Trajectories of metabolic parameters after bilateral oophorectomy in premenopausal women.

The study participants were selected from the Mayo Clinic Cohort Study of Oophorectomy and Aging-2 (MOA-2). The methodology of MOA-2 and the clinical characteristics of the women included in MOA-2 have been previously described.[ 1 , 9 – 11 ] Briefly, MOA-2 includes all premenopausal women who underwent bilateral oophorectomy before age 50 years in Olmsted County, Minnesota between January 1, 1988 and December 31, 2007 (n = 1,747). We excluded women who underwent oophorectomy at age 46–49 years (n = 622) and women who underwent oophorectomy for primary or metastatic ovarian cancer (n = 54), for the treatment of estrogen-sensitive malignancy (eg, breast cancer; n = 26), or because they carried a high genetic risk of cancer (eg, BRCA1 or BRCA2 carriers; n = 14). The medical records of these women were manually abstracted to confirm bilateral oophorectomy and to collect information on surgical indication, surgical pathology, and other clinical characteristics.[ 10 , 11 ] Bilateral oophorectomy was defined as the removal of both ovaries within the same surgical procedure or as the removal of the remaining ovary for women who had previously undergone unilateral oophorectomy. The index date was defined as the date of bilateral oophorectomy. Each woman who underwent bilateral oophorectomy was matched to a referent woman selected using simple random sampling among Olmsted County residents who were born in the same year (±1 year) and who had at least one ovary intact as of index date.[ 10 ] In the current study, in order to reduce the amount of medical record abstraction, we selected a random sample of 270 women among a total of 1,031 women from MOA-2 who underwent bilateral oophorectomy at age ≤45 years. Because most women used ET for some time after oophorectomy, the random sampling was stratified by estrogen use and was weighted to obtain a similar number of women with and without ET ( Figure 1 ; at least 1 year of ET). For each woman selected from the MOA-2 bilateral oophorectomy cohort, the corresponding age-matched referent woman was also selected, and was assigned the same index date. The sample size was derived from power calculations for detecting differences in trajectories of the metabolic parameters over 10 years of follow-up (interaction between group status and years). All data were collected using the Rochester Epidemiology Project (REP) medical records-linkage system. Extensive details about the REP have been reported elsewhere.[ 12 – 15 ] All research activities were approved by the Mayo Clinic and Olmsted Medical Center Institutional Review Boards. This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines. Metabolic parameters were manually abstracted from the medical records in the REP medical records-linkage system by two physicians (EK and LGR) and three nurse abstractors. The metabolic parameters included outpatient systolic and diastolic blood pressure, height, weight, and fasting lipids including total cholesterol, triglycerides, low-density lipoprotein cholesterol, (LDL-C), and high-density lipoprotein cholesterol (HDL-C). For height, we used the measurement closest to and before bilateral oophorectomy or index date (single adult height measure). All other parameters were abstracted from medical records, one value per year for up to 3 years before the index date, and for up to 10 years after. If multiple measurements were available in the same year, the measurement closest to July 1 st (mid-point of year) was chosen. However, we did not include any blood pressure or lipid measurements taken during an acute illness, emergency department visit, or hospitalization. Body mass index (BMI) was calculated for each weight measurement available using a single adult height measurement for each woman. Additional data were also collected on the onset of antihypertensive medication use (for any indication including migraine prevention) and of lipid lowering medication use at any time within 3 years before the index date or up to 10 years after. Hypertension onset was defined as the earliest date of systolic blood pressure >140 mmHg and/or diastolic blood pressure >90 mmHg, or the onset of antihypertensive medication use. Hyperlipidemia onset was defined as the earliest date of total cholesterol ≥200 mg/dL, triglycerides ≥150 mg/dL, HDL-C ≤50 mg/dL, or the onset of lipid lowering medication use. The complete medical records of women in both the bilateral oophorectomy and referent cohorts were manually abstracted for demographics and life habits information (eg, tobacco use). Detailed information on ET use by women in the oophorectomy cohort was also collected, including the type, dose, route, and duration of treatment.[ 10 , 11 ] Women were followed passively through the abstraction of medical records in the records-linkage system from the index date to the earliest occurring of 3 endpoints: death, last visit with a REP provider, or December 31 st of the tenth year after index. Women who started ET within one month after bilateral oophorectomy were classified as “on ET” if they were still taking ET at the date of measurement of each parameter. Women who discontinued ET were moved from the “on ET” group to the “not on ET” group for that measurement time-point, and they remained in the “not on ET group” to the end of follow-up. Each parameter (blood pressure, weight, BMI, and lipids) was recorded only once per year after the index date, and we used a simple time scale of 0 to 10 years. Generalized estimating equation (GEE) linear models were used to compare the trends in metabolic measures after the index date for bilateral oophorectomy on ET, bilateral oophorectomy not on ET, and referent women. For each model, we reported mean values for these 3 groups. Variables for group status, years after index, and the interaction of these two variables were included in the models and 3 separate p-values were reported. The models were also adjusted for age at index, years of education (≤12, 13–16, >16), and smoking status (current or former vs never). Age and smoking status were available for all women, and women missing years of education were assigned to the reference group (≤12 years). Since the metabolic measures were collected as part of routine medical care, each woman could have a different number of measurements available during the study period, and the GEE models in the primary analyses were limited to these available measurements. An autoregressive correlation structure was used to account for the repeated measurements within women. This correlation structure assumed that measurements within a single woman were correlated, with higher correlation for the measurements that were closer in time versus lower correlation for the measurements that were farther apart in time. Women with hypertension before the index date were excluded from models for systolic and diastolic blood pressure, and women with hyperlipidemia before the index date were excluded from the models for lipids. For women who developed hypertension within ten years after index and were treated, the last blood pressure value before the initiation of treatment was carried forward in the analysis (for any subsequent years). A similar approach was used for women who developed hyperlipidemia during follow-up. If pre-treatment measurements were not available, the values were imputed as 140 mmHg for systolic and 90 mmHg for diastolic blood pressure, or as 200 mg/dL for total cholesterol, 150 mg/dL for triglycerides, 50 mg/dL for HDL-C, and 100 mg/dL for LDL-C. Figures were generated to show the smoothed trajectories over time of the mean parameter values predicted from the GEE models. All analyses were performed using SAS version 9.4 (SAS Institute, Inc., Cary, NC), and tests of statistical significance were conducted at the two-tailed alpha level of 0.05. We performed 4 sets of sensitivity analyses. First, to study the effect of age at estrogen deprivation, we repeated the GEE models stratified by age at oophorectomy or index date (<40 years vs. 40–45 years). Second, to study the possible effect of missing values, we repeated the GEE models after using multiple imputation methods to impute metabolic measurements after the index date that were missing.[ 16 ] Third, to study the effect of a longer duration of ET use, we repeated the GEE models in women with ≥5 years of ET after bilateral oophorectomy, women who did not receive ET or with <5 years of ET after bilateral oophorectomy, and referent women. The length of follow-up was restricted to measurements collected 5–10 years after index. Fourth, to study the effect of differences preceding the index date, we included in the GEE models both the measurements up to 3 years before the index date and the measurements up to 10 years after. This fourth set of sensitivity analyses only involved 2 groups, women who did or did not undergo bilateral oophorectomy, because ET use did not apply before the index date. Women with hypertension or hyperlipidemia more than 3 years before index date were excluded from the corresponding analyses.

In the MOA-2 study, there were 1,031 women who underwent bilateral oophorectomy at age ≤45 years and 1,031 age-matched referent women. A weighted random sample of 270 women who underwent bilateral oophorectomy was selected, among whom 163 (60.4%) started ET within one month after oophorectomy and 107 (39.6%) did not ( Figure 1 ). Women who were randomly selected had demographic and clinical characteristics similar to women who were not selected (data not shown). The corresponding 270 age-matched referent women were also selected. Most women had data available for outpatient blood pressure and weight at some point during the 10-year follow-up ( Figure 1 ). By contrast, data for lipids were available only for a subset of women in both groups. Table 1 shows the baseline characteristics of the randomly sampled matched pairs. The women who underwent bilateral oophorectomy were less educated, were more often smokers, and were more likely to have had a prior unilateral oophorectomy. Only 6 referent women (2.2%) had undergone a unilateral oophorectomy before the index date. Finally, most women underwent hysterectomy concurrently with bilateral oophorectomy. Women who underwent bilateral oophorectomy had significantly more hyperlipidemia and significantly higher systolic blood pressure, weight, and BMI at the index date compared to the referent women. Virtually all women underwent hysterectomy either before oophorectomy or concurrent with oophorectomy. Therefore, virtually all women took ET alone. Among 163 women who started ET within one month after oophorectomy, 99 women (60.7%) used only oral ET, 20 women (12.3%) used only transdermal ET, and 44 women (27.0%) used both oral and transdermal ET. Among the 143 women (87.7%) who used oral ET, conjugated equine estrogen was the most frequent type (n = 98) and was most commonly used at a dose of 0.625 mg (n = 76). Oral estradiol was used by 18 women, most commonly at a dose of 1 mg (n = 14). Table 2 shows the mean values predicted by the GEE linear models for the metabolic parameters within 10 years after bilateral oophorectomy or the index date among the three groups. Although significant differences between the groups were observed for diastolic blood pressure, the magnitude of the differences was small, and the interaction between group and years after index was not significant. As shown in Figure 2 , the differences for diastolic blood pressure were similar over time. By contrast, a significant interaction was found for weight and BMI, and the changes occurred primarily in the initial 4–5 years. Women who underwent bilateral oophorectomy and were not taking ET at the measurement dates had higher values, compared to women who took ET. Women who took ET had modest increases over time, which were similar to the increases observed in the referent women ( Table 2 , Figure 2 ). Significant group differences were found for fasting total cholesterol and triglycerides, however, the interactions between group and years after index were not significant ( Figure 3 ). By contrast, we found a significant overall group difference and a significant interaction with years after index date for HDL-C. Once again, the changes occurred primarily in the initial 4–5 years. Women who took ET after bilateral oophorectomy experienced an increase in HDL-C over time, whereas women who did not take ET experienced a decrease over time. Supplementary Table 1 shows the predicted mean values for the metabolic parameters within 10 years after bilateral oophorectomy or the index date stratified by age at oophorectomy or index date. In this first set of sensitivity analyses, significant interactions between group and years after index were found for total cholesterol, triglycerides, and HDL-C among women age <40 years at index date; and for weight and BMI among women age 40–45 years at index date. In a second set of sensitivity analyses, we used multiple imputation methods to impute all possible values for the metabolic parameters within 10 years after index. The results were similar to the primary analysis; however, the confidence intervals were smaller due to the increased sample size (data not shown). In a third set of sensitivity analyses, we studied the mean values for the metabolic parameters among referent women, women with ≥5 years of ET after bilateral oophorectomy, and women with no ET or <5 years of ET after bilateral oophorectomy. The results were generally similar to the primary analysis. However, because the follow-up data was limited to 5–10 years after the index date, the interaction between group and years after index remained significant for only HDL-C ( p = 0.04, data not shown). Supplementary Table 2 shows the predicted mean values for the metabolic parameters within 3 years before index and within 10 years after index for referent women and women who underwent bilateral oophorectomy. In this fourth set of sensitivity analyses, significant differences between the two groups were noted for almost all measures ( Supplementary Figures 1 and 2 ), however, the interaction between group and years after index was only significant for total cholesterol and for LDL-C.

Women undergoing bilateral oophorectomy prior to the age of spontaneous menopause experience unfavorable changes in several metabolic parameters possibly related to the risk of CAD within the first few years following the surgery. Some of these changes can be reduced in magnitude by using ET. Future studies are needed to determine the effect of dose and formulation of ET on the trajectory of these metabolic parameters after bilateral oophorectomy. Recognizing the timeline of change in these parameters after bilateral oophorectomy is crucial for counseling patients and for planning timed interventions geared toward managing these risk factors, and ultimately, reducing the risk of CAD.

Over the 10 years of follow-up, women who underwent oophorectomy and took ET, women who underwent oophorectomy and did not take ET, and referent women had significantly different mean values for diastolic blood pressure, weight, BMI, total cholesterol, triglycerides, and HDL-C. However, women with and without oophorectomy were already different at the index date for hyperlipidemia, systolic blood pressure, weight, and BMI. Nevertheless, the trajectories of change over 10 years after bilateral oophorectomy were significantly different for weight, BMI, and HDL-C. The changes occurred primarily in the initial 4–5 years. Earlier age at menopause, whether spontaneous or surgically induced via bilateral oophorectomy, has been associated with an increased risk for CAD and cardiovascular mortality.[ 3 , 5 , 6 , 17 , 18 ] Premature menopause, defined as menopause occurring prior to the age of 40 years, as opposed to early menopause, defined as menopause occurring between the ages of 40–45 years, is associated with an even greater risk,[ 4 ] and is now recognized by the American College of Cardiology/American Heart Association as a risk factor for CAD.[ 19 ] Therefore, the importance of understanding and managing the risk factors for CAD in this clinical setting cannot be overstated.[ 20 , 21 ] Despite the well-recognized increase in risk of CAD after bilateral oophorectomy performed prior to spontaneous menopause, the mechanisms and the risk factors mediating this association are not known. Unfavorable changes in weight, body fat distribution, BMI, blood pressure and lipid profile that have been previously reported after bilateral oophorectomy could potentially increase the risk of CAD.[ 1 , 7 , 8 ] However, the trajectory of these changes following bilateral oophorectomy has not been examined. Even though prospective studies might provide the ideal setting to evaluate the trends in these biomarkers, such studies are difficult to conduct because of the long-term follow-up required. Data registries are therefore the next best alternative to examine such research questions. This is the first study to report on the trajectory of change in important and modifiable metabolic parameters related to risk of CAD, including weight, BMI, and dyslipidemia after bilateral oophorectomy performed in younger women. Identifying the timeline of change in these metabolic parameters is crucial for counseling patients as well as planning and timing appropriate mitigation strategies. The changes in weight and lipids seem to occur early after bilateral oophorectomy, and women should be informed of this risk, monitored for changes, and offered potential management options, including lifestyle counseling and medications (particularly statin therapy). Timely interventions can mitigate the risk of CAD and the associated mortality in this group of patients. The other novel finding from this study pertains to the effect of ET on the modifiable metabolic parameters related to risk of CAD. Even though ET has been shown to mitigate the risk of CAD in women with premature menopause,[ 4 , 17 , 22 ] its effect on the trajectory of modifiable CAD risk factors after bilateral oophorectomy is less well known. Again, even though this question is best investigated in a prospective randomized controlled clinical trial setting with comparisons drawn among different regimens of ET versus no ET, such clinical studies are difficult to conduct because of the long-term follow-up required. Therefore, a large data registry with longitudinal follow-up such as the REP medical records-linkage system provides a nearly ideal setting to investigate this question. In the current study, ET favorably impacted weight, BMI, and lipid profiles in young women after bilateral oophorectomy, confirming the results from some smaller previous studies.[ 4 , 23 ] It remains unclear whether the effect of ET on these metabolic biomarkers is the only mechanism involved in reducing the risk of CAD in this patient population. It is likely that other pathways in the pathogenesis of CAD are also favorably impacted by ET use, including improvement in insulin sensitivity[ 22 , 24 ] and vascular reactivity,[ 25 ] culminating in an overall reduced risk. The lack of an effect of ET on blood pressure points towards an alternative mechanism mediating blood pressure control in young women after bilateral oophorectomy. This study has several strengths. First, the bilateral oophorectomy cohort and the referent cohort from which the samples were randomly derived were representative of a geographically-defined population.[ 10 ] Second, details about the bilateral oophorectomy, baseline characteristics, ET, and the metabolic parameters were obtained through abstraction of medical records from a medical records-linkage system, thus limiting recall bias.[ 10 ] None of the data were self-reported. However, limitations also warrant consideration. The first and most important limitation was the source of the metabolic parameters. We considered only measures that were recorded in the medical records as part of routine medical care. Contrary to measures collected as part of a longitudinal research project, our measures may be incomplete and may lack standardization. Longitudinal data were available in most women for blood pressure and weight. However, the measures were only available on a subset of women for lipids. Unfortunately, lipid profiles are not done routinely in medical practice, and may be performed for a specific indication. On the other hand, over a long follow-up, these limitations should apply symmetrically to both cohorts of women. Therefore, any error should be non-differential and should not bias our conclusions. A set of sensitivity analyses involving imputation for missing values showed similar results. Similarly, it would have been desirable to include fasting glucose as a parameter in this study; however, fasting glucose was not routinely available for most of the participants. Second, the measures were abstracted by 2 physicians and 3 nurse abstractors. Different abstractors may have performed differently. However, the measures were mainly numerical values that did not require interpretation, and errors should be non-differential. Third, women participants were predominantly white, and all resided in Olmsted County, Minnesota. Thus, results may not be generalizable to other populations with different racial or ethnic characteristics.[ 13 ] Fourth, the observational nature of our study limits causal inference, and unknown confounding variables may have been present. For example, women who chose to use ET after bilateral oophorectomy may have had some health behavioral characteristics that were different from women who did not. Fifth, despite the initial sample size estimation, the study had limited power to study some of the group by time interactions, especially in analyses for lipids (missing data and exclusion of hyperlipidemia at the index date). On the other hand, the sample was limited to reduce the amount of medical record abstraction. Sixth, because we considered only 10 years of follow-up, the women in our study were still relatively young at the end of follow-up. It is possible that we would observe additional significant differences if the women were followed for a longer time. We plan to continue to follow our cohorts. Seventh, our study excluded women who carry high-risk genetic variants that increase the risk of ovarian cancer. Finally, we did not stratify our analyses based on the specific ET regimens and doses utilized by our cohort because there was a wide variability in ET regimens. The effect of different ET regimens on metabolic biomarkers might be best investigated in a randomized controlled clinical trial. In addition, because a majority of the participants in our study had a concurrent hysterectomy at the time of bilateral oophorectomy, progestogen therapy did not apply, and was not studied. Future studies should focus on the effect of progestogen therapy on these parameters.

We and other investigators have reported an increased risk of coronary artery disease (CAD) incidence or mortality after bilateral oophorectomy performed before the age of spontaneous menopause.[ 1 – 6 ] The effects of bilateral oophorectomy on body weight, body composition, lipid profile, and blood pressure in young women have also been reported.[ 1 , 7 , 8 ] However, it is not known how rapidly these metabolic parameters change after bilateral oophorectomy. In addition, it remains unclear which of these parameters are already different at baseline in women who undergo bilateral oophorectomy compared to women who do not.[ 9 ] Knowledge of these metabolic changes is clinically important because it may guide timely interventions. Even though estrogen therapy (ET) may reduce the risk of several adverse health consequences that result from early oophorectomy, the effects of ET on the metabolic parameters in the immediate years following bilateral oophorectomy are not well known.[ 1 ] The current study was conducted to identify trajectories of change in blood pressure, weight, and lipid profiles over a 10-year period after bilateral oophorectomy performed in premenopausal women. A secondary objective of the study was to evaluate the effect of ET on these trajectories.