Study population

We carried out a historic cohort study of all women in Quebec, Canada, who gave birth at least once in any hospital from 1989 through 2013 (N = 1,108,541), representing 99% of deliveries in the province. We extracted discharge abstracts from the Maintenance and Use of Data for the Study of Hospital Clientele registry, a database that is rigorously validated by the Quebec health ministry for research. Abstracts contain 26 diagnostic and 15 intervention codes documenting any diagnosis or procedure during each admission. We used encrypted health insurance numbers to track women over time from their first delivery to any later inpatient procedures for cataract, with follow-up extending to March 31, 2014. We excluded women without health insurance numbers who could not be tracked, as well as women who delivered <20 weeks of gestation because preeclampsia cannot be diagnosed then.

Preeclampsia

Women with preeclampsia during any delivery were identified using International Classification of Diseases, Ninth Revision ( ICD-9 ) and International Statistical Classification of Diseases, 10th Revision ( ICD-10 ) marked on the delivery discharge abstract (642.3-642.7, O11, O13-O15). The risk of preeclampsia is greatest at the first pregnancy for most women. We expressed preeclampsia as a binary exposure (yes/no), and 2 categorical exposures capturing onset time (early onset <34 weeks of gestation, late onset ≥34 weeks, no preeclampsia) and severity (severe, mild, no preeclampsia). We grouped gestational hypertension in the definition of mild preeclampsia because evidence supports that both are on the spectrum of preeclampsia, and because ICD-10 places the two in the same category. In this study, severe preeclampsia included hypertension with heavy proteinuria (≥3000 mg/24 h), evidence of hemolysis/elevated liver enzymes/low platelets, eclampsia, or preeclampsia superimposed on preexisting hypertension.

We also evaluated preeclampsia combined with metabolic disease, a potential mediator of the preeclampsia-cataract relationship. Women with preeclampsia have increased risk of metabolic diseases, which are also risk factors for cataract. We evaluated the additive effect of preeclampsia and metabolic disease using a variable for: (1) preeclampsia and any metabolic disease, (2) preeclampsia only, (3) metabolic disease only, and (4) no preeclampsia or metabolic disease. Metabolic diseases included hypertension (ICD 401-405, 642.0-642.2, 642.9, 646.2, I10-I15, O10), diabetes (249, 250, E10-E14), obesity (278.0, 649.1, E66), or dyslipidemia (272, E78) diagnosed at delivery or during any later hospitalization.

Cataract

The main outcome measure was any inpatient cataract extraction. We identified women with cataract extractions using procedure codes from the Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures (27.2-27.9) and the Canadian Classification of Health Interventions (1.CL.53-1.CL.55, 1.CL.59-1.CL.89), supplemented with ICD diagnostic codes (366, H25, H26, H28.0-H28.2). Cataract surgery is primarily carried out in hospital in Quebec. We therefore captured the majority of pregnant women who later underwent cataract extraction in the province. However, we did not know whether women had 1 or both eyes affected.

Covariates

We considered several covariates that could confound the relationship between preeclampsia and cataract. These included age at first delivery (continuous, ranging from 12-56 years), parity (1, 2, ≥3 total deliveries), asthma (ICD 493, J45), socioeconomic deprivation (poorest fifth of the population, no, unknown), and the time period at first delivery (1989 through 1993, 1994 through 1998, 1999 through 2003, 2004 through 2008, 2009 through 2013). To rule out the possibility of residual confounding, we accounted for age at first delivery as accurately as possible, using quadratic splines with knots at the 5th, 50th, and 95th percentiles. We accounted for asthma because exposure to inhaled corticosteroids is an independent risk factor for cataract.

Data analysis

We calculated the cumulative incidence of cataract extraction/1000 women with and without preeclampsia, with death as a competing event. We used Cox proportional hazard models to estimate hazard ratios (HR) and 95% confidence intervals (CI) for risk of cataract extraction, comparing women with preeclampsia to those without. For comparability, we used the number of days since the first delivery as the time scale for all women regardless of preeclampsia status. We estimated unadjusted models, and models partially adjusted for age, parity, asthma, socioeconomic deprivation, and time period. Finally, we estimated models fully adjusted for metabolic disease in addition to the covariates in the partially adjusted model. We censored deaths and women who had no cataract extraction by the end of the study: March 31, 2014. We verified the proportional hazards assumption using log (-log survival) plots.

As cataract is primarily a disease of older age, we performed sensitivity analyses in which we restricted the cohort to women who delivered during the first decade of the study (1989 through 1998), for whom we had longer length of follow-up. This group of women is more likely to undergo cataract extraction. We reran models after excluding procedures for cataracts specifically attributed to diabetes, trauma, or drug use. We also reran models after excluding gestational hypertension from the definition of mild preeclampsia, but could only do so for deliveries before 2006 when the ICD-9 was in use.

We carried out the analysis in software (SAS v9.3; SAS Institute Inc, Cary, NC), setting statistical significance to P = .05, with 2-sided hypothesis tests. As the data were deidentified, the institutional review board of the University of Montreal Hospital Center deemed that ethical approval was not required and provided a waiver.

Study population

We carried out a historic cohort study of all women in Quebec, Canada, who gave birth at least once in any hospital from 1989 through 2013 (N = 1,108,541), representing 99% of deliveries in the province. We extracted discharge abstracts from the Maintenance and Use of Data for the Study of Hospital Clientele registry, a database that is rigorously validated by the Quebec health ministry for research. Abstracts contain 26 diagnostic and 15 intervention codes documenting any diagnosis or procedure during each admission. We used encrypted health insurance numbers to track women over time from their first delivery to any later inpatient procedures for cataract, with follow-up extending to March 31, 2014. We excluded women without health insurance numbers who could not be tracked, as well as women who delivered <20 weeks of gestation because preeclampsia cannot be diagnosed then.

Preeclampsia

Women with preeclampsia during any delivery were identified using International Classification of Diseases, Ninth Revision ( ICD-9 ) and International Statistical Classification of Diseases, 10th Revision ( ICD-10 ) marked on the delivery discharge abstract (642.3-642.7, O11, O13-O15). The risk of preeclampsia is greatest at the first pregnancy for most women. We expressed preeclampsia as a binary exposure (yes/no), and 2 categorical exposures capturing onset time (early onset <34 weeks of gestation, late onset ≥34 weeks, no preeclampsia) and severity (severe, mild, no preeclampsia). We grouped gestational hypertension in the definition of mild preeclampsia because evidence supports that both are on the spectrum of preeclampsia, and because ICD-10 places the two in the same category. In this study, severe preeclampsia included hypertension with heavy proteinuria (≥3000 mg/24 h), evidence of hemolysis/elevated liver enzymes/low platelets, eclampsia, or preeclampsia superimposed on preexisting hypertension.

We also evaluated preeclampsia combined with metabolic disease, a potential mediator of the preeclampsia-cataract relationship. Women with preeclampsia have increased risk of metabolic diseases, which are also risk factors for cataract. We evaluated the additive effect of preeclampsia and metabolic disease using a variable for: (1) preeclampsia and any metabolic disease, (2) preeclampsia only, (3) metabolic disease only, and (4) no preeclampsia or metabolic disease. Metabolic diseases included hypertension (ICD 401-405, 642.0-642.2, 642.9, 646.2, I10-I15, O10), diabetes (249, 250, E10-E14), obesity (278.0, 649.1, E66), or dyslipidemia (272, E78) diagnosed at delivery or during any later hospitalization.

Cataract

The main outcome measure was any inpatient cataract extraction. We identified women with cataract extractions using procedure codes from the Canadian Classification of Diagnostic, Therapeutic, and Surgical Procedures (27.2-27.9) and the Canadian Classification of Health Interventions (1.CL.53-1.CL.55, 1.CL.59-1.CL.89), supplemented with ICD diagnostic codes (366, H25, H26, H28.0-H28.2). Cataract surgery is primarily carried out in hospital in Quebec. We therefore captured the majority of pregnant women who later underwent cataract extraction in the province. However, we did not know whether women had 1 or both eyes affected.

Covariates

We considered several covariates that could confound the relationship between preeclampsia and cataract. These included age at first delivery (continuous, ranging from 12-56 years), parity (1, 2, ≥3 total deliveries), asthma (ICD 493, J45), socioeconomic deprivation (poorest fifth of the population, no, unknown), and the time period at first delivery (1989 through 1993, 1994 through 1998, 1999 through 2003, 2004 through 2008, 2009 through 2013). To rule out the possibility of residual confounding, we accounted for age at first delivery as accurately as possible, using quadratic splines with knots at the 5th, 50th, and 95th percentiles. We accounted for asthma because exposure to inhaled corticosteroids is an independent risk factor for cataract.

Data analysis

We calculated the cumulative incidence of cataract extraction/1000 women with and without preeclampsia, with death as a competing event. We used Cox proportional hazard models to estimate hazard ratios (HR) and 95% confidence intervals (CI) for risk of cataract extraction, comparing women with preeclampsia to those without. For comparability, we used the number of days since the first delivery as the time scale for all women regardless of preeclampsia status. We estimated unadjusted models, and models partially adjusted for age, parity, asthma, socioeconomic deprivation, and time period. Finally, we estimated models fully adjusted for metabolic disease in addition to the covariates in the partially adjusted model. We censored deaths and women who had no cataract extraction by the end of the study: March 31, 2014. We verified the proportional hazards assumption using log (-log survival) plots.

As cataract is primarily a disease of older age, we performed sensitivity analyses in which we restricted the cohort to women who delivered during the first decade of the study (1989 through 1998), for whom we had longer length of follow-up. This group of women is more likely to undergo cataract extraction. We reran models after excluding procedures for cataracts specifically attributed to diabetes, trauma, or drug use. We also reran models after excluding gestational hypertension from the definition of mild preeclampsia, but could only do so for deliveries before 2006 when the ICD-9 was in use.

We carried out the analysis in software (SAS v9.3; SAS Institute Inc, Cary, NC), setting statistical significance to P = .05, with 2-sided hypothesis tests. As the data were deidentified, the institutional review board of the University of Montreal Hospital Center deemed that ethical approval was not required and provided a waiver.