Female reproductive factors in relation to risk of exfoliation glaucoma or glaucoma suspect

Abstract

Purpose

Incidence studies have reported higher rates of exfoliation syndrome and glaucoma (XFG) in women, implicating female-specific factors. We evaluated associations between female reproductive factors and incident XFG.

Design

Prospective cohort study of 121,098 females in the Nurses’ Health Study (NHS; 1980–2018) and NHSII (1991–2019).

Methods

Eligible participants were ≥40 years old, postmenopausal, had no history of XFG or cataract extraction, had data on reproductive factors, and reported ophthalmic examinations. Incident XFG or exfoliation glaucoma suspect (XFGS) cases (n=440 affected eyes) were confirmed with medical records. Per-eye Cox proportional hazards models were used to estimate multivariable-adjusted hazard ratios (MVHRs) and 95% confidence intervals (CIs). False discovery rate (FDR) was used to correct for multiple comparisons.

Results

Overall, reproductive history indicative of a greater lifetime estrogen exposure was suggestively associated with higher XFG/XFGS risk. Each 1-year increase in age at menarche was associated with 9% lower risks (MVHR=0.91; 95%CI: 0.85, 0.97; FDR-corrected p-trend=0.047). Parity was not associated, longer durations of oral contraceptive use and breastfeeding were inversely associated, and earlier age at menopause and duration of menopausal hormone therapy was adversely associated, although no FDR-corrected p-trends were significant. Among those with high genetic susceptibility for XFG, surgical menopause (bilateral oophorectomy) was inversely associated with XFG/XFGS compared to natural menopause (MVHR=0.37; 95% CI: 0.15, 0.92; FDR-corrected p-interaction=0.0007).

Conclusion

Later age at menarche was associated with a lower XFG/XFGS risk. Surgical versus natural menopause was inversely associated with XFG/XFGS risk only among those with high genetic susceptibility. These results need confirmation.

Table of Contents

This study describes the results of a prospective cohort of 121,098 females followed for 30 years evaluating female reproductive factors (age at menarche, parity, oral contraceptive use, breastfeeding, type of menopause, menopausal hormone therapy use) in relation to exfoliation glaucoma/glaucoma suspect risk (total 440 eyes with incident disease). Overall, reproductive history consistent with a greater lifetime estrogen exposure in early life, particularly age at menarche, was associated with higher risks.

INTRODUCTION

Exfoliation syndrome (XFS) is characterized by the deposition of fibrillar extracellular substances in tissue beds throughout the body.¹ XFS mainly manifests in the eye, where fibrillar substances are deposited in the conjunctiva, anterior segment, and orbital structures;² such depositions can lead to exfoliation glaucoma (XFG), the most common type of secondary glaucoma.^3,4 Additionally, the associated zonular weakness can cause complications in cataract surgery.¹ While studies have identified certain risk factors, including greater age, genetic risk factors (especially lysyl oxidase like 1 (LOXL1) variants), living away from the equator, and greater ocular ultraviolet (UV) exposure,^5–7 the etiology of this complex systemic condition remains unclear.

Investigators have long debated sex differences in XFS/XFG prevalence.^8–10 Most incidence studies^11–14 evaluating the onset of XFS and XFG, including our previous study,⁵ have reported higher rates in women, suggesting a role for sex hormones. In particular, the role of estrogen on XFG remains unclear. Biologically, 17-β estradiol, the major form of circulating estrogen, upregulates the expression of LOXL1,¹⁵ which is a component of the exfoliation material in XFS that also catalyzes the cross-linking of collagen and elastin, leading to further formation of exfoliation material¹⁶ via the profibrotic transforming growth factor beta 1 (TGF-β1) cytokine.^15,17,18 Estrogen may also have opposing beneficial effects, promoting intraocular pressure (IOP) homeostasis, improved blood flow, and neuroprotection.^19,20 Due to the uncertain role of estrogen in XFG disease and the paucity of studies on female reproductive factors linked to incident XFG/exfoliation glaucoma suspect (XFGS), we evaluated the association between various female reproductive factors and XFG/XFGS risk in a cohort of 121,098 women followed for 30 years.

METHODS

Study Population

The Nurses’ Health Study (NHS) began in 1976 with a cohort of 121,700 U.S. registered female nurses aged 30–55 years. The Nurses’ Health Study II (NHSII) began in 1989 with 116,429 US registered female nurses aged 25–42 years. In both prospective studies, participants were followed with questionnaires sent every two years to update lifestyle and health information, including glaucoma diagnoses. This study was approved by the Institutional Review Board, Human Research Committees of Brigham & Women’s Hospital, Massachusetts Eye and Ear Infirmary, and Harvard T.H. Chan School of Public Health. This study adhered to the tenets of the Declaration of Helsinki.

In this analysis, the study period was 1980 (“baseline”; when incident glaucoma cases started to be accrued) to 2018 in NHS, and 1989 (“baseline”) to 2019 in NHSII. Participants (n=121,098) were eligible to contribute person-time at risk if they: (1) were ≥40 years, (2) were postmenopausal (either through natural menopause or surgical menopause (with bilateral oophorectomy)), (3) never were previously diagnosed with XFG or XFGS or other glaucoma types based on self-reports and medical record review, (4) never had cataract extraction surgery (as detection of exfoliation material can be challenging in pseudophakic eyes), (5) had complete data on reproductive history, and (6) had an eye exam during follow-up (see Figure S1 for study population flowchart). To maximize power, we used the eye as a unit of analysis. Each eligible participant contributed two eye-year units for every year of follow-up. We allowed participants who did not meet baseline eligibility criteria to contribute eye-years of observation starting from the time they met the above eligibility criteria to the study end. We categorized eye-years based on the most recently returned questionnaire and calculated eye-years by summing the time from baseline until the development of XFG or XFGS (henceforth, collectively referred to as “XFG/XFGS”) in an eye, other types of glaucoma or glaucoma suspect, cancer, death, loss to follow-up, or the end of the study (June 2018 for NHS and June 2019 for NHSII).

Assessment of Reproductive Factors

We assessed reproductive factors, including age at menarche, oral contraceptive (OC) use, parity, breastfeeding history, age at menopause, type of menopause (surgical (bilateral oophorectomy) vs. natural), and menopausal hormone therapy (MHT) use. Using questionnaires, we asked participants about age at menarche in 1976 (NHS) and 1989 (NHSII) and the age when menstruation stopped in 1995–2008 (NHS; by 2008 all women had become menopausal) and 1995–2019 (NHSII) biennially. We assessed type of menopause in 1995–2008 in NHS, and 1995–2019 in NHSII. We asked participants to indicate whether their menopause was natural, radiation-induced, or surgically induced. We categorized the type of menopause into natural or surgically induced menopause (bilateral oophorectomy) and excluded women with other types of menopause (e.g., radiation-induced or unclear age at menopause). If participants indicated surgically induced menopause, we asked them to provide the number of ovaries or uterus removed; for women with bilateral oophorectomy, age at the time of surgery was used for age at menopause. We asked about parity (defined as pregnancies lasting at least six months) and age at first birth (AFB) in 1976–1984 and 1995 in NHS and in 1989, 2005, and 2007 in NHSII. We asked about breastfeeding history in 1986 for NHS, and in 1993 and 1996 in NHSII; participants reported the total months breastfed for all children combined.

We asked about OC use status and duration (in years) in 1976–1982 in NHS and in 1989–1995 in NHSII. We defined MHT use as taking hormones following the cessation of menstruation. We asked participants biennially from 1976 in NHS and from 1989 in NHSII about MHT status (never, past, and current) and MHT type currently used (estrogen only, estrogen plus progesterone, and other), and the duration of MHT use (in years). We updated the duration of MHT use biennially by adding two years and one year of additional use for current users and past users, respectively.

Among NHS women in 1978, we validated the self-reports of menopausal status, including age at menopause and type of menopause.²¹ More than 82% of women demonstrated accurate recall of menopausal age on repeated health questionnaires. Moreover, >99% of women had concordant menopausal age between the reported age and the age documented in their medical records. Additionally, those who had undergone surgically induced menopause showed higher accuracy in recalling menopausal age compared to participants who indicated natural menopause.²¹

XFG/XFGS Case Ascertainment

We asked participants biennially about physician-diagnosed glaucoma and the date of diagnosis. If a participant responded “yes” to this question, we asked for permission to contact their eye care provider for additional medical information to confirm the self-report. We asked providers to send all visual field (VF) tests and to either complete a glaucoma questionnaire (that included questions on the presence of exfoliation material, other secondary causes of elevated IOP, optic nerve features, and the status of filtration apparatus) or send medical records. A glaucoma specialist (LRP) systematically reviewed the records for confirmation and classification. Although XFS was the outcome of interest, as we asked participants about “glaucoma” diagnoses (and not XFS), we intentionally chose a case definition of XFG/XFGS, as participants with XFG or XFGS would likely endorse getting a glaucoma diagnosis, while those with XFS only without any glaucomatous characteristics would likely not. Also, to minimize detection bias, we censored participants whose medical records showed that they had exfoliation material without any VF loss, elevated IOP, or abnormal vertical cup-to-disc ratios (CDRs) as of the date of diagnosis because they likely would not endorse a glaucoma diagnosis.

We defined XFG/XFGS as having documentation of exfoliation material plus one of the following three glaucomatous signs in the same affected eye(s): (1) a history of IOP ≥22 mmHg, (2) vertical CDR≥0.6 or the inter-eye difference in CDR≥0.2, or (3) one or more reliable VF tests showing glaucomatous VF loss.

Genotyping, Imputation, and Genetic Risk Scores (GRSs) for XFS

To evaluate whether XFS genetic susceptibility may be an effect modifier, we derived an XFS GRS. We describe the details of the GRS calculation in the Supplementary Materials. Briefly, we used genotypes from existing genome-wide association study (GWAS) data available from five genotyping platforms within the NHS and NHSII (Affymetrix, HumanCoreExome, IlluminaHumanHap, OmniExpress, and OncoArray),^22,23 and we imputed the datasets using the 1000 Genomes phase III release.²⁴ We standardized quality control for genotyping and imputation within and across the platforms, as described previously.^22,23 For the XFS GRS, we used the GWAS by Aung et al.²⁵, which identified eight variants related to XFS in seven genes (LOXL1, CACNA1A, POMP, TMEM136, AGPAT1, SEMA6A, and RBMS3). Based on these variants, we constructed the GRS for XFS; the individual gene variants were coded for the risk allele dosage associated with higher XFS risk based on their imputation score values ranging from zero (for no alleles) up to two (signifying two alleles). We calculated the GRS as the weighted sum of the products of the effect estimate of each gene variant from the aforementioned GWAS and risk allele dosage across the eight loci.

Statistical Analysis

Primary Analysis

To maximize statistical power, we utilized per-eye Cox proportional hazards regression models accounting for inter-eye correlations to estimate multivariable-adjusted hazard ratios (MVHRs) and 95% confidence intervals (CIs). SAS was used for all analyses (v9.3; SAS Institute Inc, Cary, NC, USA). All statistical tests were two-sided with α=0.05. We tested for trends across ordinal categories by including a linear variable representing the median value of each category of reproductive factors; these tests for trends were adjusted for multiple comparisons using false discovery rate (FDR).²⁶

We stratified all analyses by cohort, period at risk, and two strong risk factors for XFG/XFGS: age (in months) and lifetime cumulative residential tier (northern [41–49°N latitude], northern or middle tier [37–49°N latitude], middle tier only [37–40°N latitude], middle or southern tier [24–40°N latitude], southern tier only [24–36°N latitude]). In Model 1, we included all reproductive factors simultaneously. Model 2 was Model 1 with additional adjustment for XFG/XFGS risk factors based on prior literature including family history of glaucoma (yes/no), ancestry (Scandinavian, Southern European, other European, other races), UV flux (Robertson-Berger units, quintiles), latitude (°N, quintiles), time spent outdoors in youth (high school to age 24 years; 1–5, 6–10, ≥11 hours/week), number of eye exams during follow-up (continuous number), self-reported history of hypercholesterolemia, hypertension, diabetes, myocardial infarction (yes/no), body mass index (BMI, kg/m²), cumulatively averaged intakes (quintiles) of total energy (kcal/day), caffeine (mg/day), folate intake (μg/day), alcohol (g/day), pack-years of smoking, and physical activity (metabolic-equivalents of task-hour/week, quartiles).

Secondary Analysis

Given that our cases included both XFGS (XFS with elevated IOP and/or glaucomatous optic disc cupping only) and XFG (XFS with glaucomatous VF loss), we have evaluated these outcomes separately in sensitivity analyses. Also, we conducted sensitivity analyses using a stricter definition for abnormal IOP (≥25 vs. ≥22 mmHg) and CDR (≥0.8 vs. ≥0.6).²⁷ Because reproductive history and exogenous hormone use profiles were strikingly different by surgical (bilateral oophorectomy) versus natural menopause, we conducted analyses separately in groups defined by type of menopause and evaluated effect modification by type of menopause. We also evaluated effect modification by age, family history of glaucoma, latitude, lifetime residential history, and a GRS for XFG/XFGS (among those with genetic data (n=20,281)). We tested for effect modifications by introducing effect modifiers in multivariable-adjusted models and testing the product term of the exposure and the effect modifier in Wald tests. These tests for interactions were adjusted for multiple comparisons using FDR.²⁶

RESULTS

We accrued 3,102,625 eye-years of follow-up from 56,031 women in NHS and 65,067 women in NHSII. When compared to excluded women (Table S1), the included participants were more likely to be White and healthier (lower prevalence of obesity, diabetes, hypertension, and smoking and greater physical activity).

To illustrate the distribution of accumulated eye-years by reproductive factors in relation to various XFG/XFGS risk factors, Table 1 shows the characteristics of those in the extreme categories of age at menarche, parity, age at menopause, and MHT use. For age at menarche, compared to women with menarche at >14 years, women with menarche at <12 years were more likely to be obese, have diabetes, and have a higher prevalence of surgical menopause (bilateral oophorectomy). Compared to women with parity of ≥4, nulliparous women were less likely to have spent a lot of time outdoors in youth, had higher BMI, were more likely to have menopause at age <45 years, and used MHT for longer durations. As expected, women with menopause at age <45 years had a high prevalence of surgical menopause and MHT use. Compared to women who never used MHT, current users had a lower diabetes prevalence, were leaner, had earlier menopause, and had a higher prevalence of surgical menopause.

Table 1:

Characteristics by extreme categories of age at menarche, parity, age at menopause, menopausal hormone use, and type of menopause in the Nurses’ Health Study (NHS) and Nurses’ Health Study II (NHSII) over the entire follow-up period (1980–2019)

	Age at menarche		Parity		Age at menopause		Menopausal hormone therapy use		Type of menopause
	<12 years	>14 years	Nulliparous	≥4 children	<45 years	≥54 years	Never user	Current user	Natural menopause	Surgical menopause
Percentage	23.1	7.3	11.5	20.5	16.5	13.8	37.7	30.5	76.5	23.5
Mean age (SD), years	60.0 (7.9)	61.3 (8.3)	58.3 (7.4)	63.2 (8.3)	56.7 (9.3)	63.6 (6.3)	60.1 (7.3)	58.0 (7.9)	61.3 (7.6)	58.4 (8.9)
Family history of glaucoma, %	17.3	16.2	19.1	14.1	15.8	17.9	16.3	16.6	16.7	16.7
Scandinavian European, %	7.1	8.4	7.8	7.1	7.3	8.3	7.3	8.6	7.9	8.0
Time spent outdoors in youth >11 hours/week, %	6.3	6.4	4.0	8.7	6.6	6.6	5.6	7.8	6.6	6.6
Self-reported diabetes, %	9.1	5.8	8.2	6.2	9.1	7.3	7.7	4.9	6.5	9.0
Self-reported hypertension, %	44.5	35.2	40.5	38.1	44.7	42.0	38.3	38.0	37.7	47.0
Self-reported BMI ≥30, %	22.8	10.0	19.4	14.2	17.3	17.5	18.5	10.4	14.6	19.1
Mean latitude (SD), °N	39.5 (3.7)	39.3 (3.8)	38.8 (3.9)	40.0 (3.4)	39.1 (3.9)	39.5 (3.4)	39.9 (3.4)	38.8 (4.0)	39.5 (3.6)	39.0 (4.0)
Mean age at menarche (SD), years	10.7 (0.6)	15.5 (0.7)	12.4 (1.5)	12.6 (1.4)	12.4 (1.5)	12.6 (1.4)	12.5 (1.4)	12.5 (1.4)	12.5 (1.4)	12.4 (1.4)
Mean parity (SD), count	2.7 (1.3)	2.8 (1.3)	0	4.7 (1.2)	2.6 (1.2)	2.8 (1.3)	2.9 (1.4)	2.7 (1.2)	2.8 (1.3)	2.7 (1.2)
Age at menopause <45 years, %	18.3	16.2	21.4	13.7	100	0	7.1	22.4	8.7	43.3
Surgical menopause (bilateral oophorectomy), %	26.1	22.1	25.4	21.0	66.4	7.5	5.1	39.5	0	100
Mean menopausal hormone therapy use duration (SD), years	4.4 (6.2)	4.1 (6.1)	5.0 (7.2)	3.4 (5.4)	10.8 (9.1)	2.0 (3.5)	0	9.2 (7.4)	2.7 (4.4)	9.9 (8.2)

BMI=Body Mass Index (kg/m2), NHS=Nurses’ Health Study, SD=Standard Deviation

A total of 315 women developed incident XFG/XFGS during follow-up, with 440 eyes affected (Table 2). Approximately 40% of the women were affected with XFG/XFGS bilaterally. Among the 440 incident XFG/XFGS affected eyes, the mean highest untreated IOP was 27.2 mmHg, and the mean vertical CDR was 0.54. Among 440 affected eyes, 35.9% had elevated IOP only, 26.1% had glaucomatous optic disc cupping only, and 38.0% had either glaucomatous optic disc cupping or VF loss.

Table 2:

Clinical characteristics of patients with incident exfoliation glaucoma or glaucoma suspect status and their affected eyes from the Nurses’ Health Study (NHS) and Nurses’ Health Study II (NHSII) over the entire follow-up period (1980–2019)

Demographic features of patients with exfoliation glaucoma / glaucoma suspect status (n = 315)
Mean (SD) age at diagnosis
NHS (n=297)	69.2 (6.8)
NHSII (n=18)	59.5 (5.5)
Laterality
Unilateral
Right Eye Only (%)	92 (29.2%)
Left Eye Only (%)	98 (31.1%)
Bilateral (%)	125
	(39.7%)
Clinical features of eyes with exfoliation glaucoma / glaucoma suspect status (n = 440)
Highest untreated IOP, mmHg, mean (SD)	27.2 (7.2)
Cup-to-disc ratio, mean (SD)	0.54 (0.21)
Categories of exfoliation glaucoma / glaucoma suspect
Exfoliation syndrome with IOP ≥22 mmHg only, n (%)	158
Exfoliation syndrome with optic disc cupping* only (with or without IOP 22 mmHg), n (%)	(35.9%) 115
Exfoliation syndrome with glaucomatous visual field loss, n (%)	(26.1%) 167 (38.0%)

IOP = intraocular pressure; NHS = Nurses’ Health Study; NHSII = Nurses’ Health Study II.

^*Refers to vertical cup-to-disc ratio of 0.60 or inter-eye difference in vertical cup-to-disc ratio≥0.20.

Table 3 presents the model estimates for the association between female reproductive factors in relation to XFG/XFGS. In general, results from the simpler Model 1 were similar to those in the multivariable-adjusted Model 2, indicating minimal confounding; thus, Model 2 results are highlighted. Later age at menarche was associated with a lower XFG/XFGS risk (MVHR_{>14 vs. <12 years}=0.59 (95% CI: 0.38, 0.92); p-trend=0.006; FDR-corrected p-value=0.048), and each additional year of later menarche was associated with a 9% lower risk (MVHR=0.91; 95% CI: 0.85, 0.97). Longer OC use duration was also associated with a lower risk of XFG/XFGS (MVHR_{≥11 years of use vs. never use}=0.64; 95% CI: 0.37, 1.12; p-trend=0.04; FDR-corrected p-value=0.10). Each additional year of OC use was associated with a 4% lower risk (MVHR=0.96; 95% CI: 0.93, 0.997). Nulliparity was associated with a non-significantly higher risk of XFG/XFGS (MVHR=1.46; 95% CI: 0.97, 2.19); however, among parous women, greater parity or higher AFB was not associated with risk (p-trends≥0.55). Longer duration of breastfeeding was inversely associated with XFG/XFGS risk (MVHR_{≥12 vs. 0 months}=0.68; 95% CI: 0.48, 0.97; p-trend=0.05; FDR-corrected p-value=0.10). Surgically induced menopause, relative to natural menopause, was non-significantly inversely associated with XFG/XFGS risk (MVHR=0.75; 95% CI: 0.54, 1.03). With age at menopause, we observed an increased XFG/XFGS risk in women with menopause <45 years compared to those with age at menopause between 50–53 years (MVHR=1.58; 95% CI: 1.12, 2.22; p-trend=0.04; FDR-corrected p-value=0.10). In addition, XFG/XFGS risk did not differ across the 45- to 60-year age range categories for menopause. For MHT use, compared to participants who have never used MHT, past MHT users (MVHR=1.53, 95% CI: 1.15, 2.04) and current users (MVHR=1.48, 95% CI: 1.09, 2.01) both had higher risks for XFG/XFGS. Longer duration of MHT use was associated with a non-significantly higher XFG/XFGS risk (p-trend=0.09; FDR-corrected p-value=0.12), with the most common duration of use, 1–4 years, showing a MVHR of 1.58 (95% CI: 1.22, 2.05).

Table 3:

Univariate and multivariable-adjusted hazard ratios (MVHRs) and 95% confidence intervals of exfoliation glaucoma or glaucoma suspect status by female reproductive factors in per-eye Cox regression models with 440 eyes affected among 121,098 women (followed from 1980 to 2019)

	Eyes affected, n (%)	Model 1a HR (95% CI)	Model 2b HR (95% CI)	P-value	FDR corrected P-value	Per 1-unit increase Model 2b HR (95% CI)
Age at Menarche				Ptrend=0.006	Ptrend=0.048	0.91 (0.85, 0.97)
<12 years	97 (22.1)	1.00 (ref)	1.00 (ref)
12 years	129 (29.5)	1.04 (0.79, 1.36)	0.96 (0.73, 1.27)
13 years	132 (30.1)	0.91 (0.70, 1.19)	0.82 (0.63, 1.08)
14 years	54 (12.3)	0.89 (0.63, 1.25)	0.79 (0.55, 1.12)
>14 years	26 (5.9)	0.68 (0.44, 1.05)	0.59 (0.38, 0.92)
OC Use Duration				Ptrend=0.04	Ptrend=0.10	0.96 (0.93, 0.997)
Never	257 (58.4)	1.00 (ref)	1.00 (ref)
≤2 years	93 (21.1)	1.21 (0.94, 1.56)	1.23 (0.95, 1.61)
3–5 years	38 (8.6)	0.80 (0.56, 1.15)	0.80 (0.55, 1.15)
6–10 years	35 (8.0)	0.84 (0.58, 1.21)	0.88 (0.60, 1.28)
≥11 years	17 (3.9)	0.61 (0.35, 1.05)	0.64 (0.37, 1.12)
Parity				Ptrend=0.73c, Ptrend=0.55d	Ptrend=0.73c, Ptrend=0.63d	Parity: 1.00 (0.93, 1.08) AFB: 0.98 (0.94, 1.01)
Nulliparous	45 (10.3)	1.38 (0.92, 2.06)	1.46 (0.97, 2.19)
AFB <25 years, parity of 1–2	44 (10.1)	0.86 (0.59, 1.25)	0.91 (0.63, 1.33)
AFB ≥25 years, parity of 1–2	88 (20.2)	0.96 (0.72, 1.29)	1.02 (0.75, 1.37)
AFB <25 years, parity of ≥3	143 (32.8)	1.04 (0.80, 1.35)	1.07 (0.82, 1.40)
AFB ≥25 years, parity of ≥3	116 (26.6)	1.00 (ref)	1.00 (ref)
Breastfeeding				Ptrend=0.05	Ptrend=0.10	0.99 (0.97, 0.999)
Nulliparous or never breastfed	184 (44.6)	1.00 (ref)	1.00 (ref)
1–6 months	131 (31.7)	0.84 (0.66, 1.09)	0.77 (0.59, 0.99)
7–11 months	41 (9.9)	0.98 (0.67, 1.41)	0.91 (0.63, 1.32)
≥12 months	57 (13.8)	0.75 (0.54, 1.05)	0.68 (0.48, 0.97)
Type of Menopause				P=0.08	P=0.12	N/A
Natural	356 (80.9)	1.00 (ref)	1.00 (ref)
Surgical (with bilateral oophorectomy	84 (19.1)	0.76 (0.56, 1.03)	0.75 (0.54, 1.03)
Age at Menopause				Ptrend=0.04	Ptrend=0.10	0.98 (0.96, 1.01)
<45 years	69 (15.7)	1.35 (0.98, 1.86)	1.58 (1.12, 2.22)
45–49 years	126 (28.6)	1.11 (0.87, 1.42)	1.22 (0.95, 1.56)
50–53 years	178 (40.5)	1.00 (ref)	1.00 (ref)
≥54 years	67 (15.2)	1.09 (0.81, 1.47)	1.12 (0.83, 1.52)
MHT				Pcurrent vs. never=0.01	Ptrend=0.12	1.02 (0.99, 1.06)
Current Status and Type Used
Never	118 (21.3)	1.00 (ref)	1.00 (ref)
Past	170 (30.6)	1.42 (1.06, 1.90)	1.53 (1.15, 2.04)
Current	137 (24.7)	1.51 (1.13, 2.02)	1.48 (1.09, 2.01)
Current – estrogen only	60 (10.8)	1.42 (0.97, 2.07)	1.43 (0.96, 2.12)
Current – estrogen plus progestin	54 (9.7)	1.58 (1.09, 2.29)	1.58 (1.07, 2.33)
Current – other	16 (2.9)	1.32 (0.75, 2.33)	1.07 (0.58, 1.96)
Duration of MHT Use				Ptrend=0.09
0 years	133 (30.5)	1.00 (ref)	1.00 (ref)
1–4 years	129 (29.6)	1.51 (1.17, 1.94)	1.58 (1.22, 2.05)
5–9 years	65 (14.9)	1.35 (0.99, 1.85)	1.36 (0.99, 1.88)
≥10 years	109 (25.0)	1.34 (0.99, 1.80)	1.31 (0.96, 1.78)

AFB = age at first birth, CI = Confidence Interval, MVHR = Multivariable-adjusted Hazard Ratio; MHT = Menopausal Hormone Therapy; OC = Oral Contraceptive

^aModel 1: multivariable-adjusted analyses were stratified by cohort, age in months, period at risk, and lifetime cumulative residential history (northern [41–49° latitude], northern or middle tier [37–49° latitude], middle tier only [37–40° latitude], middle or southern tier [24–40° latitude], southern tier only [24–36° latitude])

^bModel 2: Model 1 with additional adjustment for ancestry (Scandinavian, southern European, other European, other races), family history of glaucoma (yes/no), self-reported history (yes/no) of hypercholesterolemia, hypertension, diabetes, and myocardial infarction; body mass index (kg/m²; <22, 22–23, 2425, 26–27, 28–29, 30+), cumulatively averaged intakes (quintiles) of total energy (kcal/day), caffeine (mg/day), folate intake (ĝ/day), and alcohol (g/day); pack-years of smoking (0, 1–9, 10–19, 20–29, 30+), UV flux (Robertson-Berger units, quintiles), latitude (°N, quintiles), physical activity (metabolic- equivalents of task-hour/week, quartiles), time spent outdoors in youth (high school to age 24 years; 1–5, 6–10, >11 hours/week) and number of eye exams during follow-up (continuous number) and mutually adjusted for other reproductive/hormonal variables listed in Table 3 and Table 4

^cP for trend for parity trends across categories among parous women only

^dP for trend for AFB trends across categories among parous women only

In sensitivity analyses evaluating associations separately for XFG (with VF loss) and XFGS or analyses using stricter definitions of abnormal IOP and CDR, associations were generally similar to the main analyses, although we had lower statistical power (Table S2). While non-significant, later age at menopause tended to be associated with lower risks of XFG but higher XFGS risks. In secondary analyses evaluating differences in associations by surgical menopause versus natural menopause, we observed that associations among women with natural menopause (Table S3; eyes affected=356) were similar to those observed in women with surgical menopause (bilateral oophorectomy; Table S4; eyes affected=84), except for AFB. The association was null among women with natural menopause, while among women with surgical menopause, a higher AFB was inversely associated with XFG/XFGS risk (FDR corrected p-interaction=0.03). Among other potential effect modifiers, we observed no FDR-significant interactions with family history of glaucoma or latitude. With age, we observed interactions (FDR-corrected p-interaction=0.002) for OC use; null associations were observed for these factors in women ≥60 years, while in women <60 years, longer use of OCs were associated with a lower risk (MVHR_{>5 vs. 0 years OC use}=0.13; 95% CI: 0.03, 0.53). With an XFG/XFGS GRS (among 20,281 women and 97 XFG/XFGS eyes), suggestive effect modification was observed where associations were apparent only among those above the median score, where higher age at menopause was non-significantly associated with a higher XFG/XFGS risk (MVHR_{≥50 vs. 20–49 years}=2.14; 95% CI: 0.94, 4.87; FDR-corrected p-interaction=0.11) and surgical menopause (bilateral oophorectomy) was inversely associated with XFG/XFGS compared to natural menopause (MVHR=0.37; 95% CI: 0.15, 0.92; FDR-corrected p-interaction=0.0005).

DISCUSSION

In this large cohort study of 121,098 women with up to 30 years of follow-up, female reproductive and hormone use histories representative of lifetime lower endogenous estrogens, such as later age of menarche, longer OC use,²⁸ longer breastfeeding duration,²⁹ and surgical menopause, were generally associated with a lower XFG/XFGS risk, although later age at menarche was the only significant factor after correction for multiple comparisons. Interestingly, in a subset of women with a higher genetic susceptibility for XFG/XFGS, surgical menopause, which brings earlier menopause, was associated with lower XFG/XFGS risk. As the role of estrogen in XFG/XFGS etiology has been studied little to date, and as we lacked external validation from an independent study, our results need confirmation and should be interpreted with caution. Previous work examining LOXL1 and estrogen supported the notion that estrogen may increase the risk of developing XFG/XFGS. LOXL1 primarily catalyzes the cross-linking of collagen and elastin, a process important to the development of exfoliation material in XFS.¹⁶ Evidence suggests that LOXL1 expression is significantly upregulated in the early stages of XFS.¹⁶ Studies that examined the effect of 17-β estradiol on mice urogenital tissue found significant upregulation of LOXL1 and TGF-β1.¹⁵ TGF-β1 is a profibrotic cytokine that regulates the LOX family protein expression and is itself upregulated in expression with oxidative stress (e.g., UV exposure or aging). TGF-β1 upregulation is also implicated in the production of other extracellular matrix materials present in exfoliation material,³⁰ which can lead to blockage of the aqueous outflow and elevation in IOP.¹⁸ Thus, we hypothesize that pro-fibrotic stressors such as oxidative stress or UV exposure lead to TGF-β1 and LOXL1 activation, and that high estrogen also upregulates TGF-β1, which further activates LOXL1 expression, leading to the development of exfoliation material. The strong adverse results observed with early life female reproductive factors (i.e., earlier age at menarche, no OC use, lack of breastfeeding or having a later age at menopause in genetically susceptible women) in combination with our prior finding of greater time spent outdoors in youth being associated with nearly a doubling of risk³¹ are consistent with this hypothesis that early life stressors may be exacerbated by a high estrogen milieu in the initial development of exfoliation material. There are parallels in higher estrogen playing an adverse role in the etiology of an ocular condition with Fuchs endothelial corneal dystrophy. It is a female predominant condition associated with higher UV exposure, and lab studies^32,33 have shown that with oxidative stress, estrogen can have a detrimental effect, including the formation of genotoxic estrogen-DNA adducts on human corneal endothelial cells. Further studies of the role of estrogen in XFS etiology, particularly in the presence of stressors such as UV light, are thus warranted.

Our findings with mid- to later-life reproductive histories indicating greater estrogen exposure, such as later age at menopause and longer duration of MHT, were generally null (except current MHT status that showed adverse associations) and were inconsistent with a hypothesis of unfavorable estrogen effects. It may be that in aging eyes, especially in those with XFS, the potential adverse effects of estrogen are counterbalanced by the known beneficial anti-aging effects of estrogen^19,20 on the eye, such as lowering IOP, improving ocular blood flow, and neuroprotection. Indeed in studies from the Blue Mountains Eye Study³⁴ and our prior studies of primary open-angle glaucoma in the NHS,^35,36 factors associated with higher estrogen levels were associated with lower risk, and in sensitivity analyses, although non-significant, later age at menopause showed an inverse association with XFG, while with XFGS, the trend was the opposite. Notably, we observed borderline FDR-significant interactions with the XFS genetic susceptibility score such that among women with higher genetic susceptibility of XFG/XFGS, surgical menopause was significantly inversely associated and a later age at menopause was non-significantly positively associated. These results indicate that in these women with high genetic susceptibility, the putative adverse effects may outweigh any potential benefits of estrogen. The notable higher risk associated with menopause at age 20–44 years was unexpected. There may be pathologic mechanisms (e.g., inflammation, autoimmunity) that are common between XFG/XFGS^37,38 and premature ovarian aging.³⁹ Interestingly, Addison’s disease is a contributor to premature ovarian aging,⁴⁰ and in a metabolomics study, we previously observed that relatively lower levels of cortisone, indicative of adrenal insufficiency, were related to XFG/XFGS.⁴¹

There were several strengths to this study. The large sample size and long duration of follow-up allowed for greater statistical power. The prospective cohort study design and the availability of a rich set of covariates, which have been updated biennially, helped to minimize potential confounding.

However, our study had some limitations. First, there is limited generalizability as our participants were US women, primarily of European descent, with nursing degrees. Our findings are not generalizable to those residing in other latitudes, and to populations with different distributions of race and socioeconomic backgrounds. Second, our glaucoma confirmation method had low sensitivity, and thus, the number of cases will be much higher. To investigate XFG/XFGS, an insidious and uncommon condition, which is often missed by clinicians, we would have had to administer standardized eye exams repeatedly over 30 years on all 121,098 participants who are geographically disbursed across the US; however, this was cost-prohibitive. Therefore, we relied on participants self-reporting physician-diagnosed glaucoma and on collecting and reviewing the medical record information. Importantly, in epidemiologic methodology,⁴² it has been demonstrated that a low disease sensitivity in a cohort study does not cause biases in estimating the relative risk associated with an exposure if, as demonstrated in our study, 1) the outcome is highly specific (e.g., we required XFG/XFGS cases to have documentation of exfoliation material as well as documentation of elevated IOP, abnormal vertical CDR and reliable VFs showing VF loss) and 2) the under-ascertainment is independent of exposure information (as we included only eye-years from women during the years they reported eye exams). Our cohort data have also confirmed established risk factors for XFG/XFGS⁴³ and have contributed to international XFG/XFGS genetic consortia,²⁵ supporting our case ascertainment methods. Third, given our outcome definition of XFG/XFGS, we were unable to evaluate whether reproductive factors are related to the progression from XFS to XFG/XFGS. Fourth, we lacked data on participants’ sex hormone levels, which would have helped provide an objective measure of sex hormones in relation to XFG/XFGS. Future studies on sex hormones in the etiology of XFG/XFGS are warranted to confirm our findings. Fifth, we lacked external validation of our findings from an independent study. Lastly, as our participants began to be followed from 1976 and 1989, the exogenous hormones (OCs, MHTs) used by our participants included older formulations that differed from those currently being used, further limiting the generalizability of our findings.

In conclusion, our findings suggest that reproductive factor profiles consistent with higher endogenous estrogen exposure, particularly in early adulthood, were associated with a higher risk of XFG/XFGS. As there is very limited prior literature evaluating the relationship between female reproductive factors and XFG/XFGS, further studies are needed to confirm our findings.

Supplementary Material

1

Figure S1. Flow chart of study population selection in the Nurses' Health Studies

2

Supplementary Methods. Study population, blood collection, genotyping, imputation, and calculation of genetic risk score (GRS) for exfoliation glaucoma (XFG)

3

Table S1. Characteristics of included and excluded study participants*

4

Table S2: Sensitivity analyses evaluating exfoliation glaucoma and exfoliation glaucoma suspect separately and evaluating a stricter definition of xfoliation glaucoma or glaucoma suspect; multivariable-adjusted hazard ratios (MVHRs) and 95% confidence intervals by female reproductive actors in per-eye Cox regression models among 121,098 women (followed from 1980 to 2019)

5

Table S3: Multivariable-adjusted hazard ratios (95% confidence intervals) of exfoliation glaucoma or glaucoma suspect status by female reproductive factors in per-eye Cox regression models with 356 eyes affected among 97,907 women with natural menopause (followed from 1980 to 2019)

6

Table S4: Multivariable-adjusted hazard ratios (95% confidence intervals) of exfoliation glaucoma or glaucoma suspect status by female reproductive factors in per-eye Cox regression models with 84 eyes affected among 23,558 among women with surgical menopause with bilateral oophorectomy (followed from 1980 to 2019)

Highlights.

• Higher age at menarche was associated with a lower risk of exfoliation glaucoma

• Surgical vs. natural menopause was associated with exfoliation glaucoma risk only among those with a higher genetic susceptibility.

Abbreviations/Acronyms:

AFB

age at first birth

BMI

body mass index

CDR

cup-to-disc ratio

CI

confidence interval

GRS

Genetic Risk Scores

GWAS

Genome-Wide Association Study

IOP

intraocular pressure

LOXL1

lysyl oxidase-like 1

MHT

menopausal hormone therapy

MVHR

multivariable-adjusted hazard ratios

NHS

Nurses’ Health Study

NHSII

Nurses’ Health Study II

OC

oral contraceptives

TGF-β1

transforming growth factor beta 1

UV

ultraviolet

VF

visual field

XFGS

exfoliation glaucoma suspect

XFG

exfoliation glaucoma