Pre-diagnosis oophorectomy, estrogen therapy and mortality in a cohort of women diagnosed with breast cancer

Study participants were female residents of Wisconsin, Massachusetts, or New Hampshire with an incident invasive breast cancer reported to each state’s cancer registry during the study enrollment years. Eligibility was limited to women with listed telephone numbers, driver’s licenses (if younger than 65 years), and registry-reported dates of diagnosis. Women ages 50 to 79 were eligible in the first study (1992 to 1995); and ages 20 to 69 in subsequent studies (1997 to 2001; 2001 to 2004; 2004 to 2007). During 1992 to 2001 participants were enrolled in Wisconsin, Massachusetts, and New Hampshire; in 2001 to 2007 women were enrolled in Wisconsin only. Across the four studies, 21,713 eligible participants were identified. Physicians refused contact with 359 (1.7%), 743 (3.4%) were deceased, 624 (2.9%) could not be located and 2,794 (12.9%) refused to participate. Therefore, 17,193 (79.2%) eligible breast cancer survivors were interviewed. Thirty-eight interviews were considered unreliable by the interviewers, leaving 17,155 interviews available for analysis.

In each study, participants completed a 40-minute structured telephone interview that elicited detailed information on lifestyle and demographic factors, postmenopausal hormone use, and personal and family medical history. Study participants reported whether they had surgery to remove the uterus or ovaries, the type of surgery (hysterectomy and/or oophorectomy, including number of ovaries removed), and age at surgery. Participants were asked a standard history of postmenopausal hormone use including formulation, route of administration, age started, frequency of each episode of use, total duration, and time since last use.

In three of the four pooled studies (1992 to 1995; 1997 to 2001; 2001 to 2004), menopause was defined as the absence of menses for six consecutive months not attributable to surgery, chemotherapy, radiation, or other reasons. In the most recent study (2004 to 2007), menopause was defined as 12 consecutive months without menses. In all studies, women who reported bilateral oophorectomy prior to breast cancer diagnosis were categorized as postmenopausal. BMI was calculated as weight (kg)/tallest adult height (m)² during the one- to five-year period prior to breast cancer diagnosis.

Vital status information, including date and underlying cause of death, was collected through December 31, 2009 through linkage with the National Death Index. Cause of death was categorized according to the International Classification of Disease (ICD-10) code as breast cancer-specific mortality (C50), cardiovascular (CVD) mortality (I00 to I09, I11, I13, I20 to 51), stroke mortality (I60 to I69), or all-cause mortality [23].

For consistency in ages at enrollment across studies, this pooled analysis was limited to women age 50 and older at breast cancer diagnosis (N = 13,253). We further excluded women who were premenopausal at diagnosis (N = 1,352) or had unknown menopausal status (N = 620), or who had a previous history of cancer (except non-melanoma skin cancer) (N = 614). Ninety-three records from women with discordant ages at bilateral oophorectomy and hysterectomy and 41 records with missing ages for both procedures were excluded. To exclude gynecologic surgeries performed as part of breast cancer treatment, 84 women who reported bilateral oophorectomy or hysterectomy at the same age or after breast cancer diagnosis were excluded. Six records were excluded based on partial interview completion. After these exclusions were applied, 10,443 women were eligible for analysis. Among the eligible breast cancer survivors, 1,778 (17.0%) reported having a prior TAHBSO and 6,913 (66.2%) reported having an intact uterus and ovaries at breast cancer diagnosis; therefore, records from 8,691 breast cancer survivors contributed to this analysis. The remaining 1,752 records included 1,641 women who reported hysterectomy alone or unilateral oophorectomy with or without hysterectomy, and 111 women with missing information on previous gynecologic surgery. On average, women were interviewed 1.3 years (standard deviation (SD) = 0.5; range: 0.2 to 4.9) after breast cancer diagnosis.

All mortality comparisons were made relative to women who reported having an intact uterus and ovaries and never using postmenopausal hormones at breast cancer diagnosis. Kaplan Meier curves were produced to display mortality curves; formal statistical comparison of the difference in slopes was conducted with log rank tests [24]. For survival analyses, the time scale was defined as time since breast cancer diagnosis (years) and person-time was accrued from date of interview. Participants either developed the event of interest (breast cancer mortality/cardiovascular disease mortality/all-cause mortality) or were administratively censored on December 31, 2009. In statistical models of mortality within five years of diagnosis, person-time was accrued from date of interview to the event of interest or administrative censoring at five years after the diagnosis date. For analyses of five-year survivors, person-time was accrued from five years after diagnosis until the event of interest or administrative censoring on December 31, 2009.

Cox proportional hazards regression was used to calculate multivariate hazard ratios (HR) for all-cause mortality. The Fine and Gray method was used to fit proportional cumulative incidence-associated subhazards regression models for breast cancer and CVD-specific mortality to account for non-breast and non-CVD disease deaths as competing events [25, 26]. A priori, final estimates were adjusted for stage of diagnosis, state, study enrollment years, age at diagnosis, BMI one to five years before breast cancer diagnosis, smoking status at breast cancer diagnosis, and family history of breast cancer as potential confounders. P values ≤0.05 were considered to be statistically significant. All statistical analyses were performed with Stata 12 software (StataCorp, College Station, TX, USA).

Across the four studies, the majority of participants were enrolled in Wisconsin (69%) and were diagnosed with local invasive disease (64%). We conducted sensitivity analyses to evaluate whether inclusion of participants in Massachusetts and New Hampshire or those with regional or distant disease influenced our results.

TAHBSO was associated with a 16% decrease in mortality from all causes (HR = 0.84; 95% confidence interval (CI): 0.76, 0.92). All-cause mortality did not appear to vary according to age at surgery. The inverse association between pre-diagnosis TAHBSO and all-cause mortality was of greater magnitude among women who reported using estrogen therapy (HR = 0.78; 95% CI: 0.69, 0.89) compared to no estrogen use (HR = 0.92; 95% CI: 0.80, 1.05) (Table 2, Figure 1A). All-cause mortality risk estimates were similar across categories of estrogen duration (HR = 0.81; 95% CI: 0.65, 1.02 for 0.5 to 4.9 years; HR = 0.75; 95% CI: 0.56, 0.99 for 5 to 9.9 years; and HR = 0.77; 95% CI: 0.65, 0.92 for ≥10 years) (Table 2).

All-cause mortality among women who underwent pre-diagnosis TAHBSO and used estrogen therapy appeared to vary with follow-up time (Figure 1A): during the first four years after breast cancer diagnosis, estrogen use did not seem to influence mortality, whereas approximately five years after breast cancer diagnosis estrogen therapy appeared to confer a survival advantage. Therefore, we examined five-year mortality risk separately from deaths that occurred less than five years after diagnosis. TAHBSO alone was not associated with decreased all-cause mortality within the first five years after diagnosis (HR = 0.93; 95% CI: 0.78, 1.10). However, among five-year breast cancer survivors, we observed a 20% decrease in mortality risk among women who reported pre-diagnosis TAHBSO (HR = 0.80; 95% CI: 0.72, 0.90), particularly among women who also used estrogen therapy (HR = 0.72; 95% CI: 0.62, 0.85) (Table 3).

Mortality due to causes other than breast cancer may be more common among women diagnosed at ages 65 and older [4, 27]. With this consideration, we also evaluated HRs for cause-specific mortality according to age at diagnosis. Overall, TAHBSO was associated with decreased all-cause mortality among both women who were diagnosed at ages 50 to 64 (HR = 0.78; 95% CI: 0.66, 0.91) and ages ≥65 (HR = 0.86; 95% CI: 0.77, 0.97). When evaluated according to estrogen use, estimates remained statistically significant only among estrogen users (HR = 0.75; 95% CI: 0.61, 0.92 among women ages 50 to 64 and HR = 0.80; 95% CI: 0.67, 0.94 among women ≥65) (Table 3).

In multivariate competing-risk regression models, pre-diagnosis TAHBSO was associated with a 24% reduction in breast cancer mortality (HR = 0.76; 95% CI: 0.64, 0.90). No variation was observed according to age at surgery. When evaluated according to estrogen use, decreased breast cancer mortality risk was observed among women who reported TAHBSO and never used hormones (HR = 0.73; 95% CI: 0.55, 0.96) but not among women who used unopposed estrogens (HR = 0.97; 95% CI: 0.80, 1.18) (Table 2, Figure 1B). We did not observe a consistent pattern of breast cancer-specific mortality risk according to duration of estrogen use (Table 2).

When evaluated according to time since diagnosis, TAHBSO was associated with a suggested decrease in breast cancer mortality among five-year survivors (HR = 0.81; 95% CI: 0.64, 1.02 for TAHBSO overall; HR = 0.69; 95% CI: 0.46, 1.01 for women who reported TAHBSO and never hormone use). In analyses stratified according to age at diagnosis, TAHBSO was associated with statistically significant reductions in breast cancer mortality only among women ages 50 to 64 at diagnosis (HR = 0.78; 95% CI: 0.62, 0.98) (Table 3).

Overall, the association between TAHBSO and risk of death due to CVD was not statistically significant (HR = 0.84, 0.67, 1.06) in multivariate competing-risk regression models. When evaluated according to age at surgery, TAHBSO at ages ≤45 years was not related to CVD mortality (HR = 1.04; 95% CI: 0.75, 1.44), but decreased CVD mortality was observed among women who reported TAHBSO at older ages (HR = 0.73; 95% CI: 0.54, 0.98). Women who reported TAHBSO and estrogen therapy use had a 43% decrease in CVD mortality (HR = 0.57; 95% CI: 0.39, 0.83); and those who reported at least 10 years of estrogen use had an estimated 55% reduction in CVD mortality (HR = 0.45; 95% CI: 0.26, 0.78). Conversely, women who reported TAHBSO and did not use estrogen had a non-statistically significant 19% increase in CVD mortality risk (95% CI: 0.89, 1.58) (Table 2, Figure 1C).

Fewer CVD deaths occurred within the first five years of diagnosis or among women who were 50 to 64 years at breast cancer diagnosis compared to five-year survivors and older women. Among five-year breast cancer survivors, those who reported TAHBSO and unopposed estrogen use had an estimated 47% reduction in CVD mortality (HR = 0.53; 95% CI: 0.35, 0.81). Sample sizes were insufficient to calculate fully adjusted HRs for CVD mortality among women ages 50 to 64 at diagnosis, estimates in this group are adjusted for study enrollment years, age and stage at diagnosis only. A 68% decrease in CVD mortality was estimated among women ages 50 to 64 at diagnosis who reported previous TAHBSO and estrogen use (HR = 0.32; 95% CI: 0.13, 0.80). Among women who were ≥65 at diagnosis, we observed a 35% decrease in CVD mortality associated with TAHBSO and estrogen use (HR = 0.65; 95% CI: 0.43, 0.98) (Table 3).

In sensitivity analyses restricted to participants residing in Wisconsin (N = 5,992) or those with local disease (N = 5,604), our findings were essentially unchanged compared to the overall analysis.