Childhood or adolescent abuse and primary open-angle glaucoma in a longitudinal cohort of women

Abstract

Objective

To examine associations of childhood/adolescence abuse with primary open-angle glaucoma (POAG) and POAG subtypes.

Methods

We included US female nurse participants from the Nurses’ Health Study II (1989–2019). Participants (n = 59,712) were ≥40 years old, reported follow-up eye exams, had no glaucoma, and had early-life adverse experience data. Self-reported childhood/adolescence abuse was assessed in 2001 with the Revised Conflict Tactics Scale (CTS) and the 2-item sexual maltreatment scale of the parent–child CTS. We evaluated associations by type (any, physical, sexual) and timing (childhood, adolescence). POAG cases (n = 255) were confirmed and subtyped with medical record review. Multivariable-adjusted Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs).

Results

Severe childhood/adolescent abuse was reported by 16.5%. There was no association between any childhood/adolescent abuse (HR 1.03; 95% CI: 0.79–1.35), physical abuse, abuse experienced during childhood only or adolescence only, and POAG risk. Compared to no sexual abuse history, any sexual abuse was modestly associated with POAG (HR 1.31; 95% CI: 1.01–1.69). Specifically, we observed adverse associations with sexual abuse for the POAG subtype with paracentral vs. peripheral VF loss (P_{heterogeneity} = 0.04). For paracentral POAG, we observed a 1.80-fold higher risk (95% CI: 1.14–2.85) with any sexual abuse history, a 2.38-fold higher risk (95% CI: 1.25–4.53) with a threat of/actual forced sexual activity, and a significant dose-response relationship with increasing severity of sexual abuse (P_trend = 0.005; P_{trend_FDR corrected} = 0.045).

Conclusions

While any childhood/adolescence abuse history was not associated with POAG, the modest adverse associations with early-life sexual abuse history warrant confirmation in future studies.

Introduction

Glaucoma, an optic neuropathy characterized by progressive degeneration, is a leading cause of irreversible blindness that requires lifelong management [1]. Primary open-angle glaucoma (POAG) is the most common type, affecting 76 million people globally [2]. There is no cure for glaucoma, and treatment is focused on preventing progression [3].

While a glaucoma diagnosis may lead to anxiety/depression [4, 5], evidence suggests that stress itself may play a role in POAG development. Mental stress and stress hormones (e.g., cortisol) are associated with intraocular pressure (IOP) elevation [6–8], a glaucoma risk factor. Stress hormones, pro-inflammatory cytokines, and endothelin-1 released upon autonomic nervous system (ANS) activation influence optic nerve vascular tone and can impair vascular autoregulation, contributing to POAG pathogenesis [9]. Flammer postulated that chronic sympathetic nervous system (SNS) activation can elicit vascular system dysfunction, impairing endothelial cells and promoting vision loss [10]. Subtypes of POAG, including normal-tension glaucoma and POAG with early paracentral visual field (VF) loss, are strongly associated with autonomic dysfunction leading to chronic optic nerve ischemia [11].

Abuse in childhood and adolescence is prevalent and has been linked to multiple chronic diseases [12, 13], but has not been examined in glaucoma. Childhood abuse is associated with elevated plasma cortisol and greater pituitary-driven adrenal and autonomic responses to stress [14]. For adults, the early environment is important in the programming of the hypothalamic-pituitary-adrenal (HPA) axis; thus, the HPA axis dysregulation and hypersensitivity in response to stressors may increase vision loss risk, particularly for genetically susceptible individuals [14].

There is limited literature on emotional stress on glaucoma risk, and studies had small sample sizes, short follow-ups, and retrospective designs [7, 15]. Thus, we investigated childhood or adolescent abuse history and incidence of POAG and POAG subtypes in a 30-year longitudinal cohort study of 59,172 women. We hypothesized that childhood or adolescent abuse experience is associated with higher POAG risk. We evaluated associations by abuse type and timing with the risk of POAG and POAG subtypes defined by IOP and VF loss patterns. To elucidate potential biological mechanisms, we explored effect modification by sociodemographic, psychological, lifestyle, and genetic factors and explored mediation by psychological factors.

Materials and methods

Study design

We conducted a longitudinal cohort study (1989–2019) using data from the Nurses’ Health Study II (NHSII), a cohort of female registered nurses aged 25–42 years in 1989 followed biennially. Institutional Review Board (IRB)/Ethics Committee approval was obtained through Brigham and Women’s Hospital (IRB Protocol number: 2004P002701) and Harvard T.H. Chan School of Public Health (IRB22-1585). Participants were recruited into NHSII from June 1, 1989 to May 31, 1991; during this time, the IRB waived the requirement for formal written consent and allowed participants’ completion of biennial questionnaires to be considered as implied written consent. This study adhered to the tenets of the Declaration of Helsinki.

Inclusion/exclusion criteria were established a priori. Among 116,429 participants, we excluded 263 women with prevalent baseline glaucoma, 1388 women who were lost to follow-up within two years of baseline, 30,484 women who never reported eye exams during follow-up, 25,107 women who were missing childhood abuse data, and 15 women who died before turning 40 years old (Fig. 1). The remaining 59,172 eligible women contributed person-time of follow-up if they were ≥40 years old and reported eye exams. Women were followed until a POAG diagnosis, diagnosis of other glaucoma types, death, loss to follow-up, or end of study (May 2019), whichever came first.

Fig. 1.

Flowchart of study participants.

Measures

Ascertainment of childhood and adolescent abuse

The primary exposure was childhood (<11 years) and adolescent (11–17 years) abuse, assessed in 2001. Physical abuse was assessed using the Revised Conflict Tactics Scale (CTS) [16], which queried both the type and frequency of physical assault. Sexual abuse was measured with the 2-item sexual maltreatment scale of the Parent–Child CTS [16]. Total abuse was determined using a composite score integrating both physical and sexual abuse. Supplementary Methods have information for various definitions of abuse categories.

Ascertainment of POAG and POAG subtypes

The primary outcome was POAG onset. For confirmation of POAG, we requested medical record release permission from participants who self-reported glaucoma diagnoses on biennial questionnaires. From diagnosing eye care providers, we obtained medical records/completed glaucoma questionnaires asking about maximum IOP, optic nerve features (i.e., cup-to-disc ratio (CDR), mean deviation on VFs), filtration apparatus status, and any secondary causes for elevated IOP. Medical information and VFs were reviewed in a standardized manner by a glaucoma specialist (L.R.P.).

Our POAG case definition required documentation of open angles, reproducible VF deficits on  ≥ 2 reliable tests consistent with retinal ganglion cell axonal loss, and the absence of secondary causes of elevated IOP based on a slit lamp examination. The date of diagnosis was defined as the earliest date recorded for IOP ≥ 22 mmHg, vertical CDR ≥ 0.6/vertical CDR asymmetry ≥ 0.2, or reproducible glaucomatous VF loss. POAG cases were confirmed and subtyped by documented highest untreated IOP (high-tension [IOP ≥ 22 mmHg] or normal-tension [IOP < 22 mmHg] glaucoma) and by VF loss pattern (POAG with peripheral or paracentral VF loss based on Humphrey VFs). Specifically, POAG with paracentral VF loss (termed “para-POAG” henceforth) was defined as POAG with the earliest abnormal VF showing either isolated paracentral loss only or paracentral loss combined with VF loss in the Bjerrum and nasal step regions in the same hemifield, but without temporal wedge region loss. POAG with peripheral VF loss (termed “peri-POAG” henceforth) was defined as VF loss in the superior or inferior nasal step, temporal wedge, or Bjerrum regions without loss in the paracentral regions.

Covariates

We included pre-specified covariates that were POAG risk factors in our study [17, 18], and early-life risk factors. Covariate information was ascertained and updated biennially: age, race, glaucoma family history, early-life risk factors (parents’ occupation, whether the mother worked at home, the highest education level attained by either parent, and whether the parents owned a home during infancy, which were proxy indicators for childhood socioeconomic status (SES), and adulthood factors including self-reported number of eye exams, diabetes, body mass index (kg/m²), hypertension, pack-years of smoking, intakes of alcohol (g/day), caffeine (mg/day), and dietary nitrate (mg/day), recreational physical activity (metabolic equivalent task [MET] h/week), age at menopause, postmenopausal hormone use, and adulthood SES. Adulthood SES was derived from summing the z-scores of census tract indicators based on participants’ street addresses (median household income, home value, percentage with college degree, percentage of families with interest or dividends, percentage occupied housing, percentage living in poverty, and percentage white). Diet was assessed with validated semi-quantitative food frequency questionnaires administered every four years from 1991 [17–19]. Missing indicator variables were used if missingness for a covariate was ≥5%, and continuous covariates were imputed to the median if missingness was <5%.

Effect modifiers

Potential effect modifiers included age, glaucoma family history, race, adulthood SES, depression history, and a POAG polygenic risk score (PRS; Supplementary Methods) among a subset of women with genomic data (n = 10,833). Also, for para-POAG, we performed post hoc candidate gene-environment interaction analyses with two single nucleotide polymorphisms (SNPs) previously shown to be associated with para-POAG in women (n = 12,310: rs4236601 (CAV1/2) and n = 12,415: rs11722059 (GUCY1A3/GUCY1B3) (Supplementary methods) [20, 21].

Mediators

Potential mediators included diabetes, body mass index, hypertension, smoking, dietary factors (alcohol, caffeine, dietary nitrate intake), age at menopause, postmenopausal hormone use status, adulthood SES, depression history, and post-traumatic stress disorder (PTSD). Depression was defined as: 1) clinician diagnosis (self-reported biennially since 2003), 2) regular current antidepressant usage, or 3) elevated depressive symptoms using the validated 10-item Center for Epidemiological Studies Depression scale with a threshold of ≥10 [22]. PTSD was assessed with the Brief Trauma Questionnaire in 2008 and the 7-item Short Screening Scale for the Diagnostic and Statistical Manual (DSM) of Mental Misorders, 4th edition (DSM-IV) [23].

Statistical analysis

Primary analysis

We examined age-standardized covariate distributions by abuse categories. We conducted multivariable-adjusted analyses using Cox proportional hazards regression models stratified by age and follow-up cycle to estimate multivariable-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs). We used age as the time scale. Model 1 adjusted for race, number of eye exams, and family history of glaucoma. Model 2 (main model) additionally adjusted for early-life parental factors, diabetes, body mass index, hypertension, pack-years of smoking, intakes of alcohol, caffeine and dietary nitrate, physical activity, age at menopause, postmenopausal hormone use status, and SES. While these adulthood factors could be considered as both confounders and mediators, none were found to be significant mediators of the association based on mediation analyses (see Secondary Analyses section of the Methods) and were therefore included as covariates. Model 3 (mediation model) was Model 2 with additional adjustments for depression history and PTSD. We ran separate models for types of abuse (physical, sexual, total) and for glaucoma subtypes defined by IOP and VF loss pattern. For POAG subtype analyses, we conducted tests for heterogeneity by subtype [24]. When examining associations with physical abuse, we adjusted for sexual abuse and vice versa; when examining associations with childhood abuse, we adjusted for adolescent abuse and vice versa. We evaluated linear trends across abuse categories by testing an ordinal term (e.g., 0, 1, 2). We adjusted for multiple comparisons using the false discovery rate (FDR) method [25].

Secondary analyses

We evaluated effect modification by including interaction terms between abuse type and effect modifiers in Model 2 using Wald tests, and we estimated stratum-specific HRs and 95% CIs. We performed candidate gene-environment interaction analyses to determine whether two para-POAG SNPs modified the associations with multiplicative interactions [20, 21]. We performed mediation analyses to estimate the proportion of the association that may be explained by each mediator. Given the possible recall bias due to the abuse assessment being administered in 2001, we performed sensitivity analyses where we restricted the follow-up from 2001 to 2019. Finally, because abuse is common among women of colour, to minimize residual confounding related to SES and health disparities, we conducted sensitivity analyses excluding these women. All statistical analyses were performed using SAS version 9, and two-tailed comparisons were made with a significance threshold of α = 0.05.

Results

We accrued 1,281,463 person-years of follow-up from 59,172 participants and 255 POAG cases (Fig. 1). The 59,172 differed in baseline characteristics compared to women who were excluded and had no abuse information (n = 48,364) or compared to other excluded women with abuse information (n = 8893); the included women were more likely to be younger and had a lower prevalence of obesity, diabetes, and hypertension. They were also more likely to be White (Supplementary Table S1). Prevalence of severe abuse (Table 1, Supplementary Table S2) during childhood or adolescence was 16.5% and 36.4% for moderate abuse; 8.2% for severe physical abuse, and 10.9% for severe sexual abuse. Women who experienced severe abuse were more likely to be Black or Hispanic. We found higher abuse prevalence among women of lower childhood SES. Women who experienced abuse were more likely to smoke, have comorbidities (diabetes, obesity, hypertension, depression, and PTSD), drink more caffeine, become menopausal earlier, and use postmenopausal hormone therapy. Among 255 POAG cases (Table 2), the mean age at diagnosis was 55.9 ± 6.2 years. The mean IOP was 21.9 ± 5.2 mmHg, mean CDR was 0.65 ± 0.18, and average Mean Deviation was −3.97 ± 4.11 dB. Among POAG cases, 49% were normal-tension glaucoma, 34.2% had para-POAG, and 32.2% had POAG in both eyes. Notably, 220 of 255 cases (86.3%) occurred after 2001, when abuse history was assessed.

Table 1.

Characteristics of accrued person-time during follow-up (1989–2019) among 59,172 women in the NHSII, by total abuse categories.

	None (PY = 445,515; 34.8%)	Mild (PY = 157,814; 12.3%)	Moderate (PY = 466,454; 36.4%)	Severe (PY = 211,680; 16.5%)
Mean age, years	52.0 ± 7.7	52.0 ± 7.7	52.0 ± 7.7	52.1 ± 7.7
Any physical abuse, %	0.0	100.0	74.6	82.1
Any sexual abuse, %	0.0	0.0	56.1	78.5
History of depression, %	33.2	38.4	41.2	52.6
History of PTSDa, %	35.7	43.6	51.0	71.9
Race, %
White	97.1	97.3	95.0	95.2
Black	0.7	0.6	1.7	1.4
Hispanic	1.0	1.2	1.7	2.0
Asian	1.2	0.9	1.6	1.4
Diabetes, %	4.7	4.3	5.3	6.8
BMI ≥ 30 kg/m2, %	15.9	16.1	19.0	21.6
Hypertension, %	25.1	25.0	26.8	29.6
Family history of glaucomaa, %	27.0	27.1	27.2	28.5
Alcohol intake ≥ 15 g/day, %	5.4	6.9	6.5	6.7
Packs per day per year (pack-years) of smoking, %
Non-smoker	71.5	65.0	65.3	56.6
1–9	13.6	15.9	15.8	16.6
≥ 10	14.9	19.1	18.9	26.8
Caffeine intake, mg/day	213.9 ± 173.9	224.4 ± 174.4	224.2 ± 176.4	236.9 ± 184.5
Dietary nitrate intake, mg/day	150.6 ± 75.1	151.9 ± 70.6	151.8 ± 73.7	157.8 ± 77.5
Physical activity, MET-h/day	1.5 ± 1.1	1.6 ± 1.1	1.5 ± 1.1	1.6 ± 1.1
Number of self-reported eye examsb	1.7 ± 0.4	1.7 ± 0.4	1.7 ± 0.5	1.7 ± 0.5
Menopause < 45 yearsc, %	15.9	16.4	17.0	20.5
Postmenopausal hormone current userc, %	26.2	26.9	28.7	29.9
Parents’ highest occupationa, %
Blue-collar worker	53.5	56.9	58.4	59.9
Labourer	9.0	9.6	10.7	12.3
Farmer	7.4	5.3	6.8	6.1
White-collar worker	30.0	28.3	24.1	21.7
Mother worked at home, %	66.1	65.7	64.4	60.8
Parents’ highest education being high school graduate or lower, %	45.9	47.3	50.9	52.6
Parents owned home during infancya, %	56.1	53.8	52.3	50.4
Lowest SES quintiled, %	18.7	17.0	20.5	21.8
Among cases only (n = 255)	n = 85	n = 26	n = 102	n = 42
Mean age at diagnosis, years	55.9 ± 6.2	56.3 ± 5.8	55.8 ± 6.4	56.0 ± 6.0
Mean maximum untreated IOP, mmHg	21.6 ± 4.2	23.5 ± 5.4	21.9 ± 5.5	22.2 ± 6.0
Mean vertical cup-to-disc ratio	0.7 ± 0.2	0.6 ± 0.2	0.6 ± 0.2	0.6 ± 0.2
Mean deviation (MD; dB)	−3.6 ± 3.7	−5.0 ± 4.6	−3.8 ± 3.5	−4.4 ± 5.8
Normal-tension glaucoma (%)	45.2	48.2	51.0	52.9
POAG with paracentral VF loss (%)	34.1	29.1	23.7	39.4
Bilateral POAG with reproducible VF loss in both eyes (%)	27.3	31.5	29.0	52.2

Values are means ± standard deviation (SD) or percentages and are standardized to the age distribution of the study population.

BMI body mass index, IOP intraocular pressure, MD mean deviation, MET metabolic equivalents of task, SD standard deviation, POAG primary open-angle glaucoma, PTSD post-traumatic stress disorder, PY person-years, SES socioeconomic status.

^aAmong non-missing participants.

^bParticipants were asked about eye exams twice during follow-up.

^cAmong postmenopausal women.

^dQuintiles were derived from a score that summed the z-scores of census tract indicators based on participants’ zip codes (median household income, home value, percentage with college degree, percentage of families with interest or dividends, percentage occupied housing, percentage living in poverty, percentage White.

Table 2.

Characteristics of incident POAG cases (n = 255) among 59,172 women in the NHSII followed from 1989 to 2019.

Characteristic	Mean ± SD (range)a
Age at diagnosis, years (n = 255)	55.9 ± 6.2 (41.8–68.8)
Maximum untreated IOP, mmHgb (n = 246)	21.9 ± 5.2 (11.0–46.0)
Vertical CDRb (n = 252)	0.65 ± 0.18 (0.10–0.95)
Mean deviation, dBb (n = 246)	−3.97 ± 4.11 (−27.97–+2.00)
	n (%)
Normal-tension glaucoma	125 (49.0)
POAG with paracentral VF lossc	79 (34.2)
Bilateral POAG with reproducible VF loss in both eyes	82 (32.2)

IOP intraocular pressure, POAG, primary open-angle glaucoma, SD standard deviation.

^aValues are means ± standard deviation (SD) or percentages and are standardized to the age distribution of the study population.

^bData from the affected eye were used for unilateral disease. Data from the worse of the two eyes were used when both eyes were affected.

^cDefined as the earliest abnormal VF showing either isolated paracentral loss only (VF loss in the superior or inferior paracentral zones) or paracentral loss combined with VF loss in the Bjerrum and nasal step regions in the same hemifield, but without any loss in the temporal wedge region; among 231 cases with Humphrey VF (79 (34.2%) had paracentral loss, and 152 (65.8%) had peripheral loss).

There were no associations between total abuse, physical abuse, abuse experienced during childhood only, or abuse experienced during adolescence only with POAG (Fig. 2). However, compared to women with no sexual abuse history, we observed a 37% higher POAG risk for women who reported being sexually touched (HR = 1.37; 95% CI: 1.03–1.81; Model 2, Fig. 2) and 13% non-significant higher risk for women who experienced a threat of/actual forced sexual activity (HR = 1.13; 95% CI: 0.74–1.72; P_trend = 0.15; Model 2, Fig. 2). Any sexual abuse was associated with a significant 1.31-fold higher POAG risk (HR = 1.31; 95% CI: 1.01–1.69; Model 2, Fig. 2). No associations were observed with the number of abuse types (P = 0.47).

Fig. 2.

Adjusted hazard ratios for the association between abuse during childhood or adolescence and any incident primary open-angle glaucoma among 59,172 women in the NHSII followed from 1989 to 2019.

For POAG subtypes defined by IOP, we observed no associations with total, physical, abuse experienced during childhood only, or abuse experienced during adolescence only. However, for sexual abuse, there was a 58% higher risk of high-tension glaucoma among women who reported being touched sexually (HR = 1.51; 95% CI: 1.03–2.23; Model 2, Supplementary Fig. S1), while there was no association with having had a threat of/experienced actual forced sex at least once.

For POAG subtypes by VF loss pattern, we observed an 80% higher para-POAG risk with any sexual abuse (HR = 1.80; 95% CI: 1.14–2.85; Model 2, Fig. 3A); we did not observe any associations with peri-POAG (P_{heterogeneity} = 0.04; Fig. 3B). The association with para-POAG was mainly driven by severe sexual abuse history (HR = 2.38; 95% CI: 1.25–4.53; Model 2, Fig. 3A), and we observed a significant dose-response relationship with increasing sexual abuse severity (P_trend = 0.005; Model 2, Fig. 3A), which remained significant after multiple corrections (P_{trend_FDR corrected} = 0.045). For those who reported sexual abuse in both childhood and adolescence, the association with para-POAG was particularly strong (HR = 4.09; 95% CI: 1.47–11.32; Model 2; n = 23 para-POAG cases).

Fig. 3. Adjusted hazard ratios for the association between sexual abuse during childhood or adolescence and any incident primary open-angle glaucoma subtype defined by visual field loss pattern among 59,172 women in the NHSII followed from 1989 to 2019.

A Primary open-angle glaucoma with paracentral visual field loss. B Primary open-angle glaucoma with peripheral visual field loss.

In sensitivity analyses restricting follow-up from 2001 to 2019 (n= 220 incident POAG cases), the findings were similar to our main results (Supplementary Table S3). The association between any sexual abuse and para-POAG showed an HR of 1.74 (95% CI: 1.07–2.85; Model 2, Supplementary Table S3). Additionally, excluding racial/ethnic minorities (n = 56,818; 240 POAG cases), results were similar for a threat of/actual forced sex and para-POAG (HR = 2.19; 95% CI: 1.09–4.29; P_trend = 0.03).

We observed no effect modification by age, race, family history of glaucoma, SES, PRS, or depression for the association between abuse types and any POAG (P_interaction ≥ 0.06).

In post hoc candidate gene-environment interaction analyses (n = 12,415 with rs11722059 data and 12,310 with rs4236601 data; n = 18 para-POAG cases), we observed no interaction with the GUCY1A3/GUCY1B3 rs11722059 risk A allele (P_interaction = 0.31). However, we found a non-significant interaction between sexual abuse and the CAV1/2 rs4236601 risk A allele for para-POAG (P_interaction = 0.09): in non-carriers, no associations were observed (HR = 0.84; 95% CI: 0.09–7.63), while in carriers, there was a significant adverse association (HR = 9.73; 95% CI: 1.15–81.92), although CIs were wide.

Comparing Models 2 and 3 (in Figs. 2 and 3A, B) for mediation analyses, we observed that 8.00% (95% CI: 2.00–27.20%, P = 0.04) of the association of sexual abuse on para-POAG risk could be statistically explained by depression history, 10.40% (95% CI: 2.30–36.40%, P = 0.055) by PTSD history, and 15.90% (95% CI: 4.30–44.40%, P = 0.02) by PTSD and depression combined. None of the other variables were significant mediators of the association between abuse and POAG risk. In Model 3, additionally adjusted for depression and PTSD, the association between any sexual abuse and para-POAG was attenuated yet significant (HR = 1.74; 95% CI: 1.08–2.79; Model 3, Fig. 3A). For all POAG, adjustment for these mediators attenuated associations with any sexual abuse (HR = 1.26; 95% CI: 0.96–1.63; Model 3, Fig. 2).

Discussion

We observed no associations between total abuse or physical abuse with POAG. In exploratory analyses, we observed that compared to women with no history, women with childhood/adolescent sexual abuse history had a higher risk of POAG, particularly para-POAG. As this was one of the first studies to evaluate this association, and we evaluated multiple outcomes, additional studies are needed to confirm these findings.

Childhood/adolescent abuse has been associated with negative health outcomes in later life [26]. Early-life stressors may lead to increased psychological and physiological stress sensitivity in early life, disrupting the ANS functioning, potentially inducing chronic stress and SNS activation [27]. With SNS activation or depression [28], stress hormones are released, which have been associated with IOP elevation and glaucoma [6, 7, 15, 29]. Stress triggers pro-inflammatory cytokine and endothelin-1 release, which may influence vascular tone [9]. Chronic psychological stress and elevated cortisol may lead to endothelial dysfunction, impairing flow-mediated dilation and increasing markers of endothelial damage [30]. In the eye, such chronic SNS activation could impair vascular autoregulation in the optic nerve, such as in Flammer syndrome, contributing to glaucoma pathogenesis [9–11]. Systemic dysfunction of vascular autonomic control may be important, particularly for para-POAG [31], which is consistent with the strongest associations observed for sexual abuse history and para-POAG. We observed significant adverse associations among CAV1/2 risk variant carriers while observing null associations in non-carriers; CAV1/2 codes for caveolins important in endothelial function, providing biological support for our findings.

The specificity of the association with sexual versus physical abuse for para-POAG was notable. Compared to physical abuse, sexual abuse history has shown stronger adverse associations with more long-term health outcomes [32]. Several biological pathways may underlie the sexual abuse association, which may be mediated by depression and PTSD. Sexual abuse specifically has been linked to epigenetic changes associated with later depression, which in turn may increase POAG risk. For example, hypermethylation of the monoamine oxidase A first exon region, coding for enzymes involved in metabolizing serotonin and dopamine, was associated with sexual but not physical abuse, and mediated the sexual abuse and depression relation [33]. Also, DNA methylation of serotonin transporter (SLC6A4) genemediated the effect of sexual abuse on depression [34]. Furthermore, depression and chronic stress may lead to lower gamma-aminobutyric acid (GABA) levels and dysregulation of brain GABAergic signalling [35]; such changes may also affect GABAergic pathways in the eye, particularly the trabecular meshwork and retina [36]. Evidence suggests that sexual abuse, more than physical abuse, may dysregulate GABA signalling [37]. Further studies into the biological mechanisms of the relation between sexual abuse and glaucoma are warranted.

Our study has several strengths. We used a large cohort with biennially updated information on covariates and glaucoma diagnoses with 30 years of follow-up. The prevalence of childhood sexual abuse we observed was consistent with that from prior studies [38–40]. POAG cases had reproducible glaucomatous VF loss confirmed by a standardized review of medical information, minimizing outcome misclassification. We adjusted for several confounders and performed mediation, stratified, POAG subtype, and gene-environment interaction analyses.

There are limitations to this work. Our participants were predominately white female nurses; thus, our findings may not be generalizable to males, non-whites, and those of different socioeconomic backgrounds. In particular, our incident cases were relatively younger (mean age of 55.9 years) and had high proportions of normal-tension POAG or para-POAG. Ideally, to investigate POAG, an insidious relatively rare condition, we would have needed to conduct standardized eye exams on all participants (59,197 participants) repeatedly over 30 years; however, this was cost-prohibitive, and we needed to rely on participants first self-reporting physician-diagnosed glaucoma and second, collecting and reviewing the medical records, which is a case ascertainment method of low sensitivity. However, in epidemiologic methodology [41], it has been demonstrated that a low sensitivity for disease identification does not cause biases in estimating the relative risk related to an exposure if the outcome is highly specific (e.g., we required POAG cases to have two reliable VFs showing reproducible loss) and the under-ascertainment is independent of exposure information (as we included only person-time from those reporting eye exams). Our cohort data has confirmed established risk factors for POAG and has contributed to multiple international efforts to define the genetic architecture for POAG, supporting our case ascertainment. Additionally, the possibility of recall bias and exposure misclassification due to retrospective assessment of childhood or adolescent abuse cannot be excluded. However, previous studies reported high reliability in adults’ recall of exposure to childhood abuse [42]. We also performed a sensitivity analysis by restricting follow-up to 2001–2019 and observed similar findings, showing the robustness of our findings. Also, given the multiple subgroup and subtype analyses, our subgroup results may be due to chance, and the findings of secondary analyses need confirmation. While we were able to consider some markers of parental SES and childhood exposures, residual or unmeasured confounding is still possible as we lacked data on other childhood and parental stressors (i.e., food insecurity, housing instability, and parental income) and intra-uterine factors. For example, intimate partner violence (IPV) during pregnancy is associated with later prenatal care, missed prenatal appointments, poorer health behaviours during pregnancy, and adverse birth outcomes, such as preterm delivery and low birthweight [43]. Maternal exposure to IPV is associated with a greater risk of child maltreatment or abuse [44] and low birthweight due to IPV may contribute to the risk of POAG, such as through a higher IOP, CDR, or altered ocular organ development [45–47].

Overall, any childhood or adolescent abuse history was not associated with POAG. The modest adverse associations observed with POAG, specifically for early-life sexual abuse in relation to para-POAG, warrant confirmation in future studies.

Summary

What was known before

• While the role of emotional stress has been studied mainly after glaucoma diagnoses, important evidence suggests that stress itself may contribute to POAG development. The role of early life adverse experiences has been little studied in glaucoma risk.

What this study adds

• Our 30-year study of 59,172 women may be one of the first cohort studies to evaluate the association between childhood/adolescence abuse history in relation to risk of POAG.

Author contributions

Megan Yu—data analysis, draft of manuscript and proofreading. Hannah H. Hwang—draft of manuscript and proofreading. Andrea L. Roberts—proofreading. Karestan C Koenen—proofreading. Janey L. Wiggs—proofreading. Louis R. Pasquale—data collection, proofreading. Jae H. Kang—concept, overseeing each stage of work, data collection, data analysis, draft of manuscript, and proofreading.

Funding

This work was supported by the National Institutes of Health U01 CA176726, R01EY036460 (LRP, JHK), R01EY032559 (LRP, JLW). LRP is also supported by The Glaucoma Foundation (NYC) and an unrestricted Challenge Grant from Research to Prevent Blindness (NYC). The funders had no role in the: design or conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.

Data availability

Because of participant confidentiality and privacy concerns, data cannot be shared publicly and requests to access NHSII data must be submitted in writing. According to standard controlled access procedures, applications to use NHSII resources will be reviewed by our External Collaborations Committee to verify that the proposed use maintains the protection of the privacy of participants and the confidentiality of the data. Investigators wishing to use NHSII data are asked to submit a brief description of the proposed project (go to https://www.nurseshealthstudy.org/researchers (contact email: nhsaccess@channing.harvard.edu) for details.

Competing interests

All authors have declared no conflicts of interest. Unrelated to this work: JLW is a consultant for Allergan, Editas, Maze, Regenxbio, and Avellino. JLW also received research support from Aerpio Pharmaceuticals. LRP was a consultant for Twenty Twenty.

Supplementary information

The online version contains supplementary material available at 10.1038/s41433-025-03785-3.