Unchanged Prevalence of Preeclampsia After Implementation of Prophylactic Aspirin for All Pregnant Women With Preexisting Diabetes: A Prospective Cohort Study

Preeclampsia occurs in 10–20% of pregnant women with type 1 or type 2 (preexisting) diabetes (1,2), i.e., four times more often than in healthy Danish pregnant women, where preeclampsia occurs in ∼3% (3,4). In women with preexisting diabetes, preeclampsia often develops before 37 weeks, leading to preterm delivery (5,6).

The cause of preeclampsia is still not fully understood. Abnormal placentation with placental ischemia leading to systemic maternal endothelial dysfunction is part of the pathophysiology of preeclampsia (7,8). However, in women with diabetes, preexisting maternal endothelial dysfunction might play a pivotal pathogenetic role (9–11). In the general pregnant population, an imbalance in prostacyclin and thromboxane production has been demonstrated. Low-dose aspirin inhibits the thromboxane production (12) and hypothetically reduces abnormal placentation (13,14). Aspirin to prevent preeclampsia has been studied in randomized controlled trials (RCT), overall showing a moderately beneficial effect (13,15,16) dependent of early initiation and dose (14,15). A possible effect of aspirin on preterm delivery and fetal growth has also been suggested (13,14,17,18). In 2017 the combined multimarker screening and randomised patient treatment with ASpirin for evidence-based PREeclampsia prevention (ASPRE) trial, an RCT with 1,776 women at high risk of delivery with preeclampsia before 37 weeks (preterm preeclampsia), investigators compared aspirin versus placebo initiated before 14 weeks and found a 60% lower prevalence of preterm preeclampsia in the aspirin group (1.6% vs. 4.3%) (19). However, only 25 women with preexisting diabetes were included in the final analysis (19).

In women with preexisting diabetes, studies evaluating the effect of aspirin for preeclampsia prevention are sparse, and the effect is not convincing (17,20–22). However, no RCT with a sufficient dose of aspirin initiated in early pregnancy has been published.

The American Diabetes Association has since January 2018 recommended 60–150 mg/day of prophylactic aspirin initiated at the end of first trimester to all pregnant women with preexisting diabetes (23), followed by a similar recommendation from the American College of Obstetricians and Gynecologists (24). Accordingly, our department changed treatment strategy to prophylactic aspirin for all pregnant women with preexisting diabetes from the previous risk-based aspirin prophylaxis.

We hypothesized that prophylactic aspirin for all pregnant women with preexisting diabetes, according to international guidelines from 2018, would reduce the prevalence of preeclampsia compared with a risk-based prophylaxis in a real-world setting. We aimed to evaluate the prevalence of preeclampsia after implementation of prophylactic aspirin for all pregnant women with preexisting diabetes compared with the prevalence in a previous risk-based prophylaxis.

This study was part of a prospective observational cohort study of consecutive pregnant women with preexisting diabetes initiated in September 2015 and focusing on preeclampsia and use of home blood pressure (BP) assessment (6). In 2018 the study was extended to address the aim of the current study in investigating the prevalence of preeclampsia after implementation of aspirin prophylaxis for all pregnant women with preexisting diabetes in comparison with the prevalence in the first part of the cohort, where risk-based aspirin prophylaxis was given.

Consecutive pregnant women with preexisting diabetes and a single living fetus <20 gestational weeks referred to Center for Pregnant Women with Diabetes, Rigshospitalet, were included from September 2015 to February 2020. Exclusion criteria were age <18 years, insufficient Danish language skills, participation in the study in a previous pregnancy, and severe concomitant diseases. The women were referred from a geographically well-defined region of ∼2.7 million inhabitants. Of 565 women eligible for inclusion, 468 (83%) accepted participation (Fig. 1).

Starting on 23 February 2018, aspirin 150 mg/day was recommended for all women from 10 weeks and continued until 36 weeks (all-cohort) according for the new (2018) international recommendations (23). Prior to 23 February 2018, aspirin was only prescribed to women with the following risk factors (selected-cohort): previous preeclampsia, chronic hypertension, microalbuminuria, or nephropathy or pregnant after oocyte donation. The risk-based aspirin prophylaxis in the selected-cohort was in accordance with local guidelines for pregnant women in general, and women with diabetes without additional risk factors were not prescribed aspirin. Women with risk factors in the selected-cohort were recommended aspirin 75–150 mg/day depending on the current guideline, with the majority receiving 100 mg, from 10 weeks and continued until 36 weeks. If a woman was taking aspirin prior to pregnancy, it was recommended that she continue with the dose recommended for pregnant women with preexisting diabetes in both cohorts. Data regarding aspirin indication, prescription, whether the women filled in a prescription at the pharmacy, possible side effects, and discontinuations were retrieved from the electronic medical records. Adherence to aspirin was confirmed when the women filled in the prescription at least once or from documentation in the records.

All women followed the routine diabetes and pregnancy care program for pregnant women with preexisting diabetes as previously described in detail (6,25) and described in brief here.

Treatment targets for glycemic control changed slightly on 23 February 2018. Self-monitored plasma glucose target levels changed from 4–6 mmol/L (72–108 mg/dL) to 4.0–5.5 mmol/L (72–99 mg/dL) preprandially and from 4–8 mmol/L (72–144 mg/dL) to 4–7 mmol/L (72–126 mg/dL) postprandially. The HbA_1c target levels changed from <6.7% (50 mmol/mol) to <6.5% (48 mmol/L) before 20 weeks and from <5.8% (40 mmol/mol) to <5.6% (38 mmol/mol) after 20 weeks.

At each pregnancy visit, office BP, weight, and glycemic control including HbA1_c were registered, and a sterile urine was screened for proteinuria with a urine dipstick. If the urine dipstick was positive for protein, albumin-to-creatinine ratio (ACR) was measured. At the first visit an ACR was assessed in all women to screen for microalbuminuria (ACR 30–299 mg/g) and diabetic nephropathy (ACR ≥300 mg/g). Home BP was measured twice daily for 3 days at inclusion and, if office BP was ≥135/85 mmHg, during pregnancy (6). Antihypertensive treatment was initiated or intensified if office BP was ≥135 mmHg systolic or ≥85 mmHg diastolic and home BP was ≥130 mmHg systolic or ≥80 mmHg diastolic or if ACR was ≥300 mg/g (6). Antihypertensive treatment could also be initiated based on office BP alone if judged necessary (6).

All women were recommended 10 µg vitamin D supplementation per day. From 23 February 2018, vitamin D level was measured at first pregnancy visit as part of routine (all-cohort), while vitamin D level was not measured in the selected-cohort. Vitamin D insufficiency (serum 25-hydroxy vitamin D <50 nmol/L) was treated with additional vitamin D supplementation according to national recommendations (26). In the selected-cohort, 80 women with type 2 diabetes simultaneously participated in a cohort study where motivational interviewing was routinely used to improve adherence to healthy eating (25). Forty-four (11%) women were pregnant during the coronavirus disease 2019 pandemic, without any testing positive for severe acute respiratory syndrome coronavirus 2. These women followed the routine care program; however, telephone consultations were more commonly used.

Kidney involvement was defined as preexisting or newly detected microalbuminuria or diabetic nephropathy, based on two urine samples when possible. Diabetic microangiopathy was defined as presence of kidney involvement and/or diabetic retinopathy. Chronic hypertension was defined as preexisting hypertension or increased BP above our target for initiation of antihypertensive treatment (office BP ≥135/85 mmHg and home BP ≥130/80 mmHg), detected before 20 weeks (6), without coexisting kidney involvement. Gestational hypertension was defined as office BP ≥140 mmHg systolic and/or ≥90 mmHg diastolic measured on two occasions at least 4 h apart after 20 weeks, and home BP ≥130/80 mmHg when available, and not fulfilling the criteria for preeclampsia. Preeclampsia was diagnosed when hypertension was present with coexistence of either proteinuria ≥1+ on a urine dipstick or new onset of thrombocytopenia (<100 × 10⁹/L), impaired liver function (elevated liver enzymes to twice the normal level), renal insufficiency (serum creatinine >100 µmol/L or a doubling of the serum creatinine concentration in the absence of other renal diseases), pulmonary edema, or cerebral or visual symptoms (27). In women with diabetic nephropathy, a sudden increase in either systolic or diastolic BP ≥15% was additionally required to give the diagnosis of preeclampsia (28). Hemolysis, Elevated Liver enzymes and Low Platelets (HELLP) syndrome was also considered as preeclampsia (27). Gestational age at diagnosis of preeclampsia and gestational hypertension was noted. All medical records were carefully examined to verify the diagnoses of preeclampsia and gestational hypertension, and if doubt existed the cases were evaluated by the senior obstetrician (P.D.). Preterm preeclampsia and early-onset preeclampsia were defined as delivery with preeclampsia before 37 weeks and 34 weeks, respectively. Preterm and early preterm delivery were defined as delivery at <37 and <34 weeks, respectively.

Birth weight z score was calculated to adjust for gestational age and sex (29), and large (LGA) and small (SGA) for gestational age were defined as birth weight >90th percentile and <10th percentile, respectively. Gestational weight gain was calculated from the last weight measured at 36 weeks and self-reported prepregnancy weight (30).

Neonatal outcomes were noted from the records (31), including perinatal mortality (death occurring between 22 weeks and 1 week after delivery), admission to neonatal intensive care unit, and neonatal morbidity defined as occurrence of at least one of the following: major congenital malformations (leading to death, causing significant future handicap, or requiring surgery), neonatal hypoglycemia (plasma glucose <2.2 mmol/L within 4 h after birth), transient tachypnea (continuous positive airway pressure for >60 min), or jaundice (requiring phototherapy).

Categorical data are given as n (%) and compared with χ² test or Fisher exact test. Continuous data are given as mean (SD) or median (interquartile range) and compared using unpaired t test or Mann-Whitney U test depending on whether data were normally distributed or nonnormally distributed, respectively. If data did not follow normal distribution, logarithmic transformation was applied, and if normal distribution was obtained parametric statistics was used. Analyses were made on an intention-to-treat basis unless otherwise stated. A multiple logistic regression analysis was performed with preeclampsia (yes/no) as the outcome variable. Based on theoretical considerations aspirin prophylaxis strategy used (all-cohort/selected-cohort), parity (nulliparous yes/no), HbA_1c at inclusion, diabetic microangiopathy in early pregnancy (yes/no), and chronic hypertension (yes/no) were included as independent variables. The STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) Statement was followed. Due to the nature of the study design, a statistical power analysis was not performed a priori for this study. A two-sided P value <0.05 was considered statistically significant. All statistical analyses were performed with SPSS 25 (IBM Corp., Armonk, NY).

This cohort study was in accordance with the Declaration of Helsinki, and all participants gave written informed consent before the data collection. In addition, the study was approved by The National Committee on Health Research Ethics, Copenhagen, Denmark (H-15019186 and H-15009413), and The Danish Data Protection Agency (2012-58-0004, RH-2015-289, I-Suite: 04305).

In total, 468 women accepted participation. Abortion (n = 26, including one termination shortly after 22 weeks due to ultrasonography-detected severe fetal malformations), changing hospital before delivery (n = 1), or withdrawal of consent (n = 31) led to exclusion of 58 women, leaving 410 women for the final analysis: 207 in the all-cohort and 203 in the selected-cohort (Fig. 1).

Aspirin was prescribed for 88% (182 of 207) of the all-cohort compared with 25% (51 of 203) of the selected-cohort, with confirmed use of aspirin in 78% (161 of 207) vs. 24% (49 of 203), respectively (Table 1). Adherence to aspirin prophylaxis when prescribed was therefore at 88% (161 of 182) for the all-cohort and 96% (49 of 51) for the selected-cohort. The majority of maternal characteristics at inclusion and during pregnancy were comparable between the cohorts (Table 1). However, in the all-cohort fewer women had type 2 diabetes (P = 0.04) and BMI tended to be lower (P = 0.05), while home systolic BP was similar (P = 0.65), although office systolic BP was lower (P = 0.007) (Table 1). Among patients in the all-cohort, 59 (30%) had vitamin D insufficiency, and vitamin D substitution was prescribed for 48 of 59 (81%).

The prevalence of preeclampsia was similar among the two cohorts (12% vs. 11%, P = 0.69, relative risk 1.11 [95% CI 0.65–1.91]). Gestational age at preeclampsia diagnosis and prevalence of preterm and early-onset preeclampsia, preterm delivery, and LGA and SGA infants were also comparable between the cohorts (Table 2). Among the 161 women with confirmed aspirin use as prescribed in the all-cohort versus the 201 women in the selected-cohort who followed the risk-based guideline (either confirmed aspirin use as prescribed [n = 49] or planned no use of aspirin [n = 152]) the prevalence of preeclampsia was similar (14% [23 of 161] vs. 11% [22 of 201], respectively; P = 0.34). Likewise, a subanalysis of women in both cohorts without the risk factors justifying risk-based aspirin prophylaxis was performed, and prevalence of preeclampsia was found for the all-cohort among women with confirmed aspirin use similar to the prevalence among the untreated women in the selected-cohort (13% [16 of 124] vs. 9% [13 of 152], P = 0.24).

In multiple logistic regression analysis, aspirin prophylaxis strategy (all-cohort vs. selected-cohort) was not associated with preeclampsia with adjustment for nulliparity, HbA_1c, diabetic microangiopathy, and chronic hypertension (odds ratio 1.4 [95% CI 0.72–2.8], P = 0.31).

The prevalence of preeclampsia, preterm preeclampsia, early-onset preeclampsia, and SGA infants was similar between the two cohorts with stratification by diabetes type (Table 3). In women with type 1 diabetes, in the all-cohort gestational age at delivery was higher (P = 0.009), preterm delivery tended to be lower (P = 0.06), and early preterm delivery was lower (one vs. six cases, P = 0.03) compared with the selected-cohort. The woman in the all-cohort with early preterm delivery did not use aspirin and developed preeclampsia. Four of the six women in the selected-cohort with early preterm delivery used aspirin (100–150 mg from early pregnancy), and three of these four women developed preeclampsia, as was also the case for one of the remaining two women. In women with type 2 diabetes, pregnancy and neonatal outcomes were similar in the two cohorts apart from a slightly higher prevalence of LGA infants in the all-cohort.

In a stratified analysis the prevalence of preeclampsia was similar between the all-chort and the selected-cohort regardless of HbA_1c being above or below 53 mmol/mol at inclusion: HbA_1c <53 mmol/mol, 10% (16 of 154) vs. 12% (17 of 148), and HbA_1c ≥53 mmol/mol, 17% (9 of 52) vs. 9% (5 of 55) (P = 0.67, Cochran-Mantel-Haenszel). Among 15 women (9% [14 of 161] vs. 2% [1 of 49], symptoms (primarily epistaxis or small skin bleedings) led to short-term or persistent discontinuation of aspirin.

In this prospective observational cohort study of women with preexisting diabetes, implementation of prophylactic aspirin for all pregnant women with diabetes did not reduce the prevalence of preeclampsia compared with the previous risk-based prophylaxis.

Prophylactic aspirin in women at high risk of preeclampsia has been shown to reduce the prevalence of preeclampsia in RCTs (13,15,16,19), but only few women with preexisting diabetes were included. In the previous literature in women with diabetes, an effect of aspirin has not been convincing (17,20–22); however, in these studies an insufficient dose of aspirin (60 mg) may have been initiated too late in pregnancy. For securing optimal effect of aspirin on placentation, aspirin prophylaxis should probably be initiated in early pregnancy at a sufficient dose due to a known dose-response effect of aspirin in pregnancy (14). In our study, 150 mg was initiated in the all-cohort in early pregnancy, as in the ASPRE study (19), increasing the possibility of a clinical effect compared with the selected-cohort where the dose used was primarily 100 mg. Despite this, no decline in the prevalence of preeclampsia in the whole cohort was demonstrated. Likewise, in restriction of the analysis to women with confirmed use of aspirin in the all-cohort and an analysis of women without risk factors justifying risk-based aspirin prophylaxis no decline was demonstrated. With stratification for diabetes type or glycemic control at inclusion the prevalence of preeclampsia was still similar in the two cohorts. Furthermore, in the all-cohort office systolic BP at inclusion was lower and BMI tended to be lower, factors associated with a lower risk of preeclampsia (1,2,5). In addition to the recommendation of aspirin in the all-cohort, recommendations for vitamin D measurement and stricter glycemic goals were implemented at the same time, reflecting real-world conditions. Normal vitamin D level (32) and stricter glycemic control (9) have both been associated with a lower prevalence of preeclampsia. Thus, vitamin D supplementation in case of insufficiency and potential stricter glycemic control would therefore be expected to lower the prevalence of preeclampsia, i.e., work in the same direction as would recommending aspirin prophylaxis for all, and would not explain our findings. Likewise, no decline in preterm preeclampsia was demonstrated, as was expected from the ASPRE study, which demonstrated a 60% lower prevalence of preterm preeclampsia with aspirin use (19). The pathophysiology of preeclampsia involves both abnormal placentation and systemic maternal endothelial dysfunction. Low-dose aspirin inhibits the production of the prothrombotic vasoconstrictor thromboxane (12) and potentially lowers the risk of abnormal placentation and thereby preeclampsia in the general pregnant population (13,14,17). In women with diabetes, endothelial dysfunction is often present (2,9,11) and has been shown to be present before onset of preeclampsia (10). It might be that preexisting endothelial dysfunction, exacerbated by the pregnancy, and not only abnormal placentation leading to endothelial dysfunction (33), plays a pivotal part in the increased risk of preeclampsia in women with diabetes. This might explain why aspirin may have a limited effect when given to all women with diabetes (9). In a real-world setting among women with chronic hypertension, another condition with systemic endothelial dysfunction, a reduction in the prevalence of preeclampsia after implementation of prophylactic aspirin for all women was not documented, in line with our results (34). Likewise, in a secondary analysis of the subgroup of women with chronic hypertension in the ASPRE study an effect of aspirin was not demonstrated (35). In previous RCTs of women at high risk of preeclampsia to investigate the role of aspirin in prevention of preeclampsia in the subgroup of women with preexisting diabetes, an effect of aspirin on these women was not demonstrated (17,20–22). In the ASPRE study, where a sufficient dose of aspirin was initiated in early pregnancy, only a few women with preexisting diabetes were included, and subgroup analysis of these women was not performed (19). In line with these RCTs in women with diabetes (20–22), in our study we question whether aspirin for all pregnant women with diabetes has a beneficial effect on the prevalence of preeclampsia in this high-risk group and call for evaluation in larger RCTs with aspirin initiated in early pregnancy at a sufficient dose. Based on the prevalence of preterm preeclampsia of 7% found in the current study and the fact that aspirin could lower the prevalence of preterm preeclampsia by 60% (thus to 3%) based on the effect found in the ASPRE study (19), >450 women in each treatment arm would be required in an aspirin versus placebo trial in women with diabetes. A multicenter RCT with aspirin initiated early in pregnancy in women with preexisting diabetes with a composite end point representing placental dysfunction is ongoing in Ireland (36) and hopefully will add information about the effect of aspirin prophylaxis. Observational studies, like the present, are important contributions until the results are available.

In this study we found similar prevalence of preeclampsia among women with type 1 and type 2 diabetes, with 11% and 12%, respectively, whereas in previous literature preeclampsia was more common in women with type 1 diabetes (2). The prevalence among women with type 2 diabetes was comparable with that of previous studies (2,37), while the prevalence among women with type 1 diabetes was lower than the prevalence of 15–20% previously described among women with type 1 diabetes (1,2). Women with type 2 diabetes more often have higher BMI, chronic hypertension, and excessive gestational weight gain associated with insulin resistance (37,38), factors known to be associated with preeclampsia. These factors might all contribute to the four times higher prevalence seen in women with type 2 diabetes compared with the prevalence of 3% seen in the general Danish pregnant population (3,4). In general, the women with type 1 diabetes in the study had good glycemic control, and probably a lower prevalence of diabetic microangiopathy than previous cohorts (9), which might contribute to the relatively low prevalence of preeclampsia seen in women with type 1 diabetes compared with previous studies (1,2).

An effect of aspirin prophylaxis on preterm delivery was recently demonstrated in a large RCT in women without diabetes (18). In the women with type 1 diabetes in our study, gestational age at delivery was higher and early preterm delivery was lower in the all-cohort compared with the selected-cohort. However, the majority of women with early preterm deliveries in the selected-cohort were already using aspirin prophylaxis from early pregnancy, and an effect of aspirin prophylaxis on early preterm delivery was therefore less likely. Whether vitamin D supplementation for vitamin D insufficiency or stricter glycemic targets might play a role for this difference in early preterm deliveries remains speculative. Whether aspirin can reduce the prevalence of preterm delivery independent of a reduction in preeclampsia needs to be evaluated in future studies.

The prevalence of SGA infants was lower with aspirin use in pregnant women in general in some but not all studies (13,14,16). In line with a previous study of aspirin prophylaxis including 444 women with diabetes (39), we found no difference in the prevalence of SGA infants between the cohorts, while the prevalence of LGA infants might be increased with aspirin use (39). However, the additional focus on motivational interviewing to improve healthy eating in women with type 2 diabetes in the selected-cohort may have had a positive effect on the prevalence of LGA infants (25).

Strengths of this study are the inclusion of a large number of mainly unselected women with preexisting diabetes from a geographically well-defined area, the careful prospective data collection including glycemic control and home BP, and review of all patient records with focus on preeclampsia and aspirin use. We were able to record whether the aspirin prescriptions were filled, and aspirin use after implementation for all was higher than described in a similar study in women with chronic hypertension (34). Despite the relatively large cohort, the relatively small number of women with preeclampsia increases the risk of a type II error, and the study did not have the statistical power to detect small differences in the prevalence of preeclampsia between the groups. Due to the study design including an extension of an ongoing cohort study, a statistical power analysis was not performed a priori for this study. However, due to the lack of previous studies within this area, the size of this cohort study was judged sufficient to look for clinically relevant differences in a real-world cohort. A limitation is the observational design, not an RCT, causing a possible impact of small changes in other aspects of routine care as illustrated with glycemic control and vitamin D supplementation. Also, tablet count was not possible, and documentation of discontinuation and side effects from aspirin was not preplanned and therefore not systematically noted in the records. The prospective observational design limits the possibility of making any recommendations on the preferential use of prophylactic aspirin in pregnant women with preexisting diabetes (all women or only women at high risk)—until further RCTs are performed. The women in our study had good glycemic control, and the results might not be applicable to the general population of pregnant women with diabetes.

Implementation of prophylactic aspirin for all pregnant women with diabetes did not reduce the prevalence of preeclampsia compared with the previous risk-based prophylaxis in this cohort study. The possible effect of aspirin prophylaxis in women with diabetes on preeclampsia and preterm delivery calls for evaluation in RCTs.

Acknowledgments. The authors kindly thank midwife Maria Anna Mikkelsen and nurse Helle Løvshall for help with collection and handling of data. The authors thank the nurses at Center for Pregnant Women with Diabetes, Rigshospitalet, for help with recruitment.

Funding. N.C.D., B.Á., and E.R.M were funded by Novo Nordisk Foundation (NNF14OC0009275). M.V. was funded by Rigshospitalet’s Research Foundation.

The funding sources had no influence on the study design, handling of data, or writing of the manuscript.

Duality of Interest. D.M.J., L.R., and P.D. participate in clinical studies on the use of insulin in pregnant women with preexisting diabetes in collaboration with Novo Nordisk. B.Á. after the conduct of the study has been employed at Novo Nordisk. E.R.M participates in clinical studies and has served as a consultant for Novo Nordisk in studies focusing on insulin treatment in pregnant women. No other potential conflicts of interest relevant to this article were reported.

For the work by L.R., D.M.J., and P.D. in collaboration with Novo Nordisk, no personal honorarium is involved.

Author Contributions. N.C.D., E.R.M., and P.D. contributed to the idea of the study. N.C.D., M.V., and B.Á. researched data. N.C.D. performed the statistical analysis, contributed to the interpretation and discussion of data, and wrote the first draft of the manuscript. M.V., B.Á., S.K.N., LL.T.A, D.M.J., L.R., P.D., and E.R.M. contributed to the analysis, interpretation, and discussion of data and revised and edited the manuscript. E.R.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 81st Scientific Sessions of the American Diabetes Association, held virtually, 25–29 June 2021, and at the 53rd Diabetic Pregnancy Study Group Meeting, 2–3 September 2021.