Postmenopausal Hormone Therapy Was Associated With Later Age of Onset Among Glaucoma Cases

Abstract

Purpose

Hormonal therapy (HT) has been suggested to lower the risk of developing glaucoma. Our goal was to investigate the association between HT use and the onset of glaucoma diagnosis in postmenopausal women.

Methods

This retrospective case-only study included female veterans with open-angle glaucoma from VA records between 2000 to 2019. Propensity score matching was used to match HT (n = 1926) users to untreated (n = 1026) women on multiple covariates (e.g., age of menopause, BMI, blood pressure, antihypertensive medications, and a co-morbidity index). A simple linear regression was used to evaluate the impact of HT duration on the age of glaucoma diagnosis, and multivariate linear regression analysis was used to determine which factors contributed to the age at diagnosis of glaucoma.

Results

We found a linear relationship between the age at diagnosis of glaucoma and menopause in women with (r = 0.54) and without HT (r = 0.57) use. HT users tended to have a later diagnosis of glaucoma. Our multivariate analysis found that 0–2 years, 2–5 years, and >5 years of HT use were associated with a 2.20 [confidence interval (CI), 1.64, 2.76], 3.74 [CI, 3.02, 4.46], and 4.51 [CI, 3.84, 5.18] years later diagnosis of glaucoma. An interaction (−0.009 [−0.015, −0.003]) was observed between HT duration and age of menopause diagnosis, with the impact of HT decreasing for later menopause ages.

Conclusions

Longer duration of HT use was associated with a later diagnosis of glaucoma in postmenopausal women in this case-only analysis. The impact of HT may be modulated by menopausal age, although further study is needed. The findings support a protective role of estrogen in glaucoma pathogenesis.

Open-angle glaucoma (OAG) is a major cause of irreversible blindness in the United States. The major risk factors for OAG include age, ethnicity, family lineage, and intraocular pressure (IOP).¹ Although glaucoma can occur at any level of IOP, high tension (i.e., ocular hypertension) remains clinically relevant as a major causal risk factor for developing this disease. High tension, defined as an IOP > 21 mm Hg, is associated with a relative risk of 12.8 for developing glaucoma versus the normotensive population.^1,2 In addition, high tension in women can occur an entire decade earlier (51–60 years) than in men,³ which corresponds with the average age at which a woman enters menopause (51 years). Although previous work has suggested that the age of menopause influences the risk of developing glaucoma,^4–8 our recent work demonstrated that there was an association between the age at diagnoses of menopause and glaucoma, regardless of the age a woman became postmenopausal.⁹ We found a 0.67-year later diagnosis of glaucoma with each additional premenopausal year for women who developed glaucoma.

This is relevant, as postmenopausal women, regardless of age, have a 1.5 to 3 mm Hg higher IOP compared to aged-matched premenopausal women.^10,11 In addition, postmenopausal women taking hormone therapy (HT) containing estrogen had a lower IOP (by 0.5–3 mmHg) compared to postmenopausal women not taking HT.^12–14 Studies suggest that estrogen-based therapies may be a potential treatment for glaucoma.^15–18 In addition, a recent review of multiple clinical studies confirmed that HT modifies IOP and is a clinically relevant treatment that warrants additional investigation.⁴ Furthermore, preclinical studies from our group and others have shown that surgical menopause, which causes the rapid decline in estrogen, causes greater loss of retinal ganglion cells and visual dysfunction after retinal ganglion cell injury.^19–21 Studies have also demonstrated that estrogen-based therapy after an injury helps preserve retinal ganglion cells and visual function.^19–24 Together these studies suggest that estrogen modulates the risk for glaucoma and that estrogen-based therapy may be protective. Given this overall evidence, it is important to investigate whether HT is associated with a delayed age at diagnosis of glaucoma in postmenopausal women.

This concept is supported by several clinical studies that demonstrated that HT containing estrogen reduced the risk of glaucoma.^4,18,25,26 A study done by Newman-Casey and colleagues²⁵ found that HT containing only estrogen reduced the risk of developing OAG by a small (0.4%) but significant degree; however, this study only examined women over the age of 50 years of age. In addition, this study did not account for the duration of taking HT or the ability to document the age of menopause. Examining the Nurses’ Health Study, HT was protective when only examining high-tension OAG patients, but this protection was not observed when including both normal and high-tension OAG patients.⁴ This study grouped women into a range of onset of menopause (<45, 44–55, or >55 years of age) and glaucoma was assessed as a simple yes or no variable with no year of diagnosis. This study also did not account for other confounding factors (i.e. hypertension or endocrine metabolic conditions); therefore, additional studies examining the association between glaucoma and HT are needed. Vajaranant et al. used the Women's Health Initiative (WHI) dataset to assess if the risk of developing high-tension OAG following HT varied by race.¹⁸ This study examined women who received a placebo or HT following hysterectomy and found that age and African American descent had an increased risk of developing glaucoma. This study did not observe a benefit of HT compared to placebo when assessing all women; however, when separating women by racial background, they found that HT containing estrogen decreased the risk of OAG in women of African-American descent compared to placebo. Overall, this was a strong study with a long follow-up of patients. However, their analysis did not assess whether the use of HT would impact the onset of glaucoma.

Our previous case-only study found that the age of menopause was the most significant predictor for the diagnosis of glaucoma.⁹ Here, we aim to examine the association between HT and the onset of glaucoma in a case-only population. We hypothesize that HT was associated with a delay in glaucoma diagnosis in postmenopausal women compared to women without HT.

Methods

Study Population

This study was approved by the Emory University Institutional Review Board and the Atlanta Veterans Affairs Medical Center Research and Development Committee. Female veterans were selected from the VA Corporate Data Warehouse if they met the inclusion and exclusion criteria for the study, as detailed in Figure 1 and similar to our previous article,⁹ except for the inclusion of patients with HT treatment. Records were restricted to a period from January 2000 to December 2019. Patients were excluded if they had been diagnosed with a confounding ophthalmological disease, such as diabetic retinopathy (ICD-9: 362.0X; ICD-10: H35.X, E08.3X, E09.3X, E10.3X, E11.3X, E13.3X) and macular degeneration (ICD-9: 362.5X; ICD-10: H35.X), or confounding hormone-related disorders, such as Graves’ disease (ICD-9: 242.0X; ICD-10: E05.0X), Cushing's syndrome (ICD-9: 255.0X; ICD-10: E24.X), and polycystic ovary syndrome (ICD-9: 256.4X; ICD-10: E28.2X). In addition, to reduce heterogeneity in glaucoma phenotype, patients were excluded if they had trauma-induced glaucoma (ICD-9: 365.65X; ICD-10: H40.3X), glaucoma secondary to other eye disoders (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 diagnoses (ICD-9: 365.03X, 365.2X, ICD-10: H40.06X, H40.2X).

Figure 1.

Inclusion and exclusion criteria for the overall study. The final cohort of 3781 was selected from 31,809 patients with at least one glaucoma medication prescription or record of glaucoma treatment. Patients were excluded if diagnosis criteria for glaucoma or menopause were not met, and they presented with any excluded conditions (i.e. diabetic retinopathy or macular degeneration). See the Methods section for details regarding ICD and CPT codes.

Age of incident menopause diagnosis, the secondary variable of interest in the study, is the age at which a patient receives an ICD diagnosis of menopause (ICD-9: 256.31, 627.X, V49.81X; ICD-10: E28.31X, N95.X, Z78.X) preceded by an outpatient visit one to three years prior in which no diagnosis occurred. Women were excluded if they received their diagnosis outside the range of 35 to 65 years of age, estimated to be outside of 2 standard deviations of the mean age of menopause based on the REGARDS study.²⁷ We used a similar definition for the age of menopause diagnosis as our previous work,⁹ defining early menopause (35–45 years of age), normal menopause (45–55 years of age), and late menopause (55–65 years of age). Postmenopausal women were divided into HT (VA drug classification codes: GU500, HS300) or untreated cohorts.

The outcome measure is the age of incident glaucoma diagnosis, which is when the patient first received glaucoma treatment, either by medication (VA drug classification codes: OP100, OP101, OP102, OP103, OP105, OP107, OP109) or procedure (CPT codes: 65855, 0191T, 66174, 66175). To note, the vast majority only received medication as their incident glaucoma treatment (87%, n = 3301/3781). The initial glaucoma diagnosis had to be preceded by an ophthalmological screening (CPT codes: 920002, 92004, 92012, 92014) without intervention for glaucoma. Open-angle (ICD-9: 365.1X; ICD-10: H40.1X) diagnoses were reported for reference but not included in matching algorithms. Previous work has shown that requiring a diagnosis and intervention resulted in a more robust indication of glaucoma diagnosis.²⁸

Other covariates analyzed were racial and ethnic background (self-reported), body mass index, average systolic and diastolic blood pressure, systemic antihypertensive medications, and Elixhauser comorbidity index. Patients self-reported race as being of Asian, Black or African American, Native American or Alaska Native, Native Hawaiian or Pacific Islander, or White descent and self-reported ethnicity as being of Hispanic or non-Hispanic ethnicities. Body mass index was calculated from the average weight and height for the time period between menopause and glaucoma diagnoses. Systolic and diastolic blood pressure was averaged over the same period, and any history of systemic anti-hypertensive medication prescriptions (VA drug classification codes: CV100, CV150, CV200, CV400, CV490, CV500, CV800, CV805) was coded as a dichotomous variable. The Elixhauser co-morbidity index was determined using the coding algorithms specified by Quan et al.²⁹ and incorporated any diagnoses before incident glaucoma. A complete list of codes used are listed in Supplemental Table S1.

Analysis

To evaluate the effect of HT use on glaucoma development, we matched the patient population using HT (henceforth, the HT group, n = 1926) against the patient population without a history of HT use (the untreated group, n = 1026). We matched within different menopause age groups (early, 35–45 years; normal, 45–55 years; or late, 55–65 years) to ensure the case populations at different menopausal ages were similar. Patients were matched using propensity-score matching (with the MatchIt package)³⁰ and were matched on the age of menopause diagnosis, race, ethnicity, average BMI, previous systemic anti-hypertensive use, average diastolic and systolic blood pressure, and Elixhauser co-morbidity index. This resulted in six separate cohorts: early menopause untreated (EU), early menopause hormone therapy (EH), normal menopause untreated (NU), normal menopause (NH), late menopause untreated (LU), and late menopause HT (LH).

Because our previous study found that early, normal, and late menopausal women had a similar association between the age of menopause and the onset of glaucoma,⁹ for our initial analysis here these populations were combined for our initial linear regression analysis. Simple linear regressions of age of glaucoma diagnosis versus age of menopause diagnosis were separately developed for the HT and untreated cohorts.

To account for the other covariates found significant in our previous paper, multivariate linear regressions were created with the following terms added: HT use (using bins of 0–2 prescription years, 2–5 prescription years, and >5 prescription years with untreated serving as the reference), self-reported White descent, systemic antihypertensive medication use, and age of menopause diagnosis. We shifted the age of menopause diagnosis so that the value in the regression was the difference from a menopause diagnosis at 50 years of age. This allows the y-intercept to be more interpretable as it represents the age at which a woman who was diagnosed with menopause at 50 years of age would be predicted to develop glaucoma (provided all other covariates would be 0 as well). The validity of this model was confirmed with forward selection, looking at relevant covariates without interaction. Possible predictors for the regression model included HT use, age of menopause diagnosis, self-reported race, self-reported ethnicity, average BMI, systemic anti-hypertensive medication use, and Elixhauser co-morbidity index. Predictors were added to the model until the BIC did not decrease by at least 2 units.

To determine if race differentially affected HT response, linear models using the binned HT variable were separately developed for patients of Black or African American descent and White descent (the two largest racial groups in the study). Z-tests were used to compare the regression coefficients.³¹ Furthermore, like above, single linear regressions between the age of menopause and the age of glaucoma diagnosis for untreated and HT groups were separately developed and plotted.

Last, to inspect how HT and age of menopause interact to affect the diagnosis of glaucoma, we performed a secondary regression using HT as a continuous variable based on the number of years with a prescription and added the other covariates in addition to an interaction term between the HT use and age of menopause. To observe potential trends, we subsequently stratified by age of menopause diagnosis into three cohorts early (35–45 years of age), normal (45–55 years of age), and late (55–65 years of age). We plotted the data as violin plots that represent the age of glaucoma diagnosis of HT relative to the respective early, normal, and late untreated matched populations. We then plotted the quantile shifts between the HT and untreated groups (early, normal, and late) were calculated by subtracting each age-of-glaucoma-diagnosis percentile of the untreated group from the HT group.

All statistical testing and modeling were done in R (Boston, MA, USA). Significance was at 5%, and tests were two-sided. In comparing covariates between groups, testing was done with Kruskal-Wallis or one-way ANOVAs if the covariate was continuous. Bonferroni corrections were used as appropriate. Data are represented as mean [95% confidence interval].

Results

Propensity score matching successfully resolved covariate differences between untreated and HT groups (Table 1, unmatched data in Supplemental Table 2). There remains differences across the age of menopause groups after matching (Supplemental Table 3). Though ethnicity numbers were too low to report, proportions of patients that identified with Hispanic ethnicity did not differ between HT and untreated groups (P = 0.55).

Table 1.

Baseline Characteristics Across Matched Untreated and HT Groups Stratified by Age of Menopause Cohorts

Covariate	EU (n = 135)	EH (n = 273)	P Value	CU (n = 585)	CH (n = 1088)	P Value	LU (n = 306)	LH (n = 565)	P Value
Age of glaucoma diagnosis (yr)	49.9 [48.7, 51.1]	53.5 [52.7, 54.4]	<0.001	55.8 [55.2, 56.4]	59.3 [58.9, 59.7]	<0.001	62.5 [61.7, 63.2]	65.2 [64.6, 65.7]	<0.001
Age of menopause diagnosis (yr)	42.0 [41.6, 42.4]	42.1 [41.8, 42.4]	0.74	50.2 [50.0, 50.4]	50.1 [50.0, 50.3]	0.60	59.0 [58.7, 59.4]	59.1 [58.9, 59.4]	0.58
Race
Black or African American descent	51.1% (69)	49.5% (135)	0.75	50.4% (295)	47.5% (517)	0.26	37.6% (115)	34.2% (193)	0.31
White descent	39.3% (53)	40.3% (110)	0.84	40.7% (238)	43.2% (470)	0.32	51.3% (157)	58.2% (329)	0.05
Average systolic blood pressure (mm Hg)	126 [124, 127]	126 [125, 127]	0.89	129 [128, 130]	129 [128, 130]	0.95	131 [130, 133]	131 [130, 132]	0.66
Average diastolic blood pressure (mm Hg)	76.8 [75.6, 78.0]	76.4 [75.6, 77.2]	0.55	77.0 [76.4, 77.5]	76.5 [76.1, 76.8]	0.12	75.1 [74.3, 75.9]	74.6 [74.1, 75.1]	0.28
Systemic antihypertensive medication use	69.6% (64)	71.4% (195)	0.71	71.8% (420)	74.7% (813)	0.19	70.3% (215)	71.2% (402)	0.78
Body mass index	31.4 [30.3, 32.5]	30.3 [29.6, 30.9]	0.07	31.2 [30.7, 31.7]	30.5 [30.1, 30.8]	0.02	30.7 [30.0, 31.5]	30.1 [29.6, 30.5]	0.12
Elixhauser comorbidity index	8.49 [7.81, 9.16]	8.82 [8.38, 9.25]	0.41	7.74 [7.42, 8.05]	8.23 [8.00, 8.46]	0.01	7.62 [7.19, 8.06]	8.18 [7.86, 8.51]	0.04
HT prescription amount (years)		5.86 [4.99, 6.73]			4.79 [4.33, 5.25]			4.36 [3.77, 4.94]

EU, early menopause untreated; EH, early menopause HT; NU, normal menopause untreated; NH, normal menopause HT; LU, late menopause untreated; LH, late menopause HT.

Continuous covariates are presented as mean [95% confidence intervals]; dichotomous covariates are presented as percentages with the number of patients in parentheses. With Bonferroni correction, α = 0.00167.

To note, the average duration of HT use tended to decrease with the age of menopause diagnosis (early: 5.86 [4.99, 6.73] prescription-years, normal: 4.79 [4.33, 5.25] prescription-years, late: 4.36 [3.77, 4.94] prescription-years, P = 0.025, Table 1). Corresponding to this trend, the percentage of patients in the HT usage bin of 0–2 prescription-years increased with age of menopause diagnosis (early: 42% [115], normal: 48% [527], late: 57% [323], P < 0.001), whereas the percentage of patients in the HT usage bin of greater than 5 prescription-years decreased with age of menopause diagnosis (early: 37% [102], normal: 28% [300], late: 25% [139], P < 0.001).

The simple linear regressions for the age of glaucoma diagnosis vs. age of menopause diagnosis were performed separately in the untreated and HT cohorts. We found a slightly steeper slope in the untreated regression (0.75 [0.68, 0.81]) compared to the HT cohort (0.67 [0.62, 0.72], P = 0.055; Fig. 2A) and a lower y-intercept (untreated: 55.7 [55.3, 56.1], HT: 59.1 [58.8, 59.4], P < 0.001). We noted the two lines tended to converge, which may be due to differential HT use across the age of menopause cohorts (Fig. 2B, Table 1), as the percent of the HT group in the 0–2 year bin increases with the age of menopause (42% vs. 48% vs. 57%), while the percent in the 5+ year bin decreases (37% vs. 28% vs. 25%).

Figure 2.

Linear regression between age at diagnoses of menopausal women with and without hormone therapy (HT) and glaucoma. (A) Mean (dotted line) and 95% confidence intervals (shaded region) of the age of glaucoma diagnosis for each age of menopause for women with HT (red; r = 0.538, P < 0.001) or untreated (Black; r = 0.569; P < 0.001) groups. Solid lines represent the linear regression lines. The age of menopause was rounded down to the nearest integer. The HT and untreated slopes were 0.67 and 0.75, and y-intercepts were 59.1 and 55.7, respectively. (B) Patients were binned based on HT use into 0–2 years (light pink), 2–5 years (salmon), and 5+ years (red); the percentage of the HT group for each menopause cohort in each bin was reported followed by the numbers in parentheses.

The full multivariate linear regression model is presented in Supplemental Table 4. When using forward selection for our multivariate linear regression analysis, the age of menopause diagnosis was the initial parameter included in the model and it had the largest effect on the age of glaucoma diagnosis (Table 2). Afterward, each HT bin was subsequently included in the regression model, and we found an increased effect size with higher HT use: HT use for 0–2 prescription-years was estimated in the model with a later diagnosis of glaucoma by 2.20 years, HT use for 2–5 prescription-years with a later diagnosis of glaucoma by 3.74 years, and HT use greater than 5 prescription-years by 4.51 years (Table 2). Afterward, we included White and systemic anti-hypertensive use (Supplemental Table S4).

Table 2.

Multivariate Regressions to Predict the Age of Glaucoma Diagnosis With HT use Coded as a Binned Variable

Predictor	Regression Coefficients
Y-intercept	53.90 [53.32, 54.49]
Age of menopause diagnosis	0.70 [0.66, 0.73]
HT usage bin
0–2 years	2.20 [1.64, 2.76]
2–5 years	3.74 [3.02, 4.46]
5+ years	4.51 [3.84, 5.18]
Race: White descent	1.32 [0.85, 1.79]
Systemic antihypertensive medication use	1.82 [1.30, 2.33]
R 2	0.358

Regression coefficients are presented as mean [95% confidence intervals].

To inspect if the effect of HT on the age at diagnosis of glaucoma was different based on racial descent, we created separate linear regressions of age of glaucoma diagnosis versus age of menopause diagnosis in HT and untreated groups for patients of Black or African American descent and patients of White descent (Figure 3). We found significantly lower y-intercepts (age of glaucoma diagnosis) for both the HT and untreated regressions in patients of Black descent compared to the respective groups of the White descent patients (HT Black or African American descent: 58.4 [58.0, 58.9], HT White descent: 59.7 [59.3, 60.2], P < 0.001; untreated Black or African American descent: 55.1 [54.5, 55.7], untreated White descent: 56.6 [56.0, 57.3], P < 0.001). Notably, the slope was steeper in HT patients of Black descent (0.57 [0.49, 0.64]) compared to patients of White descent (0.70 [0.63, 0.76], P = 0.0076). We did not observe a significant difference in the slope for the untreated regressions (untreated Black or African American descent: 0.74 [0.64, 0.84], untreated White descent: 0.69 [0.60, 0.79], P = 0.54).

Figure 3.

Relationship between HT use and glaucoma diagnosis as modulated by self-reported descent. (A) Linear regressions for HT and untreated groups as stratified by self-reported race as signified by line type (solid: Black or African American descent, dashed: White descent). The HT regression slope for the Black or African American descent was 0.70 and 0.57 for HT women of White descent. (B) Age of glaucoma diagnosis for Black/African American descent and White descent divided by age of menopause: early (<45 years of age), normal (45–55 years of age), late (>55 years of age). (C) Percentage of patients in the 0–2-year bin, 2–5 years, and 5+ years of HT use in women of Black or African American and White descent cohorts across all menopause cohorts. As stratified by self-reported descent, the percentage of the HT group for each menopause cohort in each bin was reported followed by the numbers in parentheses.

However, when we performed separate multivariate regressions the main predictive parameters were of the same magnitude for each racial background: age of menopause, systemic anti-hypertensive medication usage, and HT use. While we found no significant differences between these parameters, the y-intercepts were significantly lower in patients of Black or African American descent compared to patients of White descent (P = 0.003; Table 3). Furthermore, we noted a trend that HT use was associated with a later diagnosis of glaucoma in patients of Black or African American descent compared to patients of White descent in each category of HT use (0–2 years, 2–5 years, and >5 years), but was not significantly different.

Table 3.

Multivariate Regressions to Predict the Age of Glaucoma Diagnosis Stratified by Descent

	Black or African American Descent	White Descent	P Value
Y-intercept	53.77 [52.94, 54.60]	55.51 [54.73, 56.28]	0.003
Age of menopause diagnosis	0.64 [0.58, 0.70]	0.71 [0.66, 0.77]	0.07
HT usage bin
0–2 years	2.25 [1.45, 3.06]	1.93 [1.08, 2.78]	0.59
2–5 years	3.63 [2.57, 4.69]	3.76 [2.69, 4.84]	0.86
5+ years	4.80 [3.72, 5.88]	4.08 [3.17, 5.00]	0.32
Systemic antihypertensive medication use	1.95 [1.15, 2.75]	1.59 [0.86, 2.33]	0.52
R 2	0.292	0.360

P values compare coefficients between Black or African American descent to White descent. Regression coefficients are presented as mean [95% confidence intervals].

Because of the changing effect size of HT on the age at diagnosis of glaucoma, we performed a secondary analysis assigning the HT prescription days variable as a continuous variable. This allowed us to evaluate the interaction of the HT duration and the age of menopause with the age at diagnosis of glaucoma (Table 4). Per the regression, each additional prescription-year of HT was associated with a later age at diagnosis of glaucoma by 0.18 [0.14, 0.21] years, and the interaction between age of menopause diagnosis and raw HT use was negative (−0.009 [−0.015, −0.003]), indicating that the delay in age at diagnosis of glaucoma after HT use decreased with later menopause diagnosis. Similar effect sizes were seen for other covariates, including the age of menopause diagnosis.

Table 4.

Multivariate Regressions to Predict the Age of Glaucoma Diagnosis With HT Use as a Continuous Variable and an Interaction Term Between the Age of Menopause and HT Use

Predictor	Regression Coefficients
Y-intercept	55.34 [54.83, 55.85]
Age of menopause diagnosis	0.72 [0.68, 0.76]
HT usage (yr)	0.18 [0.14, 0.21]
Age of menopause diagnosis × HT usage interaction	−0.009 [−0.015, −0.003]
Race: White descent	1.40 [0.93, 1.88]
Systemic antihypertensive medication use	1.86 [1.34, 2.39]
R 2	0.331

Regression coefficients are presented as mean [95% confidence intervals].

This differential effect across menopause group is seen in the average age of glaucoma diagnosis; the difference in glaucoma diagnosis age between the untreated and HT group is 3.6 years in the early menopause cohort (untreated: 49.9 [48.7, 51.1] years, HT: 53.5 [52.7, 54.4] years, P < 0.001), 3.5 years in the normal menopause cohort (untreated: 55.8 [55.2, 56.4] years, HT: 59.3 [58.9, 59.7] years, P < 0.001), and 2.7 years in the late menopause cohort (untreated: 62.5 [61.7, 63.2] years, HT: 65.2 [64.6, 65.7] years, P < 0.001; Figs. 4A–C). In terms of quantile shifts, EH developed glaucoma a median of 3.93 (interquartile range [IQR] = 2.86–4.44) years later than EU (one-sample sign test, P < 0.001), CH 3.71 (IQR = 3.25–3.93) years later than CU (one-sample sign test, P < 0.001), and LH 2.74 (IQR = 2.37–2.93) years later than LU (one-sample sign test, P < 0.001; Fig. 4D).

Figure 4.

Relationship between HT use and glaucoma diagnosis as stratified by age of menopause diagnosis. (A–C) Distribution of age of glaucoma diagnosis by age of menopause with the outline color signifying group (HT: red, untreated: black) and fill color signifying age of menopause: early (green: <45 years of age), normal (blue: 45–55 years of age), and late (purple: >55 years of age). (D) Distributions of percentile differences between untreated and HT groups. All distributions had significant positive shifts: EH-EU (median of 3.93 years), NH-NU (3.71 years), LH-LU (2.74 years). Points and error bars represent mean and 95% confidence intervals and significance are *P < 0.05, **P < 0.01, and ***P < 0.001. EH, early menopause + HT; EU, early menopause + untreated; NH, normal menopause + HT; NU, normal menopause + untreated; LH, late menopause + HT; LU, late menopause + untreated.

Discussion

In this study, we investigated the association between HT use and the diagnosis of glaucoma in postmenopausal women. Our findings indicate that HT use is associated with a later age at diagnosis of glaucoma, with longer durations of HT use corresponding to a later age at diagnosis for this disease.

Our multivariate analysis revealed that the age of menopause was the largest predictor for the age at diagnosis of glaucoma. This was followed by HT use, White descent, and use of antihypertensive medication. Aside from HT use, which was new to the present study, the magnitude and order of coefficients included in our multivariate analysis that were related to the age of glaucoma diagnosis agreed with our previous work.⁹ This helps confirm the association between these factors and the age of glaucoma diagnosis in women and provides additional evidence for the importance of the association between the age at diagnoses of menopause and glaucoma. For example, the age of menopause was the initial parameter included in the model and it had the largest effect on the age of glaucoma diagnosis. Afterward, each HT bin was subsequently included in the regression model and we found an increased effect size with higher HT use. Our multivariate linear regression analysis revealed that HT use for 0–2 prescription-years corresponded to a later glaucoma diagnosis of 2.20 years, while HT use for 2–5 prescription-years and greater than five prescription-years were associated with a later age at diagnosis of glaucoma of 3.74 and 4.51 years, respectively, compared to untreated women. This dose-response relationship was further supported by our secondary analysis, which treated HT duration as a continuous variable and found that each additional prescription-year of HT was associated with a 0.18 years later age of glaucoma diagnosis.

Interestingly, our results suggested that the age of menopause may modulate the protective effect of HT on glaucoma diagnosis. The negative interaction term between the age of menopause diagnosis and HT use in our regression model indicates that the effect of HT on the age of glaucoma diagnosis decreases with the later age at diagnosis of menopause. This finding is consistent with the observed differences in the magnitude of offset in the age of glaucoma diagnosis between HT users and untreated women across the early, normal, and late menopause cohorts, with the largest postponement of glaucoma diagnosis being observed in the early menopause group (3.6 years) and the smallest in the late menopause group (2.7 years).

Our study also explored potential racial differences in the association between HT and glaucoma diagnosis. Although the main predictive parameters (age of menopause, systemic antihypertensive medication usage, and HT use) were consistent across racial backgrounds, we noted a trend suggesting that HT use was associated with a further delay in the age of glaucoma diagnosis in patients of Black or African American descent compared to patients of White descent. This was not statistically significant in our dataset. Previous work by Vajaranant et al.¹⁸ found that only patients of Black or African American descent taking HT had a lower risk of developing glaucoma compared to patients of White descent. Contrary to this result, our data demonstrated that for women who developed glaucoma, HT was associated with a later age at diagnosis of glaucoma in individuals of both Black or African American and White descent, although women of Black or African American descent using HT tended to have a later age at diagnosis of glaucoma.

These findings contribute to the growing body of evidence supporting the protective role of estrogen in glaucoma pathogenesis. Previous clinical studies, such as the Nurses’ Health Study and the analysis by Vajaranant et al.¹⁸ using the WHI dataset, have reported a reduced risk of high-tension OAG in women taking HT. Our study extends these observations by demonstrating that HT use is associated with a later age of glaucoma diagnosis, rather than just a reduced risk of developing the disease. The observed later diagnosis of glaucoma with longer HT durations and the modulating effect of age of menopause suggests that estrogen supplementation may directly influence the pathophysiological processes involved in glaucoma development and progression. This notion is supported by preclinical studies demonstrating that estrogen can improve retinal ganglion cell (RGC) survival and mitigate visual dysfunction after RGC injury.^24,32

It is worth noting that our study has several limitations. First, we did not differentiate between different types of HT formulations or routes of administration, which may have varying effects on glaucoma diagnosis. Previous research using the WHI database has shown that the type of HT formulation may impact the ability to prevent developing glaucoma. For example, Vajaranant et al. found that HT containing estrogen (e.g., oral conjugated equine estrogens) reduced the risk of developing glaucoma in patients of Black or African American descent but HT containing estrogen and progesterone (e.g., oral conjugated equine estrogens and medroxyprogesterone acetate) was not beneficial.¹⁸ Based on this previous report we assessed glaucoma patients taking HT containing estrogen only, and standard HT guidelines state that HT containing estrogen only is prescribed for women without a uterus. Therefore, it would be beneficial to examine other patient cohorts that also received HT containing estrogen and progesterone for an additional analysis comparing these types of HT. These future studies would benefit from having protocols with predefinted time intervals, or follow-up, between eye exams to improve the diagnosis time of glaucoma related to menopause and the influence of HT on the progression of this disease. We also focused our analysis on Veteran females. The VINCI database is advantageous because it has been used for many years, is a national database, and has a diverse population. However, we need to investigate whether these findings apply to the civilian population. Future studies will examine how HT in postmenopausal women is related to the age at diagnosis of glaucoma in national or international databases of non-veterans.

Another limitation is that our retrospective study did not have IOP data on these patients. Both menopause and HT have been shown to contribute to modest changes in IOP (1–3 mm Hg).^10,11 IOP is a major causal risk factor for developing glaucoma, and because HT may lower IOP, it may help delay the onset of the disease. This is important to study, and future prospective studies will incorporate IOP and visual function testing in women with and without HT to observe how HT may impact the onset, progression, and management of glaucoma. Lastly, we performed a case-only retrospective study that did not examine how HT impacted the risk of developing glaucoma and was limited to women who had developed glaucoma. However, this study was able to evaluate the potential benefits of HT use and the associated delay in the age of glaucoma diagnosis. Our separate study showed that the age of menopause was the major predictor of when a woman would develop glaucoma.⁹ This initial analysis showed an association between the age at diagnoses of menopause and glaucoma that was independent of the age a woman became postmenopausal. This study builds on our previous work by demonstrating that HT was associated with a later age at diagnosis of glaucoma regardless of menopause diagnosis age in women who have developed glaucoma. We found an interaction between the age of menopause diagnosis and HT use on the age at diagnosis of glaucoma. The lesser impact of HT in older women may be related to several factors, such as the increased risk of developing glaucoma with age.

These future research studies will aim to further elucidate these mechanisms underlying the potential benefits of estrogen on glaucoma development and progression. This is key given the potential risk of systemic use of hormone therapy, we need continued research to identify potential targets based on these hormonal pathways in the management of this sight-threatening disease, while minimizing the potential risks. Prospective studies can also investigate the impact of different HT formulations, dosages, and routes of administration on glaucoma risk and progression which would be highly informative. Additionally, preclinical studies exploring the molecular and cellular pathways through which estrogen modulates retinal ganglion cell function and survival could provide valuable insights into potential therapeutic targets for glaucoma.

In conclusion, our study provides evidence that hormone therapy use is associated with a later age at diagnosis of glaucoma in postmenopausal women, with longer durations of HT corresponding to an even later age in disease diagnosis. Furthermore, our findings suggest that the protective effect of HT on glaucoma onset may be modulated by the age of menopause and potentially influenced by racial background. These results contribute to the growing understanding of the protective role of estrogen in glaucoma.