Association between hysterectomy and hypertension among Indian middle-aged and older women: a cross-sectional study.

Hysterectomy is a common non-obstetric surgical procedure among women around the world. 1 Generally, the surgery entails removing the uterus (supracervical hysterectomy), and in certain circumstances, the uterus together with the cervix (total hysterectomy). 2 Among its uses are the treatment of abnormal bleeding, endometriosis, uterine prolapse, uterine fibroids, adenomyosis and gynaecological cancer. 3 Hysterectomy provides immediate relief from the specific medical issue, but it has been found to have a number of long-term health consequences. Positive effects of surgery include relief from excessive bleeding and pelvic pain, improved quality of life and improved mental health (such as depression and anxiety). 4 5 On the other hand, possible side effects of hysterectomy include sexual dysfunction, depression, endocrine and metabolic complications, long-term risk of cardiovascular diseases and hypertension 6 7 Recently, there have been reports of an unusual rise in the number of women undergoing hysterectomy in several regions of India. These cases involve a significant number of young and early menopausal women. A 2013 article published by British Broadcasting Corporation reported that doctors in rural India tend to prioritise surgical interventions over conservative treatments in order to take advantage of the National Health Insurance Scheme and earn more money. 8 Recently conducted National Family Health Survey (NFHS-5) found that the median age for hysterectomy was 34.6 years among women aged 15–49, which is substantially lower than the age of natural menopause. 9 Numerous cases have been reported where hysterectomy was performed on women under the age of 30, with the youngest case being just 18 years old. 10 As hysterectomy rates continue to rise, it is essential to understand the potential long-term consequences, particularly for those who undergo the procedure at a young age. Hypertension is the leading preventable risk factor of cardiovascular diseases and a significant contributor to the Global Burden of Disease (GBD). 11 Globally, hypertension prevalence is rising due to ageing populations and changes in lifestyle. Over the past two decades, the number of people with hypertension among people aged 30–79 years has doubled from 648 million in 1990 to approximately 130 million in 2019. 12 According to the GBD report 2017, hypertension was accountable for 10.2 million deaths and 208 million disability-adjusted life-years (DALYs) worldwide. 13 It has emerged as a leading cause of mortality and disability in India as well, with 21% of women and 24% of men aged 15 and above affected, as revealed by NFHS-5 (2019–2021). 9 The GBD study estimated that hypertension was associated with 1.63 million deaths and 39 million DALYs in India in 2016. 13 The above-cited statistics indicate towards the significant public health burden of hypertension worldwide, including in India. While numerous studies have examined the relationship between hysterectomy and cardiovascular disease, there has been limited research on the long-term risk of hypertension following the procedure. Only a few studies have investigated the relationship between hysterectomy and hypertension, including a cohort study of Taiwanese women aged 30–49 that found hysterectomised women had a significantly higher risk of hypertension during the follow-up period. 14 Another cohort study conducted in Minnesota reported that hysterectomy with ovarian conservation was associated with an increased long-term risk of cardiovascular and metabolic conditions, including hypertension, particularly in women who underwent hysterectomy at less than 35 years of age. 15 Similar results were found in a cross-sectional study on Indian women aged 15–49, 16 and a study conducted on the French E3N cohort. 17 However, the underlying mechanism of the association is still not well established. 11 On conducting a thorough literature review, it was observed that the majority of studies investigating the association between hysterectomy and hypertension have focused on women of reproductive age. However, this association has been underexplored among elderly women. Therefore, there is a crucial need to investigate the long-term association between hysterectomy and hypertension, which would have wider implications and could greatly contribute to existing literature on this topic. In order to address this research gap, this study aims to investigate the association between hysterectomy and hypertension in middle-aged and older women in India from an ageing perspective. In addition, the study will also examine whether this association varies between different age groups.

Data from the Longitudinal Ageing Study in India (LASI), conducted in 2017–2018, were used in the study. LASI is a nationally representative survey of adults aged 45 years and above conducted by the International Institute for Population Science in collaboration with the Harvard T. H. Chan School of Public Health, the University of Southern California and the University of Washington. The survey provides reliable information on the social, economic, physical, psychological and cognitive health of older adults in India. A multistage stratified area probability cluster sampling design was used in the survey to collect data from adults aged 45 years and above and their spouses (irrespective of their age) across all the states and union territories of India. A total of 73 396 individuals were interviewed in the survey from 43 584 households. The overall non-response rate in the survey was 12.7% ranging from 25.7% in Chandigarh to 3.7% in Nagaland. Prior consent was taken from all the respondents. The Indian Council of Medical Research provided ethics approval and required guidelines to conduct the survey. LASI is envisaged to collect longitudinal data from the selected individuals every 2 years. A detailed description of the sampling design, survey implementation, instruments used and data collection procedures are published elsewhere. 18 19 This study specifically focuses on the relationship between hysterectomy and hypertension, and therefore, only women were included in the analysis. After excluding males and women aged less than 45 years, the sample size was reduced to 35 567. After removing missing values for hysterectomy, hypertension and body mass index (BMI), the final sample consisted of 32 460 women. The flow chart of the sample selection for the study is presented in figure 1 . Flow chart of the study sample selection. LASI, Longitudinal Ageing Study in India. Patients and/or the public were not involved in the study in any way. The study’s primary outcome variable was hypertension. The hypertension variable was based on the question: Have you ever been diagnosed with hypertension? Since hypertension is a chronic disease that can be controlled with medication but cannot be cured, those who reported ever being diagnosed with hypertension were considered hypertensive in the study. In LASI, women were asked: Have you undergone an operation to remove your uterus (hysterectomy)? Based on this question, a binary variable was constructed as 1 ‘undergone hysterectomy’ and 0 ‘never undergone hysterectomy’. The relation between the outcome variable and primary predictor variable was controlled for a range of background variables which are: age (45–49, 50–59, 60–69, 70–79, 80+), place of residence (urban, rural), education level (no education, primary of less than primary, middle, secondary and higher), marital status (currently married, widowed, other), religion (Hindu, Muslim, other), caste/tribe (scheduled caste, scheduled tribe, other backward class (OBC), none of the above), monthly per capita expenditure (MPCE) quantile (poorest, poorer, middle, richer, richest), geographical regions (north, east, central, west, north-east, south), family history of hypertension (no, yes), frequent physical exercise (no, yes) and BMI. Bivariate analysis was performed to assess the crude prevalence of hysterectomy and hypertension in various socioeconomic and demographic categories. The prevalence of hypertension in various socioeconomic and demographic groups was estimated by hysterectomy status. Along with that, prevalence ratios (ratio of prevalence in hysterectomy and non-hysterectomy group) were also calculated. Logistic regression models with entropy balancing weights were employed to assess the adjusted relationship between hysterectomy and hypertension. The concept, need and use of the entropy balancing approach is discussed below. Selection bias is a commonly encountered problem in observational studies, where the treatment allocation is not random, leading to systematic differences between the treatment and control groups. Such differences reduce the reliability and validity of study results. 20 To elaborate, women with higher socioeconomic status are more likely to have access to better healthcare facilities and resources, including access to doctors and specialists who may recommend a hysterectomy as a treatment option. In addition, they may have greater awareness about the benefits and risks of hysterectomy and may be more likely to choose this option. On the other hand, women with lower socioeconomic status may face economic barriers that limit their ability to access healthcare services, which may lead to a lower likelihood of undergoing a hysterectomy. Hence, the likelihood of receiving the treatment (hysterectomy) is not distributed equally among the entire study population. Hence, there is a potential risk of over-representation of certain sections of the population in the treatment group. This could make it challenging to draw accurate conclusions about the true association between hysterectomy and hypertension in the general population, as the treatment group may not be representative of the larger population. Therefore, it is important to take measures to address this potential source of bias in our study to ensure the validity and reliability of our findings. Matching techniques (such as propensity score matching (PSM)) are frequently used by researchers to address the issue of selection bias. However, this study uses a novel methodology called ‘entropy balancing’ to balance the treatment and control groups and avoid the potential risk of confounding and selection bias. To stochastically balance the covariates, PSM requires fairly large samples and correct model specification. Because of this complex and tedious search process, many studies face the problem of low balance between the treatment and control groups. According to Hainmueller, sometimes, in the matching process, improving balance on some covariates leads to decreases in balance on other covariates, and that results in ineffective bias reduction. 21 In this study, we used Hainmueller’s ‘entropy balance’ method to achieve covariate balance. 21 Entropy balancing is also a data preprocessing technique used to achieve covariate balance in observational studies with binary treatments. This covariate balance between the treatment and control group reduces model dependence for the estimation of treatment effects using regression or any other statistical technique. In contrast to matching techniques (such as PSM), entropy balancing involves using weight functions applied to the sample units to achieve covariate balance. In this method, the researcher specifies the covariates and desired moments (mean, variance, skewness) to be balanced in the treatment and control groups. After that, the entropy balancing method searches for a set of weights that satisfies the desired balancing conditions while remaining as close (in terms of entropy) to the base weights as possible to retain maximum information. Weights are then applied to sample units in order to achieve covariate balance between the two groups without losing information. These weights can be applied to any subsequent analyses (such as regression) to estimate treatment effects. 21 In general, we are interested in population average treatment effect (PATT), which is estimated as follows: P A T T = E [ Y ( 1 ) | D = 1 ] − E [ Y ( 0 ) | D = 1 ] Where D represents the treatment status (D=1 indicates treatment and D=0 indicates control). [Y(1)|D=1] denotes the expected outcome among those who received the treatment, and E[Y(0)|D=1] denotes the expected outcome among the treated if they would not have received the treatment. The first term can be easily estimated from the treatment group data. The second component, E[Y (0)|D=1], is counterfactual and cannot be directly estimated from the observed data. In experimental studies where treatment assignment is random, we can simply use E[Y (0)|D=0] (expected outcome among untreated) as the estimate of E[Y(0)|D=1]. However, because of selection bias in observational studies, E[Y (0)|D=0] cannot be used as an estimate of E[Y (0)|D=1]. In the entropy balancing method, we reweight the control group to balance the covariate moments between the treatment and control groups. Then we can calculate PATT by estimating the difference in mean outcomes between the treatment group and the reweighted control group. The counterfactual mean can be estimated as: E [ Y ( 0 ) | D ^ = 1 ] = ∑ { i | D = 0 } Y i w i ∑ { i | D = 0 w i Where w i are the entropy balancing weights calculated for every control unit. The weights are calculated using the following method that minimises the entropy balance metric: m i n W i H ( w ) = ∑ { i | D = 0 } w i l o g ( w i / q i ) subject to balance and normalising constraints ∑ { i | D = 0 } w i c r i ( X i ) = m r w i t h r ∈ 1 , … , R a n d ∑ { i | D = 0 } w i = 1 a n d wi ≥ 0 for all i such that D=0 where X i denotes exogenous pretreatment characteristics of the units, qi=1/n 0 are the base weights and c ri (X i ) = m r denotes a set of R balance constraints imposed on the covariate moments of the reweighted control group. Detailed information about the entropy balance methodology can be accessed from the original paper written by Hainmueller. 21 We calculated entropy balancing weights in STATA V.16 using the ‘ebalance’ module. 22 The weights were calculated to balance the treatment and control groups in terms of the control variables’ first three moments (mean, variance and skewness). In this study, the treatment group consists of women with hysterectomy and the control group consists of women without hysterectomy. The weights were then applied to the logistic regression analyses to estimate the adjusted relation between hysterectomy and hypertension. We also performed age-group-wise subgroup analyses to examine whether the relation between hysterectomy and hypertension varies by age. For this objective, the weights were calculated in every subgroup, and weighted logistic regression was employed in each age group. Simple unweighted logistic regression analysis was performed as part of the sensitivity analysis. We employed three logistic regression models to examine the proposed relationship. The first model was a bivariate model, including ‘hysterectomy status’ as the only independent variable. The second model adjusted the association between hysterectomy and hypertension for the background characteristics. The interaction term of hysterectomy and age was included in the third model. The results of sensitivity analyses are presented in online supplemental file 1 . Using postestimation of the interaction term in model 3, age-group-wise predicted probabilities of hypertension in the treatment and control groups were estimated. The predicted probabilities were then plotted using a line graph to visualise the interaction between hysterectomy and age.

Table 1 presents the background characteristics of the sample population, including the prevalence of hysterectomy and hypertension by hysterectomy status, as well as the prevalence ratios (PRs). The majority of participants were in the 50–59 age group (33%), lived in rural areas (65%), had no formal education (61%), were married at the time of the survey (64%), identified as Hindu (73%), belonged to the OBC caste (39%) and resided in the southern region of India (24%). Background characteristics of the sample population, prevalence of hysterectomy, and prevalence of hypertension among middle-aged and older women in India (n=32 460) Prevalence ratio: ratio of prevalence of hypertension in the treatment and control groups. BMI, body mass index; MPCE, monthly per capita expenditure. The prevalence of hysterectomy was lower in older age groups than in younger age groups. Muslim women had a lower prevalence (6.9%) of hysterectomy compared with their counterparts. The prevalence of hysterectomy was observed to be increasing with an increase in MPCE quantile. The overall prevalence of hysterectomy was 10.5% in the sample population. The prevalence of hypertension has been calculated separately for the treatment and control groups by background characteristics. Among all the categories of the background variables, the prevalence of hypertension was higher in the treatment group compared with the control group. For instance, in the age group 80+, the prevalence of hypertension was 52% (PR=1.52) higher in the treatment group compared with the control group. Overall, the prevalence of hypertension was 31.3% in the control group and 42.5% in the treatment group (PR=1.36). Table 2 illustrates the balance of the first three moments (mean, variance, skewness) of the selected covariates in the treatment and control groups before and after entropy weighting. Before weighting, the mean age in the treatment groups was 58.2 years, with a variance of 87.0 and a skewness of 0.66. On the other hand, in the control group, the mean age was 59 years, with a variance of 109.7 and a skewness of 0.68. In the control group, BMI was 23.1 kg/m 2 , while in the treatment group, it was 24.8 kg/m 2 . It can be observed from the table that following entropy weighting, the moments of the covariates in the treatment and control groups were perfectly balanced. Balance of the first three moments of the selected control variables in the treatment and control groups before and after entropy weighting Prevalence ratio: ratio of prevalence of hypertension in the treatment and control groups. BMI, body mass index; MPCE, monthly per capita expenditure. The results of entropy-weighted logistic regression and subgroup analysis are presented in table 3 . The results indicate that compared with women without hysterectomy, women with hysterectomy had 36% (OR 1.36; 95% CI 1.26 to 1.48) higher odds of hypertension. Age-group-wise analysis indicates that the magnitude (OR) of the association between hysterectomy and hypertension increases with age. In the age group 45–49, the odds of hypertension were 23% (OR 1.23; 95% CI 1.03 to 1.47) higher in women with hysterectomy compared with women without hysterectomy. This OR value increased to 1.87 (95% CI 1.18 to 2.97) in the 80+ age group. Results of entropy-weighted logistic regression analysis assessing the odds of hypertension by hysterectomy status 95%CI of OR in brackets. ***p-value<0.001, **p-value<0.01, *p-value<0.05. Predicted probabilities derived from interaction analysis (see online supplemental tables S1 and S2 ) are shown in figure 2 . In the control group, the predicted probability of hypertension was 0.20 in the age group 45–49, which increased to 0.44 in the age group 80+. In the treatment group, the predicted probability of hypertension increased to 0.61 in the age group 80+ from 0.23 in the age group 45–49. It can be observed from figure 2 that initially (in the age group 45–49), there was a modest difference (0.03 units) between the predicted probability of hypertension in the treatment and control group, but the difference increased significantly with age (0.16 units in the 80+ age group). The figure suggests that the magnitude of the association between hypertension and age varies depending on hysterectomy status. Predicted probabilities of hypertension by hysterectomy status in different age groups. Predicted probabilities are calculated using postestimation of the interaction term in the logistic regression analysis ( online supplemental tables S1 and S2 ).

Using nationally representative population-based data, this study found that hysterectomy is significantly associated with hypertension among middle-aged and older women in India. Notably, the study identified that the association between hysterectomy and hypertension was stronger in higher age groups. We applied the entropy balancing method to achieve covariate balance in the treatment and control groups. The methodology efficiently balanced the treatment and control groups up to the initial three moments of the specified covariates. Previous research on the relationship between hysterectomy and hypertension produced inconsistent results; some studies found a positive association, 14 16 17 23 while others found no relationship. 6 24 All of the above-cited studies were mere epidemiological investigations; no study has looked at the underlying biological mechanism to explain the potential association between hysterectomy and hypertension. Therefore, it is still unclear whether hysterectomy has a causal effect on hypertension or whether the observed association is due to confounding factors. For instance, BMI has been found to be associated with hysterectomy 25 as well as hypertension 26 ; hence, the association between hysterectomy and hypertension might be because of the confounding effect of BMI. However, in this study, the association was controlled for BMI in both the main and sensitivity analyses, discarding possible confounding. Hysterectomy results in the immediate termination of the reproductive function, leading to a decline in the production of sex hormones such as oestrogen and progesterone. Prior studies have suggested that sex hormones play a critical role in blood pressure regulation. 27 Oestradiol, an oestrogen steroid hormone, has a blood pressure-lowering effect in women. 27 Oestradiol levels in the body rapidly drop with the cessation of reproductive function, which may lead to an increase in blood pressure. Progesterone also has a similar depressive effect on blood pressure, and its decreased levels may lead to increased blood pressure. 27 Sex hormones also significantly affect body fat distribution, abiogenesis and adipocyte metabolism. 28 According to earlier studies, postmenopausal women who experience sex hormone decline may gain weight and develop metabolic syndrome, two of the most important risk factors for hypertension. 29 Furthermore, a reduction in the oestrogen to androgen ratio after menopause attenuates the vasorelaxant effects of oestrogens on vessel walls and stimulates vasoconstrictive effects. 30 This promotes vascular stiffness in arteries and the narrowing of blood vessels. 29 30 The mechanisms mentioned above may be some of the possible explanations for the relationship between hysterectomy and hypertension. Subgroup analysis revealed that the association between hysterectomy and hypertension was higher in higher age groups. Similar results were found in the interaction analysis. These findings pose relevant questions for future research. Reduction in sex hormone levels is given as the most common explanation for the relationship between hysterectomy and hypertension. However, it should be noted that the majority of participants in this study were menopausal women, regardless of their hysterectomy status. As a result, in higher age groups, the distribution of sex hormones is expected to be similar among all participants (both the treatment and control groups). Therefore, the above-mentioned hypothesis (reduction of oestrogen/progesterone) does not offer a sound explanation for the observed relationship in higher age groups. This leads to the question of whether women who undergo hysterectomy have different hormonal configurations than those who achieve menopause naturally. One possibility is that hysterectomised women have a longer exposure time to lower levels of sex hormones than women who achieve menopause naturally. As a result, the hysterectomy group may have a higher prevalence of hypertension due to a longer exposure time to low sex hormone levels. These findings warrant further investigation to better understand the biological mechanisms underlying the association between hysterectomy and hypertension. The popularity of hysterectomy among women in India has been on the rise in recent years. The latest NFHS-5 (2019–21) reports that 9.7% of women aged 40–49 years and 3.3% of women aged 30–39 years have undergone hysterectomy. 9 This high prevalence has generated concerns among policy-makers and experts in India. The findings of this study add to these concerns and highlight the need for further research to establish the effect of early hysterectomy on later-life health outcomes, particularly cardiovascular diseases and hypertension. Therefore, it is crucial for policy-makers to comprehend the relationship between hysterectomy and hypertension to form effective policies. Future research could adopt more sophisticated study designs such as longitudinal studies, cohort studies and clinical trials to establish robust evidence for this association. Despite the robust statistical methods employed, the study has some limitations that merit consideration. First, the use of cross-sectional data precludes establishing causal inferences. Second, some potential covariates such as glucose levels and age at hysterectomy were not controlled for, owing to data limitations. Moreover, self-reported hypertension may introduce measurement bias. Nonetheless, the study has several strengths. It is based on a large, nationally representative sample, which enhances the reliability of the estimates. The use of the entropy balancing approach enabled a perfect balance of covariates in the treatment and control groups, allowing for a robust estimation of the proposed relationship. Future research using more sophisticated study designs, such as longitudinal studies, cohort studies and clinical trials, could provide more robust evidence on the relationship between hysterectomy and hypertension.