Maternal and neonatal outcomes of antihypertensive treatment in pregnancy: A retrospective cohort study.

OBJECTIVE: To compare maternal and infant outcomes with different antihypertensive medications in pregnancy.
DESIGN: Retrospective cohort study.
SETTING: Kaiser Permanente, a large healthcare system in the United States. POPULATION: Women aged 15-49 years with a singleton birth from 2005-2014 treated for hypertension.
METHODS: We identified medication exposure from automated pharmacy data based on the earliest dispensing after the first prenatal visit. Using logistic regression, we calculated weighted outcome prevalences, adjusted odds ratios (aORs) and 95% confidence intervals, with inverse probability of treatment weighting to address confounding. MAIN OUTCOME MEASURES: Small for gestational age, preterm delivery, neonatal and maternal intensive care unit (ICU) admission, preeclampsia, and stillbirth or termination at > 20 weeks.
RESULTS: Among 6346 deliveries, 87% with chronic hypertension, the risk of the infant being small for gestational age (birthweight < 10th percentile) was lower with methyldopa than labetalol (prevalence 13.6% vs. 16.6%; aOR 0.77, 95% CI 0.63 to 0.92). For birthweight < 3rd percentile the aOR was 0.57 (0.39 to 0.80). Compared with labetalol (26.0%), risk of preterm delivery was similar for methyldopa (26.5%; aOR 1.10 [0.95 to 1.27]) and slightly higher for nifedipine (28.5%; aOR 1.25 [1.06 to 1.46]) and other β-blockers (31.2%; aOR 1.58 [1.07 to 2.23]). Neonatal ICU admission was more common with nifedipine than labetalol (25.9% vs. 23.3%, aOR 1.21 [1.02 to 1.43]) but not elevated with methyldopa. Risks of other outcomes did not differ by medication.
CONCLUSIONS: Risk of most outcomes was similar comparing labetalol, methyldopa and nifedipine. Risk of the infant being small for gestational age was substantially lower for methyldopa, suggesting this medication may warrant further consideration.

Introduction

Hypertensive disorders affect 5–10% of pregnancies [1], increasing the risk of fetal growth restriction, stillbirth and other adverse outcomes [2-5]. About 160,000 pregnant women take antihypertensive medications annually in the US [2], yet it is unclear which medication results in the best outcomes for women and infants. Current US and UK guidelines recommend labetalol and nifedipine over methyldopa, while acknowledging uncertainty [6, 7]. The International Society for the Study of Hypertension in Pregnancy has stated that both methyldopa and nifedipine are acceptable [8].

Randomized clinical trials (RCTs) have not provided definitive evidence because they have had small sample sizes and heterogeneous methods. If sufficient data were available from RCTs, a meta-analysis could be performed to compare outcomes with different medications, but unfortunately data are sparse. A 2018 Cochrane meta-analysis [9] identified 29 RCTs that compared antihypertensive medications head-to-head; taken together, these trials included a total of only 2774 women. The trials were heterogeneous, examining many different medications, which resulted in very small sample sizes for specific comparisons. The only definitive finding from the meta-analysis was that β-blockers and calcium channel blockers appeared more effective than methyldopa at preventing severe hypertension [9]. For other outcomes, there were no statistically significant differences, which is understandable because often only a few trials were included, leading to low precision and wide confidence intervals. The Cochrane meta-analysis grouped together all β-blockers, which may obscure important differences between individual medications, especially as labetalol has different receptor specificity than other commonly used β-blockers. A recent RCT reported that methyldopa was associated with significantly lower risk of small for gestational age (SGA) and NICU admission compared to labetalol, with odds ratios on the order of 0.40, and that the two medications were associated with similar risk of severe maternal hypertension or preeclampsia [10]. However, the sample size was small (~150 women per arm) and many of their risk estimates had wide confidence intervals. Due to the limitations outlined above, existing RCT data are not adequate to guide choice of medications for the treatment of hypertension in pregnancy.

When RCT data are insufficient, as is often the case for medication use in pregnancy, rigorous observational studies can provide useful information. One observational analysis [11] used data from the Control of Hypertension in Pregnancy Study [12], which randomized pregnant women to tight vs. less tight blood pressure control but did not dictate which medications were used. The post hoc observational analysis compared pregnancy outcomes with methyldopa vs. labetalol (an observational comparison, since choice of medications was not randomized) and found better outcomes with methyldopa [11]. Other antihypertensive medications were not examined. Several other observational studies have been conducted, but they had important methodologic limitations which make it difficult to draw causal inferences. Many of these studies compared women treated with an antihypertensive medication to unexposed women from the general pregnant population [3, 5, 13–16], most of whom presumably did not have hypertension. Since hypertension increases the risk of adverse pregnancy outcomes, these studies are vulnerable to confounding by indication and cannot shed light on the risks of treatment vs. those due to hypertension. Additional studies are needed using rigorous methods that can support causal inference.

Because additional evidence is needed, we sought to compare the risk of important maternal and infant outcomes with use of different antihypertensive medications using electronic health records (EHR) data for a large, diverse US population.

Methods

Overview

This retrospective cohort study was conducted at Kaiser Permanente, a US healthcare system providing health care and insurance coverage. Participating regions were Washington, Southern California, and Northern California, which together serve about 8 million people generally representative of the surrounding communities [17]. Data came from EHRs and linked birth certificate data. These data have been used in many pregnancy studies [18-21], and important variables and methods have been validated [22-25]. The study used rigorous causal inference methods [26, 27], including following recommended principles for emulating a target trial [28], using active comparators (comparing outcomes with one medication vs. another used for the same indication) [29], and addressing confounding using inverse probability weighting [27]. Study procedures were approved by the regions’ institutional review boards and those of Washington State and California, with a waiver of consent.

Study population

The population was women age 15–49 years with a singleton live or stillbirth from 2005 through 2014. Women were required to be enrolled in Kaiser Permanente from 16 weeks’ gestation through delivery, to have at least one blood pressure (BP) measured before 20 weeks, and to have chronic or gestational hypertension (defined from BP values, diagnosis codes and medication fills; our algorithm is shown in S1 Table in S1 File and has been published [30]). We included both chronic or gestational hypertension because in clinical practice, it can be difficult to determine which type of hypertension is present and because these conditions may represent different points on a continuum of disease.

Women had to have filled at least one antihypertensive medication before 36 weeks gestation, to be on monotherapy, and to have been enrolled in Kaiser Permanente for at least 150 days before their qualifying fill. They could contribute more than one pregnancy to these analyses. We excluded deliveries exposed to teratogenic medications or certain high-risk maternal medical conditions (see S1 Table in S1 File for more information). The sample size was determined by the number of eligible births.

Exposures

From computerized pharmacy data, we identified fills for labetalol, methyldopa, nifedipine and other β-blockers (S1 Table in S1 File). These data are recorded prospectively when medications are dispensed, eliminating the biases that can arise in some retrospective observational studies (for instance, studies that interview women after delivery about medications taken in pregnancy.) Unlike many prior studies, we considered labetalol separately from other β-blockers because it is a combined α and β-blocker and unlike other β-blockers, it is recommended as first-line in US guidelines [6]. Exposure was defined based on the earliest fill after the first prenatal visit (typically at 8–10 weeks’ gestation) or, if the visit date was unknown, at ≥ 10 weeks gestation; we called this the ‘index fill’. Using intent to treat principles, women’s exposure status was fixed rather than time-varying, because subsequent medication switches could be affected by the initial medication choice.

Outcomes

Outcomes included SGA, preterm delivery, neonatal intensive care unit (NICU) admission, preeclampsia, maternal ICU admission, and stillbirth or termination at > 20 weeks. SGA was defined using sex and race-specific US birthweight curves [31]. The primary outcome was birthweight <10th percentile for gestational age and a secondary outcome < 3rd percentile. Deliveries missing birthweight (n = 32) were excluded from SGA analyses. We defined preterm delivery using gestational age from the EHR (preferentially) or birth certificate data, with the primary outcome being delivery before 37 weeks gestation and a secondary outcome, delivery before 34 weeks. We considered preterm delivery a potential measure of medication effectiveness, because less effective medications could lead to higher risk of uncontrolled maternal hypertension or fetal growth restriction (a potential consequence of severe hypertension) and via these pathways, to indicated preterm delivery. The automated data available to us do not reliably distinguish spontaneous vs. indicated preterm births. We identified ICU admissions using EHR data. Preeclampsia was identified from inpatient diagnosis codes after 20 weeks’ gestation, an approach with a positive predictive value of 90% [32]. We reviewed 45 charts meeting those criteria and found a positive predictive value of 93%. We identified preeclampsia cases with “severe features” using modified criteria from the American College of Obstetricians & Gynecologists [33], drawing on BP values, laboratory results and diagnosis codes (S1 Table in S1 File).

Potential stillbirths and terminations after 20 weeks’ gestation were identified using EHR data; we included as outcomes the 76% of potential cases validated through medical record review or linkage to fetal death certificates. We grouped together stillbirths and terminations for several reasons. Terminations after 20 weeks might be done for fetal anomalies, which could in theory be affected by medication choice, as there is no definitive evidence about birth defect risk for some widely used antihypertensive medications. Also, the decision to terminate might be influenced by severe uncontrolled maternal hypertension, which could be a consequence of the initial medication choice. Finally, we hypothesized that variation in coding might lead to similar clinical scenarios being classified as either a stillbirth or termination in different instances.

Covariates

Covariates included maternal age at delivery, Kaiser Permanente region, delivery year, hypertension type (chronic or gestational), BP values, race/ethnicity, parity, maternal education, pregestational diabetes, depression, tobacco use, body mass index (BMI), and prior use of certain medications (S1 Table in S1 File). Hypertension was categorized as chronic if it was present prior to pregnancy or during the first 20 weeks gestation and as gestational otherwise. To account for hypertension severity, we identified the most recent BP value prior to the index fill and also determined whether a woman experienced one or more BPs ≥ 160/110 before pregnancy or during this pregnancy before the index fill. We categorized history of antihypertensive medication use as no use prior to the index fill, continuous use up to the index fill (allowing for 80% adherence), or prior use with a gap. Other covariates included prior use of angiotensin converting enzyme inhibitors, angiotensin receptor blockers, thiazide diuretics, diabetes medications, benzodiazepines, statins, antidepressants or antiseizure medications.

Statistical analyses

Descriptive analyses included counts and proportions for categorical variables and means and standard deviations for continuous variables. Primary analyses used logistic regression to model study outcomes, with labetalol as the referent group. Inverse probability of treatment weights (IPTW) were used to account for confounding. We calculated weighted outcome prevalences for each medication group and adjusted odds ratios (aORs) and 95% confidence intervals (CIs). We used the bootstrap to account for multiple pregnancies per woman and for the estimation of the weights [34, 35]. Treatment weights were generated from propensity scores calculated using a multinomial logistic regression model including all covariates shown in Table 1 except for BMI, education, parity, and timing of prenatal care. We omitted these variables because they were well balanced before weighting and a small proportion of deliveries had missing information for each of these characteristics. S2 Table in S1 File lists variables in the propensity score. To improve statistical efficiency, we calculated stabilized weights including some baseline covariates in both the outcome model and the numerator of the weights [36, 37]. These were Kaiser Permanente region, race/ethnicity, diabetes, type of hypertension (chronic vs. gestational), and gestational age at index fill.

Table 1

Baseline characteristics of the population before weighting, by treatment group.

Characteristic (number (%) unless otherwise stated)	Labetalol N = 3017	Methyldopa N = 1834	Nifedipine N = 1105	Other β-blockers N = 390
Maternal age, yrs, mean±SD	33.5±5.2	33.9±5.3	33.2±5.6	33.8±5.2
Nulliparousb	1109 (36.8)	699 (38.1)	422 (38.2)	139 (35.6)
Race/ethnicity
White, non-Hispanic	1011 (33.5)	551 (30.0)	369 (33.4)	175 (44.9)
Hispanic	928 (30.8)	604 (32.9)	275 (24.9)	75 (19.2)
Black, non-Hispanic	470 (15.6)	242 (13.2)	184 (16.7)	60 (15.4)
Asian	575 (19.1)	414 (22.6)	261 (23.6)	73 (18.7)
Obese (BMI ≥ 30 kg/m2)b	1796 (59.5)	968 (52.8)	581 (52.6)	223 (57.2)
Tobacco use	150 (5.0)	68 (3.7)	53 (4.8)	21 (5.4)
Chronic hypertension	2550 (84.5)	1693 (92.3)	910 (82.4)	360 (92.3)
Pre-gestational diabetes	560 (18.6)	355 (19.4)	250 (22.6)	68 (17.4)
Prenatal care in trimester 1b	2653 (87.9)	1402 (76.4)	922 (83.4)	277 (71.0)
Gestational age at index fill (weeks), mean±SD	18.8±9.6	16.5±8.0	20.8±9.6	17.1±7.9
Systolic BP, mm Hg; mean±SDc	142.8 (17.2)	138.7 (15.7)	138.9 (17.0)	132.8 (17.3)
Diastolic BP, mm Hg; mean±SDc	88.0 (12.1)	84.9 (10.9)	84.9 (12.9)	80.7 (11.9)
Prior antihypertensive medication use
Prior use, continuous	1018 (33.7)	813 (44.3)	348 (31.5)	162 (41.5)
Prior use with a gap	839 (27.8)	577 (31.5)	296 (26.8)	127 (32.6)
No prior use	1160 (38.4)	444 (24.2)	461 (41.7)	101 (25.9)
Delivery year
2005–2008	500 (16.6)	704 (38.4)	314 (28.4)	139 (35.6)
2009–2010	713 (23.6)	489 (26.7)	250 (22.6)	99 (25.4)
2011–2012	860 (28.5)	348 (19.0)	287 (26.0)	89 (22.8)
2013–2014	944 (31.3)	293 (16.0)	254 (23.0)	63 (16.2)

Abbreviations: BMI, body mass index; BP, blood pressure; SD, standard deviation.

aAll characteristics measured prior to the index medication fill, except for delivery year. The proportion with missing data across groups was as follows: for parity, 2.8 to 4.2%; for race/ethnicity, 0.4 to 1%; and for BMI, 3.7 to 13.9%. No pregnancies had missing data for other listed variables.

bCovariate not in the propensity score model.

cMost recent BP prior to index fill of antihypertensive medication.

For statistical modeling, we categorized delivery year as 2005–2008, 2009–2010, 2011–2012, and 2013–2014. We grouped together the four earliest years because very few deliveries were included from 2005–2006, when only one region had electronic BP values available. Maternal age was categorized as < 30, 30–34, 35–39 or ≥ 40 years. Gestational age at the index fill was modeled as a linear spline with knots at 140 and 210 days. The systolic and diastolic BP values closest to the index fill were modeled using linear splines, with knots at 140 mm Hg and 90 mm Hg respectively. Deliveries missing race/ethnicity (0.5%) were grouped with those with “other” race/ethnicity and treated as a category of race/ethnicity in statistical models.

To assess covariate balance, we calculated the average standardized mean differences across all treatment groups before and after IPTW [38, 39].

We excluded stillbirths/terminations from analyses of SGA, NICU and preterm delivery because they are competing events. We used inverse probability of censoring weights to account for possible bias due to excluding stillbirths; S3 Table in S1 File lists the variables used to model these weights.

In sensitivity analyses, we restricted the analysis to women with chronic hypertension (87% of the population) and excluded women with pregestational diabetes. In subgroup analyses, we examined new users separately from women with prior antihypertensive treatment. Analyses were performed using R, version 3.5.

Funding

This study was funded by the US National Institute on Child Health and Human Development grant R01HD082141. The Group Health Foundation funded Dr. Chen’s fellowship. The funders did not play a role in conducting the research or writing the paper.

Results

Among 6346 eligible deliveries, there were 3017 (48%) where the woman had taken labetalol, 1834 (29%) methyldopa, 1105 (17%) nifedipine, and 390 (6%) other β-blockers. Fig 1 shows the impact of inclusion and exclusion criteria on the study population.

Fig 1

Impact of inclusion and exclusion criteria on study population.

Abbreviations: BP, blood pressure; KPNC, Kaiser Permanente Northern California; KPSC, Kaiser Permanente Southern California; KPWA, Kaiser Permanente Washington. aA woman may meet more than one exclusion criterion within a box. Detailed information about inclusion and exclusion criteria is found in S1 Table in S1 File. bThe index fill was defined as the earliest fill after the first prenatal visit (typically at 8–10 weeks’ gestation) or, if the visit date was not known, at ≥ 10 weeks gestation.

Mean maternal age was 33.6 years, 87% had chronic hypertension, and the mean gestational age at the index fill was 18.4 weeks. Many women (37%) were taking antihypertensive medication continuously prior to the index fill, and mean BPs prior to the index fill suggest that their hypertension was on average fairly well controlled. Table 1 shows baseline characteristics by treatment group, and S4 Table in S1 File provides more detailed information for an expanded list of baseline characteristics. S5 Table in S1 File shows characteristics by treatment group after IPTW and demonstrates that overall, these were well balanced (standardized mean difference < 0.1), except for those characteristics included in the outcome model, which are not expected to be balanced by IPTW. After IPTW, the group exposed to other β-blockers looked modestly different from the other groups, likely due to this group’s small size. S6 Table and S1 Fig in S1 File describe the distributions of propensity scores and weights.

Most women did not switch medications after their index fill. The proportion of women who later filled a different medication was 15% overall, ranging from 11 to 22% for different exposure groups.

Table 2 provides crude counts of outcomes by treatment group. Fig 2 shows the risk of maternal and neonatal outcomes comparing different medications, with labetalol as the referent group. We present weighted prevalences for outcomes after accounting for confounders together with adjusted ORs and 95% CIs. For SGA < 10th percentile, risk was lower with methyldopa than labetalol (weighted prevalence 13.6% vs. 16.6%; aOR 0.77, 95% CI 0.63 to 0.92), and the association was stronger for birthweight < 3rd percentile (aOR 0.57, 95% CI 0.39 to 0.80). The mean birthweight after IPTW was 3002 ± 797 g for labetalol, 3060 ± 788 g for methyldopa, 3033 ± 798 g for nifedipine, and 2944 ± 791 g for other β-blockers.

Table 2

Counts of maternal and neonatal outcomes by treatment group.

Outcome and Medication Class	Outcomes/Exposed Pregnanciesa
	Labetalol	Methyldopa	Nifedipine	Other β-blockers
Preeclampsia	1001/3017	523/1834	338/1105	91/390
Preeclampsia with severe features	786/3017	351/1834	230/1105	56/390
Maternal ICU	61/3017	37/1834	20/1105	14/390
Stillbirth or termination	41/3017	18/1834	13/1105	2/390
SGA < 10th percentile	512/2962	231/1805	159/1091	66/382
SGA < 3rd percentile	145/2962	49/1805	41/1091	19/382
Preterm delivery < 37 weeks	812/2976	485/1816	352/1092	87/388
Preterm delivery < 34 weeks	299/2976	161/1816	112/1092	27/388
Neonatal ICU admission	726/2976	426/1816	296/1092	86/388

Abbreviations: OR, odds ratio; CI, confidence interval; SGA, small for gestational age; ICU, intensive care unit.

aActual numbers prior to inverse probability of treatment weighting. The population for different outcomes differs slightly because pregnancy losses were not included in the denominator for SGA, preterm delivery, or neonatal ICU admission, and because 32 deliveries missing infant birthweight were excluded from analyses of SGA.

Fig 2

Risk of maternal and neonatal outcomes with use of different antihypertensive medications in pregnancy*.

Abbreviations: OR, odds ratio; CI, confidence interval; SGA, small for gestational age; ICU, intensive care unit. *ORs and 95% CIs are calculated after inverse probability of treatment weighting. Labetalol is the referent group. The population for different outcomes differs slightly because pregnancy losses were not included in the denominator for SGA, preterm delivery, or neonatal ICU admission. For most outcomes, the total N is 6346, for SGA the total N is 6240, and for preterm delivery and NICU admission the total N is 6272. **Weighted prevalence in the subgroup, calculated using inverse probability of treatment weighting with unstabilized weights.

Preterm delivery was slightly more common with nifedipine than labetalol (28.5% vs. 26.0%; aOR 1.25, 95% CI 1.06 to 1.46), as was NICU admission (25.9% vs. 23.3%; aOR 1.21, 95% CI 1.02 to 1.43). β-blockers other than labetalol were associated with higher risk of preterm delivery (aOR 1.58, 95% CI 1.07 to 2.23). Methyldopa and labetalol conveyed similar risks of preterm delivery and NICU admission. After IPTW, the mean gestational age at delivery was 37.6 ± 2.8 weeks for labetalol, 37.6 ± 2.8 weeks for methyldopa, 37.4 ± 2.8 weeks for nifedipine, and 37.4 ± 2.8 weeks for other β-blockers.

There was no significant association between medication type and risk of preeclampsia (overall or with severe features), maternal ICU admission, or stillbirth/termination.

Results of sensitivity and subgroup analyses are shown in S2-S5 Figs in S1 File. Results did not change when we restricted the population to women with chronic hypertension, who made up 87% of the population. Results also did not change when we excluded women with pregestational diabetes. Some findings appeared qualitatively different when we limited analyses to new users; in this group, there was a suggestion of lower risk for many outcomes with methyldopa than with labetalol, with aORs around 0.5 to 0.7 (though most were not statistically significant).

Discussion

In this large retrospective cohort study, the prevalence of many maternal and neonatal outcomes was similar with use of different antihypertensive medications. Compared to labetalol, the risk of SGA was significantly lower with methyldopa.

Other studies have examined outcomes with use of different antihypertensive medications. Most prior studies were small, yielding inconclusive results, and many observational studies compared treated women to healthy pregnant women, making confounding likely. Our finding of lower SGA risk with methyldopa compared to labetalol (aOR 0.77, 95% CI 0.63 to 0.92) is consistent with one recent RCT, which found the prevalence of SGA was much lower with methyldopa than labetalol (21% vs. 41%;OR 0.37, 95% CI 0.23–0.61) [10]. Similar results were found by Magee et al. in a secondary analysis of RCT data [11]. The Cochrane meta-analysis of RCTs compared methyldopa to all β-blockers grouped together and reported a combined RR of 1.19 (0.76, 1.84). Grouping labetalol together with other β-blockers is problematic because it has different receptor specificity and thus may have different effects on outcomes.

Labetalol binds to β-adrenergic receptors, lowering maternal heart rate and cardiac output, while also acting on α-adrenergic receptors in peripheral blood vessels to block the adrenergic stimulation that causes vasoconstriction. In contrast, methyldopa lowers blood pressure by binding to α2-adrenergic receptors as an agonist, reducing sympathetic outflow that causes peripheral vasoconstriction. Methyldopa crosses the placenta and recently, α2 receptors have been found on the placenta where they may regulate placental cell syncytialization and migration [40]. Methyldopa and labetalol may have differing effects on placental uptake of folate, a critical nutrient. Keating et al. found that labetalol exposure reduced the uptake of folate by placental cells and also decreased these cells’ viability, while exposure to methyldopa did not [41]. Another mechanism through which antihypertensive medications could affect fetal growth is via methylation of placental DNA. Studies have found that alterations in placental DNA methylation are associated with maternal BP levels [42] and with infants being small for gestational age [43]. Placental genes that were affected included genes associated with cardiometabolic disease [42] and with cell proliferation, protein transport, and inflammation [43]. We were not able to find studies that examined placental DNA methylation in relation to specific antihypertensive medications; this topic warrants further investigation.

We found a slightly higher risk of preterm delivery with nifedipine compared to labetalol in an analysis including over 4000 women. The Cochrane review found only one relevant RCT, a study of 112 women yielding an RR of 1.61 that was not statistically significant [44]. For NICU admission, we observed slightly higher risk with nifedipine than labetalol (aOR 1.21, 95% CI 1.02 to 1.43). Similarly, in a recent RCT, NICU admission was more frequent with nifedipine (18%) than labetalol (10%), yielding a risk difference of 7.8 (95% CI 2.2 to 13.4) [45]. The Cochrane meta-analysis reported a summary RR of 1.14 with a 95% CI of 0.63 to 2.05, which is wide enough to be consistent with our finding. Still, since our study was not randomized, our findings could reflect confounding, including by indication for use, since nifedipine is also used for tocolysis.

Current US guidelines recommend labetalol and nifedipine above other medications and state that methyldopa is less preferred because of possible lower effectiveness and adverse effects [6]. UK guidelines recommend labetalol, followed by nifedipine and then methyldopa [7]. There is little actual evidence to support this order of priority, and several older RCTs suggested that labetalol and methyldopa are equally effective in lowering BP [46-48]. While recognizing the potential for unmeasured confounding, our large observational study suggests that outcomes are very similar between methyldopa and labetalol, except for SGA. We suggest that when there is substantial concern for SGA, it may be reasonable to give more consideration to methyldopa.

This study has several strengths. The large population improves precision and allowed more granular analyses, including examining labetalol and nifedipine as individual agents and directly comparing antihypertensive medications. We studied a diverse population in community practice and adjusted for many covariates including BP. We had precise measures of gestational age and birthweight, not available in administrative (claims) datasets. We also had information about confounders not readily available in many large datasets, such as smoking, race/ethnicity and BMI. We conducted a validation study which demonstrated that our algorithm for preeclampsia had very high positive predictive value, and we validated potential stillbirths and terminations, reducing outcome misclassification.

The study also has limitations. There is potential for residual confounding because treatment was not randomized. Because we studied medication use in real world clinical practice, there were not uniform criteria for initiating or intensifying antihypertensive medications. It is possible that women filled medications but did not take them, leading to misclassification of exposure. All women had health insurance and access to care and in general, their hypertension was well controlled at the time of the index fill, which may affect generalizability. Our data did not allow us to distinguish between spontaneous and indicated preterm birth, which on average would be expected to bias findings toward the null. The subgroup of women with gestational hypertension was too small to analyze separately. We did not have information about use of low dose aspirin, which the US Preventive Services Task Force recommended for women with chronic hypertension in 2014 [49]. The mean difference in birthweight between medications was small, and it could be argued that a difference this small is not clinically important. However, even a small shift of the birthweight curve to the left could result in a large relative increase in infants born SGA or with low birth weight, which may have important consequences for their long term health.

Our findings suggest a need for future research. We observed that labetalol appeared to convey higher risk of SGA. Infants born SGA may remain small, return to a normal growth curve, or experience compensatory weight gain leading to obesity, increasing future cardiometabolic risk. Future studies should examine child growth and development in relation to the use of specific antihypertensive medications during pregnancy. Other causal inference methods could be used to examine the associations we studied, including instrumental variable approaches such as Mendelian randomization. These observational analyses and designs rely on different untestable assumptions than the methods we utilized [50], and so if they found results similar to ours, this would further support a causal association.

In conclusion, in this large retrospective study, the prevalence of most maternal and infant outcomes was similar with different antihypertensive medications. A significantly lower risk of SGA was seen for methyldopa than labetalol, which is noteworthy because methyldopa is not preferred in US or UK guidelines [6, 7]. Our results suggest that methyldopa may warrant additional consideration, especially when there is heightened concern about growth restriction.

(DOCX)

Click here for additional data file.

7 Oct 2021

7 Apr 2022

Dublin et al.: Response to Reviewer Comments

Thank you for the opportunity to revise our manuscript. We appreciate the comments from the editor and reviewers and have sought to thoroughly address them. Please see our responses below. We first provide responses to the editor and reviewers’ substantive comments and at the bottom address additional journal requirements. All page and paragraph numbers below refer to the location in the “tracked changes” version of the manuscript.

Additional Editor Comments:

1. Please clarify the conflicts of interest with respect to the medications in the study, and whether any biases may arise.

Response: We do not have any direct conflicts of interest. The sponsor of the study, the US National Institutes of Health, has no financial interest in the results. We did not receive funding for this study from any pharmaceutical company. As we reported, several authors have received research funding from pharmaceutical companies for topics unrelated to the current work. More specifically, none of the projects funded by these companies involved the medications studied in the current paper. Moreover, the medications being studied in the current paper are older medications that are no longer under patent, and so no single company is expected to benefit exclusively if one drug were to be found to be safer or more effective than the others. Thus, we do not believe that any biases will arise related to these potential interests we have reported. Below I will provide detail about specific funding received by different co-authors.

Dr. Dublin has received research funding from GSK for unrelated work. GSK manufactures a combination medication used to treat hypertension in the general population, hydrochlorothiazide/triamterene. This drug is not recommended to treat pregnant women and was not examined in this study.

Dr. Reynolds has received research funding from Amgen, Merck and Novartis. Amgen does not make medications used to treat hypertension. Both Merck and Novartis make antihypertensive drugs in the family of angiotensin converting enzymes inhibitors (ACEIs) or angiotensin receptor blockers (ARBs). It is widely recommended that these classes of medications not be used in pregnancy due to risk of birth defects. These drugs were not studied in our analysis.

Dr. Easterling has received funding from Alnylam Pharmaceuticals and Ferring Pharmaceuticals. Neither company makes antihypertensive medications. He also has had funding from DiabetOmics, a company that makes point of care tests. One is a test for early detection of preeclampsia, which is an adverse outcome that can occur in women with hypertension in pregnancy. While this topic is tangentially related to our paper (we studied rates of preeclampsia with different medications), we do not feel there is any potential for a conflict.

Dr. Chen currently works for Roche and Ms. Bider-Canfield for Genentech (a member of the Roche group of companies.) These companies do not make any medications to treat hypertension. These investigators did not work for Roche or Genentech at the time these analyses were conducted nor has their work at Roche or Genentech involved this subject matter.

Dr. Avalos has had funding from Bausch Health Companies. Bausch makes several medications that are diuretics that can be used to treat hypertension. These include chlorothiazide, diltiazem, and ethacrynic acid. None of these drugs were studied in our paper. They are considered to be not favored in pregnancy according to current guidelines.

Dr. Shortreed has received funding from Syneos Health for a study of safety of opioid medications. This study was required by the US Food & Drug Administration. Syneos Health was in charge of overseeing projects funded by the many different pharmaceutical companies that were mandated to fund studies on the safety of opioid medications. It is likely that some of these companies also make medications to treat hypertension. We are not aware of what medications each of these companies make that could be relevant to the current paper. Regardless, the research she participated in with funding from Syneos in had no overlap with the topic of the current paper. For the reasons discussed above, we do not believe this research funding has any potential for a conflict of interest regarding the current study.

2. Please try to be more concise in Discussion.

Response: We have revised the Discussion to remove material and make it more concise.

Reviewer #1:

1. A retrospective cohort study with 6346 pregnant women is not enough to draw a robust conclusion, I am wondering if the authors may conduct a trans-ethnic meta-analysis based on similar published data with subgroup analyses according to age, ethnicity, geographical region, antihypertensive drug types. For this reason, the following papers can be cited and followed for the meta-analytic procedures (if the data is not enough available, at least DISCUSSION should be added as the LIMITATION of this study with enough citation to support the viewpoints):

Ref 1: Wu Y, et al. Multi-trait analysis for genome-wide association study of five psychiatric disorders. Transl Psychiatry. 2020 Jun 30;10(1):209.

Ref 2: Jiang L, et al. Sex-Specific Association of Circulating Ferritin Level and Risk of Type 2 Diabetes: A Dose-Response Meta-Analysis of Prospective Studies. J Clin Endocrinol Metab. 2019 Oct 1;104(10):4539-4551.

Response: We are unsure why the reviewer believes the current study is insufficient – whether the concern is that it is retrospective, or that the sample size is too small. We assume that the primary concern here is sample size, since that is a problem that a meta-analysis could overcome.

The reviewer proposes that we conduct a trans-ethnic meta-analysis of “similar published data” including subgroup analyses according to many characteristics including age, ethnicity, geographic region, etc. We agree with the author that it would be a major benefit to the field if a meta-analysis could be conducted including a large sample size, drawing on data from rigorous randomized or observational studies, with detailed subgroup analyses. Unfortunately, the needed data do not yet exist, for several reasons.

First, the body of literature from randomized trials is not sufficient to support a meta-analysis of this type. We have added new material in the introduction to make this clear. According to the most recent Cochrane review (2018)[1], a total of 29 studies have compared antihypertensive medications head to head for the treatment of hypertension in pregnancy; these studies included a total of 2774 women. These studies have been generally very small (including < 100 women, on average). Moreover, they have studied a large number of different medications. This means that for comparisons of specific medications or classes, the focus of this paper, often only 1 or 2 studies were available. In this situation, meta-analysis is not helpful. Our study population of 6346 women is over twice the size of the total number of women studied to date in randomized trials. This gave us larger numbers than the meta-analysis for class-specific comparisons, and we were able to compare individual medications in a more granular way. For instance, the Cochrane review made comparisons of all beta blockers vs. methyldopa for the outcome of small for gestational age, an outcome for which we had our most noteworthy finding. They identified a total of 6 RCTs including studies with a total of 577 women. They grouped all beta blockers together. We believe it is critical to study labetalol separately (because it has different receptor specificity and also is recommended as first line by current US guidelines, while other beta blockers are recommended not to be used in pregnancy). In our comparison of methyldopa vs. labetalol we were able to study 4851 women – more than 8 times as many as the Cochrane review. In summary, because of the dearth of randomized trials, and the small size and heterogeneity of existing trials, a meta-analysis of RCTs could not address the questions we examined.

Perhaps the reviewer means that we should conduct a meta-analysis of prior observational studies. However, we note that the reviewer expresses concern about retrospective studies, which are the predominant type of study in the literature. Also, heterogeneity in these studies and potential for bias due to confounding would make it inappropriate to summarize them through meta-analysis. The prior studies focused on different medication exposures and used very different comparator groups. See the summary table below. Most studies compared a specific medication or class of medications to no exposure—meaning that most of the control group did not have hypertension. Because hypertension itself leads to adverse pregnancy outcome, comparing women treated for hypertension to healthy women is not helpful for understanding the risks and benefits of a medication, as this comparison is plagued by confounding by indication. Combining these studies into a meta-analysis would not advance the field.

We agree that in the future, if more large and rigorous studies are published, such an effort would be very worthwhile. Unfortunately, at the current time the literature does not exist to support such a meta-analysis.

Changes made to manuscript: We have added material to the introduction explaining why the literature is not sufficient to support a meta-analysis. We have added more material about the methodologic limitations of prior observational studies. See p. 4, paragraph 2, and p. 5, paragraphs 1-2.

Table: Exposure and comparator groups in prior observational studies

Author & year Exposure medication (N) Comparator group

Meidahl Petersen 2012[2]

All beta blockers (N=2459)

Labetalol (1452)

Unexposed pregnant women

Orbach 2013[3]

Methyldopa (340), atenolol (107) Unexposed women; specifically excluded women with hypertension

Magee 2015[4]

Methyldopa (224) Labetalol (433)

Su 2013[5]

Many different classes were studied including beta blockers (414) and calcium channel blockers (303). Limited to women with chronic hypertension. Women without chronic hypertension

Xie 2014[6]

Labetalol (300) Methyldopa (923)

2. As is known, A retrospective cohort study is not reliable compared to a prospective cohort study or Mendelian Randomization analysis that would be help for to disclose the causality.

But I strongly suggest to do causal inference analysis to see if the different antihypertensive treatment in pregnancy are causally triggering the different maternal and neonatal outcomes. If cannot, please discuss the limitations in the Discussion section in detail with additional citations to support the viewpoints. For these reasons, the following papers regarding causal inference in the Mendelian Randomization framework can be cited and followed.

Ref 1: Wang X, Fang X, Zheng W, Zhou J, Song Z, Xu M, Min J, Wang F: Genetic support of a causal relationship between iron status and type 2 diabetes: a Mendelian randomization study. J Clin Endocrinol Metab 2021.

Ref 2:Zhang, F. et al. Causal influences of neuroticism on mental health and cardiovascular disease. Hum. Genet. DOI: https://doi.org/10.1007/s00439-021-02288-x (2021).

Ref 3:Zhang, F. et al. Genetic evidence suggests posttraumatic stress disorder as a subtype of major depressive disorder. J. Clin. Investig. 27, 145942, DOI: https://doi.org/10.1172/jci145942 (2021).

Ref 4: Hou L, et al. Exploring the causal pathway from ischemic stroke to atrial fibrillation: a network Mendelian randomization study.Mol Med. 2020 Jan 15;26(1):7

Response: We agree with the reviewer that it can be challenging to infer causality from observational study results because of potential for bias, including confounding. However, we disagree with the assumption that a prospective cohort study is inherently superior to a retrospective study. In our study, information about medication exposures came from computerized pharmacy data that are recorded prospectively, at the time of drug dispensing, well before any study outcomes occurred. Thus, using this approach to ascertain exposure avoids many of the limitations of other retrospective study designs (such as those that interview women after delivery about past exposures). Computerized pharmacy data are considered the “gold standard” for large scale pharmacoepidemiologic studies because they are prospectively recorded and shown to be more accurate than self-report or medication inventory. We have added a statement in the Methods about this (p. 7, paragraph 2).

We agree with the reviewer that Mendelian randomization is an approach with potential to overcome some types of bias. There are multiple potential approaches to strengthen causal inference, of which Mendelian randomization is one, but not the only one. In designing this study, we selected methods that are considered rigorous causal inference methods by statisticians and epidemiologists.[7, 8]

First, we designed our study to emulate a target trial,[9] defining covariates before exposure and outcomes after exposure, something made possible by the electronic pharmacy data used to define exposure in our study. Second, we used inverse probability of treatment weighting to account for measured confounding, which is specifically designed to estimate the average treatment effect, that is, to provide a comparison of outcomes that would be expected if the whole population were to be treated with one medication versus another. Our target estimands are based on contrasts of counterfactual outcomes, which is the essence of causal inference.[10]

We agree that alternative approaches such as Mendelian randomization or another type of instrumental variable analysis could be used. And just as our study relies on the assumption of no unmeasured confounding, these alternative methods also rely on untestable assumptions.[11, 12] There is no one correct way to analyze data from observational studies, but a multitude of approaches each with their own set of assumptions. We considered the data available to us and several different possible approaches before deciding on the causal inference approach of inverse probability weighting.

More specifically, responding to the request that we conduct an analysis using Mendelian randomization: We are not able to conduct such a study with the data we have available. We do not have access to genetic information for the over six thousand women in our cohort. In response to this review, we thought about how we would design such as study if genetic information on women were available. Designing such a study seems challenging, in particular because it seems possible that relevant mutations affecting genes involved in hypertension or its treatment (for instance, the receptors that beta blockers target) might prevent the development of hypertension in the first place, and so women with these mutations might simply not develop hypertension and might never require treatment in pregnancy. Our goal in this work was to help guide the choice of medications for women who do require treatment in pregnancy. Thus it seems possible that Mendelian randomization might not be suitable to address the specific clinical question we focused on.

Changes made to manuscript: we have added material in the Introduction about the need for rigorous methods that can support causal inference (p. 5, paragraph 2, last sentence). In the Methods, we have added material in the Overview section stating that we attempted to emulate a target trial and chose methods that can support causal inference (p. 6, paragraph 1). We also added material to point out that the medication exposure data were recorded prospectively and thus not subject to recall bias (p. 7, paragraph 2). We added material in the Discussion saying that Mendelian randomization is a complementary approach that could be explored to strengthen causal inference (p. 20, paragraph 1).

3. In the Discussion section, the authors should discuss the potential mechanisms that is behand the conclusion, including antihypertensive drug targets and biological pathway, genetic and drug-induced epigenetic/epi-transcriptomic prenatal origins of small for gestational age (SGA), preterm delivery, preeclampsia, and stillbirth.

Response: We appreciate the reviewer’s suggestion. Following this suggestion could lead to a very lengthy discussion because of the broad range of topics that the reviewer asks us to address. We have studied multiple drugs and multiple outcomes and the reviewer asks for discussion of multiple levels of mechanisms (genetic, epigenetic, epi-transcriptomic). This would be an interesting topic for a review article but is outside the scope of the current paper. In addition, the editor has asked us to make the Discussion more concise.

We have added a paragraph in the Discussion focused on mechanisms by which labetalol and methyldopa might affect risk of SGA, because that is the outcome for which we have the most notable finding. See p. 17, paragraph 1.

Reviewer #2:

1. An excellent work that shed light on some of the effects of antihypertensive drugs used in pregnancy.

As the authors stated it is very difficult to get such information from randomized controlled presepective studies with such large numbers. This study help this gap in literature as well as it paves the way for other studies that may prospectively look at the effect of methydopa on SGA compared to other antihypertensive drugs.

Response: Thank you. We appreciate your interest in our work.

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Response: Thank you, we have reviewed carefully and made changes if needed.

2. Please include your tables as part of your main manuscript and remove the individual files. Please note that supplementary tables (should remain/ be uploaded) as separate "supporting information" files.

Response: Thank you. We have confirmed that we are meeting those requirements.

3. Thank you for stating the following in the Competing Interests section:

[I have read the journal's policy and the authors of this manuscript have the following competing interests:

SD received a grant to support this work from the National Institute on Child Health and Human Development. She has also received grant support from GSK for unrelated work.

LC received a postdoctoral fellowship from the Group Health Foundation. She is now employed by Genentech (a member of Roche Group).

ZBC is now employed by Roche Pharmaceuticals.

TRE has consulted for Alnylam Pharmaceuticals, DiabetOmics, and Ferring Pharmaceuticals.

SMS has received grant funding through her institutions from Syneos Health.

LAA received funding through her institution from Bausch Health Companies and KR from Novartis and Merck & Co.].

Please confirm that this does not alter your adherence to all PLOS ONE policies on sharing data and materials, by including the following statement: "This does not alter our adherence to PLOS ONE policies on sharing data and materials.” (as detailed online in our guide for authors http://journals.plos.org/plosone/s/competing-interests). If there are restrictions on sharing of data and/or materials, please state these. Please note that we cannot proceed with consideration of your article until this information has been declared.

Please include your updated Competing Interests statement in your cover letter; we will change the online submission form on your behalf.

Response: we have added the updated Competing Interests statement in the new cover letter.

4. We noted in your submission details that a portion of your manuscript may have been presented or published elsewhere. [Results were presented as an oral presentation at the International Conference on Pharmacoepidemiology, 35th annual meeting, in Philadelphia, Pennsylvania, from August 24-28, 2019 and additional results at the virtual International Conference on Pharmacoepidemiology, 36th annual meeting, September 16-17, 2020. ] Please clarify whether this conference proceeding was peer-reviewed and formally published. If this work was previously peer-reviewed and published, in the cover letter please provide the reason that this work does not constitute dual publication and should be included in the current manuscript.

Response: We submitted an abstract to each conference, not a full manuscript. The abstracts were reviewed and rated by anonymous peer reviewers. The abstracts were published. Each abstract is very brief and does not contain anything approaching the depth of information contained in the manuscript. Thus, this work does not constitute dual publication. We have added a statement in the cover letter about this.

5. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially identifying or sensitive patient information) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. Please see http://www.bmj.com/content/340/bmj.c181.long for guidelines on how to de-identify and prepare clinical data for publication. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

Response: In the cover letter, we have described the restrictions on sharing a de-identified dataset and the process through which we can share data, and we provide contact information for the study data access committee, an institutional official, and an institutional review board.

6. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

Response: We have reviewed the reference list and no changes were needed. No papers were retracted.

REFERENCES

1. Abalos E, Duley L, Steyn DW, Gialdini C. Antihypertensive drug therapy for mild to moderate hypertension during pregnancy. The Cochrane database of systematic reviews. 2018;10(10):Cd002252.

2. Meidahl Petersen K, Jimenez-Solem E, Andersen JT, Petersen M, Brødbæk K, Køber L, et al. β-Blocker treatment during pregnancy and adverse pregnancy outcomes: a nationwide population-based cohort study. BMJ open. 2012;2(4).

3. Orbach H, Matok I, Gorodischer R, Sheiner E, Daniel S, Wiznitzer A, et al. Hypertension and antihypertensive drugs in pregnancy and perinatal outcomes. American journal of obstetrics and gynecology. 2013;208(4):301.e1-6.

4. Magee LA, von Dadelszen P, Singer J, Lee T, Rey E, Ross S, et al. Do labetalol and methyldopa have different effects on pregnancy outcome? Analysis of data from the Control of Hypertension In Pregnancy Study (CHIPS) trial. BJOG : an international journal of obstetrics and gynaecology. 2016;123(7):1143-51.

5. Su CY, Lin HC, Cheng HC, Yen AM, Chen YH, Kao S. Pregnancy outcomes of anti-hypertensives for women with chronic hypertension: a population-based study. PloS One. 2013;8(2):e53844.

6. Xie RH, Guo Y, Krewski D, Mattison D, Walker MC, Nerenberg K, et al. Association between labetalol use for hypertension in pregnancy and adverse infant outcomes. European journal of obstetrics, gynecology, and reproductive biology. 2014;175:124-8.

7. Hernán MA, Robins JM. Estimating causal effects from epidemiological data. Journal of epidemiology and community health. 2006;60(7):578-86.

8. Hernán MA, Robins JM. Causal Inference: What If. 1st ed. Boca Raton: Chapman & Hall/CRC; 2020.

9. Hernán MA, Robins JM. Using Big Data to Emulate a Target Trial When a Randomized Trial Is Not Available. American journal of epidemiology. 2016;183(8):758-64.

10. Pearl J. Causality: models, reasoning and inference. New York, N.Y.: Cambridge University Press; 2000.

11. Swanson SA, Tiemeier H, Ikram MA, Hernán MA. Nature as a Trialist?: Deconstructing the Analogy Between Mendelian Randomization and Randomized Trials. Epidemiology. 2017;28(5):653-9.

12. Hernán MA, Robins JM. Instruments for causal inference: an epidemiologist's dream? Epidemiology. 2006;17(4):360-72.

Submitted filename: Response to Reviewers_2022 02 17.docx

Click here for additional data file.

27 Apr 2022

Maternal and neonatal outcomes of antihypertensive treatment in pregnancy: A retrospective cohort study

PONE-D-21-28668R1

Dear Dr. Dublin,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Zhong-Cheng Luo

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

6 May 2022

PONE-D-21-28668R1

Maternal and neonatal outcomes of antihypertensive treatment in pregnancy: A retrospective cohort study

Dear Dr. Dublin:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Zhong-Cheng Luo

Academic Editor

PLOS ONE