Association between hysterectomy and dementia risk in Korean women aged 40–59: a nationwide retrospective cohort study
Article information
Abstract
Objective
To determine whether hysterectomy for benign diseases is associated with the risk of dementia in middle-aged women.
Methods
We conducted a retrospective cohort study using data from the Korean National Health Insurance Service database (2002–2020). Women aged 40–59 years who underwent hysterectomy for benign indications (n=16,818) were propensity score-matched (1:1) to controls who had not received hysterectomy. Subjects were followed up until the diagnosis of dementia, death, or the end of the study period (2020). Dementia (all types), Alzheimer’s disease (AD), and vascular dementia (VaD) were identified by International Classification of Diseases, 10th revision codes. Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for dementia associated with hysterectomy.
Results
Median follow-up was 11.4 years. Dementia occurred in 302 (1.8%) women who did not undergo hysterectomy and 257 (1.5%) women who underwent hysterectomy (P=0.061). Cox analysis revealed that hysterectomy was not significantly associated with all-cause dementia (HR, 0.865; 95% CI, 0.724–1.033), with a non-significant trend towards reduced risk. Subgroup analysis also failed to identify any significant association; AD (HR, 0.696; 95% CI, 0.463–1.048) and VaD (HR, 0.625; 95% CI, 0.284–1.377) were not significantly associated with an increased hazard.
Conclusion
In this large Korean cohort, hysterectomy for benign diseases in women aged 40–59 years was not associated with a significant change in the subsequent risk of dementia. Collectively, our results indicate that hysterectomy was not associated with an increased incidence of dementia, including in analyses stratified by adnexal surgery.
Introduction
Dementia is a major and growing public health concern in aging populations worldwide. Alzheimer’s disease (AD) and vascular dementia (VaD) are the most common subtypes [1–4]. As life expectancy increases, the prevalence of dementia is projected to rise dramatically, particularly among women, who account for nearly two-thirds of all global cases [5]. This disparity has drawn attention to the potential influence of hormones and reproductive factors on brain aging. The impact of hysterectomy, with or without oophorectomy, on long-term cognitive outcomes has been debated for decades. Because surgical removal of the reproductive organs induces abrupt changes in estrogen exposure, hysterectomy has been hypothesized to accelerate cognitive decline or increase the risk dementia, especially when performed before natural menopause [6,7].
Experimental evidence indicates that estrogen supports synaptic plasticity, maintains cerebral perfusion, and reduces the accumulation of beta-amyloid, exerting both neuroprotective and vasoprotective effects [6,8]. Therefore, early or surgical menopause may diminish these protective effects and predispose individuals to dementia. In the present study, we were focused specifically on the use of hysterectomy to treat benign gynecological disease, particularly in terms of whether this type of surgery was associated with a subsequent risk of dementia in middle-aged women, rather than directly measuring estrogen exposure. Animal studies further suggest that hysterectomy, even with ovarian conservation, may alter cognitive processes via the non-hormonal signaling pathways linking the uterus to the brain [9,10]. These hormonal and non-hormonal mechanisms provide biological plausibility for the potential association between hysterectomy and dementia, the central focus of the present study.
Currently, there is mixed epidemiological evidence for an association between hysterectomy and dementia. In Denmark, a nationwide cohort study reported a significantly increased risk of early onset dementia, defined as an onset prior to 50 years-of-age, in women who underwent hysterectomy, particularly when accompanied by unilateral or bilateral oophorectomy [7]. In the USA, a tertiary center cohort study reported that surgical menopause involving oophorectomy was associated with a higher risk of dementia [11]. More recently, large-scale investigations have reported weak or no associations. In Finland, a case-control study found evidence for an increased risk of AD in women who had undergone benign hysterectomy [12]. Furthermore, an international pooled analysis of five cohorts revealed that after menopause was taken into account, hysterectomy itself did not independently influence the risk of dementia [13]. Similarly, analyses in the United Kingdom using the UK Biobank cohort found no direct association after adjusting for menopausal timing and comorbidities [14].
Despite these prior findings, robust evidence from Asian populations is scarce and few studies have used carefully matched comparison groups, although cultural, genetic, and clinical factors may influence the patterns of hysterectomy and risk of dementia. Therefore, the primary objective of this study was to determine whether hysterectomy for benign gynecological disease was associated with an increased risk of dementia in Korean women aged 40–59 years using a large nationwide retrospective cohort with propensity score-matched controls. By clarifying this association, our study aimed to provide clinicians and patients with the evidence needed for informed decision-making with regards to conducting hysterectomy in mid-life and its potential implications for long-term cognitive health.
Materials and methods
1. Database
In this study, we utilized data obtained from the National Health Insurance Corporation (NHIS) database in South Korea which provides comprehensive information relating to health insurance claims, including diagnostic codes, prescription records, surgical procedure data, sex, age, type of insurance coverage, and details of hospital admissions and outpatient visits [15]. The NHIS database provides comprehensive, population-based information relating to health insurance claims-including diagnostic codes, prescription records, and surgical procedures-and has been widely validated for epidemiological research [15,16]. For the present retrospective cohort study, population-based data were extracted from the NHIS database covering the period from January 1, 2002, to December 31, 2020.
2. Subject selection
Participant inclusion was determined by specific criteria validated through the Korea Health Insurance Medical Care Expenses (2016, 2020 editions) and the International Classification of Diseases, 10th revision. The target population comprised women aged 40–59 years who underwent hysterectomy for benign conditions between January 1, 2002 and December 1, 2011. These patients constituted the hysterectomy group. The comparison group included women in the same age range who underwent a general health examination during the same period, but did not undergo hysterectomy.
To match the processing capacity of the NHIS server, a random sampling approach was applied to include 20% of the eligible women. The exclusion criteria included hysterectomies or health examinations performed in 2002, which served as the washout period. Women who underwent a hysterectomy after the initial medical examination were excluded from the non-hysterectomy group. Participants with a history of cancer or dementia within 365 days of enrollment were also excluded. The observation period for each participant lasted until December 31, 2020.
Data relating to a broad range of variables were collected, including age, hyperlipidemia, diabetes mellitus, hypertension, physical activity, smoking status, alcohol consumption, socioeconomic status (SES), age at menarche and menopause, residential area, body mass index (BMI), parity, Charlson comorbidity index (CCI), prior adnexal surgery, prior menopausal hormone therapy, uterine fibroids, and endometriosis. One-to-one propensity score matching (PSM) was performed to balance covariates between the hysterectomy and control groups.
3. Outcomes
Participants were defined as having dementia if they had at least four medical claims containing diagnostic codes for AD (F00, G30), VaD (F01), or total dementia (F00, F01, F02, F03, G30, and G31).
4. Variables
In this study, we investigated a range of demographic, behavioral, and clinical variables. Age was categorized into 5-year intervals. Menarche was classified as occurring before or after 13 years-of-age, and parity was grouped as ‘no response or 0’, ‘1’, ‘2’, or ‘3 or more’. Menopausal status was determined by self-reported questionnaires, and menopausal hormone therapy was defined as lasting more than 6 months before cohort entry. SES was considered low for women eligible for national medical aid or basic insurance. Self-reported variables included smoking, alcohol consumption, and physical activity. BMI was calculated using Asia-Pacific classification criteria [17]. Residential areas were categorized as urban or non-urban.
Surgical history, including hysterectomy and adnexal procedures, was identified by procedural codes. The presence of hyperlipidemia (E78), diabetes (E10–E14), hypertension (I10–I15), uterine fibroids (D25), and endometriosis (N80) was determined when patients had at least two medical visits for the relevant condition before study entry. The CCI was computed based on diagnosis codes recorded within 1 year before enrollmen [18].
5. Statistics
Statistical analyses were performed using R version 3.5.1 (The R Foundation for Statistical Computing, Vienna, Austria) and statistical significance was set at a two-sided P-value <0.05. Following PSM, categorical variables were compared using the Cochran-Mantel-Haenszel test, whereas continuous variables were analyzed using paired t-tests and Wilcoxon signed-rank tests. Standardized mean differences (SMDs) were used to evaluate the balance between matched covariates. SMDs ≤0.2 were considered to indicate a good balance. The normality of continuous variables was examined using the Anderson-Darling test.
Stratified Cox regression analyses were performed to estimate the risk of dementia associated with hysterectomy. Because PSM balanced all measured baseline covariates between groups, hazard ratios (HRs) from the stratified Cox models are presented without additional covariate adjustment, as further adjustment was neither necessary nor recommended after successful matching. In the hysterectomy group, the index date was defined as the date of surgery. The comparison group was defined as the first day of health examination. The censoring date corresponded to the earliest occurrence of a confirmed dementia diagnosis, death, or last recorded medical service use. Missing data in the Cox model were managed by listwise deletion. A separate stratified Cox regression analysis was conducted to compare women who underwent laparoscopic hysterectomy with those who did not.
6. Ethics
All raw data obtained from the NHIS were fully de-identified prior to analysis to ensure that identifiable personal information was removed. Analyses were conducted exclusively on a secure NHIS server with restricted access. Only summary outputs, such as aggregated statistics, figures, and tables, were permitted for export, thus maintaining the complete confidentiality of participants. In accordance with the South Korean Bioethics and Safety Act, the requirement for individual informed consent was waived because of the use of anonymized administrative data.
Adhering to NHIS privacy protocols, only the final research results were allowed to leave the closed server, thus ensuring the ongoing protection of sensitive information. Ethical approval for this study was granted by the Institutional Review Board of Sanggye Paik Hospital (approval number SGPAIK 2021-12-005).
Results
Our analysis included 4,580,240 women who underwent a health checkup or hysterectomy using the NHIS between 2002 and 2011. PSM identified 16,818 women in the hysterectomy group and 16,818 women in the non-hysterectomy group (Fig. 1). The median age of all participants was 47 years (interquartile range, 45–50), and the median follow-up period was 11.4 years (interquartile range, 10.0–13.3). The characteristics of the study population are summarized in Table 1. There were no missing data for baseline covariates, as presented in Table 1. Following PSM, all SMDs for baseline covariates were <0.12 (maximum of 0.111 with most values <0.05), thus indicating that there was an excellent balance between the two groups (Table 1).
Selection of participants to investigate the risk of dementia for women in the hysterectomy vs. non-hysterectomy groups using Korean National Health Insurance Data 2002–2020 (after propensity score matching).
Baseline characteristics of the study subjects with and without hysterectomy according to Korea National Health Insurance Data 2002–2020 (after propensity score matching)
A total of 302 (1.8%) women in the non-hysterectomy group and 257 (1.5%) in the hysterectomy group were diagnosed with dementia; this difference was not statistically significant (P=0.061). Data relating to the incidence of AD and VaD according to hysterectomy status are summarized in Table 2.
Incidence of dementia in women with and without hysterectomy: an analysis of Korea National Health Insurance Data 2002–2020 (after propensity score matching)
Supplementary Table 1 shows the incidence of dementia per 100,000 person-years in women who underwent hysterectomy and those who did not. The mean incidence rates were 153 per 100,000 person-years in the non-hysterectomy group and 131 per 100,000 person-years in the non-hysterectomy and hysterectomy groups, respectively. Fig. 2 depicts Kaplan-Meier curves for both groups; the stratified log-rank test yielded a P-value of 0.08.
Kaplan-Meier analysis of the incidence of dementia for women in the hysterectomy vs. non-hysterectomy groups: insights from Korea National Health Insurance Data 2002–2020 (after propensity score matching). Stratified log-rank test P-value=0.08. HTT, hysterectomy.
In the Cox proportional hazards model, there was no significant association between hysterectomy and dementia (HR, 0.865; 95% confidence interval [CI], 0.724–1.033) (Table 3). Subgroup analysis, according to dementia subtype, showed that neither AD (HR, 0.696; 95% CI, 0.463–1.048) nor VaD (HR, 0.625; 95% CI, 0.284–1.377) was significantly associated with hysterectomy. A similar trend was observed when analyses were restricted to hysterectomies with or without adnexal surgery, thus indicating no significant association with any form of dementia. There was a non-significant tendency towards a lower risk of AD and VaD for women who had undergone hysterectomy.
The risk of dementia following hysterectomy according to Korean Health Insurance Data 2002–2020 (after propensity score matching)
Table 4 summarizes the results of the subgroup analysis stratified by age. We investigated whether this risk differed between women who underwent surgery in their 40s (40–49 years-of-age) and those in their 50s (50–59 years-of-age). Hysterectomy was not associated with an increased risk of all-cause dementia in the younger (HR, 0.982; 95% CI, 0.751–1.283) or older age groups (HR, 0.79; 95% CI, 0.599–1.041). This lack of an association was consistent for both AD and VaD across the two age groups. These findings indicate that hysterectomy for benign indications did not significantly alter the subsequent risk of dementia, regardless of this procedure was performed during the premenopausal or perimenopausal period.
Age-stratified comparison of the risk of dementia for women with and without hysterectomy according to Korea National Health Insurance Data 2002–2020 (after propensity score matching)
To further assess the robustness of these findings, we performed sensitivity analysis comparing laparoscopic hysterectomy with no hysterectomy; this further demonstrated that laparoscopic hysterectomy was not associated with the risk of dementia (HR, 0.959; 95% CI, 0.691–1.331) (Supplementary Table 2). This result was consistent with our main analysis.
Discussion
In this large population-based cohort of Korean women aged 40–59 years-of-age, hysterectomy for benign gynecological diseases was not associated with an increased risk of subsequent dementia. This null association was consistent across dementia subtypes and in both subgroup and sensitivity analyses. Although Kaplan-Meier curves visually suggested a modest separation favoring the hysterectomy group, this difference was not statistically significant (P=0.08). We interpret this pattern cautiously because it likely reflects limited event accumulation and statistical power, given the relatively young age of the cohort and the low absolute incidence of dementia, rather than a true protective effect.
Our results are consistent with those reported by recent large prospective studies, indicating that hysterectomy itself does not materially influence the risk of dementia when ovarian function is preserved or when the timing of menopause is considered. Across international cohorts, pooled analyses reported that the type of menopause (natural vs. surgical) was not an independent predictor once the age at menopause was modeled [13]. In the United Kingdom, analyses of the UK Biobank cohort revealed that early menopause was the key determinant of a risk of dementia, and not hysterectomy or oophorectomy [14]. These findings suggest that a shorter lifetime exposure to estrogen, rather than hysterectomy, underlies the association between reproductive surgery and cognitive aging.
In contrast, studies reporting an increased risk of dementia typically involved younger women or concurrent oophorectomies. In Denmark, registry-based research found that when performed before the age of 50 years, hysterectomy substantially elevated the risk of early onset dementia, particularly when accompanied by bilateral oophorectomy [7]. In the USA, a cohort from the Mayo Clinic revealed stepwise increases in the risk of dementia with more extensive gynecological surgery [11]. These results align with the hypothesis that the abrupt loss of ovarian hormones during early adulthood disrupts estrogen-mediated neuroprotection. However, our present cohort primarily included perimenopausal women undergoing hysterectomy for benign indications, and most retained their ovaries; this may explain the absence of an elevated risk.
From a biological perspective, estrogen deficiency may accelerate neurodegenerative changes by reducing cerebral blood flow, impairing synaptic signaling, and increasing oxidative stress [6,8]. The critical window hypothesis posits that estrogen supplementation is most protective when initiated close to the onset of menopause, whereas late initiation may be ineffective [19,20]. As most women in our study underwent hysterectomy close to the natural menopausal transition, the consequential reduction in endogenous estrogen exposure may have been too small to influence the onset of dementia during the follow-up period. Although the HR for all-cause dementia was numerically higher in women who underwent hysterectomy with adnexal surgery (HR, 1.058) compared to those without (HR, 0.831), this difference was not statistically significant. This contrasting trend might theoretically align with the hypothesis that ovarian loss aggravates the risk of dementia; however, given the smaller sample size and fewer events in the adnexal surgery subgroup, this finding likely reflects limited statistical power (type II error) rather than a true aggravating effect.
Our current findings have important clinical implications. Concerns relating to the risk of dementia sometimes influence surgical decision-making for benign gynecological conditions. The present study provides reassurance that among women in midlife, hysterectomy without premature ovarian failure does not significantly alter cognitive outcomes. In addition, menopausal hormone therapy after surgery, when clinically appropriate, could potentially help mitigate hormonal effects on the brain, as suggested by prior studies.
This study has several strengths, including a large nationally representative sample, a long follow-up period, and robust PSM, which minimized confounding by comorbidities and lifestyle factors. The use of comprehensive NHIS data allowed for the accurate identification of hysterectomies, oophorectomies, and dementia diagnoses. However, this study had several limitations that need to be considered. The diagnosis of dementia was based on administrative claims rather than direct neurocognitive assessments, potentially underestimating mild cases. Information relating to education, diet, and physical activity was unavailable, and residual/unmeasured confounding factors (e.g., cognitive reserve, family history, genetic risk, and depression) could not be ruled out. Finally, given the age range of the participants (40–59 years at baseline), longer follow-up into older age may be needed to capture late-onset dementia. Because most dementia cases in Korea and other Asian populations are diagnosed after the age of 70 years, our cohort predominantly captured early- and early late-onset dementia rather than the full lifetime burden.
In conclusion, this nationwide cohort study found no evidence that hysterectomy for benign diseases increased the risk of dementia in middle-aged Korean women. Future prospective studies, incorporating cognitive testing and hormonal biomarkers, as well as longer follow-up periods, are now warranted to confirm these results and elucidate the underlying mechanisms.
Supplementary Information
Notes
Conflict of interest
The authors have declared they have no conflicts of interest.
Ethics approval
According to the National Health Insurance Service (NHIS)'s privacy policy, raw data cannot be provided.
Patient consent
Patient consent was waived by the Institutional Review Board due to the retrospective nature of this study and the use of de-identified population-based data.
Funding information
This study was supported by a research grant from the Research Year of Inje University in 2025.