test; p = 0.167
Deep vein thrombosis—Yes36892.44.3**2.7
Deep vein thrombosis—No20,69993.
 χ2 = 4.29; df = 3; p = 0.232
 χ2 = 12.51; df = 3; p = 0.006
Heart disease—Yes2,05295.
Heart disease—No19,01593.
 χ2 = 19.77; df = 3; p < 0.001
Osteoporosis or bone fractures—Yes32095.02.5**2.2
Osteoporosis or bone fractures—No20,74793.
 χ2 = 3.04; df = 3; p = 0.386
Pulmonary embolism—Yes34595.42.3**1.4
Pulmonary embolism—No20,72293.
 χ2 = 2.57; df = 3; p = 0.463

The survival analysis described above was then repeated for the matched cohort. The optimum Royston–Parmar model was on the normal scale with 2 degrees of freedom. Variables used in the matching process were not included as covariates; rather, Austin[34] suggests that survival models should be stratified on the matched groups to account for the matched nature of the cohort. As it is not possible to stratify a parametric model when the numbers in each strata are so small (n = 4), we divided the matched groups into deciles based on the propensity score of the treated case, and the model was then stratified by these deciles (n ~ 340 in each strata).

All data analyses were performed using Stata/SE version 12.1 for Windows. Human research and ethics approval for this study was obtained from the Human Research Ethics Committee at the Royal Brisbane and Women's Hospital (HREC/10/QRBW/425).


Of the 25,536 patients diagnosed with primary female breast cancer in QLD between 1997 and 2008, 21,067 (82%) were eligible. The remaining 4,469 women were excluded owing to not having a matching hospital record (2,736 cases, 11%), being younger than 20 years or older than 79 years at the time of diagnosis (1,726 cases, 7%) or where the basis of diagnosis was either autopsy or death certificate only (seven cases, 0.03%). Those who were eligible amassed a total of 119,340 years at risk (median follow-up of 4.6 years; interquartile range 3.0–8.6 years). Overall, 3,165 (15%) women died during follow-up, including 2,195 (10%) from breast cancer. Key demographic, clinical and treatment characteristics of the study cohort are summarised in Table 2.

Overall, 1,426 women (7%) underwent risk-reducing gynaecological surgery (Table 2). However, this varied by age, with 13% of breast cancer patients in the 20–39 age group having risk-reducing gynaecological surgery compared to only 3% who were aged 70–79 years at diagnosis. Apart from younger age, women were more likely to have risk-reducing gynaecological surgery if they were nonindigenous, diagnosed with infiltrating ductal and lobular carcinoma, if they had positive axillary lymph nodes and attended both a public and private hospital for breast cancer treatment. Women who lived in a major city or who had cerebrovascular disease, diabetes mellitus or heart disease were less likely to undergo risk-reducing gynaecological surgery.

A total of 171 women developed gynaecological cancer subsequent to breast cancer, all in women who did not have risk-reducing gynaecological surgery (p = 0.006, Table 2). Of those, 23 cancers developed in premenopausal women (including eight ovarian cancers) and 148 in postmenopausal women, respectively. In addition, 1,006 women developed new primary breast cancers and 868 were diagnosed with at least one other cancer following their initial breast cancer. There were no significant differences in the distribution of subsequent new breast cancers (p = 0.094) or other cancers (p = 0.123) by final risk-reducing surgery status.

After adjustment for the covariates listed in Table 3, breast cancer patients who had both a hysterectomy and BSO had a significantly higher survival rate 10 years after diagnosis for all causes of mortality (85%) compared to those who did not have any risk-reducing gynaecological surgery (79%, p = 0.002; Table 4 and Fig. 1). The differential was similar for breast cancer-specific mortality (adjusted 10 year survival of 89 and 85%, respectively, p = 0.005). However, for both all-cause and breast cancer-specific mortality, there was no statistically significant evidence of a survival benefit among women who had either a risk-reducing hysterectomy only or BSO only compared to the nonsurgery group. There was also no disparity in survival by risk-reducing gynaecological surgery group due to causes other than breast cancer, including other types of cancer (Table 4 and Fig. 1) or for noncancer deaths only (data not shown).

Figure 1.

Adjusted survival curves by cause of death and prophylactic gynaecological surgery groupa,b. Abbreviations: BSO: bilateral salpingo-oophorectomy. Notes: aAn individual woman may contribute survival time to more than one prophylactic surgical procedure group. bThe covariates used to adjust the model are listed in Table 3.

Table 3. Covariates included in flexible parametric survival models by cause of death and age group
CovariateAll causes—all age groupsBreast cancer-specific—all age groupsBreast cancer-specific—0–49 years oldBreast cancer-specific—50 years and olderNonbreast cancer—all age groupsBreast cancer-specific—0–49 years old matched cohort
  1. Abbreviations: *: included in model; **: included in model as a time-dependent covariate.

Prophylactic gynaecological procedure******
Age group****** 
Indigenous status***** 
Area-based socioeconomic status**   *
Locality of residence***** 
Tumour size******** 
Lymph node ratio********* 
Laterality** *  
Type of breast cancer surgery*******
Hospital type******
Second primary breast cancer*  *** 
Gynaecological cancer*** *** 
Other cancer********
Atherosclerosis*** *  
Cerebrovascular disease**   **
Dementia**   ** 
Deep vein thrombosis******* **
Heart disease*** *** 
Osteoporosis or bone fractures******
Pulmonary embolism********
Table 4. Ten-year survival estimates by cause of death and prophylactic gynaecological surgery group
Prophylactic gynaecological surgical procedurenaUnadjusted 10-year survival estimates (95% CI)Adjusted 10-year survival estimates (95% CI)Model coefficients (β, 95% CI)p
  1. The covariates used to adjust each model are listed in Table 3.

  2. a

    Includes all women who contributed survival time to that prophylactic surgical procedure. An individual woman may contribute survival time to more than one prophylactic surgical procedure.

  3. Abbreviations: BSO: bilateral salpingo-oophorectomy; 95% CI: 95% confidence interval.

All-cause mortality
Neither20,65076.4 (75.6–77.2)78.5 (77.8–79.3)1.00 
Hysterectomy only65079.2 (74.4–83.3)79.7 (75.9–83.5)−0.04 (−0.21, +0.13)0.610
BSO only28776.2 (68.1–83.0)78.4 (72.2–84.6)+0.01 (−0.25, +0.28)0.914
Hysterectomy and BSO49784.4 (79.8–88.2)85.0 (81.4–88.6)−0.31 (−0.50, −0.11)0.002
Overall effect: χ2 = 9.84; degrees of freedom = 3; p = 0.020
Breast cancer-specific mortality
Neither20,65084.3 (83.6–85.0)84.0 (83.4–84.7)1.00 
Hysterectomy only65084.6 (80.2–88.2)85.1 (81.7–88.5)−0.05 (−0.24, +0.13)0.581
BSO only28779.9 (72.0–86.2)83.9 (78.6–89.3)+0.01 (−0.26, +0.29)0.929
Hysterectomy and BSO49789.2 (85.1–92.4)89.3 (86.2–92.5)−0.31 (−0.52, −0.09)0.005
Overall effect: χ2 = 8.02; degrees of freedom = 3; p = 0.046
Nonbreast cancer mortality
Neither20,65090.4 (89.7–91.0)92.5 (92.0–93.0)1.00 
Hysterectomy only65094.2 (90.3–96.6)92.8 (89.6–95.9)−0.05 (−0.67, +0.56)0.862
BSO only28796.6 (90.1–98.9)94.2 (88.8–99.6)−0.37 (−1.64, +0.90)0.570
Hysterectomy and BSO49794.8 (91.1–97.0)93.1 (89.9–96.3)−0.13 (−0.77, +0.52)0.701
Overall effect: χ2 = 0.49; degrees of freedom = 3; p = 0.921

Further analysis by age at diagnosis for breast cancer-specific survival indicated that the improvement in prognosis among those who had both a hysterectomy and BSO was only significant among younger women (Table 5, Fig. 2). Premenopausal women (20–49 age group) had significantly better survival after 10 years (93%) compared to women of the same age who had neither procedure (83%, p = 0.001). In contrast, there were no significant differences in breast cancer-specific survival by type of risk-reducing gynaecological surgery for women 50–79 years.

Figure 2.

Adjusted breast cancer-specific survival curves by type of prophylactic gynaecological surgical procedure and age groupa,b. Abbreviations: BSO: bilateral salpingo-oophorectomy. Notes: aAn individual woman may contribute survival time to more than one prophylactic surgical procedure group. bThe covariates used to adjust the model are listed in Table 3.

Table 5. Ten-year survival estimates for breast cancer-specific mortality by type of prophylactic gynaecological surgery and age group
Prophylactic gynaecological surgical procedurenaUnadjusted 10-year survival estimates (95% CI)Adjusted 10-year survival estimates (95% CI)Model coefficients (β, 95% CI)p
  1. The covariates used to adjust each model are listed in Table 3.

  2. a

    Includes all women who contributed survival time to that prophylactic surgical procedure. An individual woman may contribute survival time to more than one prophylactic surgical procedure.

  3. Abbreviations: BSO: bilateral salpingo-oophorectomy; 95% CI: 95% confidence interval.

Breast cancer-specific mortality—20−49 years old
Neither5,90483.2 (81.8–84.5)83.0 (81.7–84.3)1.00 
Hysterectomy only29882.6 (75.4–88.3)82.7 (76.8–88.5)+0.03 (−0.24, +0.31)0.812
BSO only18478.8 (68.9–86.6)83.6 (76.7–90.5)−0.01 (−0.34, +0.32)0.949
Hysterectomy and BSO24492.5 (87.0–96.0)92.9 (88.9–97.0)−0.61 (−0.97, −0.26)0.001
Overall effect: χ2 = 11.58; degrees of freedom = 3; p = 0.009
Breast cancer-specific mortality—50–79 years old
Neither14,74684.8 (83.9–85.5)84.6 (83.9–85.4)1.00 
Hysterectomy only35285.8 (80.1–90.3)86.9 (82.8–91.1)−0.13 (−0.38, +0.12)0.317
BSO only10382.8 (69.7–91.6)84.0 (74.7–93.3)+0.04 (−0.45, +0.52)0.878
Hysterectomy and BSO25386.7 (80.5–91.4)86.7 (82.0–91.3)−0.11 (−0.39, +0.16)0.425
Overall effect: χ2 = 1.63; degrees of freedom = 3; p = 0.652

When we repeated the breast cancer-specific survival analysis for women aged 20–49 using the matched sample, results were similar (Supporting Information Table). Again, a significant survival advantage was only seen for women who had hysterectomy plus BSO compared to those who did not have any risk-reducing gynaecological surgery (p = 0.002).


In premenopausal women diagnosed with primary breast cancer, risk-reducing hysterectomy and BSO increased breast cancer-specific survival from 83 to 93% after 10 years. This effect remained after matching for some characteristics that are known to influence prognosis. In contrast, no significant survival benefit of risk-reducing gynaecologic surgery was observed for postmenopausal women.

It is generally accepted that oestrogen can stimulate breast cancer growth.[7] Endocrine treatments suppressing circulating oestrogens via action on the hypothalamic-pituitary-ovarian axis improve survival outcomes in premenopausal hormone receptor-positive breast cancer patients.[28] Ovarian ablation either by radiation treatment or through surgical removal of the ovaries has been advocated in the past but has become less commonly used owing to the availability of a modern array of noninvasive endocrine treatment options.[35] These modern treatments are widely thought to be at least as effective as surgical removal of the ovaries.[36]

Our findings may provide a challenge to this belief. The main effect of hysterectomy and BSO on breast cancer-specific survival limited to premenopausal women suggests that hysterectomy plus BSO provides advantage by combined hormone ablation.[28] In Australia, endocrine treatment is well accepted and established in hormone receptor-positive breast cancer patients. Before the introduction of antioestrogenic medication in the late 1970s, ovarian ablation was performed through surgical removal of the ovaries, radiation treatment, GnrH analogues and chemotherapy. Silencing of the ovaries using radiation treatment resulted in a 25% benefit compared to patients who had no adjuvant treatment.[35] As has been highlighted elsewhere,[35] we can also assume that BSO had a smaller impact in terms of hormonal ablation on breast cancer patients who were given chemotherapy. However, chemotherapy is variable in its effectiveness of silencing the ovaries with reported rates ranging between 10 and 98%[37, 38] and BSO may thus have an effect in addition to either chemotherapy or hormonal treatment. Although our study does not answer this important question, a three-arm randomised controlled clinical trial (SOFT) that assigned patients to receive either oral tamoxifen (control) or tamoxifen plus ovarian function suppression through triptorelin, surgical oophorectomy or ovarian irradiation is in progress.[39] It remains to be researched further as to why only patients who had a BSO plus hysterectomy benefitted from improved survival but patients who had a BSO or hysterectomy alone did not. On the other hand, a pattern of care study compared Australian (Perth, Western Australia), Canadian and Scottish treatment patterns based on hospital data. In Australia, 29% of breast cancer patients received chemotherapy, and 59% received endocrine treatment. These authors assessed the treatment received compared to current guidelines, and found that in all jurisdictions patients with Stage II, node positive, hormone receptor-negative tumours and Stage IV hormone receptor-positive tumours may not receive chemotherapy to the full extend recommended; however, no shortfalls in endocrine treatment were noted. Indeed, Australian treatment was exceeding “optimal” levels[40]; however, these data are likely reflecting the prescribed rather than the actual received medication. Bell et al.[41] assessed self-reported tamoxifen or aromatase use in 1,683 women for 5 years after diagnosis of a hormone receptor-positive breast cancer in Victoria between 2004 and 2006. It found that 7.8% of women self-reported no endocrine medication, 10.2% reported using oral adjuvant endocrine therapy up to 2 years, 15.6% 3 years and two-thirds of women for at least 4 years. This indicates that although coverage of endocrine treatment in Australia is good, there may be a significant proportion of women who do not start endocrine treatment, or do not persist with such treatment for the required length of time, and could therefore particularly benefit from risk-reducing surgery.

We did not find any difference in survival after 10 years from causes other than breast cancer by risk-reducing gynaecological surgery status. Our results therefore indirectly suggest that the effect of combined risk-reducing surgery on menopause-related risk factors such as cardiovascular health was minimal and appear to have been heavily outweighed by the survival advantages due to a decrease in breast cancer-specific mortality. However, a significantly higher proportion of women in the “no surgery” group were identified as having cerebrovascular and/or heart disease comorbidities, and this may be part of the reason why they were not offered risk-reducing surgery; only women with low risk of cardiovascular disease may have elected for prophylaxis.

Although population-based studies reflect “real”-world scenarios, they do not provide definitive proof of mechanism of action leading to the observed outcomes. Overall, within the 10-year observation period of our study, 171 women who did not have risk-reducing surgery developed gynaecological cancer. Of those only 23 patients were premenopausal (14 developed uterine cancer and eight ovarian cancer). In contrast, none of the women who had risk-reducing gynaecological surgery developed gynaecological cancer. The relatively small number of prevented cancers in premenopausal women indicates that it is unlikely that the significant survival advantage among premenopausal women is mainly a result of surgical prophylaxis of these potential gynaecological cancers. Our data did not provide details of women's BRCA1/2 status and family history. Given that only eight premenopausal breast cancer patients who did not have risk-reducing surgery developed a new primary ovarian cancer during the observation period, it is also unlikely that the results were largely driven by patients at high risk due to genetic mutations. However, the possibility remains that the majority of those who were BRCA1/2 positive may have been offered risk-reducing prophylactic gynaecological surgery.

As noted in the Introduction, for women from the general population, the effect of hysterectomy plus BSO on overall survival is controversial. The prospective Nurses' Health Study cohort study included 29,380 women who had a hysterectomy for benign disease (mean age at surgery = 45 years; 28 years follow-up).[21-23, 42] Women who additionally had a BSO had significant reductions in ovarian cancer incidence and mortality and reduced risk of breast cancer incidence for premenopausal women following hysterectomy and BSO. However, BSO at the time of hysterectomy was associated with increased overall mortality in women younger than 50 years who never used oestrogen therapy, and at no time was BSO associated with increased overall survival.[42, 43]

The prospective Women's Health Initiative Observational Study included 25,448 women who had a hysterectomy for a benign condition (average age 49 years; follow-up 8 years).[20] Women in this study were initially invited to participate in the Women's Health Initiative randomised trial that evaluated postmenopausal hormone therapy, but were either found ineligible or declined participation in the trial. Women who had a BSO during hysterectomy had significant reductions in ovarian cancer incidence and mortality compared with women who conserved their ovaries. In contrast to the Nurses' Health Study, breast cancer incidence was not reduced for women who had a BSO, nor was there an increased risk in all-cause mortality among premenopausal or postmenopausal women who had a BSO at the time of hysterectomy.

The retrospective population-based Mayo Clinic Cohort Study of Oophorectomy and Aging enrolled 2,365 women who underwent USO or BSO for benign disease in conjunction with hysterectomy.[26] Every member of the cohort was matched by age to a referent woman in the same population who had not undergone oophorectomy. The median age at time of surgery was 44 years among premenopausal women who had a BSO and 62 years among postmenopausal women (average follow-up 25 years). Overall mortality was significantly higher in women who had received prophylactic BSO before the age of 45 years compared to referent women, while having a BSO made no difference to all-cause mortality in postmenopausal women.

The differences in outcomes of these studies compared to the results presented here are likely explained by the different groups of women enrolled. In particular, the three studies outlined above enrolled women from the general population who required a hysterectomy for benign conditions, whereas our study enrolled only patients diagnosed with primary breast cancer. The latter population clearly has a significantly increased risk of death as well as a significantly increased risk of developing gynaecological cancers.[9]

Although this population-based study uses innovative new statistical models, which better handle nonproportional effects, the design used within our study inherits limitations that need to be acknowledged. First, the follow-up duration available to us was limited to a maximum of 10 years, due to the fact that health administrative data became available in Queensland only in 1997. Second, we were unable to determine whether preexisting comorbidities were present at the time of breast cancer diagnosis; in most cases these could only be subsequently ascertained if they were recorded in the hospital chart during treatment. On that basis we were unable to take comorbidities into account in the propensity score matching, which leaves open the possibility of some bias remaining in the matched cohort analysis. There was also some potential for misclassification of women regarding the prophylactic gynaecological surgery groups due to procedures that may have been performed before matched records being available. Further, reasons for surgery were not recorded in the information provided by Queensland Health. Information on postoperative, adjuvant treatment as well as hormonal replacement therapy could not be obtained because these treatments do not require a hospital admission. Finally, data on hormone receptor status were not available, which would have been valuable to examine if prophylactic gynaecological surgery was effective in hormone receptor-positive patients only or if the effect also extended to hormone receptor-negative breast cancer patients. Similarly, we were not able to obtain patients' BRCA status.

In summary, the results indicate that premenopausal women with breast cancer may benefit from hysterectomy plus BSO in addition to the ovarian ablation provided by the adjuvant treatment they commonly receive. Although the results of our study are promising and important, the decision to undergo prophylactic gynaecological surgery obviously has major ramifications for younger women. Therefore, our findings need to be replicated in at least one other independent dataset and tested in a randomised trial before current treatment recommendations for premenopausal women diagnosed with breast cancer are reconsidered.


Peter Baade and Monika Janda were supported by an Australian National Health and Medical Research Council Career Development Fellowships (#1005334 and #1045247, respectively).