Hypertensive Disorders of Pregnancy Increases Risk of Developing Neovascular Age-related Macular Degeneration in Later Life

Abstract

Background.

The link between hypertensive disorders of pregnancy (HDP) and short-term adverse outcomes has long been recognized. However, there is now evidence that survivors of HDP remain at risk of adverse biological events long after giving birth, including ocular complications. We sought to determine whether HDP is associated with development of the choroidal neovascular (CNV) form of age-related macular degeneration (AMD), a leading cause of blindness.

Methods.

We identified women who experienced HDP (preeclampsia/eclampsia, HELLP syndrome, and gestational hypertension) as recorded in Utah birth certificates from 1939-2013. Exposed subjects were matched 1:2 to unexposed women on: mother’s age at delivery, and on child’s birth year, sex, and birth order. An AMD outcome from 1992-2015 was determined from electronic medical records. Risk of CNV AMD was estimated using stratified Cox regression with a competing risk of death.

Findings.

We included 31,454 HDP-exposed and 62,908 unexposed women for study. Of 681 exposed and 1,355 unexposed with subsequent AMD, 158 and 284 developed choroidal neovascularization, respectively. Women with HDP exhibited an 80% higher risk for early-onset CNV AMD before age 70 [adjusted hazard ratio (aHR) 1.8, 95%CI 1.23-2.58; p<0.002] and were 20% more likely to develop CNV AMD at any age (aHR 1.2, 95%CI 1.01-1.42; p=0.035) compared to unexposed. HDP was not associated with non-CNV AMD.

Conclusion.

Women with HDP have an increased risk for developing CNV AMD compared to unexposed women, particularly for earlier-onset disease. Our findings may have important implications for earlier screening and detection of choroidal neovascularization.

Summary statement:

Do women with a history of hypertensive disorders of pregnancy have a higher incidence of neovascular age-related macular degeneration (AMD) later in life? In a population-based study of 94,362 women, those who experienced hypertensive disorders of pregnancy (HDP) were 80% more likely to develop early-onset choroidal neovascularization AMD.

Introduction

Pregnancy-related maternal deaths – primarily from the cardiovascular complication of preeclampsia and eclampsia – are unacceptably high worldwide, even in developed countries(1–3). The association between hypertensive disorders of pregnancy (HDP) and short-term adverse outcomes has long been recognized. However, there is now substantial evidence that survivors of HDP remain at risk of adverse events long after giving birth including ocular complications and the subsequent development of chronic diseases(4–7). While the majority of visual disturbances and ophthalmic exam findings associated with preeclampsia resolve completely following delivery, there are reports of persistent ophthalmic changes for several months postpartum(8–11) and recent population-based studies have established an association between preeclampsia and the development of both diabetic and non-diabetic retinopathy, as well as retinal detachment later in life(12–14). To date, an association between HDP and the subsequent development of age-related macular degeneration has not been reported, despite changes in choroidal vasculature on exam among women with HDP(15–17). Thus we sought to determine whether women with a history of HDP have increased risk for the subsequent development of choroidal neovascular (“wet”) age-related macular degeneration, a leading cause of blindness worldwide(18), in a large population-based cohort.

Methods

Study setting.

The Utah Population Database (UPDB) is a research resource housed at the University of Utah that contains computerized data records for over nine million deceased and living individuals who formerly or currently reside in the state(19). The UPDB was originally created in the 1970’s from genealogic data of 1.6 million Utah settlers and their descendants who were primarily of Northern European descent and was later expanded to incorporate statewide birth and death certificates from the early 1900’s. The UPDB was further enriched by linking of electronic medical records (EMR) beginning in 1992 of 2 million patients seen in the University of Utah Health statewide system of hospitals and clinics, including the John Moran Eye Center and its satellite facilities, as well as Medicare claims data for 275,000 Utah beneficiaries enrolled in parts A and B. These data are routinely updated and probabilistic record-linking is performed with individuals in UPDB as described(20). Approvals to conduct this human-subjects study was received from the University of Utah Institutional Review Board (IRB 00010201) and the Resource for Genetic Epidemiologic Research, the body that governs the use of UPDB data.

Study design and participants.

A retrospective cohort design was used in our investigation. Exposed participants were women born in 1900-1975 with a history of HDP recorded on a Utah birth certificate (1939-2013), including: preeclampsia/eclampsia (PE), hemolysis, elevated liver enzymes, and low platelets (HELLP) syndrome, and gestational hypertension (GHTN). Prior to 1989, the birth certificate included free-text fields for complications of pregnancy in which an HDP event could be recorded and later converted to electronically available data. In 1990 and later, the birth certificate included check boxes for complications of pregnancy including HDP. Example formats of Utah birth certificates from 1939 to current are included in an online supplement.

An unexposed comparison group was comprised of women with no evidence of HDP on a Utah birth certificate, randomly selected from the Utah population and individually matched 2:1 to the exposed women based on: mother’s age at child’s birth (<21y, 21-25y, 26-30y, 31-35y, 36-40y, >40y); birth year and sex of the child; and, the child’s birth order (1^st, 2^nd, 3^rd, 4^th, 5^th or higher). Women who had >1 pregnancy affected by HDP were included once for the single-most severe event to the least severe event as follows: eclampsia, HELLP, preeclampsia, and GHTN. If a woman had multiple HDP-affected births of equal severity, the earliest occurrence of a preterm birth was retained; if all births were term, the event associated with the earliest birth was retained.

Exposed and unexposed women met the following criteria for study inclusion: a) singleton live birth in Utah between 1939 and 2013; b) at least one year of follow-up in Utah based on UPDB records; c) alive on or after January 1, 1992; d) age 40y or older on December 31, 2012 or if deceased, death occurred in Utah; and, e) no prior history of age-related macular degeneration (AMD) at the time of the child’s birth. Of 1,796,341 singleton birth certificates evaluated, 94,362 women met inclusion criteria: 31,454 case subjects and 62,908 unaffected matched population controls.

Outcome.

The outcome of interest was a subsequent diagnosis of choroidal neovascular age-related macular degeneration (CNV AMD) occurring between 1992-2015 in patient electronic medical records, linked to individuals in the UPDB, from: (a) all Utah inpatient and ambulatory facilities discharges; (b) the University of Utah Healthcare’s Moran Eye Center and statewide affiliated clinics; and, (c) Utah Medicare beneficiary claims data. A diagnosis of CNV AMD was based on International Classification of Diseases, Revision 9 (ICD-9) code 362.52 (exudative senile macular degeneration) appearing in the patient record in any diagnostic position. In a large Veterans Administration database, Latkany, et al. performed a validation study by chart review of exudative AMD and reported accuracy for ICD-9 code 362.52 of 98%(21). For comparison, a non-CNV AMD outcome was defined as any diagnosis of ICD-9 code 362.51 (non-exudative senile macular degeneration, the “dry” form of AMD) with no CNV AMD diagnosis in the patient record. In a review of academic medical center records, Muir and colleagues reported 92% of non-exudative AMD encounters were coded correctly(22). In either case of CNV or non-CNV AMD, a diagnosis of unspecified AMD (ICD-9 362.50) or drusen AMD (ICD-9 362.57) may have also appeared in the diagnostic record.

Statistical analyses.

In HDP-exposed and unexposed women, time was measured from the birth date of a child until: study end, December 31, 2015; date of last follow-up or death, if prior to year-end 2015; or an index date of a CNV AMD outcome, at which time observations were right-censored. The estimated risk of CNV AMD (or non-CNV AMD, in the comparison scenario) in exposed compared with unexposed women was determined from the adjusted hazard ratio (aHR) using a Cox proportional hazards model. Analogous to a conditional logistic regression, separate hazard functions (stratified Cox regressions) were incorporated for each group: an exposed subject with HDP (one or more PE, HELLP, and/or GHTN event) and two individually-matched unexposed subjects to account for mother’s age at birth, child’s birth year and sex, and birth order of the child. As AMD is typically a later-onset disorder, a competing-risks framework was incorporated to include death as a competing risk(23).

Factors previously associated with HDP(24–26) were examined for inclusion as model covariates: mother’s race/ethnicity (as white and non-Hispanic, white and Hispanic, or non-white); body mass index (BMI) from mother’s pre-pregnancy weight categorized as underweight (<18.5), normal weight (reference; 18.5-24.9), overweight (25-29.9), obese (30 and over), or unknown; history of diabetes (type I or II); history of chronic hypertension; parity (one child, two children, 3 children, or 4 or more children born to a mother); mother’s highest level of education (less than high school, high-school graduate, post-high school education or college graduate, or not reported); child’s birthweight categorized as very-low (<1500kg), low (1500 to <2500kg), normal (reference; 2500-4500kg), high (>4500kg), or not reported; child’s gestational age as very early pre-term (<29 weeks), early pre-term (29 to <33 weeks), pre-term (33 to 36 weeks), term (reference; >36 to <42 weeks), and post-term (42 or more weeks); and county of residence at time of birth as primarily urban (reference), mixed urban/rural, primarily rural, and outside of Utah. A stepwise model (P<0.25 to enter or stay) was used to select a set of covariates that included: mother’s race/ethnicity and pre-pregnancy BMI, parity, child’s gestational age, and county of residence. An unadjusted model was also estimated. Study analyses were conducted using SAS®, version 9.4 (SAS Institute).

Results

Characteristics of the study population and the distribution of HDP and AMD outcomes among 94,362 women (31,454 exposed to HDP and 62,908 matched unexposed) are shown in Tables 1 and 2, respectively. Mothers exposed to HDP were more likely to be overweight or obese prior to pregnancy, have a history of diabetes or hypertension, and to have fewer children over their reproductive lives than were mothers with no HDP exposure (Table 1). Study follow-up times were lengthy (generally >20y). Unexposed women were less likely to be right censored due to death than HDP exposed (8.7% and 7.8%, respectively). Although women with HDP may be predisposed to earlier death, our study required subjects to be living at the time electronic medical records became available in 1992. Therefore, the most severe cases of HDP resulting in early mortality were not included. Children of exposed and unexposed mothers did not differ on matching variables of child’s sex, birth year, and birth order in an unmatched comparison (Table 1). Children born to mothers with HDP were more likely to be early pre-term (<33 weeks gestation) and low birth weight (<2500 kg) compared to their respective matched children born to unexposed mothers.

Table 1.

Characteristics of the study population for hypertensive disorders of pregnancy (HDP) exposed and unexposed women.

Characteristic	HDP exposed		2:1 unexposed
	N	%	N	%	P*
Total births	31,454	100.0	62,908	100.0	–
Characteristics of mother:
Age at child’s birth (mean, ±SD)	27.6	±6.38	27.4	±6.36	<0.001
12 - 21y	6,012	19.1	12,716	20.2
22 - 26y	9,005	28.6	17,860	28.4
27 - 33y	10,123	32.2	20,288	32.3
34 - 55y	6,314	20.1	12,044	19.1	0.22
Birth year of mother
1900-1949	5,027	16.0	10,030	15.9
1950-1965	11,452	36.4	22,711	36.1
1966-1975	14,975	47.6	30,167	48.0	0.45
Race/ethnicity of mother
White and non-Hispanic	26,906	85.5	54,054	85.9
White and Hispanic	2,644	8.4	5,217	8.3
Non-white	1,904	6.1	3,637	5.8	0.07
Pre-pregnancy body mass index†
Underweight, <18.5	760	2.4	3,251	5.2
Normal, 18.5-24.9	12,568	39.9	35,960	57.1
Overweight, 25.0-29.9	8,333	26.5	13,494	21.5
Obese, ≥30.0	8,830	28.1	8,560	13.6
Unavailable	963	3.1	1,643	2.6	<0.001
History of type I or II diabetes
No	29,453	93.6	60,331	95.9
Yes	2,001	6.4	2,577	4.1	<0.001
History of hypertension
No	30,732	97.7	62,688	99.7
Yes	722	2.3	220	0.3	<0.001
Maximum number of children
1 child	3,290	10.5	5,512	8.8
2 children	7,671	24.4	13,929	22.1
3 children	7,597	24.2	14,706	23.4
4 or more children	12,896	41.0	28,761	45.7	<0.001
Highest level of education
Less than high school graduate	494	1.6	799	1.3
High school graduate	2,541	8.1	5,035	8.0
More than high school graduate	24,746	78.7	49,714	79.0
Not reported	3,673	11.7	7,360	11.7	0.56
County of residence
Urban counties	26,521	84.3	54,236	86.2
Mixed urban/rural	1,245	4.0	2,415	3.8
Rural	3,435	10.9	5,769	9.2
Outside of Utah	253	0.8	488	0.8	0.28
Vital status
Alive at last follow up	28,922	92.2	57,466	91.3
Died prior to study end/diagnosis	2,462	7.8	5,442	8.7	<0.001
Years of follow up (mean, ±SD)	23.2	±13.87	23.9	±13.54	<0.001
Characteristics of child:
Female	15,037	47.8	30,072	47.8
Male	16,417	52.2	32,836	52.2	0.99‡
Birth year of child
1939-1979	5,750	18.3	11,504	18.3
1980-1999	18,310	58.2	36,623	58.2
2000-2013	7,394	23.5	14,781	23.5	0.97‡
Birth order of child
1st	16,480	52.4	32,953	52.4
2nd	5,472	17.4	10,951	17.4
3rd or higher	9,502	30.2	19,004	30.2	0.99‡
Birth weight
Very low (<1500 gm)	1,044	3.3	420	0.7
Low (1500 to <2500 gm)	4,181	13.3	2,523	4.0
Normal (2500 to 4500 gm)	25,311	80.5	58,216	92.5
High (>4500 gm)	388	1.2	674	1.1
Not reported	530	1.7	1,075	1.7	<0.001
Gestation (weeks)
Very early preterm (<29 wk)	1,244	4.0	744	1.2
Early preterm (29 to <33 wk)	4,597	14.6	3,368	5.4
Preterm (33 to <37 wk)	1,160	3.7	3,078	4.9
Term (37 to 41 wk)	23,715	75.4	53,989	85.8
Post-term (>41 wk)	738	2.3	1,729	2.7	<0.001

^*Mantel-Haenszel chi-square test of association, categorical variables; paired t-test, continuous variables.

^†Weight in kilograms divided by height in meters squared.

^‡Matching variable of exposed to unexposed subjects in the study design.

Table 2.

Distribution of hypertensive disorders of pregnancy (HDP) events and age-related macular degeneration (AMD) outcomes.

Events and outcomes	HDP exposed women		2:1 unexposed women
	N	%	N	%	P*
Total number of HDP events:	36,182	100.0	—	–
Eclampsia	849	2.3	—	–
HELLP	369	1.0	—	–
Preeclampsia	16,053	44.4	—	–
Pregnancy-induced hypertension (GHTN)	18,911	52.3	—	–
Total number of births:	31,454	100.0	62,908	100.0
Single HDP event at birth:	26,842	85.3	—	–
Eclampsia	781	2.5	—	–
HELLP	144	0.5	—	–
Preeclampsia	11,555	36.7	—	–
GHTN	14,362	45.7	—	–
Multiple HDP events at birth:	4,612	14.7	—	–
Eclampsia and: HELLP or preeclampsia	32	0.1	—	–
Eclampsia and GHTN	34	0.1	—	–
HELLP and: preeclampsia or GHTN	104	0.3	—	–
Preeclampsia and GHTN	4,326	13.8	—	–
Total number of AMD outcomes:	681	100.0	1,355	100.0
CNV AMD	158	23.2	284	21.0
Non-CNV AMD	310	45.5	650	48.0
Unspecified if CNV or non-CNV	213	31.3	421	31.1	0.44
Age at index CNV AMD (mean, ±SD)	74.7	8.01	75.8	7.65	0.28
Under 70 years	41	25.9	51	18.0
70 years or older	117	74.1	233	82.0	0.005
Age at index non-CNV AMD (mean, ±SD)	75.3	7.39	75.5	7.53	0.69
Under 70 years	70	22.6	130	20.0
70 years or older	240	77.4	520	80.0	0.40
Years from HDP to CNV AMD (mean, ±SD)	45.7	8.76	46.8	8.52	0.22

ABBREVIATIONS: HELLP, hemolysis, elevated liver enzymes, and low platelets. GHTN, gestational hypertension. CNV, Choroidal neovascularization.

^*Mantel-Haenszel chi-square test of association, categorical variables; paired t-test, continuous variables.

In HDP case births to exposed mothers, GHTN was the most common event (46%), followed by preeclampsia (37%) and having both GHTN and preeclampsia (14%; see Table 2). In most instances (85%), a single HDP was recorded in the birth certificate while 15% had multiple hypertensive diagnoses noted. Of 681 exposed and 1,355 unexposed women with subsequent AMD, 158 (23.2%) and 284 (21.0%) were diagnosed with choroidal neovascularization, respectively (Table 2). Our data suggest women with HDP were significantly more likely to be diagnosed with CNV AMD at an earlier age (under 70 years vs. 70 years or older) than their unexposed counterparts (Table 2). Given AMD is a later-onset disorder, few children of mothers with an HDP event had an AMD diagnosis themselves; 32 total in 94,362 births.

The risk of a mother developing a subsequent CNV AMD diagnosis among the HDP exposed compared to unexposed women (covariate-adjusted model) is shown in Figure 1. Women exposed to HDP were 20% more likely to develop CNV AMD compared to mothers who did not have HDP (aHR=1.20, 95%CI 1.01-1.42; p=0.035). For the non-CNV AMD phenotype, we did not observe a significant difference in risk between HDP-exposed and unexposed mothers. The risk of earlier-onset CNV AMD (diagnosed before age 70) appeared to be of greater magnitude, 80% higher in mothers exposed to HDP compared to unexposed (aHR=1.78, 95%CI 1.23-2.58; p<0.002). Results unadjusted for covariates (other than accounting for individual matching of unexposed to exposed subjects) are shown in eFigure 1 (available online).

Figure 1.

Mothers exposed to hypertensive disorders of pregnancy (preeclampsia/eclampsia, HELLP, and GHTN) and risk of choroidal neovascularization age-related macular degeneration (CNV AMD) or non-CNV AMD compared to unexposed mothers, adjusted for mother’s race/ethnicity, pre-pregnancy body mass index, parity, county of residence, and child’s gestational age.

As eclampsia, HELLP, and preeclampsia (with or without GHTN in the background) arguably represent a more severe phenotype than GHTN only, we examined risk of subsequent AMD after excluding HDP births in which the only exposure recorded on the birth certificate was GHTN. Our findings for women exposed to eclampsia, HELLP, or preeclampsia (and not GHTN exclusively) compared to respective unexposed women were consistent with HDP (Figure 1), and inclusion or exclusion of GHTN-only births did not appear to impact risk of AMD (see online Supplement, eFigure 2).

Discussion

In a statewide population investigation, we observed a greater risk of CNV AMD (particularly for onset before age 70) in women who had eclampsia, HELLP, preeclampsia, GHTN, or multiple HDP events from Utah birth certificates compared to unexposed women after controlling for the birth year, mother’s age at birth, and child’s sex and birth order, and adjusting for mother’s race/ethnicity, obesity, parity, and urban/rural residence. While a recently published Canadian cohort study(12) reported a higher cumulative incidence of retinal diseases (traction and serous detachments, retinal breaks, and diabetic retinopathy) among women with a history of preeclampsia compared to women without a history of preeclampsia, the authors included AMD in a composite ‘other retinal diseases’ outcome and did not report their findings for AMD as an individual outcome. Our study is the first large, population-based report of a CNV AMD and HDP association to our knowledge.

An association between CNV AMD and HDP is plausible, given CNV is characterized by growth of abnormal vessels from the choroidal circulation into the sub-RPE or subretinal space, and HDP develops following superficial placentation in the setting of abnormal placental angiogenesis(26). Like many other complex diseases, genetics play a role in the pathogenesis of both CNV and HDP: the STOX1 gene within the 10q22 locus has been evaluated as a candidate ‘preeclampsia gene’(27, 28), and variants of the HTRA1 gene within the chromosome 10q26 locus(29) and of the CFH, CFHR1 and CFHR3 genes on chromosome 1 are associated with increased risk and/or protection for developing AMD(30, 31). The genetic basis of AMD is well-established, and our study has access to genotype information for AMD risk loci in over 12,000 patients who are linked to the UPDB. We plan to identify a subset of women in our cohort who have AMD risk genotypes and, if sufficiently powered, examine HDP and risk of chromosome 1-driven and chromosome 10-driven AMD in a background of risk or no risk alleles at chromosome 10 and 1, respectively, and in non-risk AMD genotypes to inform any shared etiology.

These findings have potential clinical relevance both for ophthalmologists and for obstetricians as AMD is the leading cause of adult-onset blindness in the United States(32), and visual impairment has significant effects on functional status and quality of life among older adults(33–35). Although there is no known effective treatment for the non-exudative form of AMD characterized by late-stage geographic atrophy, CNV AMD is amenable to intervention with intravitreous injection of a vascular endothelial growth factor inhibitor in order to limit progression of CNV AMD and stabilize or reverse vision loss(36). Early recognition and treatment of CNV AMD increases the likelihood of visual recovery(37). If the findings of this study are confirmed by others, ophthalmologists may be advised to obtain pregnancy histories on their midlife and older female patients. Likewise, pregnancy has been clearly shown to be a window to women’s future health, and obstetricians will increasingly be called on to make appropriate postpartum referrals for women at increased risk for various chronic diseases. Women whose pregnancies are complicated by HDP have increased risks for cardiovascular disease and diabetes, and should be followed by their primary care providers for screening and management. Thus, our study suggests regular ophthalmology evaluations may be added to these preventive health recommendations for women with a history of HDP.

Strengths of our approach include ascertainment of HDP events from Utah birth certificates spanning several decades and linkage to statewide AMD diagnoses. As a majority of women in Utah are Caucasians primarily of northern European descent, our results may not be generalizable to non-Caucasian populations. We acknowledge HDP recorded on the birth certificates may contain inaccurate diagnoses, particularly as the criteria for diagnosing subtypes of HDP have changed over time. Although previous validation studies of large administrative databases have demonstrated that billing codes captured a clinical diagnosis of AMD with a high degree of accuracy(21, 22), the outcomes of CNV AMD or non-CNV AMD subtypes were based on ICD-9 diagnosis codes and not a clinical diagnosis of AMD, thus misclassification error may be present. We acknowledge that some women with non-CNV AMD in 1992-2015 may have developed the exudative form of AMD after that time. Given that the Moran Eye Center’s system of hospitals and clinics provides comprehensive eye services by experienced ophthalmologists throughout Utah representing more than 325,000 patient visits, we believe any bias from misclassification is minimal.

Women have an increased risk of developing cardiovascular disease after pre-eclampsia based on systematic reviews and a meta-analysis of several studies(38, 39) and may be at risk of cataract extraction later in life(40). In a meta-analysis including several prospective studies, previous cataract surgery is strongly associated with late AMD and to a modest degree, cardiovascular disease as well(41). As data on previous cataract surgery was not available to our study and comprehensively assessing cardiovascular disease in our cohort was beyond the scope of this project, we were limited in our ability to analyze these intermediate risk factors in relation to HDP exposure and CNV outcome. However, as a competing risk of death was incorporated in our analysis, we accounted for severe cardiovascular morbidity. As our study is retrospective, it is descriptive in nature and well-designed prospective cohort studies are needed to confirm our findings; however, given the long lead times between a pregnancy complicated by HDP and onset of AMD, often 50 or more years, a prospective design is problematic.

Conclusions

Women exposed to hypertensive disorders of pregnancy (HDP) have an increased risk of developing the CNV phenotype of AMD, with highest risk for onset before age 70y compared to women without an HDP history. Because early recognition of CNV AMD increases the likelihood of visual recovery with treatment, our findings will likely have important clinical implications.