Maternal hypertensive disorders in SLE pregnancy and future cardiovascular outcomes

Abstract

Background:

Hypertensive disorders of pregnancy (HDP) increase cardiovascular disease (CVD) risk. Pregnancy morbidities, including preeclampsia, and CVD are common in systemic lupus erythematosus (SLE). Possible connections are important to explore. In a population-based cohort, we investigated whether HDP is associated with a higher risk of cardiovascular outcomes separately in SLE and non-SLE to examine the role of SLE.

Methods:

We identified first singleton births in the Medical Birth Register (1987–2012) among mothers with SLE and a large general population comparison group. Discharge diagnoses for HDP, cardiovascular outcomes, and hypertension in the Patient Register were identified using ICD codes. We estimated adjusted hazard ratios and 95% confidence intervals (HR, 95% CI) of the association between HDP and outcomes, in separate models in women with and without SLE. We then evaluated additive and multiplicative effect modification using relative excess risk due to interaction and Cox models jointly accounting for SLE and HDP, respectively. Mediation analysis estimated the proportion of the association between SLE and outcome explained by HDP.

Results:

HDP were more common in SLE pregnancies (20% vs 7%). In SLE, HDP were associated with a two-fold higher rate of cardiovascular outcomes and three-fold higher rate of incident hypertension. HDP mediated 20% of the latter association. In women without SLE, HDP was associated with higher hypertension incidence later in life.

Conclusion:

In women with and without SLE, HDP were associated with a three-fold higher rate of hypertension. In SLE, women with HDP developed cardiovascular outcomes twice as often as women without HDP.

Introduction

Hypertensive disorders of pregnancy (HDP), including preeclampsia, are associated with diffuse endothelial dysfunction, irreversible vascular damage, and increased arterial stiffness. Pre-gestational vascular health, including hypertension, contributes to preeclampsia risk.[1] HDP broadly appears to increase cardiovascular disease (CVD) incidence in women, and preeclampsia, specifically, has been associated with a higher incidence of cardiovascular events, including stroke later in life.[2]–[4] Whether pregnancy is a stress test that unmasks endothelial vulnerability which manifests as HDP or whether the maternal hypertensive disorder itself causes damage leading to CVD is unknown.[5]

Women with systemic lupus erythematosus (SLE) are at increased risk of CVD, including stroke.[6] Hypertension and systemic inflammatory disease, such as SLE, are well established risk factors for preeclampsia.[1], [7] The relationship between HDP and future CVD is complex and may differ between women with and without SLE. In a population-based longitudinal cohort we investigated whether HDP is associated with a higher incidence of cardiovascular disease and incident hypertension later in life in women with SLE compared with women without SLE.

Methods

Study population

We identified all singleton pregnancies in the Swedish Medical Birth Register (MBR) among women with prevalent SLE and women from the general population comparator (without SLE) from a population-based Swedish register linkage (SLINK; 1987–2012).[8] Briefly, SLINK identified each individual with SLE in the National Patient Register and randomly selected five individuals from the general population matching on sex, birth year, calendar time, and county of residence. The MBR includes almost all births at 22 completed gestational weeks or later in Sweden starting July 2008 (and at a minimum of 28 weeks before that). SLE was defined as ≥2 SLE ICD-coded visits with at least 1 code before pregnancy from a specialist who typically manages or diagnoses SLE (rheumatology, internal medicine, nephrology, dermatology). Women from SLINK without SLE were included as non-SLE comparators. The primary analysis restricted to first singleton pregnancies among nulliparous women for comparability with other studies.[2], [4] Missing maternal identification number and women with a pre-gestational history of the cardiovascular outcomes were also excluded. Furthermore, women with pre-gestational hypertension were excluded only for analyses of incident hypertension.

Hypertensive Disorders of Pregnancy

Discharge diagnoses for HDP (including preeclampsia) in the National Patient Register (NPR) were identified using ICD codes (Supplementary Table 1). HDP exposure was based on the first pregnancy. Sensitivity analyses examining subsequent pregnancies used HDP as a time-dependent covariate such that once a woman received an HDP diagnosis, future person-time was classified as exposed.

Cardiovascular Outcomes

Incident outcomes included fatal and non-fatal acute myocardial infarction (AMI), fatal and non-fatal stroke (ischemic, hemorrhagic, subarachnoid hemorrhage, or unspecified), transient ischemic attacks (TIA), unstable angina, and heart failure in the NPR, Swedish Stroke Register (Riks-Stroke), and Cause of Death Register (Supp Table 1). In addition to this primary composite outcome, we evaluated AMI and stroke separately. A secondary outcome, incident hypertension, was evaluated among those with no history of pre-gestational hypertension. We required ≥1 visit for hypertension or ≥2 dispensations of antihypertensive medication ≥3 months postpartum (Supp Table 1). We performed two additional sensitivity analyses with this secondary outcome. First, we required at least one ICD code for hypertension at least three months postpartum to reduce potential misclassification associated with antihypertensives prescribed to treat renal disease in SLE patients with nephritis. Second, we assessed whether antihypertensives prescribed to treat HDP in a future pregnancy might be misclassified as our outcome of incident hypertension.

Covariates

We extracted maternal age at delivery, maternal country of birth (Nordic vs. Non-Nordic), and completed years of education at delivery (≤9, 10–12, ≥13 years, or missing) from the MBR. Maternal body mass index (BMI) was derived using height and weight from the first antenatal visit and categorized according to the WHO classification. Maternal smoking during the first trimester (yes, no, or missing) was self-reported in the MBR. Pregestational hypertension, diabetes, renal disease, and deep venous thrombosis and pulmonary embolism were defined as ≥1 discharge diagnosis from inpatient and outpatient care or, as appropriate, ≥1 related medication dispensation from Swedish pharmacies from the Prescribed Drug Register (Supp Table 1).

Follow-up

Individuals contributed person-time from first delivery date (1987 onwards) until the first of: outcome, death, emigration, or end of study (Dec 31, 2016). Emigration data were available through 2013. Sensitivity analyses restricting until the end of 2013 with and without emigration data for censoring yielded nearly identical results. Therefore, we incorporated available emigration data and assumed that more recent missing emigration data were unlikely to bias the results.

Statistical analysis

Maternal characteristics including medication use and pregestational comorbidities were compared by SLE and HDP jointly. Cox models estimated hazard ratios and 95% confidence intervals (HR, 95% CI) of the association between HDP and CV outcomes separately for SLE and non-SLE pregnancies. They were adjusted for confounders and stratified by calendar year (1987–1994, 1995–2002, and 2003–2012).

We present results on effect measure modification using the approach recommended by Knol and VanderWeele.[9] To investigate effect measure modification on the multiplicative scale, we estimated the ratio of the effect of SLE within strata of HDP (ratio of hazard ratios) using an adjusted Cox model. We also assessed interaction on the additive scale by estimating the relative excess risk due to interaction (RERI) using the formula ${HR}_{{SLE}^{+}{HDP}^{+}}–{HR}_{{SLE}^{+}{HDP}^{-}}–{HR}_{{SLE}^{-}{HDP}^{+}}+1$.

Mediation analysis estimated the proportion of the association between SLE and CV outcomes that is explained by HDP in first pregnancy. In other words, we suspect that there are at least two ways in which SLE increases the risk of CV outcomes, and one of those paths is via HDP. Total effects were decomposed into direct and indirect effects using a Cox model for the outcome and logistic regression for the mediator.[10] Proportion mediated was calculated using the formula: $\frac{{HR}_{NDE}\times \left({HR}_{NIE}-1\right)}{{HR}_{NDE}\times {HR}_{NIE}-1}$. All models were adjusted for maternal age at delivery, country of birth, education, BMI (categorical including missing), and first trimester smoking, and exposure-mediator (SLE-HDP) interaction. This analysis was performed for our primary and secondary outcomes.

We used SAS version 9.4 (SAS Institute, Cary, NC, USA) and R version 3.5 (R Foundation for Statistical Computing, Vienna, Austria) for data management and statistical analysis.

Results

Between 1987 and 2012 we identified 450 singleton first deliveries among women with prevalent SLE and 2890 singleton first deliveries among nulliparous women without SLE. Overall, the women with and without SLE were comparable with respect to age, BMI, and maternal country of birth. Women with SLE were more likely to have pre-pregnancy comorbidities including hypertension, renal disease, and thrombotic events (Table 1). Among women with SLE we observed an unadjusted incidence rate (IR) of 50.0 cases per 10,000 person-years (95%CI 33.4, 71.3) for incident CV outcomes compared to 7.2 cases per 10,000 person-years (95%CI 2.8, 17.6) among women without SLE.

Table 1.

Characteristics of women with a first pregnancy identified between 1987 and 2012 in the Medical Birth Register, stratified by systemic lupus erythematosus (SLE) status and hypertensive disorders of the pregnancy (HDP).

	SLE			No SLE
	HDP	No HDP	Overall	HDP	No HDP	Overall
No. women	90	360	450	200	2690	2890
Age at delivery, mean (SD)	30 (5.4)	30 (4.4)	30 (4.6)	31 (5.3)	30 (5.0)	30 (5.0)
Born in Nordic country*, %	90	90	90	92	90	90
Education, %
≤9 years	11	8	9	2	7	7
10‒12 years	42	33	35	42	39	39
≥13 years	40	54	51	48	47	47
Missing	7	5	5	8	7	7
First trimester BMI, %
Underweight (<18.5 kg/m2)	3	5	4	2	3	3
Normal weight (18.5–24.9 kg/m2)	46	60	57	38	56	55
Overweight (25.0–29.9 kg/m2)	13	14	14	24	18	18
Obese (≥30.0 kg/m2)	11	8	8	16	7	8
Missing	27	14	17	20	16	16
First trimester smoking, %
Yes	11	8	8	6	10	10
No	76	88	86	90	85	85
Missing	13	4	6	4	5	5
Pregestational disease, %
Hypertension	20	8	11	3	1	1
Diabetes	4	1	2	<1	<1	<1
Renal disease	21	10	12	<1	<1	<1
DVT/PE	9	3	4	<1	<1	<1
Year of delivery, %
1987‒1994	23	14	16	18	18	18
1995‒2002	22	21	21	18	22	22
2003‒2012	54	66	63	64	60	61

SLE = systemic lupus erythematosus; HDP = hypertensive disorders of the pregnancy; BMI = body mass index;

DVT/PE = deep venous thrombosis and pulmonary embolism.

^*Nordic countries: Sweden, Denmark, Finland, Norway, and Iceland.

CV outcomes among women with and without HDP

Among the women without SLE who experienced HDP during their first pregnancy, we observed an unadjusted rate of CV outcomes of 12 per 10,000 person-years compared to 89 per 10,000 person-years among those with SLE and HDP (Table 2). HDP was associated with an approximately two-fold higher rate of CV outcomes among women with SLE across multiple sensitivity analyses. (Table 2) When we included first and subsequent pregnancies with HDP as a time-dependent exposure we found similar adjusted HRs (SLE: 2.3 (95%CI 1.0, 5.3) and general population: 1.7 (95%CI 0.6, 4.7)). The HR for CV outcomes in women with SLE and HDP was about 8 times higher than the HR for women without SLE but with HDP (Table 3). However, we found little evidence of effect measure modification on the multiplicative (ratio of HRs (95% CI): 1.1 (0.3, 4.6)) and additive (RERI (95% CI): 5.2 (−4.1, 14.6)) scales.

Table 2.

Incidence rates and hazard ratios (HRs) for cardiovascular/cerebrovascular outcomes in hypertensive disorders of the pregnancy, stratified by systemic lupus erythematous (SLE) status.

	SLE		No SLE
	HDP	No HDP	HDP	No HDP
Cardiovascular outcomes *
Events/Person-years	10/1127	17/4275	3/2596	25/36 054
Rate per 10 000 py (95% CI)	88.8 (44.5, 155.6)	39.8 (9.3, 158.0)	11.6 (2.9, 30.0)	6.9 (0.6, 75.5)
Crude HR (95% CI)	2.0 (0.9, 4.5)	1.0 [Ref]	1.7 (0.5, 5.7)	1.0 [Ref]
Adjusted HR (95% CI)	1.9 (0.8, 4.3)	1.0 [Ref]	1.7 (0.5, 6.0)	1.0 [Ref]
Acute myocardial infarction (only)
Events/Person-years	2/1199	3/4491	1/2613	4/36 234
Rate per 10 000 py (95% CI)	16.7 (2.8, 51.5)	6.7 (0.2, 156.5)	3.8 (0.2, 16.8)	1.1 (0.0, 95.0)
Stroke (only)
Events/Person-years	7/1156	7/4368	0/2615	13/36 149
Rate per 10 000 py (95% CI)	60.6 (26.0, 117.1)	16.0 (2.4, 90.5)	Not estimable	3.6 (0.6, 18.5)
Hypertension **
N	67	333	181	2665
Events/Person-years	36/687	69/3683	39/2203	181/34 836
Rate per 10 000 py (95% CI)	523.7 (370.7, 713.9)	187.3 (89.3, 385.9)	177.0 (127.1, 238.5)	52.0 (26.7, 100.3)
Crude HR (95% CI)	2.9 (1.9, 4.4)	1.0 [Ref]	3.4 (2.4, 4.9)	1.0 [Ref]
Adjusted HR (95% CI)	3.0 (2.0, 4.6)	1.0 [Ref]	3.1 (2.2, 4.4)	1.0 [Ref]
Hypertension-NPR ONLY γ
N	73	335	194	2669
Events/Person-years	16/946	23/4127	18/2492	40/35 901
Rate per 10 000 py (95% CI)	169.1 (99.2, 266.0)	55.7 (17.4, 168.9)	72.2 (43.8, 110.9)	11.1 (3.9, 30.6)
Crude HR (95% CI)	2.7 (1.4, 5.2)	1.0 [Ref]	6.7 (3.8, 11.8)	1.0 [Ref]
Adjusted HR (95% CI)	2.6 (1.3, 5.1)	1.0 [Ref]	6.3 (3.5, 11.3)	1.0 [Ref]

HDP = hypertensive disorders of the pregnancy; HR = hazard ratio; CI = confidence interval; py = person-years; NPR = National Patient Register.

Crude and adjusted hazard ratio models were stratified by period of delivery. The adjusted model was controlled for maternal age at delivery, county of birth, education, BMI (categorical including missing), and first trimester smoking.

^*Includes fatal and non-fatal acute myocardial infarction, unstable angina, stroke, transient ischemic stroke, and heart failure.

^**Incident hypertension defined as ≥1 visit for hypertension or ≥2 dispensations of antihypertensive medication ≥3 months postpartum. Exclusions include women who had pregestational hypertension that was defined using visit and mediciation dispensation data.Women with history of the outcome at start of follow-up were excluded: SLE/HDP n=24, SLE/non-HDP n=33, non-SLE/HDP n=19, non-SLE/non-HDP n=24.

^γIncident hypertension defined only on the basis of ICD-coded visits for hypertension. Women with history of the outcome at start of follow-up were excluded: SLE/HDP n=11, SLE/non-HDP n=23, non-SLE/HDP n=5, non-SLE/non-HDP n=24. Exclusions include women who had pregestational hypertension that was defined using visit and mediciation dispensation data.

Table 3.

Interaction between hypertensive disorders of pregnancy (HDP) and SLE on the risk of CV outcomes and hypertension (restricted to first pregnancies).

Primary outcome: Cardiovascular outcomes
	Hypertensive Disorder of Pregnancy
	No HR (95% CI)	Yes HR (95% CI)	HR (95%CI) for HDP stratified by SLE
No SLE	1.0 [Ref]	1.7 (0.5, 5.7)*	1.7 (0.5, 5.7)
SLE	7.2 (3.8, 13.6)*	13.2 (6.2, 27.7)*	1.8 (0.8, 4.1)
HR (95%CI) for SLE stratified by HDP	7.2 (3.8, 13.6)	7.7 (2.1, 28.6)
Multiplicative interaction, ratio of HRs (95% CI)			1.1 (0.3, 4.6)
Additive interaction, RERI (95% CI)			5.2 (−4.1, 14.6)
Secondary outcome: Incident hypertension
	Hypertensive Disorder of Pregnancy		HR (95%CI) for HDP stratified by SLE
	No HR (95% CI)	Yes HR (95% CI)
No SLE	1.0 [Ref]	3.2 (2.2, 4.5)*	3.2 (2.2, 4.5)
SLE	4.0 (3.1, 5.3)*	12.4 (8.6, 17.9)*	3.1 (2.0, 4.6)
HR (95%CI) for SLE stratified by HDP	4.0 (3.1, 5.3)	3.9 (2.5, 6.2)
Multiplicative interaction, ratio of HRs (95% CI)			1.0 (0.5, 1.7)
Additive interaction, RERI (95% CI)			6.1 (1.7, 10.6)

SLE = systemic lupus erythematosus; HDP = hypertensive disorders of the pregnancy; HR = hazard ratio;

CI = confidence interval; RERI = relative excess risk due to interaction.

All models were adjusted for maternal age at delivery, country of birth, education, BMI, smoking, and stratified by year of delivery. CV outcomes includes fatal and non-fatal acute myocardial infarction (AMI), fatal and non-fatal stroke (ischemic, hemorrhagic, subarachnoid hemorrhage, or unspecified), transient ischemic attacks (TIA), unstable angina, and heart failure.

^*In reference to the no SLE-no HDP group for each outcome

We also found that HDP was associated with a higher incidence of hypertension later in life in both women with and without SLE. The unadjusted incidence rate was considerably higher in women with SLE and HDP (524 incident cases of hypertension per 10,000 person-years) compared to 177 incident cases of hypertension per 10,000 person-years for women with HDP from the general population. Sensitivity analyses suggested that this was not due to misclassification of antihypertensive use for renal disease in women with SLE nor antihypertensive use for possible HDP in subsequent pregnancies. Additionally, all women with incident hypertension (outcome) were not pregnant at the time they developed the outcome. Therefore, any dispensation of antihypertensive medications at the time was not related to pregnancy-related complications. There was evidence of effect modification on the additive scale for the incident hypertension outcomes (RERI (95% CI): 6.1 (1.7, 10.6)).

Sensitivity analyses

The above unadjusted IRs for CV outcomes were relatively consistent across analyses although with small numbers. Only 1 non-SLE HDP-exposed pregnancy had an AMI after pregnancy, and none had a stroke during post-partum follow-up. Although the SLE-HDP pregnancies contributed less post-partum person-time, there were 2 AMIs in 1199 person-years of follow-up and 7 strokes over 1156 person-years (Table 2). Among the SLE and non-SLE pregnancies with a post-partum stroke during follow-up, none had evidence of pregestational hypertension.

When incident hypertension was redefined solely on the basis of ICD-coded visits at least three months post-partum, results were comparable among the women with SLE (HR=2.6 (95%CI 1.3, 5.1)). However, the hazard ratio of the association between HDP and hypertension increased in magnitude among women without SLE (HR=6.3 (95%CI 3.5, 11.3)). Results were similar when we restricted to the 2006–2013 time period during which anti-hypertensive data were available.

Mediation analysis

When examining the association between SLE and CV outcomes after pregnancy, HDP contributed to approximately 10% of these outcomes. However, HDP explained a larger proportion (20%) of the association between SLE and development of hypertension later in life.

Discussion

Among women with SLE, HDP (including preeclampsia) in their first pregnancy was associated with a two-fold higher rate of CV outcomes (including stroke, AMI, TIA) during a median 10.8 years of follow-up compared to women whose first pregnancy was not complicated by HDP. Although power was limited due to sparse events in the general population, we estimated hazard ratios between 1.5 and 2.0, which is consistent with the published literature.[2], [11] Consistent with prior studies, we also found that HDP was associated with a three-fold higher incidence of hypertension among women with and without SLE.[2], [4]

Other studies have noted a link between SLE and stroke after pregnancy.[12], [13] Despite methodological differences among studies, including how SLE was defined, which outcomes were evaluated, and whether preeclampsia or HDP was specifically studied, our findings are consistent with their observations.

Our study is based on register data collected in conjunction with clinical care, birth registration, and population-based record-keeping. The data are collected prospectively for record-keeping and research purposes. We identified women with SLE before delivery using a register-based algorithm with low misclassification and linked these with the MBR.[8], [14] Compared to definitions that require that SLE is recorded as a diagnosis at delivery, our approach reduces misclassification.[15] Positive predictive values of conditions included in our study from Scandinavian registers have found generally good validity ranging from 74% in Denmark to 84–99% in Swedish and Norwegian data for preeclampsia and HDP, respectively.[16]–[18]

Despite identifying 3340 women with their first singleton delivery during the study period, including 450 women with SLE[2], there were few events during a median of 10.8 years follow-up, particularly among women without SLE (median follow-up 11.8 years). We could not confirm established associations between HDP and CVD, possibly due to the relatively short follow-up time given that pre-menopausal CVD is rare among women free of SLE. The mean age at end of follow-up was only 42 years in the SLE exposed and 43 years in the women without SLE. Due to limited power, we could not evaluate early-onset HDP nor preeclampsia specifically.

When evaluating the incidence of hypertension later in life, a CVD risk factor that typically presents before CV events, our results were consistent with the published literature in the general population. Among women without SLE, we found that relying solely on ICD codes to define incident hypertension led to a higher hazard ratio associated with HDP. This could be, in part, due to women without HDP and SLE being less likely to interact with secondary healthcare at this “young” age. In other words, without access to primary care data and excluding anti-hypertensive use to identify incident hypertension, we may be more likely to underestimate the outcome in women with neither SLE nor HDP. In SLE, we expected more CV outcomes compared to the general population because it is known that women with SLE are affected by CVD at a younger age than women from the general population. Both accelerated atherosclerosis and a prothrombotic state associated with lupus itself, as well as presence of antiphospholipid antibodies (aPL), may contribute to premature CVD in SLE. Unfortunately, data on aPL were not available. When prior deep vein thrombosis and pulmonary embolism were used as a proxy to identify high-risk for stroke, we found that of the 27 strokes observed only one had this noted (woman with SLE but no HDP). Further we cannot exclude the possibility that women with SLE who were able to get pregnant and carry to at least 22 weeks (28 weeks through 2008) to be captured in the Birth Register are less likely to have CVD at a younger age.

Premature CVD is a well-documented complication in women with SLE which is likely, at least in part, due to renal disease, prothrombotic aPL, and systemic inflammation. Our data confirm that women who experience a hypertensive disorder in pregnancy are at greater risk of developing hypertension after pregnancy, and that this association is also evident for women with SLE. Women with SLE and HDP were also at increased risk of CVD, particularly stroke, at young ages and should be monitored closely and consider treatment to attenuate risk.

Significance and Innovations.

This study revealed that women with SLE and hypertensive disorders of pregnancy have a higher rate of cardiovascular outcomes after pregnancy compared with women with SLE without maternal hypertension.

Additionally, women with SLE experienced a higher incidence of hypertension later in life following hypertensive disorders of pregnancy.

We could not confirm established associations between HDP and postpartum cardiovascular outcomes in the general population, which may be due to the relatively short follow-up time given that pre-menopausal cardiovascular disease is uncommon.