All‐cause mortality after self‐reported vs. hospital‐reported gynecological surgery: The HUNT study

Abstract

Introduction

The impact of hysterectomy and bilateral oophorectomy on future mortality risk is still debated. Self‐reported exposure may be a major source of bias in many studies. We aimed to determine mortality following hospital‐reported hysterectomy and bilateral oophorectomy and compare to results based on self‐reported surgeries.

Material and Methods

We included 37 140 women from the second (1995–1997) and third (2006–2008) HUNT surveys. HUNT (the Trøndelag Health Study) is a population‐based Norwegian cohort study, and we linked to the Norwegian Cause of Death Registry, the Cancer Registry of Norway, and hospital surgical codes. The participants were followed from inclusion in HUNT until December 2020. Women were classified as unexposed, exposed to hysterectomy alone, exposed to bilateral oophorectomy alone, or exposed to both surgeries. Hospital‐reported exposure was determined using surgical procedure codes, whereas self‐reported exposure was obtained from the HUNT questionnaires. To be classified as exposed based on self‐report, a woman had to indicate that she had undergone the surgery and provide her age at the time of the procedure. We compared the hazard of death in unexposed vs. exposed groups using Cox regression analysis, adjusting for relevant covariates.

Results

Compared with unexposed women, hospital‐reported hysterectomy was associated with a reduced hazard of death (HR 0.72 [95% CI 0.62–0.84]), while self‐reported hysterectomy showed no association (HR 1.06 [95% CI 0.97–1.16]). For bilateral oophorectomy, hospital‐reported cases showed an HR of 0.83 (95% CI 0.64–1.07), while self‐reported cases indicated an HR of 1.14 (95% CI 0.97–1.33). Women who had both hysterectomy and bilateral oophorectomy demonstrated no significant association with mortality, regardless of whether the surgeries were hospital‐reported (HR 1.13 [95% CI 0.96–1.34]) or self‐reported (0.93 [95% CI 0.82–1.06]). Early hospital‐reported bilateral oophorectomy was associated with an increased hazard of death, particularly when performed before the age of 50.

Conclusions

Results differed between hospital‐reported and self‐reported exposure. This could explain the heterogeneity observed in previous studies on mortality risk after hysterectomy and oophorectomy. In this study, hospital‐reported hysterectomy and bilateral oophorectomy were not associated with increased mortality, whereas early bilateral oophorectomy was.

Mortality risk after hysterectomy and oophorectomy varied between hospital‐reported and self‐reported exposure. This could explain the heterogeneity between previous studies. In this study, hospital‐reported hysterectomy and bilateral oophorectomy were not associated with increased mortality risk, but early bilateral oophorectomy was.

Abbreviations

BMI

Body mass index

BSO

Bilateral oophorectomy

CI

Confidence interval

HDL

High‐density lipoprotein

HNT

Nord‐Trøndelag Hospital Trust

HR

Hazard ratio

HUNT

the Trøndelag Health Study

ICD‐9

International Classification of Diseases, Ninth Revision

ICD‐10

International Classification of Diseases, Tenth Revision

Key message.

Estimates of mortality after hysterectomy and bilateral oophorectomy differed substantially depending on whether exposure was defined using self‐report or hospital procedure codes.

1. INTRODUCTION

Associations between hysterectomy and future incident and fatal cardiovascular disease have been extensively researched. Despite this, a recent meta‐analysis of 17 studies could not firmly conclude regarding these associations, due to a high level of heterogeneity between the studies included. ¹ A changing trend over time, with a greater variety of treatments in recent years, shifts in indications for surgery, and changes in surgical methods might explain some of this heterogeneity. As for hysterectomy, the impact of bilateral oophorectomy (BSO) on future cardiovascular mortality is still controversial, but there is increasing evidence that premenopausal bilateral oophorectomy increases the risk in women who are not given estrogen replacement therapy. ² , ³ , ⁴ , ⁵ , ⁶ Most studies assessing the associations between these surgeries and future morbidity and mortality are observational and use self‐reported data to define the exposure, as described in the meta‐analysis by Chen et al., where more than half of the included studies were based on self‐report/interview. ¹ This can potentially be a major cause for bias and explain the inability to conclude, even in meta‐analyses.

A few studies have evaluated the validity of self‐reported hysterectomy and found it satisfactory for epidemiological purposes. ⁷ , ⁸ , ⁹ , ¹⁰ However, caution was advised when interpreting results based on self‐reported oophorectomy, as the one study evaluating it found low sensitivity. ⁷ In a previous study, our research group found that self‐reported hysterectomy was associated with increased all‐cause and cardiovascular mortality, while BSO was not. ¹¹ We saw the need to evaluate the validity of self‐reported gynecological surgery in our cohort, and in a recent study, we could similarly conclude that self‐reported hysterectomy corresponded quite well with hospital procedure codes, while 1/3 of women reporting a history of bilateral oophorectomy had no corresponding surgical procedure code. ¹² To the best of our knowledge, no previous studies have directly assessed how self‐reported exposure affects the estimated risks after hysterectomy and BSO.

The primary aim was to assess how the estimated association between gynecological surgery and mortality differs when exposure is defined using hospital procedure codes versus self‐report. A secondary aim was to examine cardiovascular mortality using the same exposure definitions. We hypothesized that we would find similar associations between hysterectomy and mortality, but a stronger link between bilateral oophorectomy and mortality when using hospital data.

2. MATERIAL AND METHODS

This study is reported following the STROBE guidelines for observational studies.

The HUNT study is a large Norwegian population‐based cohort study that has previously been described elswhere. ¹³ , ¹⁴ In summary, all inhabitants in the Nord‐Trøndelag region above 20 years of age (HUNT1‐4) was invited to participate. In HUNT4, the Sør‐Trøndelag region was included, and in this region, all inhabitants above 18 years of age were invited. The surveys include two questionnaires, blood sampling and clinical measurements. The study started with HUNT1 in 1984–1986 and have been repeated in approximately 10‐year intervals. The questionnaires have had adaptations in each decade. In the second questionnaire in HUNT2 (1995–1997) and HUNT3 (2006–2008), women were asked about previous gynecological surgery, and therefore, we included women only from these two HUNT surveys. Details about the questionnaire that included history of gynecological surgery were described in our previous article. ¹² Briefly summarized, women were asked if they ever had removed their uterus or both ovaries, and at what age. Most covariates are from HUNT and recorded at the time of inclusion, with a few important exceptions. Using the Norwegian 11‐digit unique personal identification number, we linked data from HUNT to the following registries: The Norwegian Cause of Death Registry for our main outcome, all‐cause mortality and for cardiovascular mortality, the Cancer Registry of Norway for a history of cancer, and the electronic patient administrative systems of Nord‐Trøndelag Hospital Trust (HNT). Our data from HNT included surgical procedure codes from the two local hospitals, Levanger and Namsos, and the regional hospital, St. Olav's University Hospital. We also linked to HUNT4 for details on education.

In total, 30 263 women attended HUNT2 and answered both questionnaires, whereas 6877 additional women entered the study in HUNT3. Date of inclusion was set to date of questionnaire 1 in HUNT2 or HUNT3, whenever a woman first entered the study. To ensure anonymity, a woman's birthday was set to the 15th of the month and year she was born. Covariates from HUNT were defined based on questionnaire answers, medical examination, and lab results at the time of inclusion. Level of education was an exception to this rule, as we used the information in the last participation in HUNT, including HUNT4 when applicable. In HUNT3, women were not asked about education. We classified level of education into four groups: Primary school or less, 1–3 years of high school, <4 years of higher education, 4 years or more of higher education. Age was defined as obtained age at inclusion. Total and HDL cholesterol, systolic blood pressure, weight, height, and waist and hip circumference were all measured by trained staff at the HUNT examination stations. Waist and hip circumference were used to calculate waist/hip ratio, and BMI was calculated from height and weight. Smoking status was categorized based on self‐report as never, previous, or current. In HUNT3, smoking status differentiated between daily and sometimes smoking, but not in HUNT2. Therefore, we combined daily and sometimes smoking into currently smoking. Previous cardiovascular disease was defined as self‐reported history of angina, myocardial infarction, and/or stroke. Parity, defined as number of births, was handled as a categorical variable ranging from 0 to 8. Age at first menstruation (menarche) and a history of diabetes was self‐reported. Self‐reported health was categorized as bad, not quite good, good, or very good.

By law, all deaths of Norwegians should be registered in The Norwegian Cause of Death Registry, and the coverage has been >98% for many years. Doctors complete a death certificate that can include up to three diagnoses in the causal chain leading to death (Ia–c) and other relevant diseases (II). In addition, the registry itself reports the main underlying cause of death based on the death certificate and autopsy reports, following international rules and guidelines from the World Health Organization. ¹⁵ We defined all‐cause mortality as any death registered in the death registry during follow‐up (from inclusion in HUNT to December 31, 2020). For the secondary analyses on mortality related to cardiovascular disease, we included any death that had an ICD‐9 (390–459) or ICD‐10 (I00‐I99) code for cardiovascular disease in the death certificate (Ia–c or II) or registered as the underlying cause of death from the registry.

Cancer diagnoses in Norway have been registered in the Cancer Registry of Norway since January 1, 1953. For included women, we obtained any cancer diagnoses from 1953 until the end of follow‐up. A woman was defined as having had a gynecological cancer if she had at least one incidence of cancer of the cervix, uterus, ovaries, or other and unspecified female genital organs (ICD‐10 codes C51–C57). Women who had gynecological cancer before inclusion in HUNT or before surgery were excluded.

Electronic registration of surgical procedure codes in the Nord‐Trøndelag Hospital Trust (HNT) started on January 1, 1988. We obtained all procedure codes from that day and until end of follow‐up. Women were defined as unexposed or exposed based on ever having the corresponding procedure code from January 1988 until December 31, 2020. Based on the surgical procedure codes, we defined 4 groups:

1. No hysterectomy or oophorectomy in HNT

2. Hysterectomy alone in HNT

3. BSO alone in HNT

4. Hysterectomy & BSO in HNT

We defined the same four groups based on self‐report. A woman had to report having had the surgery and her age at that time to be classified as exposed to the surgery. Women with incomplete information on gynecological surgery were excluded. Women could contribute to both exposure definitions. Hospital‐reported and self‐reported surgeries were analyzed in separate models and not combined. See Figure 1. In addition, we conducted subgroup analyses based on age at surgery. Each of the three exposed groups was subdivided into surgery before 45, 45–49, 50–54, and ≥ 55 years of age. These age groups were chosen a priori as the impact of hysterectomy and BSO is expected to lessen the closer to menopause it is performed. ¹⁶ , ¹⁷ , ¹⁸ We were unable to categorize surgery prior to 40 years of age due to low numbers. We also conducted a sensitivity analysis where we excluded those who reported gynecological surgery prior to 1988 to better compare the results to those based on surgical procedure codes. To assess the potential immortal time bias, we conducted a sensitivity analysis where women who had their surgery after inclusion in HUNT were excluded.

FIGURE 1.

(A) Inclusion and exclusion of participants, with exposure based on hospital‐reported procedure codes. (B) Inclusion and exclusion of participants, with exposure based on self‐report.

Given the variability in previous studies on hysterectomy and mortality, we assessed time trends to determine whether mortality risks associated with hysterectomy depended on the time period of surgery. We stratified the mortality analyses by the decade in which the surgery was performed. Based on hospital procedure codes, we defined hysterectomy before year 2000, year 2000–2009, and after year 2010. Based on self‐report, we defined hysterectomy before 1960, 1960–1969, 1970–1979, 1980–1989, 1990–1999, and after year 2000.

2.1. Statistical analyses

All statistical analyses were done in STATA 18. The present study included all eligible women in the cohort (Figure 1). Women with missing data on any of the variables used in the model were excluded in the fully adjusted analyses. Number of included participants in the fully adjusted model is listed in parentheses in Tables 2 and 3. Categorical variables are presented as numbers and percentages while continuous covariates are presented as mean (SD). Interpretation focused on the direction, magnitude, and consistency of hazard ratios across exposure definitions, with 95% confidence intervals used to describe statistical precision.

TABLE 2.

Associations of hospital‐reported hysterectomy (Hyst) and bilateral oophorectomy (BSO) with all‐cause mortality.

	N	Person‐years	Deaths	Age‐adjusted HR (95% CI)	Fully adjusted HR (95% CI)
Ovaries and uterus intact	32 408 (27 666)	630 000 (560 000)	8730 (7032)	Ref	Ref
Hysterectomy	2273 (2063)	50 000 (46 000)	188 (169)	0.70 (0.60–0.81)	0.72 (0.62–0.84)
Hyst <45 y	821 (723)	18 000 (16 000)	16 (15)	0.58 (0.36–0.95)	0.57 (0.34–0.95)
Hyst 45–49 y	631 (585)	14 000 (13 000)	24 (22)	0.63 (0.42–0.95)	0.62 (0.41–0.95)
Hyst 50–54 y	319 (296)	7200 (6700)	25 (24)	0.84 (0.57–1.25)	0.94 (0.63–1.40)
Hyst ≥55y	502 (459)	11 000 (10 000)	123 (108)	0.71 (0.59–0.85)	0.74 (0.61–0.89)
BSO	385 (346)	8100 (7500)	70 (58)	0.82 (0.65–1.03)	0.83 (0.64–1.07)
BSO <45 y	23 (17)	467 (365)	2 (2)	2.91 (0.73–11.64)	4.08 (1.02–16.38)
BSO 45–49 y	39 (35)	758 (694)	5 (5)	2.65 (1.10–6.39)	2.50 (1.04–6.03)
BSO 50–54 y	59 (53)	1300 (1200)	5 (5)	1.54 (0.64–3.71)	1.72 (0.72–4.15)
BSO ≥55 y	264 (241)	5600 (5200)	58 (46)	0.73 (0.56–0.94)	0.71 (0.53–0.95)
Hyst & BSO	707 (648)	15 000 (14 000)	155 (142)	1.09 (0.93–1.28)	1.13 (0.96–1.34)
Hyst & BSO <45 y	72 (65)	1500 (1400)	6 (6)	2.29 (1.03–5.12)	2.22 (0.99–4.96)
Hyst & BSO 45–49 y	118 (106)	2500 (2300)	17 (15)	2.64 (1.64–4.26)	2.50 (1.50–4.16)
Hyst & BSO 50–54 y	137 (122)	3000 (2700)	12 (9)	1.02 (0.58–1.80)	0.94 (0.49–1.81)
Hyst & BSO ≥55y	380 (355)	7900 (7400)	120 (112)	0.99 (0.83–1.19)	1.05 (0.87–1.26)

Note: Age adjustment: Adjusting for achieved age at inclusion in HUNT. Full adjustment: Age at inclusion, level of education, BMI, waist/hip ratio, smoking, total & HDL cholesterol, systolic blood pressure, diabetes, cardiovascular disease (angina, MI or stroke), parity, menarche and self‐reported health. Number (n), person‐years and deaths are shown for the age‐adjusted analysis, and for the fully adjusted analysis, in parentheses.

TABLE 3.

Associations of self‐reported hysterectomy (Hyst) and bilateral oophorectomy (BSO) with all‐cause mortality.

	n	Person‐years	Deaths	Age‐adjusted HR (95% CI)	Fully adjusted HR (95% CI)
Ovaries and uterus intact	27 518 (24 125)	560 000 (500 000)	6328 (5225)	Ref	Ref
Hysterectomy	1995 (1783)	39 000 (36 000)	642 (544)	1.06 (0.97–1.15)	1.06 (0.97–1.16)
Hyst <45 y	937 (830)	19 000 (17 000)	246 (208)	1.17 (1.03–1.33)	1.09 (0.95–1.25)
Hyst 45–49 y	544 (499)	11 000 (10 000)	171 (149)	1.07 (0.92–1.25)	1.09 (0.92–1.28)
Hyst 50–54 y	263 (240)	5200 (4800)	98 (86)	1.13 (0.93–1.38)	1.20 (0.97–1.48)
Hyst ≥55y	251 (214)	4500 (4000)	127 (101)	0.84 (0.70–1.00)	0.91 (0.75–1.11)
BSO	390 (341)	7400 (6700)	163 (137)	1.14 (0.97–1.33)	1.18 (1.00–1.40)
BSO <45 y	226 (192)	4300 (3800)	92 (78)	1.11 (0.91–1.37)	1.20 (0.96–1.50)
BSO 45–49 y	52 (49)	1000 (977)	21 (18)	1.39 (0.91–2.14)	1.29 (0.81–2.05)
BSO 50–54 y	36 (32)	650 (580)	17 (15)	2.20 (1.37–3.54)	2.06 (1.24–3.42)
BSO ≥55 y	76 (68)	1400 (1300)	33 (26)	0.87 (0.62–1.23)	0.87 (0.59–1.28)
Hyst & BSO	691 (610)	13 000 (12 000)	290 (239)	0.97 (0.87–1.10)	0.93 (0.82–1.06)
Hyst & BSO <45 y	184 (161)	3600 (3300)	75 (60)	1.04 (0.83–1.31)	0.92 (0.71–1.18)
Hyst & BSO 45–49 y	172 (158)	3500 (3200)	63 (54)	0.94 (0.73–1.20)	0.95 (0.72–1.24)
Hyst & BSO 50–54 y	152 (133)	3100 (2700)	47 (41)	0.90 (0.67–1.19)	0.85 (0.62–1.15)
Hyst & BSO ≥55 y	183 (158)	3200 (2900)	105 (84)	0.99 (0.82–1.20)	0.99 (0.80–1.23)

Note: Age adjustment: Adjusting for achieved age at inclusion in HUNT. Full adjustment: Age at inclusion, level of education, BMI, waist/hip ratio, smoking, total & HDL cholesterol, systolic blood pressure, diabetes, cardiovascular disease (angina, MI or stroke), parity, menarche, and self‐reported health. Number (n), person‐years, and deaths are shown for the age‐adjusted analysis and for the fully adjusted analysis in parentheses.

We specified one main Cox proportional hazards model including all four exposure groups. Prespecified subgroup analyses stratified exposure groups by age at surgery. Secondary analyses used cardiovascular mortality as the event. We performed regression analysis with age (rather than time in study) as the timescale and death as the event. In our main analysis, women in the three exposed groups were compared to women with their uterus and ovaries intact. The subgroup analysis compared women with surgeries in different age groups to women with their uterus and ovaries intact. Analyses were performed separately using hospital‐reported and self‐reported exposure definitions. The same analyses were performed with cardiovascular death as the event in our secondary analyses. Women were considered at risk and under observation from the time they entered the HUNT study (inclusion date from HUNT2 for most women, and from HUNT3 for those who did not attend HUNT2) and were followed until death or censored on December 31, 2020. We did two analyses: one model with adjustment for age only and one fully adjusted model. In the fully adjusted model, we adjusted for age, level of education, BMI, waist/hip ratio, smoking status, total & HDL cholesterol, systolic blood pressure, diabetes, cardiovascular disease (angina, myocardial infarction, or stroke), parity, menarche, and self‐reported health. Obtained age at inclusion was included as a covariate because it strongly influences several baseline risk factors, such as BMI, cholesterol, and blood pressure, while age was used as the underlying timescale in the Cox models.

The assumption of proportional hazard was assessed graphically from log–log survival curves and Schoenfeld residual plots and was considered satisfactory.

3. RESULTS

3.1. Hospital procedure codes defining exposure

In our main analysis, we defined exposure based on hospital procedure codes. Inclusion and exclusion are summarized in Figure 1A.

The median follow‐up time was 23.8 years (min–max 0.04–25.4 years) for unexposed women, 24.2 years (min–max 0.9–25.4 years) for women with hysterectomy alone, 24.1 years (min–max 0.2–25.4 years) for those with BSO alone, and 24.1 years (min–max 0.8–25.4 years) for participants with hysterectomy & BSO. See Table 2 for details on person‐years and number of deaths.

Clinical characteristics are presented in Table 1. Despite statistically significant differences in most covariates, we found only a few of them clinically relevant. Still, we would like to point out the difference in age at inclusion, where women with hysterectomy alone on average were almost 6 years younger than the controls.

TABLE 1.

Characteristics of the participants.

	No hysterectomy/BSO	Hysterectomy alone	BSO alone	BSO and hysterectomy
	n = 32 408	n = 2273	n = 385	n = 707
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
Age at inclusion (y)	49.2 (18.0)	43.3 (12.1)	50.4 (12.8)	49.3 (12.4)
Systolic blood pressure (mm Hg)	133.8 (23.6)	128.3 (18.8)	133.4 (20.9)	133.0 (21.7)
BMI (kg/m2)	26.2 (4.7)	26.4 (4.4)	26.5 (4.4)	27.0 (4.9)
Waist/hip ratio	0.81 (0.07)	0.80 (0.07)	0.81 (0.07)	0.81 (0.07)
HDL cholesterol (mmol/L)	1.49 (0.38)	1.46 (0.36)	1.51 (0.39)	1.47 (0.42)
Total cholesterol (mmol/L)	5.86 (1.35)	5.58 (1.16)	6.02 (1.33)	5.92 (1.20)
Age at menarche (y)	13.4 (1.5)	13.1 (1.4)	13.3 (1.4)	13.2 (1.4)
Parity	2.2 (1.5)	2.4 (1.2)	2.4 (1.4)	2.3 (1.3)
Education	n (%)	n (%)	n (%)	n (%)
Primary school or less	10 910 (33.7%)	515 (22.7%)	113 (29.4%)	212 (30.0%)
1–3 years of high school	11 097 (34.2%)	1069 (47.0%)	161 (41.8%)	302 (42.7%)
Higher education <4 years	4288 (13.2%)	329 (14.5%)	54 (14.0%)	93 (13.2%)
Higher education >4 years	4449 (13.7%)	292 (12.8%)	43 (11.2%)	81 (11.5%)
Self‐reported health
Bad	555 (1.7%)	30 (1.3%)	8 (2.1%)	13 (1.8%)
Not quite good	8435 (26.0%)	557 (24.5%)	108 (28.1%)	211 (29.8%)
Good	17 624 (54.4%)	1323 (58.2%)	202 (52.5%)	384 (54.3%)
Very good	5380 (16.6%)	334 (14.7%)	62 (16.1%)	93 (13.2%)
Smoking
Never smoked	15 746 (48.6%)	939 (41.3%)	148 (38.4%)	297 (42.0%)
Previously smoked	7168 (22.1%)	578 (25.4%)	102 (26.5%)	194 (27.4%)
Current smoker	8715 (26.9%)	728 (32.0%)	131 (34.0%)	208 (29.4%)
Self‐reported diabetes	966 (3.0%)	43 (1.9%)	7 (1.8%)	23 (3.3%)
Self‐reported CVD	1986 (6.1%)	46 (2.0%)	14 (3.6%)	22 (3.1%)
Self‐reported myocardial infarction	550 (1.7%)	7 (0.3%)	0 (0.0%)	7 (1.0%)
Self‐reported stroke	612 (1.9%)	19 (0.8%)	8 (2.1%)	9 (1.3%)
Self‐reported angina	1258 (3.9%)	28 (1.2%)	7 (1.8%)	10 (1.4%)

Self‐reported use of systemic estrogen had more than 25% missing values and therefore was not included as a variable in our analyses.

3.2. Self‐reported exposure

Inclusion and exclusion based on self‐report is summarized in Figure 1B.

Median follow‐up time was 24.0 years (min–max 0.04–25.4 years) for unexposed women, 23.8 years (min–max 0.13–25.4 years) for women with hysterectomy alone, 23.6 years (min–max 0.15–25.4 years) for those with BSO alone, and 23.3 years (min–max 0.91–25.4 years) for participants with hysterectomy & BSO. See Table 3 for details on person‐years and number of deaths.

3.3. Hysterectomy and mortality

Our main analyses, based on exposure defined from hospital procedure codes, are summarized in Table 2. Hazard of all‐cause mortality was decreased by 28% in women with hysterectomy alone. Age‐adjusted and fully adjusted analyses gave approximately the same hazard ratio. In the subgroup analyses based on age at hysterectomy, all‐cause mortality was decreased in all age groups except 50–54 years at hysterectomy. Younger age at hysterectomy was associated with lower hazard ratio.

Analyses based on self‐reported exposure showed different results from those with hospital‐reported exposure and are summarized in Table 3. The results indicated no significant impact on all‐cause mortality. These estimates shifted to a statistically significant protective association with all‐cause mortality when excluding women who reported hysterectomy before 1988 (Table S1), which aligned more closely to the hospital‐reported results. In subgroup analyses, we found an increased age‐adjusted all‐cause mortality after hysterectomy performed before 45 years of age, but this was not statistically significant after full adjustment. Hysterectomy after 55 years of age was associated with decreased age‐adjusted all‐cause mortality but was not significant after full adjustment.

3.4. BSO and mortality

Hospital‐reported BSO alone was not significantly associated with mortality, but in subgroup analyses, we observed a strong association between BSO at a young age and increased all‐cause mortality. The hazard attenuated with increasing age at surgery and was no longer increased after 55 years of age (Table 2).

By contrast, self‐reported BSO seemed to be associated with increased hazard of mortality, but barely not statistically significant. BSO at ages <45 and 45–49 years had lower mortality hazard than BSO at 50–54 years, which was the only age group with a statistically significantly increased HR (Table 3). When excluding those who reported BSO before 1988, there was still not increased hazard at age <45, but a significant increased hazard in ages 45–49 and 50–54 (Table S1).

3.5. Hysterectomy & BSO and mortality

Hospital‐reported hysterectomy & BSO was not associated with all‐cause mortality. In subgroup analyses, we found a similar but less pronounced trend as with BSO alone, where younger age at surgery was associated with increased mortality. There was a neutral association from 50 years of age at surgery (Table 2).

Self‐reported hysterectomy & BSO was not associated with mortality in any of the analyses (Table 3). Differences in results compared to hospital‐reported cases were less pronounced in the group as a whole, but much more evident in the subgroup analyses.

3.6. Cardiovascular mortality (exploratory analyses)

The results of our secondary analyses on cardiovascular mortality are presented in Table 4. As for all‐cause mortality, the results varied between self‐reported and hospital‐reported exposure. Analyses of cardiovascular mortality were exploratory and limited by few events, particularly in subgroup analyses. The cardiovascular mortality HRs for women with hysterectomy was nearly identical to those observed for all‐cause mortality. For women with BSO, the results were not statistically significant, as was also seen for all‐cause mortality. Notably, the HR for cardiovascular mortality following hysterectomy & BSO was 0.73, while the HR for all‐cause mortality was 1.09. It is important to highlight that nearly all women in the combined hysterectomy & BSO group who died from cardiovascular causes had their surgery after 55 years of age.

TABLE 4.

Associations of hysterectomy (Hyst) and bilateral oophorectomy (BSO) with cardiovascular mortality.

	Hospital‐reported exposure			Self‐reported exposure
	Deaths	Age‐adjusted HR (95% CI)	Fully adjusted HR (95% CI)	Deaths	Age‐adjusted HR (95% CI)	Fully adjusted HR (95% CI)
Ovaries and uterus intact	4515 (3614)	Ref	Ref	3147 (2588)	Ref	Ref
Hysterectomy	77 (68)	0.68 (0.54–0.85)	0.69 (0.54–0.88)	332 (276)	1.10 (0.99–1.24)	1.09 (0.96–1.24)
Hyst <45 y	5 (4)	0.76 (0.31–1.83)	0.62 (0.23–1.68)	120 (99)	1.22 (1.01–1.46)	1.09 (0.89–1.34)
Hyst 45–49 y	6 (6)	0.53 (0.24–1.19)	0.55 (0.25–1.23)	83 (74)	1.05 (0.85–1.31)	1.12 (0.89–1.42)
Hyst 50–54 y	6 (6)	0.59 (0.27–1.32)	0.71 (0.32–1.59)	52 (41)	1.22 (0.93–1.61)	1.09 (0.80–1.49)
Hyst ≥55y	60 (52)	0.70 (0.54–0.90)	0.71 (0.54–0.94)	77 (62)	0.95 (0.76–1.20)	1.05 (0.82–1.35)
BSO	30 (25)	0.74 (0.51–1.06)	0.78 (0.53–1.16)	84 (69)	1.15 (0.93–1.43)	1.20 (0.94–1.52)
BSO <45 y	0	‐	‐	47 (38)	1.11 (0.83–1.48)	1.18 (0.86–1.63)
BSO 45–49 y	1 (1)	1.81 (0.25–12.87)	1.57 (0.22–11.16)	9 (9)	1.18 (0.62–2.28)	1.34 (0.69–2.58)
BSO 50–54 y	2 (2)	2.13 (0.53–8.55)	2.41 (0.60–9.67)	8 (7)	2.22 (1.11–4.44)	2.14 (1.02–4.51)
BSO ≥55 y	27 (22)	0.69 (0.47–1.01)	0.72 (0.48–1.10)	20 (15)	1.03 (0.66–1.60)	0.98 (0.59–1.62)
Hyst + BSO	47 (43)	0.71 (0.53–0.95)	0.73 (0.54–0.99)	156 (131)	1.02 (0.87–1.20)	0.99 (0.83–1.18)
Hyst + BSO <45 y	1 (1)	1.59 (0.22–11.30)	1.53 (0.21–10.91)	42 (35)	1.12 (0.83–1.52)	1.03 (0.73–1.44)
Hyst + BSO 45–49 y	1 (1)	0.54 (0.08–3.85)	0.58 (0.08–4.10)	27 (24)	0.79 (0.54–1.16)	0.87 (0.58–1.30)
Hyst + BSO 50–54 y	2 (1)	0.51 (0.13–2.04)	0.32 (0.05–2.29)	23 (21)	0.90 (0.60–1.35)	0.81 (0.53–1.24)
Hyst + BSO ≥55y	43 (40)	0.72 (0.53–0.97)	0.75 (0.55–1.03)	64 (51)	1.14 (0.89–1.46)	1.15 (0.87–1.52)

Note: Age adjustment: Adjusting for achieved age at inclusion in HUNT. Full adjustment: Age, level of education, BMI, waist/hip ratio, smoking, total & HDL cholesterol, systolic blood pressure, diabetes, cardiovascular disease (angina, MI or stroke), parity, menarche, and self‐reported health. Deaths are shown for the age‐adjusted analysis, and for the fully adjusted analysis in parentheses.

4. DISCUSSION

Our study revealed marked differences in mortality associated with hysterectomy and bilateral oophorectomy depending on how exposure was defined.

Using hospital procedure codes, hysterectomy was associated with a lower hazard of death (hazard ratio 0.72), whereas self‐reported hysterectomy showed no such association. For bilateral oophorectomy, hospital‐reported procedures performed at a young age were associated with an increased hazard of death, with attenuation of this association at older ages, while self‐reported bilateral oophorectomy demonstrated inconsistent patterns across age groups. Given the large sample size of this population‐based cohort, small differences in hazard ratios may reach statistical significance despite limited clinical relevance. Therefore, our interpretation emphasizes the direction, magnitude, and consistency of association across exposure definitions rather than statistical significance alone. In particular, hazard ratios close to unity should be interpreted cautiously, even when confidence intervals exclude 1. Importantly, excluding participants who reported hysterectomy prior to 1988 resulted in HRs that aligned more closely with hospital‐derived estimates, suggesting that differences in exposure definition and timing contributed to the observed discrepancies. We identified a temporal shift in mortality hazard among self‐reported cases occurring before and after the year 2000 (Table S2). Similarly, mortality trends shifted for hospital‐defined cases before and after year 2010. To our knowledge, this research is pioneering in evaluating the impact of differing hysterectomy definitions on mortality within the same cohort as well as examining how the timing of surgery influences this relationship. Our findings indicate that the timing of hysterectomy is an important factor, and that self‐reported data aligned with hospital records over time.

Our study further demonstrated that hospital‐reported BSO performed at a young age was associated with an increased hazard of all‐cause mortality. However, this risk gradually decreased with advancing age at the time of surgery, transitioning towards a neutral association at older ages. Specifically, the association between BSO and mortality attenuated with increasing age at surgery and was no longer increased at age 55 or older, which is in line with previous studies. In papers by Gierach et al. and Cusimano et al., ⁴ , ⁶ they show a gradually decreasing risk with older age at surgery, ending up with a neutral to protective association from 50 years of age. Also, in the revised Markov model by Rush et al. ¹⁹ from 2022, they analyzed age at surgery in one‐year increments and found 53 years to be the age at which there was no increased risk of death associated with BSO. By contrast, self‐reported BSO showed an inconsistent pattern across age groups, with the highest HRs observed at 50–54 years of age. This pattern is not biologically plausible and further supports the presence of exposure misclassification when bilateral oophorectomy is defined using self‐report.

There are both strengths and limitations to our study. A major strength is that we utilized a large population‐based cohort and additionally linked to robust national medical registries for important covariates. A unique strength is that we had access to both self‐reported and hospital‐reported gynecological surgery in the same cohort. This allowed us to shed light on why there was substantial heterogeneity between previous studies assessing mortality after hysterectomy.

An important limitation is that we were unable to adjust for hormone therapy due to a high degree of missing values. This is one of the inherent limitations of using a historical cohort for secondary analyses. Secondly, hospital procedure codes were only available from 1988, but we corrected for this by also comparing to women who self‐reported surgery after 1988. Thirdly, we did not have access to indications for surgery. To compensate for this, we excluded women with gynecological cancer before inclusion or before surgery to ensure that we estimated hazard of death after surgery on benign indication. Still, important comorbidities like BRCA mutations could potentially explain the increased mortality hazard for women who had bilateral oophorectomy at an early age, and we were unable to adjust for this. Fourthly, the minority of women who were included from HUNT3 might have some variations in measurements compared to those from HUNT2, as 10 years had passed and the research personnel were probably new. Still, this does not affect the comparison between self‐reported and hospital‐reported exposure. Finally, there is a risk of immortal time bias for women who had surgery after inclusion in HUNT. Some hypothesize that while hysterectomy may be linked to increased mortality, this is due to pre‐surgical risk factors rather than the surgery itself. To address this, we included women who had surgery at any time during follow‐up. We conducted a sensitivity analysis excluding women who had surgery after inclusion in HUNT. The HR for all‐cause mortality changed by 0.06–0.19 points across the three groups (Table S3), but our overall conclusion remained unchanged. On the contrary, for women who had surgery before inclusion, there is a risk of overadjustment due to cardiovascular risk factors acting as both confounders and mediators, particularly in subgroups with a limited number of events. Therefore, we have included the results from age‐adjusted analyses side‐by‐side with the fully adjusted analyses in the tables for easy comparison.

This study was not designed to estimate absolute mortality risks or numbers needed to harm or benefit. Our primary objective was to compare relative effect estimates derived from different exposure definitions within the same cohort to assess the potential impact of exposure misclassification on observed associations.

The marked differences in estimated associations between hospital‐reported and self‐reported hysterectomy within the same cohort suggest that exposure definition is a likely contributor to the heterogeneity observed in previous studies. Even within the same cohort, the different definitions of hysterectomy completely flipped our results. Further, we found a shift from increased to decreased hazard when comparing the decades within which the women had their surgery. This implies that indications for surgery, surgical approach, and/or who are considered eligible for surgery have changed over time. This can further explain the previous contradictions between studies.

Younger age at hysterectomy was associated with lower hazard ratios for all‐cause mortality in hospital‐reported analyses. This contrasts with other studies which found neutral to protective associations, except for those very young at surgery. ⁴ , ²⁰ Seeing how we had surgical procedure codes from 1988 and onwards, those who had surgery at young ages were also younger at inclusion (data not shown). We hypothesize that the younger age in this group causes bias that we are not fully able to adjust for. Immortal time bias may also be a small contributing factor, and we were unable to adjust for changes in use of hormone therapy over time.

In our previous validation study, ¹² we concluded that self‐reported hysterectomy was good enough for epidemiological purposes, while self‐reported BSO was not. At first glance, our findings in this study contradict that with a big shift in mortality HRs between self‐reported and hospital‐reported hysterectomy. On further scrutiny, this shift was explained by self‐report including surgeries done in earlier decades than the hospital‐reported ones, and when we excluded those with self‐reported hysterectomy before 1988, we found a similar association with mortality as the procedure‐based exposure. This is in line with our previous conclusion that self‐reported hysterectomy can be used in epidemiological studies. Taken together, these findings highlight the importance of carefully considering exposure definition and timing when interpreting associations between gynecological surgery and long‐term mortality in observational studies. Also in line with our validation study, we found different associations between BSO and mortality based on how exposure was defined, even when excluding those who reported surgery before 1988. Self‐reported BSO was illogically associated with greater hazard at ages 50–54, than before 45 years, while when defining BSO from procedure codes, we found greater risk the younger the age at surgery, which is in line with previous studies and biologically plausible.

5. CONCLUSION

Estimates of mortality following bilateral oophorectomy differed substantially depending on whether exposure was defined by self‐report or hospital procedure codes. For hysterectomy the results also varied, but when excluding those who self‐reported hysterectomy prior to 1988, the HRs closely aligned between self‐reported and hospital‐reported exposure. These findings highlight exposure misclassification as a likely contributor to the heterogeneity observed in previous studies. While early bilateral oophorectomy was associated with increased mortality when defined using hospital data, the primary contribution of this study is methodological and supports cautious interpretation of studies relying solely on self‐reported gynecological surgery.

AUTHOR CONTRIBUTIONS

Tina Ellinor Rosland: Planned and carried out the data acquisition, interpreted the data, and wrote the first draft. Nora Johansen: Planned and carried out the data acquisition, interpreted the data, and revised the article. Astrid Helene Liavaag: Planned the data acquisition, interpreted the data, and revised the article. Bjørn Olav Åsvold: Planned the data acquisition, contributed to the interpretation of the data, and revised the article. Are Hugo Pripp: Planned the data acquisition, contributed to the interpretation of the data, and revised the article. Trond Melbye Michelsen: Planned the data acquisition, interpreted the data, and revised the article. All authors revised the manuscript critically for important intellectual content, approved the final version of the manuscript, and agreed to be accountable for all aspects of the work.

FUNDING INFORMATION

South‐Eastern Norway Regional Health Authority, grant number 2021009 (TER), funded this study. The funders were not involved in the study design, data collection, analyses, or interpretation. Further, they were not involved in the writing or the decision to publish this paper.

CONFLICT OF INTEREST STATEMENT

The authors report no conflicts of interest.

ETHICS STATEMENT

Participation in HUNT2 and HUNT3 was voluntary, and each participant signed a written consent regarding participation in the study and use of their data for later research, including linking to other registries. Our study was evaluated and approved on June 2, 2021 by The Regional Committee of Health and Research Ethics (REC) of Mid Norway (Reference number 199707).

Supporting information

Figure S1. Kaplan–Meier plots.

Table S1. Associations of self‐reported hysterectomy (Hyst) and bilateral salpingo‐oophorectomy (BSO) from 1988 with all‐cause mortality.

Table S2. Associations of hysterectomy (Hyst) in different decades with all‐cause mortality.

Table S3. Associations of hospital‐reported surgery before inclusion with all‐cause mortality.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from HUNT Research Centre. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the author(s) with the permission of HUNT Research Centre.