Association of Age of Menopause and Glaucoma Diagnosis in Female Veterans

Abstract

Purpose: Age of menopause has been associated with the risk of developing glaucoma; however, it is unclear if the onset of menopause is directly associated with the development of glaucoma. Our objective was to determine if there is an association between the age at diagnosis of menopause and glaucoma.

Methods: This retrospective, case-only analysis was performed using the Veterans Affairs (VA) Corporate Data Warehouse of female veterans from 2000 to 2019. Women with both menopause and glaucoma diagnoses were matched based on covariates. The two matched cohorts were early menopause–early comparative (EM-EC; n = 1075) and late menopause–late comparative (LM-LC; n = 1050) women. We used a Pearson correlation to examine the linear relationship between age at diagnosis of menopause and glaucoma. Afterward, we used a multivariate linear regression model with age at diagnosis of glaucoma serving as the outcome variable to account for the covariates.

Results: We found that EM women developed glaucoma 6.0 years (interquartile range [IQR], 5.1–6.5) earlier than the EC group (P < 0.001), and LM women developed glaucoma an average of 5.2 years (IQR, 4.8–5.7) later than the LC group (P < 0.001). There was a modest linear relationship between the age of menopause and glaucoma diagnoses in the EM-EC (r = 0.40) and LM-LC (r = 0.46) cohorts. In our multivariate analysis, age at diagnosis of menopause was the largest factor related to age at diagnosis of glaucoma while accounting for our covariates. Our models predicted a 0.67-year delay in age at diagnosis of glaucoma with each additional premenopausal year.

Conclusions: This case-only analysis elucidates a temporal association between menopause and glaucoma, highlighting the need to characterize the role of menopause in the onset of glaucoma for women.

Glaucoma is the second leading cause of blindness in the world and is projected to affect 112 million people by 2040.¹ The etiology of glaucoma is unknown but Higginbotham revisited the potential role of sex in glaucoma onset in a 2004 editorial.² This is relevant, as 59% of the glaucomatous population is female.³ The higher proportion of women developing glaucoma is often ascribed to longer lifespans; however, when examining prevalence by age, females have a higher prevalence rate of glaucoma over 80 years compared to men within the same age range (8.84% vs. 6.22%; data from the National Eye Institute), which suggests that older women may have sex-specific factors playing an unknown role in the development of glaucoma. 

Several clinical studies have examined other factors related to female health such as estrogen (a sex hormone that decreases after menopause), parity, and oral contraception with developing glaucoma. These studies have been aggregated and well-summarized in a recent systematic review.⁴ In brief, menopause causes modest elevations in intraocular pressure (IOP), which is the major modifiable risk factor associated with glaucoma.^5,6 Further, hormone replacement therapy containing estrogen use in postmenopausal women caused modest reductions in IOP.^7,8 Polymorphisms in the estrogen metabolic pathway are also associated with an increased primary open-angle glaucoma (POAG) risk in females.⁹ In addition, preclinical studies have found that retinal ganglion cell loss, which is the primary cause of vision loss in glaucoma, was ameliorated with estrogen treatment in multiple ocular hypertension animal models.^10–12 Mechanistically, surgical menopause in animal models reduces aqueous humor outflow facility, and ocular stiffness,^13,14 which may be associated with the elevation in IOP and the risk of developing glaucoma. 

Most notably, the Rotterdam study found an increased risk for POAG with early menopause (under 45 years of age onset).¹⁵ Several studies have also found early menopause increased or tended to increase, the risk of POAG in women,^16–19 although early menopause has not been uniformly found to increase the risk of glaucoma. Other studies have found that women experiencing late menopause (>55 years of age) had a lower risk of developing POAG.^20,21 These studies lay the fundamental groundwork to suggest that the timing of menopause plays a role in the risk of developing glaucoma. However, these studies could not assess if there was a temporal association between menopause and glaucoma. This was largely because many of these studies were cross-sectional evaluations. These studies often required confirmation of a glaucoma diagnosis by a later clinical examination and relied on self-reported information from questionnaires, sometimes decades later, to recall the age of menopause. To account for these factors, studies binned menopause into age ranges (e.g., <45, 45–55, and >55 years of age) and glaucoma as a binary outcome (yes/no) without a year of diagnosis. In addition, these studies did not control for confounding risk factors between groups or exclude participants with potentially confounding pathologies related to hormones (e.g., polycystic ovaries) or postmenopausal use of hormone replacement therapy. Thus, a study that can use a national database, with a diverse population, of patients who have received health care longitudinally would provide additional insight into how menopause and glaucoma are associated. 

Overall, there appears to be an association between menopause and glaucoma, and this connection may be partially mediated through estrogen signaling. However, no studies have determined if there is an association between age at diagnosis of menopause and glaucoma. Our goal was to examine the association between menopause and glaucoma. We anticipated finding a positive correlation between age at diagnosis of menopause and glaucoma in this retrospective study of females. 

Methods

Study Population

Here, we used the electronic health records of female veterans to evaluate the temporal association between age at diagnosis of menopause and glaucoma. The Veterans Affairs (VA) electronic health record is a national database that allows the inclusion of women from across the country, and females are the fastest growing group within veterans.²² To assess the association between age at diagnosis of menopause and glaucoma, we conducted a case-only retrospective study of U.S. female veterans with medical records in the VA Corporate Data Warehouse from January 2000 to December 2019. This study was approved by the Emory University Institutional Review Board and the Atlanta VA Medical Center Research and Development Committee. We were specifically interested in the association between menopause and glaucoma; thus, patients included in the study were women with diagnoses of both glaucoma and menopause. Figure 1 provides an overview of the criteria for inclusion/exclusion of the patient population used in the present study. 

Figure 1.

 

Inclusion and exclusion criteria for the overall study. From the 31,487 female patients in the VA Healthcare system with at least a single prescription or procedure record for glaucoma treatment, the final study population of 1825 women was selected. Gray boxes show patients who were excluded by criterion, and the black boxes show the number of patients that met each criterion. See Methods for detailed information about ICD, CPT, and VA drug classification codes.

In brief, the variable of interest was the age of menopause diagnosis. In this study, age at diagnosis of menopause was classified by the first occurrence of a menopause-related International Classification of Diseases (ICD) diagnosis (ICD-9: 256.31, 627.X, V49.81X; ICD-10: E28.31X, N95.X, Z78.X). All women had to have a preceding visit to the VA Healthcare System within 1 to 3 years of the menopause diagnosis without a menopause diagnosis occurring at this prior visit. The diagnosis of menopause had to occur between the ages of 35 and 65 years, a range that represents the ages for 95% of the female population entering menopause in the United States (within 2 SD of the mean age of menopause reported in the REGARDS study²³). Patients without a menopause diagnosis were excluded from this study. To investigate the effect of menopause, patients with any record of hormone therapy in the VA Healthcare System (VA drug classification codes GU500 and HS300) were not included in this study. Patients were then grouped into early, normal, and late menopause by their age of menopause diagnosis. The normal, or comparative, cohort group consisted of women who underwent menopause between 45 and 55 years of age, or within 1 SD of the mean age of menopause based on the REGARDS study. The early and late menopause groups were women who underwent menopause between 35 and 45 years of age or between 55 and 65 years of age, respectively.²³ 

The main outcome measure of this study was the age of glaucoma diagnosis. Glaucoma diagnosis is defined as the initiation of glaucoma treatment medication (VA drug classification codes: OP100, OP101, OP102, OP103, OP105, OP107, OP109) or procedure (codes: 65855, 0191T, 66174, 66175). Each glaucoma patient required a prior negative ophthalmological screening (Current Procedural Terminology [CPT] codes: 92002, 92004, 92012, 92014) within 1 to 3 years of the glaucoma diagnosis. The initiation of pharmacological treatment or procedure was used to determine the diagnosis of glaucoma, as these codes have been shown to robustly indicate glaucoma diagnosis.²⁴ In short, only patients receiving treatment for glaucoma were included in this study. In this study, we focused on open-angle glaucoma and excluded patients with trauma-induced glaucoma (ICD-9: 365.65X; ICD-10: H40.3X), glaucoma secondary to other eye disorders (ICD-9: 365.6X; ICD-10: H40.5X), glaucoma secondary to inflammation (ICD-9: 365.62X; ICD-10: H40.4X), glaucoma secondary to anomalies or systemic syndromes (ICD-9: 365.4X), steroid-induced or steroid responder glaucoma (ICD-9: 365.06X, 365.3X; ICD-10: H40.04X, H40.6X), and primary angle-closure glaucoma (ICD-9: 365.03X, 365.2X, ICD-10: H40.06X, H40.2X). Patients who had diabetic retinopathy (ICD-9: 362.0X; ICD-10: H35.X, E08.3X, E09.3X, E11.3X, E12.3X, E13.3X), macular degeneration (ICD-9: 362.5X; ICD-10: H35.X), Graves’ disease (ICD-9: 242.0X; ICD-10: E05.0X), Cushing's syndrome (ICD-9: 255.0X; ICD-10: E24.X), or polycystic ovary syndrome (ICD-9: 256.4X; ICD-10: E28.2X) diagnoses were also excluded from the study. 

We recorded demographic characteristics, biometric measurements, and anti-hypertensive medication usage. Patients were identified as having an Asian, black or African American, Native Hawaiian or Pacific Islander, American Indian or Alaska Native, or white racial background based on self-reported responses. In addition, patients were similarly classified as Hispanic or non-Hispanic ethnicities. Weight, height, and blood pressure measurements that were taken in the period between menopause and glaucoma diagnoses were averaged, and the average body mass index (BMI) during this period was calculated with the average weight and height measurements. Previous studies indicated a protective role of systemic anti-hypertensive medication against glaucoma,²⁵ and patients were categorized by use of systemic anti-hypertensive medication (VA drug classification codes: CV100, CV150, CV200, CV400, CV490, CV500, CV800, CV805). A complete list of codes used is provided in Supplementary Table S1. 

As an assessment of the patient's overall health, we also calculated the Elixhauser comorbidity index using the diagnosis records for the patients. The Elixhauser comorbidity index was developed to account for comorbidities in patients across 31 dichotomous categories, including neurological, cardiovascular, and renal diseases; the presence of each category was determined by the ICD-9 and ICD-10 codes established by Quan et al.²⁶ In short, each unit increase (a higher Elixhauser comorbidity index) indicates an additional disease category diagnosis. 

Analysis

Due to differences in the covariates (e.g., race, ethnicity) of women diagnosed as having experienced early and late menopause, we performed k:1 propensity score matching through the MatchIt package in R (R Foundation for Statistical Computing, Vienna, Austria).²⁷ To use well-established case-comparative cohort matching algorithms rather than the weighting algorithms frequently used when there are multiple case groups, the early and late menopause groups were independently matched against the entire normal menopause group. This approach generated two matched cohorts: (1) the EM-EC cohort, consisting of early menopause (EM) women (n = 215) and early comparative (EC) women (n = 860); and (2) the LM-LC cohort, consisting of the late menopause (LM) women (n = 525) and late comparative (LC) women (n = 525). Note, that the EC and LC groups were two different cohorts of women in the normal menopause population that were matched on the characteristics (e.g., race, ethnicity) of the EM and LM groups, but it is possible that there were women in both EC and LC groups. To prevent the identifiability of the patients due to the small number of individuals of certain racial backgrounds, race was divided into three categories: black or African American descent, white descent, and grouped descent (which included Asian descent, Native American or Alaska Native descent, and Native Hawaiian or Pacific Islander descent). 

First, we used a Pearson correlation and univariate linear regression to examine the linear correlation between the ages of menopause and glaucoma diagnoses in the EM-EC and LM-LC cohorts without accounting for the covariates. Regression coefficients and intercepts between the two linear regressions were compared utilizing a Z-test established by Clogg et al.^28,29 Note that the reference value for the age of menopause diagnosis was 50 years of age (e.g., a 40 year age of onset is equal to –10 years, and a 60 year age of onset is +10 years) to increase the interpretability of the intercept.³⁰ Then, we examined how the age at diagnosis of glaucoma changed after menopause by calculating the quantile shift. A quantile shift is the difference between two groups at each percentile. The quartile shifts are reported as median (interquartile range [IQR]). We used a density plot to show the shift in years for the diagnosis of glaucoma as distributions among the (1) EM-EC, (2) EC-LC, and (3) LM-LC groups. 

In our next analysis, we used a multivariate linear regression model on each matched cohort with the age at diagnosis of glaucoma serving as the outcome variable. The covariates were iteratively included in the regression model with the covariate leading to the greatest reduction in the Bayesian information criterion (BIC) added for that iteration. The potential coefficients in our multivariate regression were (1) age of menopause diagnosis; (2) race (binaries for self-reported Asian, black or African American, Native American or Alaska Native, Native Hawaiian or Pacific Islander, or white descent); (3) ethnicity (binary for self-reported Hispanic ethnicity); (4) BMI; (5) binary for previous systemic anti-hypertensive usage; and (6) Elixhauser comorbidity index. For the present study, we used systemic anti-hypertensive usage as a covariate in our regression in place of systolic and diastolic blood pressures due to the confounding relationship between blood pressure and anti-hypertensive medicine usage with glaucoma.²⁵ We terminated the forward selection of the coefficients into our multivariate model when the BIC did not decrease by more than two units. Similar to above, we used Z-tests to compare the coefficients between our regression models for the EM-EC and LM-LC cohorts, respectively. 

To assess if race differentially modulates the relationship between the ages of menopause and glaucoma diagnoses, we performed two multivariate linear regressions of each matched cohort for patients of black or African American and white descent, which were the two largest backgrounds represented in our dataset. Excluding race as a covariate, this multivariate linear regression used the same factors as the model selected above. Afterward, we again compared the coefficients of the multiple linear regression for the black or African American and white descent through Z-tests. 

We performed statistical tests using R software with a 5% level of significance, and tests were two sided. Dichotomous covariates are reported as the number and percentage of patients with a positive binary. For continuous covariates, we used a one-way analysis of variance (ANOVA) to compare across all cohorts and between matched cohorts, and the data are reported as means (95% confidence intervals [CIs]). In scenarios with multiple tests (e.g., comparing covariates across groups or comparing regression coefficients between subcohorts), we used Bonferroni corrections of our P values to determine significance. 

Results

The mean age of a menopause diagnosis for all women in the early, normal, and late menopause groups before matching in our study population was 51.8, which is similar to the range of 49 to 51 years of age reported by previous studies^23,31–33 (Supplementary Table S2). After propensity score matching, there were no differences in the covariates between the EM and LM groups with their respective comparative cohorts (EC and LC) (Table 1). However, we observed significant differences between the EC and LC groups, specifically in the age of glaucoma diagnosis (EC: 55.3 years; 95% CI, 54.9–55.8; LC: 57.1 years; 95% CI, 56.5–57.7; P < 0.001), the proportion of patients of black or African American descent (EC: 54.2%; LC: 45.1%; P = 0.001), the proportion of patients of white descent (EC: 37.6%; LC: 46.7%; P < 0.001), and average systolic blood pressure (EC, 126 mm Hg; 95% CI, 126–127; LC: 132 mm Hg; 95% CI, 131,133; P < 0.001). Though not reported due to low numbers in the cohorts, no significant differences in the proportion of patients of Hispanic ethnicity were seen between the EC and EM groups (P = 0.86) or between the LC and LM groups (P = 0.88). 

Table 1.

 

Baseline Characteristics of EM-EC and LM-LC Cohorts After Matching

 

We found a modest linear relationship when examining the age of diagnosis for glaucoma versus menopause in the EM-EC cohort (r = 0.396; P < 0.001) and LM-LC cohort (r = 0.462; P < 0.001) (Fig. 2). We did not find a significant difference in the slopes of these linear regressions (P = 1.00). However, we did observe a difference in the age of glaucoma diagnosis (y-intercepts; P < 0.001) in the EM-EC cohort (55.14 years; 95% CI, 54.72–55.56) and LM-LC cohort (56.49 years; 95% CI, 55.95–57.03) (Fig. 2). When looking at the mean age of glaucoma diagnosis between the EM and LM cohorts, we observed a significant difference in their age at diagnosis relative to their respective matched cohorts (Figs. 3A, 3B). Examining the quantile shifts, EM women developed glaucoma an average of 6.0 ± 0.97 years (IQR, 5.1–6.5) earlier (one-sample sign test, P < 0.001) than the EC group, whereas LM women developed glaucoma an average of 5.2 ± 0.80 years (IQR, 4.8–5.7; P < 0.001) later than the LC group (Fig. 3C). Comparing the two comparative cohorts, we found that the LC cohort developed glaucoma 1.8 ± 0.41 years (IQR, 1.5–1.9) later (P < 0.001) compared to the EC group (Fig. 3C), reflecting the difference in age at diagnosis of glaucoma as shown in Table 1. 

Figure 2.

 

Positive linear relationship between ages of menopause and glaucoma onset in the study population. Means and 95% CIs for the ages of glaucoma onset for each age of menopause integer are indicated by dotted lines and shading, respectively, between the EM-EC (green; r = 0.40, P < 0.001) and LM-LC (purple; r = 0.46; P < 0.001) cohorts with linear regression of age of glaucoma onset versus continuous age of menopause onset.

Figure 3.

 

Age of menopause diagnosis was associated with a delay in glaucoma diagnosis. (A, B) Distribution of age of glaucoma diagnosis for women in the EM (green) and the EC (light blue) cohorts (A) and women in the LM (purple) and LC (dark blue) cohorts (B). (C) Distribution of percentile differences between the EC and EM groups (green), EC and LC groups (gray), and LC and LM groups (purple). All distributions were significantly non-zero, with the LC-EC (median of 1.8 years later; IQR, 1.5–1.9) and LM-LC (median of 5.2 years later; IQR, 4.8–5.7) distributions having a positive shift and the EM-EC distribution having a negative shift (median of 6.0 years earlier; IQR, 5.1–6.5). *P < 0.05, **P < 0.01, ***P < 0.001.

Next, we performed multivariate linear regressions of the EM-EC and LM-LC matched cohorts. The full models for each cohort are presented in Supplementary Table S3. The BIC-based forward selection approach derived a model for the LM-LC matched cohort with age of menopause diagnosis, anti-hypertensive medication usage, and a binary representing identification of white descent, which we then extended to the EM-EC matched cohort. The EM-EC model had selected age of menopause diagnosis and white descent (Supplementary Figures S1 and S2, Supplementary Tables S4 and S5). We note that black or African American descent was the next covariate to be included in the regression after the BIC cutoff. The final models estimated that, with each additional year before menopause diagnosis, the age of glaucoma diagnosis was delayed by 0.67 year (95% CI, 0.58–0.76) and 0.66 year (95% CI, 0.59–0.74) in the EM-EC and LM-LC cohorts, respectively (Table 2). Our multivariate analysis also predicted a delay in glaucoma diagnosis in patients of white descent and the use of anti-hypertensive medication in the EM-EC and LM-LC cohorts. In our multivariate analysis, none of the coefficients between the EM-EC and LM-LC models was significantly different (y-intercept, P = 0.07; age of menopause diagnosis, P = 0.91; systemic anti-hypertensive usage, P = 0.57; white descent, P = 0.67). 

Table 2.

 

Multivariate Regressions to Predict the Age of Glaucoma Diagnosis for EM-EC and LM-LC Cohorts

 

As race is known to be associated with the development of glaucoma, we next evaluated if white or black and African American descent modified the relationship between the ages of menopause and glaucoma diagnoses. Here, we performed two stratified multivariate linear regressions on patients of black or African American descent and patients of white descent within each matched cohort (EM-EC and LM-LC). We did not find any significant differences in the regressions of the two stratified subcohorts in the EM-EC cohort for the y-intercept (P = 0.15), age of menopause diagnosis (P = 0.68), or systemic anti-hypertensive medication usage (P = 0.50) (Table 3). We also did not observe any differences in the stratified regressions in the LM-LC cohort (y-intercept, P = 0.32; age of menopause diagnosis, P = 0.08; systemic anti-hypertensive medication usage, P = 0.80) (Table 3). Further, the coefficients for the age of menopause predictor of glaucoma diagnosis for both the black or African American patients (EM-EC: 0.67; 95% CI, 0.54–0.79; LM-LC: 0.57; 95% CI, 0.45–0.68) and the white patients (EM-EC: 0.63; 95% CI, 0.47–0.78; LM-LC: 0.71; 95% CI, 0.60–0.82) were consistent with the full cohorts’ coefficients (EM-EC: 0.67; 95% CI, 0.58–0.76; LM-LC: 0.66; 95% CI, 0.59–0.74). 

Table 3.

 

Multivariate Regressions to Predict the Age of Glaucoma Diagnosis for EM-EC and LM-LC Cohorts Stratified by Race

 

When looking at the outcome variable distribution based on the included predictors, the average age of glaucoma diagnosis was later in patients of white descent (55.2 years; 95% CI, 54.5–55.9) than in patients of black or African American descent (53.7 years; 95% CI, 53.1–54.3) in the EM-EC matched cohort (Fig. 4A). A similar relationship was found in the LM-LC matched cohort: black or African American descent, 59.0 years (95% CI, 58.4–59.6); white descent, 60.5 years (95% CI, 59.9–61.2) (Fig. 4B). Systemic anti-hypertensive medications were associated with a delay in age at diagnosis of glaucoma in the EM-EC cohort (54.5 years; 95% CI, 54.0–55.1 vs. 53.5 years; 95% CI, 52.7–54.2) (Supplementary Fig. S3A) and in the LM-LC cohort (60.2 years; 95% CI, 59.6–60.7 vs. 58.6 years; 95% CI, 57.8–59.4) (Supplementary Fig. S3B). 

Figure 4.

 

Relationship between age at diagnosis of menopause and glaucoma as modulated by race. (A, B) Distribution of age of glaucoma diagnosis by race, with the color signifying group and sample size: EC (light blue) and EM (green) (A) and LC (dark blue) and LM (purple) (B). (C, D) Univariate linear regression of age of glaucoma diagnosis versus age of menopause diagnosis for patients of black or African American descent (solid) or white descent (dashed) versus age of menopause diagnosis in early (C) and late (D) menopausal cohorts. Points and error bars represent mean and 95% CIs. *P < 0.05, **P < 0.01, ***P < 0.001.

Discussion

We identified a modest linear relationship between the age of menopause and glaucoma diagnoses using matched cohorts of early and late menopausal women. There was no difference in the association of age at diagnosis of glaucoma after menopause in the EM and LM cohorts (Fig. 2, Table 2). Previous studies identified an increased risk of glaucoma development with early menopause and reduced risk with late menopause.^20,21 Many of these studies did not include the age of glaucoma in their analyses, binned menopause into age ranges from self-reported data, and did not control for confounding parameters. Our results suggest that the age of menopause had a similar impact on the age of glaucoma regardless of when women entered menopause. Interestingly, we only assessed the mean age of menopause within the EM and LM cohorts relative to respective normal menopausal cohorts. In brief, EM women had an average 6-year earlier (IQR, 5.1–6.5) diagnosis of glaucoma relative to a matched group (EC) of women who experienced normal menopause, whereas LM women had a 5.2-year later (IQR, 4.8–5.7) diagnosis of glaucoma compared to their matched comparative group (LC). The shifts in mean glaucoma diagnosis age in the EM and LM cohorts agreed with the trends in these earlier works. However, the linear association we observed between the age of menopause and glaucoma diagnoses indicates that this association may be independent of early, normal, or late menopause classifications and may be due to time from menopause. One study highlighted that there was a 5% decrease in developing POAG for each year a woman was exposed to endogenous sex hormones (e.g., years of menstruation).¹⁵ Although this study was commenting on years of menstruation and we assessed age at diagnosis of menopause, our data agree with this previous study, as entering menopause later was associated with a delayed age at diagnosis of glaucoma. Further, it may indicate that it is not the number of years of menstruation, but the later cessation of menstruation, that influences the diagnosis of glaucoma in these women. These results indicate that there is a modest association between menopause and glaucoma and that this impacts all women regardless of menopausal age. Combined, these results signify that menopause may be a factor to consider for monitoring eye health in aging women. 

Our work also expanded on previous reports of menopause and glaucoma by accounting for confounding factors (e.g., race, ethnicity, BMI, anti-hypertensive medication usage, comorbidities) through the matching of clinical records. We performed a multivariate linear regression to account for the role that these covariates may have in the association between menopause and glaucoma in the EM-EC and LM-LC cohorts. In these analyses, the age of menopause diagnosis was the first covariate included in the multivariate regression, which estimated a 0.67-year delay in glaucoma diagnosis with each additional year before menopause in the EM-EC and LM-LC cohorts. Another covariate in our multivariate linear regression was the use of systemic anti-hypertensive medication. The relationship between blood pressure and glaucoma is complicated, as both hypotension and hypertension are associated with an increased risk of developing glaucoma.^34–38 However, the use of anti-hypertensive medications in hypertensive patients has been shown to decrease the risk of developing glaucoma.^25,39,40 In our population, it was challenging to independently assess the impact of blood pressure, as ∼69% of our population was receiving anti-hypertensive medication. However, we did find that the use of anti-hypertensive medication was associated with a delayed diagnosis of glaucoma, which agrees with previous studies on the use of anti-hypertensive medication and the risk of developing glaucoma.^25,39,40 

Race was shown to be predictive of the age at diagnosis of glaucoma in our study. Previous studies have found a connection between race and glaucoma with higher prevalence and worse outcomes in patients of black or African American descent and a lower prevalence in individuals of white descent.^41,42 However, it is worth noting that race is multifactorial, and differences do not necessarily reflect genetic factors but could also implicate cultural, social, economic, and environmental factors, as well.⁴³ Due to an inability to directly assess the range of these factors, we were only able to include race as recorded from patients’ self-reports in the medical records. Both patients of black or African American descent (49.6%) and white descent (41.8%) were sufficiently represented for further analysis. We found no interaction between race and age of menopause diagnosis when predicting the age at diagnosis of glaucoma. To confirm these findings, we performed separate multivariate linear regressions on patients of black or African American and white descent within each matched cohort. Notably, we did not detect any differences between these two subcohorts in the coefficients of our multivariate regressions between patients of black or African and white descent using Z-tests (Table 3). This suggests that the association between menopause and glaucoma is comparable between these two races (i.e., the relationship between menopause and glaucoma is independent of racial background). One limitation of this study, though, is its smaller population of patients of Asian, Native American or Alaska Native, and Native Hawaiian or Pacific Islander descent, which may limit the generalizability of our results to those populations. According to the Department of Veterans Affairs recent Veteran Population Projection Model (VetPop2020), the patient population for the VA Healthcare System was predominantly patients of white descent (80%) and black or African American descent (12%) in 2019, the final year of the study period; hence, these studies may have to be replicated in other racial backgrounds as more data are collected. Although we chose to include them in our analyses, we were unable to divide them into individual racial backgrounds due to the limited number of patients currently in the system. 

Naturally, some limitations come with doing a case-only retrospective study. Ideally, we would be able to simultaneously match women experiencing early, normal, and late menopause. However, our initial covariate analysis identified significant differences in the characteristics (e.g., race, ethnic background, BMI, blood pressure, and Elixhauser comorbidity index) of the three menopause groups, indicating the need for matching among groups. To utilize the more interpretable case-comparative cohort matching algorithms rather than weighting algorithms, we divided our populations into EM-EC and LM-LC cohorts for matching. We found significant differences in the average age of glaucoma diagnosis between the EC and LC groups, which were anticipated due to the differences between the EM and LM groups. Compared to the EC group, the LC cohort had a higher percentage of individuals of white descent (46.7% vs. 37.6%) and a slightly later age of menopause diagnosis (50.5 years vs. 50.3 years). The differences in race and age of menopause diagnosis in our multivariate model would predict a delayed diagnosis of glaucoma in the LC group compared to the EC group. This is exactly what we observed; however, as the multivariate regressions can also account for our other covariates, we do not expect any bias from the different comparative cohorts to affect the predictor coefficients. Further, when we performed separate multivariate regressions on individuals with black or African American and white descent, the predictor coefficients were not significantly different (as described above). Although we did not directly compare the EM and LM groups, we were able to find consistent results with the regressions, indicating there is veracity to the findings. 

Our study was performed using a VA database. This database has many advantages, such as a large and diverse population, a national database, and consistent coding. These findings should be confirmed in civilian or other non-veteran databases. Thus, it would be beneficial for future studies to investigate the association between the ages at diagnosis of menopause and glaucoma in other veteran and general population databases (e.g., Million Vet Program, All-of-US, UK Biobank) that include the type of glaucoma, social, biological, and environmental factors. 

Our cohorts only included women with a diagnosis of menopause and glaucoma. Using this case-only design allowed us to assess the temporal association between these two diagnoses and is often used to understand how a disease is associated with another event or the environment. Our study highlights that the age at diagnosis of glaucoma typically occurs within 5 to 6 years after menopause, suggesting that this is a period in which women should be more carefully monitored for the onset of glaucoma. We will also use this information when designing future prospective case-comparative studies for tracking the onset of glaucoma after menopause. We were also limited to assessing factors present in a patient's electronic health record. Although this allowed us to better estimate the association between the age at diagnosis of menopause and that of glaucoma, it did not allow us to consider perimenopause or the menopausal transition, which may have a relevant role in the menopause–glaucoma relationship. Unfortunately, perimenopause is not reliably coded, but future prospective studies focused on perimenopause and glaucoma could potentially investigate this relationship. For our present study, we surveyed the population to understand the overall diagnoses of glaucoma and menopause within the VA. Overall, 80% of women were diagnosed with glaucoma after menopause, and 8.6% were diagnosed with menopause within 2 years of developing glaucoma, suggesting that glaucoma presents mainly after menopause or near the perimenopause transition. Another limitation is that we also cannot verify diagnoses. To minimize this limitation, we used stringent inclusion criteria. For glaucoma diagnosis, a previous study showed that ICD diagnoses within a single medical center had a specificity of 82.6% for glaucoma.²⁴ Based on this previous study,²⁴ we used the initiation of treatment for glaucoma to define the age at diagnosis of glaucoma in our multiple-center study. In our datasets, the first treatment for glaucoma was primarily medication (86% of glaucoma patients), but some patients received surgical intervention as the first treatment for glaucoma (14% of glaucoma cases). It would have also been beneficial to perform a subanalysis on the relationship between menopause and various types of open-angle glaucoma; however, without the ability to confirm each glaucoma subtype, we avoided this analysis. Here, our goal was to examine the overall temporal association between menopause and glaucoma. We did find an association between these two diagnoses, which likely benefited from using the VA database, where coding across multiple centers is similar, thus improving consistency across the nation and reducing potential biases or inconsistency associated with different institutions. Another limitation of using a retrospective dataset is the potential underlying health conditions that influence the age at diagnosis of menopause or glaucoma. To reduce this bias, we used the Elixhauser comorbidity index as part of our matching criteria. 

These stringent criteria limited our analysis to when menopause and glaucoma became symptomatic in patients; however, each condition likely existed before diagnosis. This is a limitation of clinical studies (retrospective or prospective), as procedures, disease onset, or biological events are only recorded when patients seek medical care. To improve our specificity for each condition, we required that each woman have a clinical visit without a menopause diagnosis (within 1–3 years of the menopause diagnosis) and a negative ophthalmological screening (within 1–3 years of the incident medication usage). We believe future long-term prospective studies would help us better understand the relationship between menopause and glaucoma. This study also did not directly assess the risk of developing glaucoma related to age at diagnosis of menopause. 

We also excluded patients with other known ocular pathologies (i.e., diabetic retinopathy and related macular degeneration). Previous studies have not specifically examined the association of menopause and glaucoma without these other pathologies; therefore, our primary analysis only included patients with glaucoma while excluding these ocular pathologies. This was important, as diabetic retinopathy potentially increases the risk of other ocular diseases, including glaucoma and age-related macular degeneration.^44,45 However, we ran an additional multivariate analysis and found that the association between menopause and glaucoma in the EM-EC and LM-LC cohorts did not significantly change by including ocular diseases (Supplementary Table S6). This showed that there was a strong and consistent association between the diagnoses of menopause and glaucoma. 

Another limitation was that we only included women who had developed glaucoma (e.g., a case-only study design). Previous works have shown that the age at diagnosis of menopause influences the risk of developing glaucoma.^16–21 However, our goal was to determine if there was an association between age at diagnosis of menopause and glaucoma, which to date has not been investigated. We found an analogous association between the age at diagnosis of menopause and glaucoma across different ages of menopause, where the average age at diagnosis of glaucoma increased with age at diagnosis of menopause. If no association between menopause and glaucoma existed, we would see that the average age of glaucoma would not vary consistently with the age of menopause. This indicates that for women suffering from glaucoma, their age of menopause is associated with its onset. This association makes menopause an important life event to consider when monitoring for the diagnosis of glaucoma. 

Overall, this study identified a linear relationship between the age of menopause and the age at diagnosis of glaucoma. This association was consistent through different permutations (simple regression and multivariate analyses) of the data and between two distinctly matched cohorts, regardless of differences in covariates. An advantage of this study was our ability to use a national database, which increases our applicability, as these trends are not limited to a region and thus increase the diversity of our populations. In addition, this association was important to identify in a veteran population, as veterans have a higher prevalence of glaucoma compared to the general population.⁴⁶ These results also found that females who experienced early or late menopause had different ages of diagnoses of glaucoma, which agreed with trends in previous data indicating that there is a higher risk for glaucoma with early menopause and protection for late menopause.^20,21 Although these previous studies assessed discrete changes in the risk of developing glaucoma, our findings demonstrate a direct association between the age of menopause diagnosis and the age of glaucoma diagnosis. This is particularly important, as preclinical animal studies have shown that surgical menopause leads to a faster decline and more severe loss of visual function.^47,48 This highlights the relevance of menopause timing to the age of glaucoma diagnosis, which may also impact glaucoma progression and later-stage threats to vision. These findings suggest that menopause, independent of its onset, is a relevant factor in determining screening frequency and management of glaucoma. 

Acknowledgments

Disclosure: K. Hogan, None; X. Cui, None; A. Giangiacomo, None; A.J. Feola, None