Introduction

Sleep-disordered breathing (SDB) encompasses a spectrum of disorders characterised by increased upper airway resistance during sleep, and ranges from snoring to obstructive sleep apnoea (OSA). OSA occurs when the soft tissues of the upper airway collapse repeatedly during sleep, causing obstruction and cessation of airflow, often resulting in falls in blood oxygen saturation. SDB is being increasingly recognised as a potential risk factor in pregnancy, with reports of significant associations with gestational diabetes mellitus,[1] impaired fetal growth[2] and preterm birth.[3] Moreover physiological changes in pregnancy may increase the risk of SDB: these include weight gain, oedema, reduced upper airway dimensions,[4] and increased upper airway collapsibility due to hormonal changes.[5]

In the non-pregnant population, moderate-to-severe OSA is associated with hypertension and cardiovascular disease[6,7] due to a nocturnal cyclic pattern of hypoxia and hypercapnia, and recurrent surges of vasoconstriction. The degree of OSA appears to be linearly related to cardiovascular risk.[8] Accumulating evidence suggests this relationship may also exist in pregnancy with several studies reporting an association between SDB and gestational hypertensive disorders,[3,9,10] although obesity can act as a confounder to this relationship.[11] Hypertensive disorders of pregnancy (HDP), particularly preeclampsia, are associated with placental dysfunction resulting in impaired fetal growth and increased risks of acute compromise.[12–15] Whether co-existing SDB may increase these risks is currently unknown.

SDB may plausibly impact on fetal health in two ways. Firstly, recurrent episodes of hypoxaemia and cortical arousal lead to sympathetic activation and inflammation, leading to endothelial dysfunction.[16] The resultant placental dysfunction may lead to impaired fetal growth.[2] Secondly, such fetuses may be particularly sensitive to maternal hypoxaemia, which may lead directly to fetal hypoxia by reducing the availability of oxygen in the intervillous space. If sufficiently severe, this may result in fetal acidosis and asphyxial injury with potential long-term neurodevelopmental impairment and stillbirth.[17–19]

Few studies have investigated the impact of objectively-measured SDB on the health of the fetus, and results to date have been conflicting.[2,20–25] This may relate, in part, to challenges adjusting for comorbidities such as increasing BMI, which may mask FGR among women with SDB. For example, infants deemed appropriate for gestational age by standard growth curves may in fact be growth restricted after customisation for maternal obesity.[26,27] No studies to date have investigated the effect of SDB on the wellbeing of the fetus among mothers diagnosed with HDP. This is important because these fetuses might be expected to be more vulnerable to the sympathetic activation, endothelial dysfunction and hypoxemia that accompanies SDB. Given that HDP is a leading contributor to iatrogenic preterm birth, it would be helpful if we could identify reversible co-morbidities such as SDB that might enable safer prolongation of gestation. In this study, we proposed to determine whether the presence of objectively-confirmed SDB compromised the growth and well-being of the fetus among women with, and without, HDP.

Materials and methods

This data is from a single-centre prospective matched case-control study conducted between October 2012 and October 2015, comparing women with the diagnosis of HDP to healthy women with uncomplicated pregnancies. The Human Research Ethics Committees at Austin Health, Mercy Hospital for Women and University of Melbourne approved the study and written informed consent was obtained from all participants.

Cases included women with a singleton pregnancy diagnosed with gestational hypertension (GH) or preeclampsia (PE) not requiring immediate delivery between 26 and 37 weeks’ gestation, and were recruited from the Pregnancy Day Assessment Centre or were inpatients at Mercy Hospital for Women, Melbourne, Australia. Hypertension in pregnancy was defined as systolic BP greater than or equal to 140mmHg and/or diastolic BP greater than or equal to 90mmHg, confirmed by a minimum of two readings over several hours. GH was defined as the new onset of hypertension after 20 weeks of gestation.[28] We used the most recent International Society for the Study of Hypertension in Pregnancy (ISSHP) definition of PE: new hypertension after 20 weeks gestation and one or more of the following new-onset conditions: i) proteinuria, ii) other maternal organ dysfunction (renal insufficiency, liver involvement or neurological complications), and iii) uteroplacental dysfunction–fetal growth restriction.[28] Women with chronic hypertension who developed superimposed PE were also eligible to participate as a PE case.

Control participants with normotensive uncomplicated pregnancies were one-to-one matched by BMI to each of the cases (within ±4 kg/m², measured at the first antenatal appointment), and recruited from the antenatal outpatient clinic. Exclusion criteria included <18 years of age, multiple gestation, fetal abnormality or other maternal/fetal condition likely to mandate early or imminent delivery, and previous diagnosis of a sleep disorder.

GH and PE participants underwent overnight full polysomnography (PSG, ‘sleep study’) at their earliest convenience, to establish the presence or absence of SDB. Control women underwent PSG within ±4 weeks of gestational age to their matched hypertensive case participant. Attended overnight PSG was conducted in the Austin Health sleep laboratory using the Compumedics E series (Abbotsford, Victoria, Australia), or if preferred, unattended in the participant’s home with the Somté (Compumedics) portable sleep-monitoring device. Portable sleep monitoring systems are commonly used in clinical settings, and have been shown to have a high level of agreement with standard laboratory-based systems.[29,30] Inpatients were studied using the portable device. As per the American Academy of Sleep Medicine (AASM) criteria,[31] respiratory events were categorised as apnoeas (a decrease in airflow of ≥90% from baseline for ≥10 sec); hypopnoeas (decrease in airflow ≥30% from baseline for ≥10 sec and followed by either an oxygen desaturation of ≥3% or an EEG cortical arousal); and respiratory event related arousals (RERAs; a sequence of breaths lasting ≥10 sec characterised by increasing respiratory effort leading to an arousal from sleep). The number of apnoeas and/or hypopnoeas and/or RERAs per hour of sleep was calculated as the respiratory disturbance index (RDI). SDB was defined as an RDI of ≥5 events per hour, with secondary analyses performed with more significant SDB defined as RDI ≥15. SDB severity is classified as mild (RDI 5–14.9/hr), moderate (RDI 15–29.9/hr) and severe (RDI ≥30) in sleep literature.[32] The oxygen desaturation index (ODI) ≥3% was defined as the number of arterial oxygen desaturations of ≥3% from baseline, per hour of sleep.[31]

Time-synchronised cardiotocography (CTG) was performed using the Monica AN24 fetal heart rate (FHR) monitor (Monica Healthcare Ltd.) during the PSG to enable correlation of any FHR abnormalities with respiratory events and/or oxygen desaturations. All FHR traces were reviewed the following morning to ensure no abnormalities were present that mandated further management. CTG success was calculated as ‘minutes of FHR trace during sleep/minutes of sleep*100’ to give a value between 0 and 100%. The criteria for CTGs to be included in analysis was i) an overall CTG success rate of >50%, and ii) at least 3 x 1-hour blocks with >50% asleep and >80% successful FHR trace. This inclusion criterion was chosen as the FHR was difficult to interpret where there were large sections of missing data, and potential for FHR events to be missed.

A priori criteria for an abnormal FHR event potentially signifying fetal distress were as follows: Event 1 –prolonged bradycardia (defined as >15 bpm below baseline for ≥90 seconds and <5 minutes), Event 2 –recurrent severe variable decelerations (defined as a fall of >60 bpm from previous baseline and of >60 seconds duration, and at least 2 per 2 hours), Event 3 –repeated unprovoked or late decelerations accompanied by tachycardia or loss of variability, and Event 4 –decelerations lasting between 60–90 seconds and dropping below the baseline by >15 bpm.[33]

Fetal growth was monitored with ultrasound in the third trimester soon after recruitment, to enable sufficient time between third trimester growth assessment and delivery to identify slowing of fetal growth trajectory. The ultrasound estimate of fetal weight along with birthweight following delivery was customised for maternal height, pre-pregnancy (or if unknown, early pregnancy) weight, ethnicity, parity and fetal sex using the Australian dataset of the GROW software (www.gestation.net).[34]

Fetal growth restriction (FGR) was defined as a customised birthweight <10th centile for gestational age. Evidence of slowed third trimester growth was defined as a fall in customised centile of greater than a third (>33% decrease) between the third trimester ultrasound and birth. Slowed third trimester growth was calculated only for those participants where at least four weeks had lapsed between the third trimester ultrasound and birth. To facilitate standardised comparison, a rate of change in customised centile per day was calculated for each participant, and then multiplied by 42 to generate the estimated change in customised centile that would occur over exactly six weeks. This was chosen based on our previous studies examining maternal SDB and its impact on fetal growth,[2] and the importance of standardising assessment of fetal growth velocity.[35] Medical records were reviewed after birth to collect relevant delivery and perinatal outcomes.

Venous cord blood was collected at delivery and analysed for the fetal growth regulators IGF-1, IGF-2, IGFBP-1 and IGFBP-2. Free IGF-1 and IGF-2 determinations were performed using MILLIPLEX MAP Human IGF-I, II Magnetic Bead Panel kit (Millipore, Billerica, MA, USA). To separate IGFs from their binding proteins, all plasma samples underwent an acid-ethanol extraction procedure before measurement according to the manufacturer's instructions. All assays were read using the Bio-Plex workstation (Bio-Rad Laboratories, Hercules, CA, USA) and results analysed with Bio-Plex Manager (version 4.1.1) software. Intraassay and interassay coefficients of variation were less than 10%.

Results

A total of 87 pregnant women participated; a flowchart illustrating the data available at each step of analysis is shown in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flow chart of participant data available for each fetal outcome.

FDIU = fetal death in utero, FGR = fetal growth restriction, GH = gestational hypertension, PSG = polysomnography, CTG = cardiotocography, T3 = third trimester, US = ultrasound, PE = preeclampsia, CPAP = continuous positive airway pressure.

https://doi.org/10.1371/journal.pone.0229568.g001

Due to the case-control matched study design, there were no differences in age, parity, gestational diabetes, BMI and gestation at sleep study for the HDP versus normotensive control group (Table 1). SDB was diagnosed in 52.5% of the HDP group compared to 38.1% of the control group (p = .19) and there was no difference in RDI between the groups (Table 1). Once allocated into groups based on SDB, all demographic variables remained well-matched except for BMI at the time of the sleep study which was higher amongst those with SDB compared to No SDB in the HDP group. At the time of the sleep study, 23 of the hypertensive participants were taking anti-hypertensive medication. After undergoing PSG, three participants diagnosed with severe SDB commenced continuous positive airway pressure (CPAP) treatment and were subsequently excluded from all analyses apart from CTG analysis. As shown in Fig 1, perinatal outcomes were available on all other participants, but only 58 (73%) had greater than four weeks between growth scan and delivery enabling assessment of growth trajectory. Fifty-nine (72%) had satisfactory CTG recording which met our inclusion criteria and 56 (71%) provided a cord blood sample.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Demographics for hypertensive disorders of pregnancy cases and normotensive controls stratified by SDB status.

https://doi.org/10.1371/journal.pone.0229568.t001

The impact of SDB on acute measures of fetal well-being: Cardiotocography

Twenty (33.9%) of the CTGs showed one or more of the four events indicating a potential episode of fetal compromise during sleep, but these were confined to Events 1 (prolonged bradycardia) and 4 (deceleration lasting between 60–90 sec with >15 bpm fall from baseline). The number of overall CTG events overnight per participant ranged from 0–5. As detailed in Table 2, there was no impact of HDP on the number of participants displaying any type of FHR event, or the total number of events per night (p = .78 and p = .19 respectively). Further, there was no impact of SDB, expressed as a dichotomous variable, on the number of HDP or normotensive participants displaying fetal heart rate events, nor the total number of events overnight.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Cardiotocography variables for hypertensive disorders of pregnancy cases and normotensive controls stratified by SDB.

https://doi.org/10.1371/journal.pone.0229568.t002

Significant univariate predictors of FHR decelerations overnight were nulliparity, earlier gestational age, diagnosis of PE and FGR at birth, but not presence or severity of SDB in terms of RDI or ODI3 (S1 Table). A stepwise selection model confirmed FGR at birth as the strongest predictor of number of FHR events per hour on CTG (χ² = 18.63, p = .001, R² = 0.33; Table 3), with the diagnosis of PE no longer a significant factor. Over 60% of mothers with a growth-restricted infant displayed evidence of CTG abnormalities during sleep, versus only 26% of well-grown fetuses (p = .04). SDB did not exacerbate FHR events in FGR fetuses (number of events—SDB = 2.7 ± 1.5 vs. No SDB = 2.8 ± 1.6, p = .23) overall, although severity of SDB became a significant predictor of FHR deceleration in the multivariate model (Table 3). This relationship was limited to the control group (r = .44, p = .02, n = 28); examination of the linear relationship revealed that only normotensive women with an RDI ≥3.5 were observed to have adverse nocturnal FHR events.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Factors associated with number of fetal heart rate events per hour on cardiotocography on stepwise ordinal regression modelling.

https://doi.org/10.1371/journal.pone.0229568.t003

The impact of SDB on fetal growth and perinatal outcomes

As expected, women with HDP had infants with a lower birthweight, a higher incidence of FGR at birth and evidence of lower customised birthweight centiles and slowed fetal growth between the third trimester ultrasound and birth, compared to normotensive controls (Table 4). However, the co-existence of SDB with HDP did not appear to impact further on birthweight or fetal growth. In fact, among the women with HDP, co-existing SDB was associated with significantly better perinatal outcomes: larger customised birthweight centiles and fewer FGR at birth than those without SDB. This association persisted even after adjustment for the higher proportion of PE participants in the No SDB group (p = .052). This observation was not seen among control participants where there were no significant differences in birthweight centile or fetal growth between the SDB and No SDB groups (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Measures of fetal growth and perinatal outcomes for those with and without SDB in the hypertensive disorders of pregnancy and control groups.

https://doi.org/10.1371/journal.pone.0229568.t004

Similarly, women with HDP had a significantly higher risk of preterm birth, caesarean delivery and SCN or NICU admission compared to normotensive controls (Table 4). Again, co-existing SDB did not amplify the risks for those with HDP, rather SDB was associated with a lower incidence of pre-term birth. Apart from a higher rate of SCN admission, there were no adverse perinatal outcomes for normotensive women with SDB.

Secondary analysis of participants with an RDI ≥15 showed no negative effects of moderate to severe SDB on fetal growth or perinatal outcomes, again showing evidence of larger customised birthweight centile and fewer FGR at birth, compared to mothers with an RDI <15 (S2 Table).

The impact of SDB on biomarkers of fetal growth

After adjusting for gestational age, there was no difference in markers of fetal growth between HDP and normotensive women (Table 5). The gestational age at which the cord blood sample was taken did not differ by SDB status. As shown in Fig 2, amongst those with HDP there was no impact of SDB on the IGF axis. However, in the control group, the participants with SDB had significantly higher IGF-1, IGF-2 and lower IGFBP-2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Mean ± SEM insulin-like growth factor 1 and 2 (IGF-1, IGF-2) and insulin-like growth factor binding protein 1 and 2 (IGFBP-1, IGFBP 2) for the SDB vs. No SDB group within the HDP and control group (*p value = .02).

HDP group–SDB n = 17, No SDB n = 13. Control group SDB n = 8, No SDB n = 17, except for IGFBP-1 where HDP group–No SDB n = 12, Control group No SDB n = 16. SDB = sleep-disordered breathing, HDP = hypertensive disorders of pregnancy.

https://doi.org/10.1371/journal.pone.0229568.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Biomarkers of fetal growth in cord blood across hypertensive groups.

https://doi.org/10.1371/journal.pone.0229568.t005

Discussion

This is one of the few prospective studies that has combined comprehensive assessment of fetal growth and wellbeing with objectively-measured SDB, and the first study thus far to have examined this relationship in women with HDP. We found that women with HDP were more likely to have smaller infants with a higher incidence of FGR, however the presence of co-existing SDB did not adversely impact on fetal growth or size at birth. Similarly, fetuses known to have FGR were more likely to display nocturnal FHR decelerations but this was not exacerbated by the presence of SDB. Furthermore, no adverse impact on perinatal outcomes was seen in normotensive women with SDB.

FGR at birth was the strongest risk factor for FHR decelerations at night. Given that FGR is associated with severe underlying placental dysfunction, and strongly associated with adverse perinatal outcome including stillbirth,[36,37] it might be expected that FGR fetuses have the least oxygen reserve to withstand further physiological insults at night. This also underscores why night time may be a particularly risky time for FGR fetuses—the importance of maternal sleep has been highlighted by Warland et al.[38] finding that more than half of mothers experiencing stillbirth believed that their baby died at night. For this important reason, we had a particular interest in seeing whether maternal SDB would be more poorly tolerated by a fetus already suffering uteroplacental insufficiency, as demonstrated previously by Fung et al.[2] In our study however, FGR fetuses were not further adversely impacted by maternal SDB. This may be because the more severely growth-restricted fetuses were most difficult to obtain a successful CTG trace, and SDB in this cohort was generally mild. Nevertheless, this association between customised centile of <10% at birth and fetal heart rate events in late pregnancy provides an avenue to explore with regard to nocturnal FHR assessment as an early predictor of FGR risk at birth.

SDB severity, expressed as a continuous measure, was independently associated with nocturnal FHR events, mainly in normotensive women. Despite this, it should be kept in mind that the FHR events observed were mild, with the two most ‘severe’ types of FHR decelerations not observed at all (recurrent severe variable decelerations and repeated unprovoked decelerations or late decelerations). It is likely that the appropriately-grown fetus is well protected from transient maternal hypoxaemia, due to a combination of fetal adaptive behaviours such as increasing heart rate and cardiac output to increase O₂ uptake from its placenta,[39] and the high affinity fetal haemoglobin-oxygen dissociation curve.[40] Olivarez et al.[25] demonstrated that FHR events among 20 women diagnosed with SDB were variable decelerations appropriate for gestational age and therefore not pathologic. Two early case studies[41,42] demonstrating substantial FHR decelerations in association with maternal sleep apnoea presented women with very severe SDB–yet the seven women with the most severe disease (RDI>30) in our study still only demonstrated minor fetal events.

The fact that fetal outcomes were not worse in HDP associated with SDB is unexpected. The absence of any relationship may reflect relatively mild SDB disease; however the lack of association between SDB and poor fetal health remained even across the more severe disease subtype. While we may have been underpowered for many of the individual adverse outcomes, there were no apparent differences across fetal outcomes including fetal growth restriction, preterm birth, mode of delivery and NICU admission, when taken collectively.

Previous studies by Fung et al.[2] and Kneitel et al.[43] found that SDB was associated with slowed fetal growth across pregnancy rather than low birthweight. While we found SDB had no pathological effects on fetal growth trajectory, this may have been due to the population targeted and study design. A limitation of monitoring third trimester growth in HDP, particularly PE, is that the maternal and fetal condition often mandates very pre-term delivery. Hence, the third trimester growth trajectory analysis only included fetuses well enough to complete a third trimester ultrasound and remain in utero for at least another two weeks before delivery. Consequently a number of FGR fetuses, particularly within the No SDB group, were excluded.

One interesting result from this study was that hypertensive women with at least mild SDB had larger newborns even after fetal weight was customised for maternal characteristics such as obesity, and they were less likely to be premature. While this may be a chance finding, it is intriguing to consider whether the presence of SDB may confer a better prognosis in HDP. An association between SDB and increased birthweight has been demonstrated recently in women with mild OSA,[44] with a potential explanation relating to an association between increased placental weight and maternal OSA severity, possibly mediated by placental leptin overexpression.[45] Despite no increase in birthweight, we also demonstrated heightened levels of important fetal growth regulators IGF-1 and IGF-2 with a corresponding decrease in IGFBP-2 in normotensive women with SDB. Among women with preeclampsia, the intermittent hypoxic challenge of SDB may result in a less severe placental phenotype than the chronic placental dysfunction and hypoxia/ischaemia model classically associated with PE. HDP linked to SDB may have a more benign course with placental and maternal endothelial re-oxygenation daily, whereas PE resultant from early uteroplacental insufficiency would be chronically hypoxic. While we acknowledge this interpretation is speculative, it would be amenable to further investigation in appropriately designed laboratory studies.

The contrasting findings in the hypertensive and normotensive mothers on the impact of SDB on fetal growth measures contribute to the current uncertainties reported in the literature. Studies have reported varying results: ranging from a link between diagnosed maternal SDB and small for gestational age infants[21,46] markers of placental tissue hypoxia[47] and lower levels of estriol, a marker of fetoplacental unit wellbeing,[48] through to no relationship between SDB and fetal growth.[20,23,24,49] These inconsistencies may relate to inadequate adjustment for confounding variables, the predominance of mild SDB or differing methodologies, however it is clear the answer remains elusive and further study is warranted.

The lack of association between SDB and adverse perinatal outcomes in our study helps inform the debate regarding treatment recommendations. Even if it were feasible to screen and treat all pregnant women affected by SDB, adherence to CPAP may be poor and any benefit to the fetus may only be modest. In our study, of the 15 women with moderate to severe SDB who attended follow-up, only three women accepted CPAP treatment. This problem has been identified in other studies[23] and suggests that stronger links between SDB and poor pregnancy outcomes need to be demonstrated before pregnant women regard treatment as an acceptably high priority.

Supporting information

Acknowledgments

Thank you to the staff at the antenatal clinics at Mercy Hospital for Women and the Sleep Laboratory at Austin Health for their support of this project. In particular, thank you to Ms. Gabrielle Pell for her assistance with participant recruitment, Ms. Pavlina Toman for polysomnography analysis and Dr. Hannah Skrzypek for cardiotocography analysis.