Is Gestational Hypertension Protective against Perinatal Mortality in Twin Pregnancies?

Background Pregnancy-induced or gestational hypertension is a common pregnancy complication. Paradoxically, gestational hypertension has been associated with a protective effect against perinatal mortality in twin pregnancies in analytic models (logistic regression) without accounting for survival time. Whether this effect is real remains uncertain. This study aimed to validate the impact of gestational hypertension on perinatal mortality in twin pregnancies using a survival analysis approach. Methods This was a retrospective cohort study of 278,821 twin pregnancies, using the U.S. 1995–2000 matched multiple birth dataset (the largest dataset available for multiple births). Cox proportional hazard models were applied to estimate the adjusted hazard ratios (aHR) of perinatal death (stillbirth and neonatal death) comparing gestational hypertensive vs. non-hypertensive pregnancies controlling for maternal characteristics and twin cluster-level dependence. Results Comparing births in gestational hypertensive vs. non-hypertensive twin pregnancies, perinatal mortality rates were significantly lower (1.20% vs. 3.38%), so were neonatal mortality (0.72% vs. 2.30%) and stillbirth (0.48% vs. 1.10%) rates. The aHRs (95% confidence intervals) were 0.34 (0.31–0.38) for perinatal death, 0.31 (0.27–0.34) for neonatal death, and 0.45 (0.38–0.53) for stillbirth, respectively. The protective effect of gestational hypertension against perinatal death became weaker over advancing gestational age; the aHRs in very preterm (<32 weeks), mild preterm (32–36 weeks) and term (37+ weeks) births were 0.29, 0.48 and 0.76, respectively. The largest risk reductions in neonatal mortality were observed for infections and immaturity-related conditions. Conclusions Gestational hypertension appears to be beneficial for fetal survival in twin pregnancies, especially in those ending more prematurely or for deaths due to infections and immaturity-related conditions. Prospective studies are required to rule out the possibility of unmeasured confounders.


Introduction
Pregnancy-induced hypertension or gestational hypertension is a common pregnancy complication affecting 5% to 10% of all pregnancies [1]. It occurs more frequently in multiple pregnancies most of which (about 97%-98%) are twin pregnancies [2,3]. Multiple pregnancies have become increasingly common in recent decades [4], but few studies have addressed the potential differential needs in the management of pregnancy-specific complications in multiple vs. singleton pregnancies. In the presence of multiple fetuses, the maternal/placental burdens to support more fetuses create a substantial physiological difference between multiple vs. singleton pregnancies that may explain the elevated risks of some gestational-specific complications in multiple pregnancies. In case of gestational hypertension, the maternal elevation in blood pressure may be considered a developmental competing survival mechanism to increase blood/nutrients supply to the fetus, but this is a delicate risks/benefits balance because high blood pressure itself poses a risk to the mother and fetus [5]. This developmental need for higher blood pressure to increase fetal blood/nutrients supply would be likely stronger in the presence of multiple fetuses. Therefore, it is a plausible hypothesis that the development of gestational hypertension may be beneficial to fetal survival in multiple pregnancies with a much greater need for blood/nutrients supply. Indeed, gestational hypertension has been associated with a reduced risk of neonatal death in twin but not singleton pregnancies [6]. A reduced risk of fetal and infant mortality has been reported in gestational hypertensive vs. nonhypertensive twin pregnancies ending in preterm births [7]. In contrast, gestational hypertension has been generally associated with adverse birth outcomes in singleton pregnancies [6,8]. If this paradoxically beneficial effect of gestational hypertension on fetal survival in multiple pregnancies is real, there may be a need for differential clinical management recommendations in multiple vs. singleton gestational hypertensive pregnancies. That is, the optimal policy for the management of gestational hypertension in singleton pregnancies may be not the optimal choice in multiple pregnancies. However, previous studies used logistic regression to model the risk of perinatal mortality without accounting for the duration of fetal survival [6,7]. In estimating the risk of perinatal mortality, survival analysis accounting for the duration of survival would be a more efficient and appropriate model. This study aimed to validate the impact of gestational hypertension on the risk of perinatal mortality in twin pregnancies using a survival analysis approach.

Methods
This was a retrospective cohort study using the U.S. matched multiple birth dataset 1995-2000 provided by the National Center for Health Statistics (NCHS) [9]. It is the largest linked dataset available for multiple births. Subjects included in the present study must meet the following inclusion criteria: 1) births in twin pregnancies with non-missing value for gestational hypertension; 2) pregnancies without chronic hypertension; 3) gestational age at delivery between 20 and 42 weeks inclusive (twin births outside of this gestational age range are extremely rare, and likely recording/ transcription errors). In cases one twin had valid data but the other twin did not or was missing in a twin set, the twin with valid data remained in the analysis cohort. Therefore, the total number of births in the final study cohort is slightly less than 2 times the total number of pregnancies. We did not exclude births with missing data on birth weight (n = 8615) (1.6% of all births) because a significant proportion of perinatal deaths (7187 of 17779) (40.4%) had missing data on birth weight. The main interest is the risk of perinatal death which could be estimated irrespective of missing data on birth weight. Birth weight was set to missing for births at very extreme birth weights (,500 g, or .6000 g) (probably recording or transcription errors). Births with missing data on birth weight were allowed to drop out in birth weight-specific analyses. A total of 278,821 twin pregnancies (555,457 twin births) constituted the final study cohort. Research ethics approval was waived by the Sainte-Justine hospital research ethics board because the study was based on the anonymized matched multiple birth dataset downloadable from the NCHS website.
During the study period in the U.S. 1995-2000, the diagnosis of pregnancy-induced or gestational hypertension was according to the then commonly accepted criteria: two or more occasions in blood diastolic pressure $90 mmHg or systolic pressure $ 140 mmHg, taken $4 hours apart and occurring after 20 weeks of gestation without proteinuria [10]. In the NCHS multiple birth database, it is impossible to distinguish between gestational hypertension (without proteinuria) and preeclampsia (gestational hypertension with proteinuria).
The primary outcome was perinatal death including stillbirth (fetal deaths at 20 or more weeks gestation) and neonatal death (deaths during the first 4 weeks of postnatal life). We also examined cause-specific neonatal mortality (congenital anomalies, asphyxia, immaturity related conditions, infections, sudden infant death syndrome, others) according to the International Collaborative Effort (ICE) on Perinatal and Infant Mortality [11]. Causes of death are missing for all stillbirths in the NCHS birth data.
The data analysis unit was the mother for maternal/pregnancy variables (e.g. gestational hypertension), but the fetus/newborn for perinatal outcomes (perinatal death, stillbirth, neonatal death). Marginal Cox regression models for clustered data [12] were applied to estimate the crude and adjusted hazard ratios (aHR) and 95% confidence interval (CI) of perinatal death, stillbirth and neonatal death comparing gestational hypertensive vs. nonhypertensive pregnancies accounting for twin set-cluster level dependence. The survival time (time to event or outcome) variable in the Cox models was gestational age (in weeks) for stillbirth, and gestational age at delivery plus the number of postnatal surviving weeks for neonatal deaths, and gestational age at delivery plus 4 weeks for all surviving babies (right-censored). The aHR were adjusted for maternal race, age, marital status, parity, smoking, other reported maternal illnesses, fetal sex and mode of delivery.
To gain insight into clinical characteristics of risk changes, the HRs of perinatal death were examined by important clinical characteristics: mode of delivery (caesarean, vaginal), gestational age (very preterm ,31 weeks, mild preterm 32-36 weeks, term $ 37 weeks), birth weight (very low ,1500 g, low 1500-2499 g, normal $2500 g), and fetal growth -small, appropriate or large for gestational age (SGA ,10 th percentile, AGA 10-90 th percentiles, LGA .90 th percentile) according to sex-and gestational age-specific birth weight percentiles for non-malformation births to mothers without smoking and without any reported major maternal illness in the study cohort.
Births with missing data on co-variable were allowed to drop out in all multivariate adjustment models, except for smoking and mode of delivery for which the proportion of missing was high (17.8% for smoking, 36.4% for mode of delivery); the missing was included as a valid category to avoid large numbers of drop-outs in the adjusted risk models. For gestational age specific analysis, fetus-at-risk approach was applied to avoid a potential ''collider'' effect of stratification by gestational age at birth [13]. The fetusesat-risk denominator is the number of all fetuses at risk of death (both born and yet unborn babies). For example, in calculating perinatal mortality for births at 32-36 weeks gestation, those yet unborn babies (births at .36 weeks) were considered at risk and included in the denominator. All data management and analyses were carried out using Statistical Analysis System, version 9.2 (SAS Institute, Cary, North Carolina) [14]. Two-tailed p values ,0.05 were considered statistically significant.
Stratified analyses by gestational age revealed that the protective effect of gestational hypertension against perinatal death diminished as pregnancy approached term. The aHR of perinatal death comparing gestational hypertensive vs. non-hypertensive pregnancies was 0.29 for very preterm births, 0.48 for mild preterm births, and 0.76 for term births, respectively (all p,0.001) ( Table 2). Similarly, the aHR for neonatal death was 0.26 for very preterm births, 0.47 for mild preterm births, and 0.76 for term births, respectively (Tables 3). For stillbirth, the risk difference became non-significant for term births in hypertensive vs. non-hypertensive pregnancies (Table 4). Stratified analyses by birth weight or birth weight for gestational age also showed a tapering protective effect of gestational hypertension against perinatal death with increasing birth weight; low birth weight, SGA and AGA births experienced a significant protective effect, while the risk difference became non-significant for newborns reaching normal birth weight ( Table 2). Stratified analyses by mode of delivery indicated that the protective effect was stronger for vaginal births (aHR = 0.20) than for caesarean births (aHR = 0.55).
Cause-specific neonatal mortality analyses showed that gestational hypertension was associated with a significantly lower risk of neonatal death due to any cause (congenital anomalies, immaturity related conditions, asphyxia, infections, other causes) other than sudden infant death syndrome ( Table 5). The strongest protective effects were observed for death due to infections (aHR = 0.23) and immaturity-related conditions (aHR = 0.26).
Among the 45,272 births in gestational hypertensive pregnancies, 2945 births (6.5%) were in pregnancies complicated by both gestational hypertension and diabetes. The results were very similar if these 2945 births were excluded from the analyses. If births with any reported congenital anomalies (n = 12,905) were excluded from the analyses, similar protect effects were observed for gestational hypertension against perinatal death, stillbirth and neonatal death. Compared gestational hypertensive vs. nonhypertensive twin pregnancies, the aHR (95% CI) were 0.33

Main Findings
To our knowledge, this is the first report on the association of gestational hypertension with perinatal mortality in twin pregnancies using a survival analysis approach. The results suggest a protective effect of gestational hypertension against perinatal death in twin pregnancies. This protective effect appears to diminish as pregnancy advances towards term, and become non-significant for newborns reaching normal birth weight. Also, we first observed that the strongest risk reductions in neonatal mortality comparing gestational hypertensive vs. non-hypertensive pregnancies were for deaths due to immaturity-related conditions and infections.

Comparisons with Findings in Previous Studies
Although gestational hypertension is generally associated with adverse neonatal outcomes in singletons [6,8], the same condition has been associated with a beneficial impact on the survival of newborns in twin pregnancies [6,7]. The current study has overcome a major limitation in previous studies based on logistic regression to model the mortality risk ignoring a crucial piece of information -survival time. In time-dependent circumstances, Cox proportional hazards models are more efficient than logistic regression models in ascertaining the true risk differences [15]. A previous study reported that the risk reductions in stillbirth and neonatal death may be limited to preterm twins comparing gestational hypertensive vs. non-hypertensive pregnancies [7]. In contrast, we observed moderate and statistically significant risk reductions in overall perinatal mortality and neonatal morality in both term and preterm twins.
The protective effect of gestational hypertension against perinatal death in twins seems to be counterintuitive, since hypertension is a known risk factor of adverse perinatal outcomes in singletons [6,8]. The reasons and mechanisms underlying the lower risk of perinatal death in gestational hypertensive vs. nonhypertensive twin pregnancies are unclear. One may speculate that non-hypertensive twin pregnancies may experience a higher perinatal mortality because they may be more frequently complicated by other serious maternal illnesses or birth defects than gestational hypertensive twin pregnancies. However, this is not the case. To the contrary, we observed that other maternal illnesses were more frequent in gestational hypertensive vs. nonhypertensive pregnancies, while reported birth defects were similarly frequent (Table 1). Alternatively, one may speculate that twin fetuses may benefit from better blood/nutrients supply to promote better fetal growth in gestational hypertensive pregnancies. This appears not to be the case too. We observed a higher rate of SGA birth in gestational hypertensive vs. non-hypertensive pregnancies. Perinatal mortality rate was lower despite higher rates of preterm and SGA births in gestational hypertensive pregnancies, suggesting that higher blood pressure may carry some benefits to twin fetuses other than promoting fetal growth. Interestingly, cause-specific mortality risk analyses revealed that the strongest protective effects of gestational hypertension against neonatal death were for deaths due to immaturity-related conditions and infections. This raises the possibility that higher blood pressure may promote fetal maturity and immune functional development. However, this interpretation is purely speculative. Another possibility is that anti-hypertensive medications may be somewhat beneficial for fetal survival. Indeed, it has long been suspected that labetalol, a common drug in the treatment of gestational hypertension, may promote fetal lung maturation to lower perinatal mortality [16]. Of various anti-hypertensive drugs, it remains uncertain which agent is the best in terms of risks/ benefits in the treatment of hypertension in pregnancy [17]. More research data are needed on the impact of anti-hypertensive agents on perinatal mortality.
The diminishing benefit of gestational hypertension against perinatal death with advancing gestational age in twin pregnancies could be a consequence of persistent high blood pressure, because high blood pressure may strain the placenta over time. Unfortunately, we had no data on the timing of onset and duration of gestational hypertension and blood pressure levels. Irrespective of gestational hypertension, pregnancies ending in preterm births are unhealthy by definition compared to those ending in term births. Alternatively, one may speculate that non-hypertensive pregnancies ending in preterm births may be even sicker than hypertensive pregnancies ending in preterm births, resulting in a stronger observed protective effect of gestational hypertension against perinatal death in preterm twins. However, even among preterm births, we observed a higher rate of other serious maternal illnesses in gestational hypertensive vs. non-hypertensive twin pregnancies. There is a possibility that other unreported serious maternal illnesses or fetal problems might have been more common in nonhypertensive pregnancies ending in preterm births. Large prospective cohort studies are required to clarify this question and rule out the possibility of unmeasured confounders.
The observed protective effect of gestational hypertension against perinatal mortality is similar to the recently reported protective effect of diabetes in pregnancy (mostly, gestational diabetes) against perinatal mortality in twin pregnancies [18]. We speculate that gestational-specific elevations in both blood pressure and glucose may represent a competing developmental survival mechanism for fetuses to extract more nutrients from the mother that may confer a perinatal survival benefit in multiple pregnancies.

Limitations
The NCHS birth database could not allow a distinction between gestational hypertensive (without proteinuria) and preeclampsia (with proteinuria). We had no data on the severity of gestational hypertension and other associated adverse conditions. Severe hypertension or preeclampsia is a more serious condition that is associated with adverse outcomes such as placental abruption and stillbirth [19]. It is unknown but likely that the protective effect of gestational hypertension against perinatal death in twins may be restricted to mild and wellcontrolled hypertensive pregnancies. This question may be answered in large prospective cohort studies with data to distinguish gestational hypertension from preeclampsia.