Gestational hypertensive disorders and retinal microvasculature: the Generation R Study

This study was embedded in the Generation R Study, a population-based prospective cohort study from early pregnancy onwards in Rotterdam, the Netherlands [10, 11]. The study has been approved by the Medical Ethics Committee of the Erasmus University Medical Center (Erasmus MC), Rotterdam, the Netherlands, and the procedures followed were in accordance with institutional guidelines [12]. Written informed consent was obtained from all participants. For the present study we included women with a live born singleton with available information on GHD and on postnatal follow-up data. Women were excluded from the main analyses when they had a history of chronic hypertension prior to enrollment in the Generation R Study. We also excluded women who were pregnant during follow-up and women without information on retinal vascular calibers during the follow-up visit 6 years after the index pregnancy. The final population for analysis comprised 3391 women (Fig. 1).

The presence of doctor-diagnosed PE or GH was retrieved from hospital charts and was determined on the basis of the former criteria described by the International Society for the Study of Hypertension in Pregnancy of 2001 [13, 14]. GH was defined by a systolic blood pressure ≥140 mmHg or a diastolic blood pressure ≥90 mmHg after 20 weeks of gestation in previously normotensive women. PE was defined as de novo GH with concurrent new onset proteinuria in a random urine sample with no evidence of urinary tract infection [13].

Blood pressure was measured at study enrollment (median 13.2 weeks of gestation [90% range 10.6–17.0 weeks]) and 6 years after the index pregnancy (90% range, 5.7–7.4 years) as the average of two systolic and diastolic blood pressure readings, with the validated Omron 907 automated digital oscillometric sphygmomanometer (OMRON Healthcare Europe B.V., Hoofddorp, the Netherlands) [15].

All participants were in a standardized supine position to prevent differences due to position. A cuff was placed around the right upper arm. In case of an upper arm exceeding 33 cm, a larger cuff (32–42 cm) was used. The mean value of two blood pressure readings over a 5-min interval was documented for each participant. Blood pressure was measured by trained research assistants wearing normal clothing (i.e., no white coat). Mean arterial pressure (MAP) 6 years after pregnancy was calculated as the average systolic blood pressure plus two times the average diastolic blood pressure, divided by 3. Hypertension 6 years after the index pregnancy was defined by the average of two consecutive blood pressure readings, with systolic blood pressure ≥140 mmHg and/or diastolic blood pressure ≥90 mmHg or the use of antihypertensive medication [16]. Using an alternative cut-off, with systolic blood pressure ≥130 mmHg and/or diastolic blood pressure ≥80 mmHg, did not change the direction of our results but attenuated our results to non-significant levels.

At study enrollment maternal height (in centimeters) and weight (in kilograms) without shoes were measured, after which body mass index (BMI) (in kilograms/meter squared) was calculated. Identical measurements were obtained during follow-up 6 years after pregnancy. Pre-pregnancy BMI was established at enrollment through a questionnaire. Pre-pregnancy weight was highly correlated with the measured early pregnancy weight (Pearson’s correlation coefficient r 0.95 [P < 0.001]) [17].

Retinal vascular calibers were assessed by taking digital retinal photographs 6 years after the index pregnancy. Unilateral and unfractionated photographs were taken of the left eye by a Topcon digital retinal camera (model TRC, NW300) while centered on the optic disk. The resolution of the images was set to 4096 and 3072 pixels. Additionally, a semi‐automatic computer‐imaging program was used to measure the six largest retinal arteriolar and venular calibers of the photos. These calibers were located one half to one disk diameter from the optic disk margin [18]. The averages of the six largest retinal arteriolar and retinal venular calibers were then summarized as central retinal arteriolar and central retinal venular equivalents [19]. The semi‐automatic computer‐imaging program that was used for computation of the central retinal arteriolar and central retinal venular equivalents was operated by two graders who were blinded to participants’ characteristics. Grader-specific standard deviation scores (SDSs) were used for both central retinal arteriolar and central retinal venular equivalents. Intraclass correlation coefficients between both graders were excellent for both retinal arteriolar calibers (0.77; 95% confidence interval [CI]: 0.69–0.84) and retinal venular calibers (0.87; 95% CI: 0.81–0.91). This suggests adequate reproducibility.

Information on maternal characteristics during pregnancy, including maternal age, self-reported pre-pregnancy weight, gravidity, parity, ethnicity, educational level, smoking, and chronic hypertension, was available from questionnaires repeatedly applied during pregnancy. Information on gestational age at birth, birth weight, and child sex was obtained from medical records [20, 21]. Six years after the index pregnancy, we obtained information, through questionnaires, on gravidity and parity at follow-up, medication intake, and smoking.

We performed four types of analyses. First, a non-response analysis was performed by comparing subject characteristics between mothers with and mothers without available follow-up data 6 years after the index pregnancy (Additional file 1). Women with available follow-up data, but without information on retinal vascular calibers, were left out of the non-response analysis. Second, to reduce the possibility of potential bias associated with missing data, missing values in covariates were imputed using multiple imputation procedures. Five draws for each missing value were performed providing five substituted data points, which in turn created five completed datasets. Analyses were performed separately on each completed dataset and thereafter combined into one global result [22]. In the total population for analysis, 19.4% had missing information on pre-pregnancy BMI, 2.3% on ethnicity, 6.7% on education, 11.3% on smoking during pregnancy, 29.5% on gravidity at follow-up, 29.9% on smoking during follow-up, 2.3% on blood pressure during follow-up, and 0.2% on medication intake during follow-up. Third, differences in maternal characteristics during pregnancy and follow-up were compared between women with GHD and women with normotensive pregnancies using one-way analysis of variance (ANOVA) for continuous variables and chi-square tests for categorical variables. Fourth, associations between GHD, normotensive pregnancies, and retinal vascular calibers were assessed through linear regression. The linear regression model included covariates selected based on their associations with the outcome of interest based on previous studies or a change in effect estimate of >10% (maternal age at enrollment, ethnicity, educational level at enrollment, smoking during pregnancy, and pre-pregnancy BMI; lastly, when assessing retinal arteriolar caliber, we additionally adjusted for retinal venular caliber and vice versa. GHD are known to increase blood pressure, and blood pressure is known to be associated with smaller retinal arteriolar calibers [23]. To examine the mediating role of mean arterial blood pressure at the time of retinal imaging in the association of GHD with retinal vascular calibers, we analyzed the direct and indirect causal mediation effects through mediation analyses [24]. We used the full model, as was used for linear regression analysis, to adjust for confounding. Statistical analyses were performed using the Statistical Package for the Social Sciences version 21.0 for Windows (SPSS Inc., Chicago, IL, USA) and R version 3.3.2 (R foundation for Statistical Computing, Vienna, Austria [packages ‘foreign’, ‘rms’ and ‘mediation’]) [25].

Several limitations of the present studies need to be discussed. First, we did not obtain retinal vascular imaging from 33.6% of all women who came for follow-up visits 6 years after pregnancy, because retinal vascular imaging was introduced into the Generation R Study after recruitment of study subjects had already started. As this was independent of any subject characteristics, we do not expect any additional selection bias. However, there may be some loss of power due to a smaller sample size available for our analysis and hence larger confidence intervals of the reported associations. Second, compared to non-responders (43%) study participants were on average older at study enrollment, more often primiparous and of European descent, higher educated, more often non-smokers, and more often had GH. This may have led to some degree of selection bias, as the included women were relatively healthy, and may have led to an underestimation of the association between GHD and retinal microvasculature. Third, due to unavailability of pre-pregnancy data on blood pressure (as is also the case in most other studies focusing on pregnant women and cardiovascular outcomes after pregnancy), we cannot exclude the possibility that microvascular changes and hypertension preceded the onset of GHD. However, we compensated for this by excluding women with chronic hypertension. Information on chronic hypertension before pregnancy was obtained through a questionnaire during pregnancy, which was cross-checked with information from the original medical records and the Dutch obstetric database [43]. Our results did not change significantly after we performed a sensitivity analysis excluding women with hypertension in early pregnancy (13.2 weeks of gestation, 95% CI; 11.1–17.0) or women with hypertension at the time of retinal imaging. Hypertension in early pregnancy was defined as a systolic blood pressure ≥140 mmHg and/or a diastolic blood pressure ≥90 mmHg. Hypertension at the time of retinal imaging was defined as the intake of antihypertensive medication and a systolic blood pressure ≥140 mmHg and/or a diastolic blood pressure ≥90 mmHg. Lowering the cut-off for hypertension at the time of retinal imaging (systolic blood pressure ≥130 mmHg and/or diastolic blood pressure ≥80 mmHg) did not change the direction of our results, but did attenuate our results to non-significant levels. Fourth, retinal vascular calibers were not assessed before pregnancy. Therefore, microvascular changes might be due to GHD or might have predated pregnancy. Fifth, the observational nature of this study does not allow for inference of causality and does not preclude the existence of residual confounding. Sixth, information on pregnancies and GHD occurring after the index pregnancy was incomplete. The absence of these data might have affected our results. Finally, our study also has several strengths. First, this is a prospective cohort study from early pregnancy onwards with a large sample size of 3391 participants. Second, retinal images were taken and graded following standardized protocols.