CKD in Patients with Bilateral Oophorectomy

Abstract

Background and objectives 

Premenopausal women who undergo bilateral oophorectomy are at a higher risk of morbidity and mortality. Given the potential benefits of estrogen on kidney function, we hypothesized that women who undergo bilateral oophorectomy are at higher risk of CKD.

Design, setting, participants, & measurements 

We performed a population-based cohort study of 1653 women residing in Olmsted County, Minnesota who underwent bilateral oophorectomy before age 50 years old and before the onset of menopause from 1988 to 2007. These women were matched by age (±1 year) to 1653 referent women who did not undergo oophorectomy. Women were followed over a median of 14 years to assess the incidence of CKD. CKD was primarily defined using eGFR (eGFR<60 ml/min per 1.73 m² on two occasions >90 days apart). Hazard ratios were derived using Cox proportional hazards models, and absolute risk increases were derived using Kaplan–Meier curves at 20 years. All analyses were adjusted for 17 chronic conditions present at index date, race, education, body mass index, smoking, age, and calendar year.

Results 

Women who underwent bilateral oophorectomy had a higher risk of eGFR-based CKD (211 events for oophorectomy and 131 for referent women; adjusted hazard ratio, 1.42; 95% confidence interval, 1.14 to 1.77; absolute risk increase, 6.6%). The risk was higher in women who underwent oophorectomy at age ≤45 years old (110 events for oophorectomy and 60 for referent women; adjusted hazard ratio, 1.59; 95% confidence interval, 1.15 to 2.19; absolute risk increase, 7.5%).

Conclusions 

Premenopausal women who undergo bilateral oophorectomy, particularly those ≤45 years old, are at higher risk of developing CKD, even after adjusting for multiple chronic conditions and other possible confounders present at index date.

Podcast 

This article contains a podcast at https://www.asn-online.org/media/podcast/CJASN/2018_10_11_CJASNPodcast_18_1

Visual Abstract

Introduction

Estrogen has shown protective effects on the kidneys in both animal models and observational human studies. In particular, estrogen reduced glomerulosclerosis and glomerular permeability in ischemia-reperfusion injury in various animal models (¹). In one model of sclerosis-prone mice, ovariectomy caused severe glomerulosclerosis and kidney dysfunction that was then corrected by continuous estrogen replacement (²,³). Younger women have a lower incidence of ESKD and slower progression of CKD than men; however, that advantage disappears after menopause (^4–6). In addition, women may have a steeper increase in risk of cardiovascular and all-cause mortality for a given lower GFR or higher albuminuria compared with men (⁷). Observational studies of estrogen therapy in women have had variable results, with some studies showing beneficial effects of estrogen on markers of kidney dysfunction, such as albuminuria (⁸,⁹), and some other studies showing potential harm (¹⁰,¹¹). These discrepant studies have had significant methodologic differences, including the ages of the women studied and the methods of assessing kidney function.

Women who undergo bilateral oophorectomy before natural menopause (often performed concurrently with hysterectomy) may have a particularly high risk of harm from prolonged estrogen deprivation (¹²). There is increasing evidence that premenopausal women who undergo bilateral oophorectomy may be at a higher risk for significant morbidity and mortality, particularly if they are younger than age 46 years old at the time of oophorectomy and do not receive adequate estrogen therapy (^13–16). Studying this population of women who become acutely and surgically menopausal offers a unique opportunity to identify the precise onset of estrogen deprivation as well as adjust for the comorbidities and other possible confounders that are present at the index date (time of oophorectomy).

The long-term risk of CKD in women undergoing bilateral oophorectomy has not been explored previously. In this population-based cohort study, we evaluated the long-term risk of developing CKD in women who underwent bilateral oophorectomy compared with age-matched referent women without oophorectomy after adjusting for potential confounders, including cardiovascular, metabolic, and other chronic conditions present at the index date. We also assessed whether age at the time of oophorectomy and the use of estrogen therapy may modify this risk.

Materials and Methods

Cohort Definition and Comorbidities at Index Date

Women were identified as part of the Mayo Clinic Cohort Study of Oophorectomy and Aging-2 (MOA-2) as described extensively elsewhere (^15–17). All of the data used for this study were derived from the medical records linkage system of the Rochester Epidemiology Project, which has been described elsewhere (^18–21). All research activities were approved by the Mayo Clinic and Olmsted Medical Center Institutional Review Boards. The cohort study included women who underwent bilateral oophorectomy or second unilateral oophorectomy from January 1, 1988 to December 31, 2007 and age-matched referent women from the same population. The oophorectomy had to be performed before the onset of menopause and before age 50 years old. The date of the surgical procedure was considered the index date, and simple random sampling was used to identify referent women from the same Olmsted County, Minnesota population born in the same year (±1 year) who had not undergone bilateral oophorectomy before the index date. Referent women did not have to be menopausal at the index date and remained eligible if they underwent bilateral oophorectomy after the index date.

Medical comorbidities present at the index date were electronically abstracted using selected International Classification of Diseases (ICD) diagnosis codes (ICD-8 or ICD-9) (²²). Women needed to have at least two diagnostic codes in a given category separated by >30 days to avoid false positive diagnoses (¹⁷). Medical records were manually reviewed, and oophorectomy status was confirmed by trained study personnel (primarily L.G.R.). In addition, extensive clinical information was manually abstracted from the medical records, including demographic and reproductive characteristics and systemic estrogen therapy after the index date.

Assessment of CKD during Follow-Up

CKD was defined in two different ways—primarily using eGFR values and in a secondary analysis, using adjudicated diagnostic codes. The primary definition required an eGFR<60 ml/min per 1.73 m² on at least two occasions >90 days apart (²³). Serum creatinine (Cr) measurements were obtained from the electronic indexes of the Rochester Epidemiology Project and were available from the Mayo Clinic starting in 1994 and from Olmsted Medical Center in 1998, the two major providers of clinical care in Olmsted County, Minnesota. The Chronic Kidney Disease Epidemiology Collaboration equation was used to calculate eGFR from all electronically available, isotope dilution mass spectrometry-calibrated serum Cr levels (²³). The onset of CKD was defined as the date of the second reduced eGFR measurement (meeting the time gap specified), and CKD was considered incident if the onset was on or after the index date. Women without eGFR results available were assumed to not have CKD.

In the secondary definition, a nephrologist (A.G.K.) reviewed and adjudicated the medical records of women who had received diagnostic codes for CKD. We first compiled a list of ICD diagnosis codes for CKD and then identified the women with two or more codes >30 days apart assigned at any time (before or after the index date). The complete medical records of these women were then reviewed, and the Kidney Disease: Improving Global Outcomes criteria for CKD stages 1, 2, and 3b–5 were applied (²⁴). These criteria required an eGFR<45 ml/min per 1.73 m² and/or the presence of markers of kidney damage (proteinuria or abnormal urinary sediment) over a 3-month period. We excluded stage 3a (defined as an eGFR of 45–60 ml/min per 1.73 m² in the absence of proteinuria or abnormal urinary sediment), because during the pilot phase of the study, we noted that patients with stage 3a often went undiagnosed by providers and therefore, would have been missed by our targeted chart review. The rationale for adding this secondary definition was that CKD as defined by eGFR could be evaluated using laboratory values extracted electronically alone, whereas review of urine protein and sediment requires the clinical context for interpretation. Unfortunately, reviewing the medical records of the entire cohort was not feasible. We wanted to include CKD stages 1 and 2 in our analysis, because many of the studies suggesting a relationship between estrogen and the kidney evaluated proteinuria as an outcome.

Statistical Analyses

Each definition of CKD (eGFR and adjudicated diagnostic codes) was considered separately for the cohort analyses, and women with CKD onset before the index date were excluded from the corresponding outcome analysis (only incident outcomes were assessed). Women were followed from index date to the date of CKD onset, or they were censored at the earliest of death, time of last visit with a Rochester Epidemiology Project provider (lost to follow-up), or end of the study (December 31, 2014). We estimated unadjusted and adjusted hazard ratios (HRs) and 95% confidence intervals (95% CIs) using Cox proportional hazards models, with age as the timescale. Cumulative incidence curves were estimated using the Kaplan–Meier method, and absolute risks were obtained from the Kaplan–Meier curves at 20 years after oophorectomy or index. Differences between the two cohorts were also measured using the absolute risk increase (ARI) or absolute risk reduction obtained by subtracting the two absolute risks.

The Kaplan–Meier estimates and Cox models were adjusted for potential confounders using inverse probability weights derived from a logistic regression model including 17 preexisting chronic conditions, years of education (≤12, 13–16, or >16), race (white versus nonwhite), body mass index (BMI; <30 versus ≥30 kg/m²), smoking status (current or former versus never), and age and calendar year at baseline (continuous) (²⁵). Inverse probability weights were calculated overall and separately within each stratum to maximize the balance of characteristics at index date. Analyses were performed using SAS version 9.4 (SAS Institute, Inc., Cary, NC), and tests of statistical significance were conducted at the two-tailed α-level of 0.05.

Additional details on cohort selection, assessment of kidney function, methods for statistical analyses in strata by age at index date and use of estrogen therapy, and methods for sensitivity analyses are reported in Supplemental Material. The strata by age reflect the definition of premature or early menopause (≤45 years old).

Results

Characteristics at Index Date

There were 1653 women in the bilateral oophorectomy cohort and 1653 age-matched referent women. The median length of follow-up was 14.5 years in women who underwent bilateral oophorectomy (interquartile range [IQR], 10.3–19.1) and 14.4 years in the referent cohort (IQR, 10.4–19.3). The characteristics at index date of women without prevalent CKD are summarized in Table 1. The majority of women in both cohorts were white; however, the percentage of whites was higher in the bilateral oophorectomy group. Women who underwent bilateral oophorectomy had higher BMIs and were less educated. The majority of women had some estrogen therapy after bilateral oophorectomy (n=1494, 90%), whereas only 479 referent women (29%) had estrogen therapy after the index date (Supplemental Material has more details).

Table 1. - Baseline characteristics of women without prevalent CKD in the Mayo Clinic Cohort Study of Oophorectomy and Aging-2 from Olmsted County, Minnesota

Characteristic a	Referent, n=1644	Bilateral Oophorectomy, n=1638
Age at oophorectomy, median (IQR)	44 (40–47)	44 (40–47)
Age at oophorectomy, yr, n (%)
≤45	1028 (63)	1024 (63)
46–49	616 (37)	614 (37)
Calendar year of oophorectomy, n (%)
1988–1992	317 (19)	317 (19)
1993–1997	404 (25)	406 (25)
1998–2002	549 (33)	546 (33)
2003–2007	374 (23)	369 (23)
Race, n (%)
White	1561 (95)	1596 (97)
Black	29 (2)	18 (1)
Asian	49 (3)	18 (1)
Other	5 (0.3)	6 (0.4)
Years of education, n (%)
<9	31 (2)	8 (0.5)
9–12	443 (27)	517 (32)
13–16	858 (52)	883 (54)
>16	277 (17)	227 (14)
Unknown b	35 (2)	3 (0.2)
Smoking status, n (%)
Never	949 (58)	891 (54)
Past	377 (23)	388 (24)
Current	318 (19)	359 (22)
BMI, kg/m2, median (IQR)	25.9 (22.7–30.4)	27.2 (23.3–32.4)
BMI, kg/m2, n (%)
<25.0	697 (42)	590 (36)
25.0–29.9	486 (30)	477 (29)
≥30.0	438 (27)	571 (35)
Unknown b	23 (1)	0 (0)
Hysterectomy, n (%)
No	1482 (90)	24 (1)
Before	162 (10)	156 (10)
Concurrent with oophorectomy	—	1458 (89)
Oophorectomy indication, n (%) c
No ovarian indication	—	968 (59)
Benign condition	—	670 (41)
Concurrent hysterectomy indication, n (%) d
Cancer	—	11 (0.7)
Suspicion of cancer	—	308 (19)
Bleeding	—	823 (50)
Pain	—	588 (36)
Fibroids or polyps	—	412 (25)
Prolapse	—	335 (20)
Endometriosis	—	200 (12)
Other	—	225 (14)

IQR, interquartile range; BMI, body mass index; —, not applicable.

^aWomen with prevalent CKD defined using eGFR were excluded from this table (nine referent women and 15 women who underwent bilateral oophorectomy).

^bIn the regression models used to derive inverse probability weights, women with unknown education were assigned to the ≤12-year group, and women with unknown BMIs were assigned to the <30-kg/m² group.

^cThe indication was listed by the gynecologist in the medical record at the time of oophorectomy. Benign conditions include benign tumors, cyst, or endometriosis in either ovary. No ovarian indication includes women without a benign ovarian condition in either ovary. Historically, the terms “prophylactic,” “elective,” or “incidental” oophorectomy were used; however, we avoided these terms.

^dThe indication was listed by the gynecologist in the medical record at the time of hysterectomy. Each woman may have more than one indication recorded.

Comorbidities Present at Index Date

There were significant differences in the burden of chronic diseases at index date in the two cohorts. Women who underwent bilateral oophorectomy were more likely to have previous diagnoses of depression, anxiety, substance abuse disorders, hyperlipidemia, hypertension, diabetes mellitus, cardiac arrhythmias, asthma, and chronic obstructive pulmonary disease. Supplemental Figure 1 shows the odds ratios and 95% CIs from case-control analyses for the 17 chronic conditions. Because of these differences at index date, inverse probability weighting was used in all adjusted analyses of CKD outcomes to balance the two cohorts with respect to these chronic conditions and other potential confounders. As an example, Supplemental Figure 2 shows the adjustments obtained in the overall sample and the stratum ≤45 years old at index date (using the eGFR-based definition of CKD; other results are not shown).

Primary Analyses: CKD Defined Using eGFR

Figure 1 shows the number of women in each cohort who developed CKD using the eGFR-based definition. The median eGFR level before and closest to the index date was slightly lower in women who underwent bilateral oophorectomy (95.6 ml/min per 1.73 m²; IQR, 82.7–105.7) than in referent women (96.9 ml/min per 1.73 m²; IQR, 83.9–106.5; P=0.003). There were 15 women in the bilateral oophorectomy cohort and nine women in the referent cohort with prevalent CKD at the index date (Supplemental Figure 1). After the index date, 211 women in the bilateral oophorectomy cohort and 131 women in the referent cohort developed incident CKD (unadjusted HR, 1.67; 95% CI, 1.35 to 2.07) (Table 2, Figure 2). The HR remained significant after adjusting for education, race, BMI, smoking, age, calendar year, and the presence of 17 chronic conditions at index date using inverse probability weights (adjusted HR, 1.42; 95% CI, 1.14 to 1.77). The absolute risk of incident CKD 20 years after the index date was 20.2% (95% CI, 17.3% to 23.5%) in the bilateral oophorectomy cohort and 13.6% (95% CI, 11.2% to 16.4%) in the referent cohort (ARI, 6.6%). The HR was higher in women who underwent oophorectomy at age ≤45 years old (adjusted HR, 1.59; 95% CI, 1.15 to 2.19; ARI, 7.5%) compared with those at age 46–49 years old (adjusted HR, 1.33; 95% CI, 0.98 to 1.81; ARI, 6.1%). However, the HRs were not significantly different across the two strata. In women who underwent oophorectomy at age ≤45 years old, the HR was higher in women who did not take estrogen or stopped it before their 46th birthday (adjusted HR, 2.07; 95% CI, 0.72 to 5.92; ARI, 11.9%) compared with women who received estrogen through their 46th birthday (adjusted HR, 1.52; 95% CI, 1.05 to 2.20; ARI, 12.5%). However, the HRs were not significantly different across the two strata.

Figure 1.: Incident CKD defined using eGFR values was more common in women who underwent bilateral oophorectomy. Plasma or serum creatinine (Cr) tests were extracted from the Rochester Epidemiology Project electronic indexes (available back to 1994 from the Mayo Clinic and back to 1998 from the Olmsted Medical Center). eGFR was calculated from Cr values using the Chronic Kidney Disease Epidemiology Collaboration equation. CKD was defined as eGFR values <60 ml/min per 1.73 m² on two occasions >90 days apart. CKD present at the index date was considered prevalent, and CKD that developed on or after the index date was considered incident.

Table 2. - Associations of oophorectomy with incident CKD defined by eGFR (primary analyses)

Chronic Condition and Strata	Bilateral Oophorectomy				Referent Women				Unadjusted Models		Adjusted Models a
	N at Risk	Person-yr	N of Events	Absolute Risk, % (95% CI) b	N at Risk	Person-yr	N of Events	Absolute Risk, % (95% CI) b	Hazard Ratio (95% CI)	P Value	Hazard Ratio (95% CI)	P Value
Primary analyses
Overall	1638	22,723	211	20.2 (17.3 to 23.5)	1644	22,967	131	13.6 (11.2 to 16.4)	1.67 (1.35 to 2.07)	<0.001	1.42 (1.14 to 1.77)	0.002
Age ≤45 yr	1024	14,439	110	17.6 (14.2 to 21.7)	1028	14,373	60	10.1 (7.6 to 13.4)	1.88 (1.38 to 2.56)	<0.001	1.59 (1.15 to 2.19)	<0.01
ET>45 yr c	638	7795	75	25.0 (19.1 to 32.3)	597	7298	45	12.5 (8.9 to 17.5)	1.58 (1.10 to 2.27)	0.01	1.52 (1.05 to 2.20)	0.03
No ET or ≤45 yr	177	1613	17	27.0 (13.5 to 49.6)	161	1628	6	15.1 (4.6 to 43.2)	2.87 (1.15 to 7.21)	0.02	2.07 (0.72 to 5.92)	0.18
Age 46–49 yr	614	8283	101	25.4 (20.2 to 31.6)	616	8593	71	19.3 (14.9 to 24.8)	1.50 (1.11 to 2.03)	<0.01	1.33 (0.98 to 1.81)	0.07
ET>49 yr d	440	5622	71	29.0 (22.5 to 36.9)	422	5592	56	25.9 (19.3 to 34.4)	1.30 (0.92 to 1.85)	0.14	1.18 (0.83 to 1.69)	0.36
No ET or ≤49 yr	152	1408	20	39.5 (21.3 to 65.1)	153	1485	12	25.2 (11.7 to 49.3)	1.73 (0.85 to 3.52)	0.13	1.34 (0.65 to 2.77)	0.43
First set of sensitivity analyses e
Overall	1230	15,290	124	23.8 (18.0 to 31.0)	1132	14,399	60	13.7 (9.6 to 19.3)	2.02 (1.49 to 2.74)	<0.001	1.61 (1.17 to 2.21)	0.003
Age ≤45 yr	753	9440	64	21.4 (14.0 to 31.7)	677	8577	26	11.4 (6.5 to 19.9)	2.28 (1.45 to 3.59)	<0.001	1.71 (1.05 to 2.78)	0.03
ET>45 yr c	434	4714	38	40.0 (15.6 to 78.5)	383	4210	19	11.1 (6.3 to 19.2)	1.77 (1.02 to 3.07)	0.04	1.60 (0.90 to 2.83)	0.11
No ET or ≤45	144	1178	12	12.5 (5.5 to 27.3) f	125	1105	4	6.3 (2.5 to 15.7)	2.85 (0.92 to 8.76)	0.07	1.54 (0.46 to 5.16)	0.49
Age 46–49 yr	477	5850	60	28.4 (19.8 to 39.6)	455	5821	34	17.1 (11.1 to 26.0)	1.79 (1.18 to 2.72)	<0.01	1.51 (0.98 to 2.31)	0.06
ET>49 yr d	330	3789	41	32.8 (21.8 to 47.4)	304	3615	26	11.9 (7.9 to 18.0) f	1.55 (0.95 to 2.53)	0.08	1.36 (0.81 to 2.27)	0.25
No ET or ≤49 yr	127	1055	10	12.0 (5.9 to 23.4) f	119	1034	5	8.2 (3.7 to 17.8) f	2.01 (0.68 to 5.92)	0.20	1.32 (0.44 to 4.03)	0.62
Second set of sensitivity analyses g
Overall	656	9860	68	14.4 (11.1 to 18.8)	888	12,829	55	8.8 (6.5 to 12.0)	1.46 (1.03 to 2.08)	0.03	1.41 (0.99 to 1.99)	0.06
Age ≤45 yr	420	6452	37	14.1 (10.0 to 19.6)	592	8503	27	7.3 (4.7 to 11.1)	1.63 (1.00 to 2.67)	0.05	1.58 (0.97 to 2.59)	0.07
Age 46–49 yr	236	3407	31	15.5 (10.1 to 23.3)	296	4326	28	12.0 (7.7 to 18.6)	1.30 (0.78 to 2.16)	0.32	1.27 (0.76 to 2.12)	0.36

95% CI, 95% confidence interval; ET, estrogen therapy.

^aHazard ratios were calculated using Cox proportional hazards models with age as the timescale and adjusted using inverse probability weights derived from a regression model including 17 chronic conditions present at baseline, years of education (unknown, ≤12, 13–16, or >16), race (white versus nonwhite), body mass index (unknown or <30 versus ≥30 kg/m²), cigarette smoking (current or former versus never), age at baseline (continuous), and calendar year at baseline (continuous). These adjustments were performed separately in each stratum to maximize the balance at baseline. None of the interactions by age were significant.

^bAbsolute cumulative risk at 20 years after bilateral oophorectomy (or index) calculated using the Kaplan–Meier method. The estimates were adjusted using inverse probability weights derived from a logistic regression model including 17 chronic conditions present at baseline, years of education (unknown, ≤12, 13–16, or >16), race (white versus nonwhite), body mass index (unknown or <30 versus ≥30 kg/m²), cigarette smoking (current or former versus never), age at baseline (continuous), and calendar year at baseline (continuous). These adjustments were performed separately in each stratum to maximize the balance at baseline.

^cWomen who were taking ET on their 46th birthday after bilateral oophorectomy (only oral or transdermal). Women who developed CKD before their 46th birthday, died or were lost to follow-up before their 46th birthday, or had not reached age 46 years old as of December 31, 2014 were not included in the corresponding analysis. Follow-up for these analyses was started at age 46 years old. None of the interactions by ET were significant in the ≤45-years-old age stratum.

^dWomen who were taking ET on their 50th birthday after bilateral oophorectomy (only oral or transdermal). Women who developed CKD before their 50th birthday, died or were lost to follow-up before their 50th birthday, or had not reached age 50 years old as of December 31, 2014 were not included in the corresponding analysis. Follow-up for these analyses was started at age 50 years old. None of the interactions by ET were significant in the 46- to 49-years-old age stratum.

^eExcluding women with oophorectomy or index date before January 1, 1994 and women with no serum creatinine measurements available.

^fThe absolute risk was reported at 15 years after bilateral oophorectomy (or index) rather than at 20 years, because the follow-up was shorter.

^gExcluding women with any of the 17 chronic conditions at the index date or with onset of CKD defined by eGFR or adjudicated diagnostic codes before the index date.

Figure 2.: Higher cumulative incidence of CKD by eGFR-based criteria (upper panels) and adjudicated diagnostic codes (lower panels) in women who underwent bilateral oophorectomy. Cumulative incidence curves estimated using the Kaplan–Meier method and adjusted using inverse probability weights are shown in red for the bilateral oophorectomy cohort and black for the referent cohort. The hazard ratios (HRs) and corresponding 95% confidence intervals were calculated using Cox proportional hazards models with age as the timescale and adjusted using inverse probability weights. Analyses are shown overall (left panels), for women age ≤45 years old at the index date (center panels), and for women age 46–49 years old at the index date (right panels).

A first sensitivity analysis limited to women with an index date in the years 1994–2007 (when electronic Cr measurements were available) and excluding 142 women who did not have any serum Cr measurements (19 from the bilateral oophorectomy cohort and 123 from the referent cohort) yielded even stronger associations (overall adjusted HR, 1.61; 95% CI, 1.17 to 2.21) (Table 2). In a second sensitivity analysis restricted to women without any chronic conditions at baseline (excluding 997 women from the bilateral oophorectomy cohort and 765 women from the referent cohort), the HR was significant in the unadjusted model (HR, 1.46; 95% CI, 1.03 to 2.08) but was not significant in the adjusted model (HR, 1.41; 95% CI, 0.99 to 1.99; ARI, 5.6%). The results were virtually unchanged from the primary analysis in a third sensitivity analysis, which censored 84 referent women who underwent bilateral oophorectomy after the index date but before age 50 years old (results not shown).

Secondary Analyses: CKD Defined Using Adjudicated Diagnostic Codes

Figure 3 shows the number of women in each cohort who developed CKD defined using the adjudicated diagnostic codes. CKD was present at the index date in 11 women in the bilateral oophorectomy cohort and six women in the referent cohort (Supplemental Figure 1). Incident CKD developed in 61 women who underwent bilateral oophorectomy (28 by eGFR and 33 by urine protein) and 43 referent women (25 by eGFR, 17 by urine protein, and one by abnormal urinary sediment; adjusted HR, 1.17; 95% CI, 0.79 to 1.74) (Table 3, Figure 2). The HR was also not statistically significant in a first sensitivity analysis restricted to women with no chronic conditions at baseline (adjusted HR, 1.25; 95% CI, 0.67 to 2.32) (Table 3). Similar results to the secondary analysis were observed in an additional sensitivity analysis that censored referent women who underwent bilateral oophorectomy after the index date but before age 50 years old (results not shown). The causes of CKD obtained from medical record abstraction are described in Supplemental Material.

Figure 3.: Incident CKD defined using adjudicated diagnostic codes was more common in women who underwent bilateral oophorectomy. The electronic indexes of the Rochester Epidemiology Project were screened for a list of International Classification of Diseases (ICD) diagnosis codes for CKD (ICD-8 or ICD-9). The medical records for all women with at least two of these codes separated by >30 days were then reviewed by a nephrologist. CKD was defined as an eGFR<45 ml/min per 1.73 m² or evidence of kidney damage (proteinuria or active urinary sediment) on at least two occasions >90 days apart. CKD present at the index date was considered prevalent, and CKD that developed on or after the index date was considered incident. ^aWe excluded CKD stage 3a, because it often went undiagnosed by the care providers. ^bStructural abnormalities without evidence of kidney dysfunction were not included (e.g., atrophic kidney, medullary sponge kidney, hydronephrosis, and partial or complete nephrectomy).

Table 3. - Associations of oophorectomy with incident CKD defined by adjudicated diagnostic codes (secondary analyses)

Chronic Condition and Strata	Bilateral Oophorectomy				Referent Women				Unadjusted Models		Adjusted Models a
	N at Risk	Person-yr	N of Events	Absolute Risk, % (95% CI) b	N at Risk	Person-yr	N of Events	Absolute Risk, % (95% CI) b	Hazard Ratio (95% CI)	P Value	Hazard Ratio (95% CI)	P Value
Secondary analyses
Overall	1642	23,467	61	5.3 (3.8 to 7.3)	1647	23,530	43	4.7 (3.3 to 6.6)	1.44 (0.98 to 2.12)	0.06	1.17 (0.79 to 1.74)	0.44
Age ≤45 yr	1024	14,754	39	5.8 (3.9 to 8.6)	1028	14,542	26	4.4 (2.8 to 6.9)	1.50 (0.92 to 2.44)	0.10	1.20 (0.73 to 1.99)	0.48
ET>45 yr c	644	8086	22	5.8 (3.3 to 9.8)	600	7459	19	7.8 (4.6 to 13.3)	1.07 (0.58 to 1.97)	0.84	0.97 (0.52 to 1.81)	0.92
No ET or ≤45	176	1639	9	25.1 (10.7 to 52.3)	161	1644	2	3.4 (0.9 to 12.2)	4.70 (0.99 to 22.31)	0.05	2.08 (0.39 to 11.10)	0.39
Age 46–49 yr	618	8713	22	4.7 (2.8 to 7.7)	619	8988	17	5.2 (3.0 to 8.9)	1.34 (0.71 to 2.53)	0.36	1.23 (0.64 to 2.35)	0.54
ET>49 yr d	446	5959	14	5.4 (2.7 to 10.5)	425	5855	14	4.1 (2.3 to 7.1)	1.00 (0.47 to 2.10)	0.99	0.83 (0.39 to 1.78)	0.63
No ET or ≤49 yr	155	1489	5	2.6 (0.9 to 7.2)	154	1599	3	4.4 (0.7 to 24.5)	1.80 (0.46 to 6.99)	0.40	1.20 (0.30 to 4.74)	0.80
Sensitivity analyses e
Overall	656	10,113	22	4.2 (2.5 to 6.9)	888	13,004	19	3.7 (2.3 to 6.1)	1.37 (0.74 to 2.56)	0.31	1.25 (0.67 to 2.32)	0.48
Age ≤45 yr	420	6565	14	4.5 (2.4 to 8.5)	592	8572	11	2.7 (1.4 to 5.2)	1.50 (0.68 to 3.30)	0.31	1.32 (0.60 to 2.94)	0.49
Age 46–49 yr	236	3548	8	3.6 (1.6 to 7.9)	296	4432	8	6.0 (2.9 to 12.5)	1.15 (0.42 to 3.14)	0.78	1.11 (0.41 to 3.03)	0.84

95% CI, 95% confidence interval; ET, estrogen therapy.

^aHazard ratios were calculated using Cox proportional hazards models with age as the timescale and adjusted using inverse probability weights derived from a regression model including 17 chronic conditions present at baseline, years of education (unknown, ≤12, 13–16, or >16), race (white versus nonwhite), body mass index (unknown or <30 versus ≥30 kg/m²), cigarette smoking (current or former versus never), age at baseline (continuous), and calendar year at baseline (continuous). These adjustments were performed separately in each stratum to maximize the balance at baseline. None of the interactions by age were significant.

^bAbsolute cumulative risk at 20 years after bilateral oophorectomy (or index) calculated using the Kaplan–Meier method. The estimates were adjusted using inverse probability weights derived from a logistic regression model including 17 chronic conditions present at baseline, years of education (unknown, ≤12, 13–16, or >16), race (white versus nonwhite), body mass index (unknown or <30 versus ≥30 kg/m²), cigarette smoking (current or former versus never), age at baseline (continuous), and calendar year at baseline (continuous). These adjustments were performed separately in each stratum to maximize the balance at baseline.

^cWomen who were taking ET on their 46th birthday after bilateral oophorectomy (only oral or transdermal). Women who developed CKD before their 46th birthday, died or were lost to follow-up before their 46th birthday, or had not reached age 46 years old as of December 31, 2014 were not included in the corresponding analysis. Follow-up for these analyses was started at age 46 years old. None of the interactions by ET were significant in the ≤45-years-old age stratum.

^dWomen who were taking ET on their 50th birthday after bilateral oophorectomy (only oral or transdermal). Women who developed CKD before their 50th birthday, died or were lost to follow-up before their 50th birthday, or had not reached age 50 years old as of December 31, 2014 were not included in the corresponding analysis. Follow-up for these analyses was started at age 50 years old. None of the interactions by ET were significant in the 46- to 49-years-old age stratum.

^eExcluding women with any of the 17 chronic conditions at index date or with onset of CKD defined by eGFR or adjudicated diagnostic codes before the index date.

Discussion

We found that premenopausal women who underwent bilateral oophorectomy before age 50 years old were at higher risk of developing CKD compared with a cohort of age-matched referent women. This population-based study offers several advantages over prior efforts to understand the association of menopause and estrogen therapy with kidney function, because we identified women who experienced an abrupt onset of menopause. This design allowed us to adjust for chronic conditions present at the time of menopause onset that may affect the risk of CKD. We found consistent elevations and patterns of risk using both definitions of CKD, with a notably higher risk of CKD in women ≤45 years old at the time of oophorectomy and women who did not receive estrogen therapy; however, our findings were not statistically significant using the adjudicated diagnostic codes, likely because of limited statistical power. Similarly, the statistical power may have been limited in some of the stratified analyses due to small numbers of events.

Animal models support the hypothesis that estrogen deprivation may have important direct harmful effects on kidney structure and function. Mesangial cells that are critical to the pathogenesis of glomerulosclerosis express both estrogen receptor subtypes (α and β) in mice and humans (²⁶,²⁷). Estrogen has been found to prevent the accumulation of extracellular matrix and decrease the synthesis of type 1 and type 4 collagen (²⁶,²⁸,²⁹). Ovariectomy caused significant kidney dysfunction, accumulation of extracellular matrix, and glomerulosclerosis in one study of sclerosis-prone mice (²). In addition, glomerulosclerosis and albuminuria were both prevented by continuous estradiol administration in a subsequent study using the same mouse model (³⁰).

Previous studies in women have suggested that women have a lower incidence of ESKD and slower progression of CKD before menopause than men (⁵,⁶). This observed sex difference is often attributed to the beneficial effects of estrogen on the kidney, because the sex difference attenuates after menopause. A post hoc analysis in the Modification of Diet and Renal Disease Study showed a slower decrease in GFR in women than in men with CKD, particularly in women younger than age 52 years old (⁶). However, the effect of sex was attenuated after the investigators adjusted the analyses for BP, proteinuria, and HDL. Therefore, the difference in progression may have been mediated by these intervening conditions rather than it being due to a direct effect of estrogen on kidney function.

At least three observational studies in women and a recent meta-analysis have suggested that estrogen therapy may be associated with a lower risk of albuminuria (⁸,⁹,³¹,³²). Estrogen can stimulate the release of nitric oxide and cause vasodilation (³³,³⁴). Impaired vasodilation and endothelial dysfunction are seen in CKD, and there is mounting evidence that nitric oxide deficiency may be linked to acceleration of kidney damage (³⁵). Loss of estrogen after bilateral oophorectomy may, therefore, affect kidney structure and function in several ways, and it may increase the risk of developing CKD.

One challenge in studying the effects of estrogen therapy in humans is that the age of a woman and the time since menopause are important factors (timing hypothesis). For example, although the Women’s Health Initiative clinical trials showed an overall higher risk of coronary heart disease with estrogen therapy (³⁶), a post hoc analysis showed a trend toward reduced cardiovascular events in women who initiated hormone therapy closer to the time of menopause compared with those who initiated hormone therapy later (³⁷). Similarly, a large Canadian study showed that hormone therapy in older women (age >66 years old) was associated with a greater rate of decline in eGFR compared with in nonusers (¹⁰). By contrast, we observed a higher risk of CKD after bilateral oophorectomy after adjusting for relevant comorbid conditions in our cohort of younger women. Our findings suggest that there may be a direct beneficial effect of estrogen on kidney function. This is further supported by our observation that women younger than age 46 years old at the time of oophorectomy and who did not take estrogen or stopped it before their 46th birthday had a higher HR.

Observational evidence continues to mount suggesting that women who undergo bilateral oophorectomy at a younger age, particularly those without adequate estrogen therapy, have a higher risk of long-term morbidity and mortality (¹³,³⁸,³⁹). Findings from the MOA-2 have recently shown a significant accumulation of multimorbidity after bilateral oophorectomy that may be in part ameliorated by estrogen therapy (¹⁵,¹⁶). Levine et al. (⁴⁰) showed that bilateral oophorectomy may be associated with accelerated aging as measured by the DNA methylation level in blood and saliva, a biologic marker considered an “epigenetic clock.” These findings have prompted the suggestion that women who become prematurely menopausal after bilateral oophorectomy should be treated with hormone therapy at “replacement” doses of estrogen to protect cardiovascular, bone, and neurologic health and slow the pace of aging as opposed to the lowest possible dose to control symptoms (⁴¹). Our study suggests that the kidney may be an additional organ that can suffer from premature estrogen deprivation. The kidney damage may be due to a combination of direct effects of estrogen deprivation and secondary damage mediated by the accumulation of multimorbidity in other organs and systems (e.g., kidney damage secondary to hypertension or diabetes). It remains unknown whether the effects of exogenous estrogen are similar to the effects of endogenous estrogen. In addition, exogenous estrogen used as therapy may vary by composition, dose, and route of administration.

Our study has some limitations. First, our study focused on a single geographically defined United States population, and the observed associations may differ in other populations in the United States and worldwide. Second, we were only able to access serum Cr and urine protein measurements that were taken as part of routine clinical care; therefore, some women did not have any measurements. Women who underwent bilateral oophorectomy had more frequent Cr tests and medical visits than referent women after the index date. These differences were small in absolute terms, but they were statistically significant (because of the large sample size) (Supplemental Material). In our adjudication of diagnostic codes, we found that 85% of the women with incident CKD had urine protein and Cr values at baseline to confirm absence of CKD; however, the absence of CKD at baseline could not be confirmed for the remaining 15% of the women. We tried to define CKD using two sets of criteria, a more inclusive set of criteria on the basis of eGFR alone and a set of criteria on the basis of targeted chart review using diagnostic codes. We observed a similar pattern of risk in the predefined age and estrogen therapy strata using the two definitions.

In conclusion, women who undergo bilateral oophorectomy before natural menopause are at higher risk of CKD, as defined by reduced eGFR, after adjustments for potential confounding comorbidities. Therefore, women considering bilateral oophorectomy for the prevention of ovarian cancer, particularly those at ages ≤45 years old, need to be counseled regarding the potential risks of multimorbidity, which may include the risk of CKD. There is a trend toward conserving the ovaries at the time of hysterectomy unless there is a compelling indication to remove them; however, the practice continues (^42–44). Adequate hormone therapy in this group of women is essential, but there are limited data regarding the correct dose and route of administration (⁴¹). In addition, although estrogen has been shown to have multiple beneficial effects on the kidney in animal models, human studies have had contradictory results. To resolve these contradictions, future studies of the effects of hormone therapy on kidney function should be guided by the timing hypothesis (i.e., the effects of estrogen vary with the age of women).

Disclosures

None.