Rodent models of gestational SDB show metabolic aberrancies in offspring growth and development

Rodent models (mice and rats) of the effects of gestational SDB allow for mechanistic investigation of the independent effects of SF and IH as separate exposures to the developing offspring. In the human SDB condition, it is not possible to isolate the effects of IH vs SF due to co-occurrence and interacting effects. While all of the studies in rodents were randomized control trials, most of the reports lack description of the key bias indicators such as similarity of groups at baseline, sequence generation and random housing (Table S1).

Intermittent hypoxia: Investigators reported either no difference in BW, ^{Reference Khalyfa, Cortese and Qiao49,Reference McDonald, Dempsey and O’Halloran50} significantly lower BW ^{Reference Gozal, Reeves, Row, Neville, Guo and Lipton51} or growth restriction ^{Reference Iqbal and Ciriello52} in offspring exposed to gestational IH during pregnancy compared to control animals. In two studies, IH-exposed offspring had lower BW but rapid catch-up growth, ^{Reference Gozal, Reeves, Row, Neville, Guo and Lipton51,Reference Iqbal and Ciriello52} which has been shown to be related to increased risk of childhood obesity in humans. ^{Reference Ong, Ahmed, Emmett, Preece and Dunger53} At 6–12 weeks of life, male offspring exposed to gestational IH had greater fat mass compared to controls. ^{Reference Iqbal and Ciriello52} Khalyfa et al. found that male, but not female, offspring exposed late gestational IH (days 13–18 of gestation) had higher body weight and visceral adipose tissue at 24 weeks compared to controls. ^{Reference Khalyfa, Cortese and Qiao49} HOMA-IR and plasma leptin were higher, and plasma adiponectin was lower in these male offspring exposed to maternal IH compared to controls at 24 weeks, ^{Reference Khalyfa, Cortese and Qiao49} suggesting that there may be sex differences in exposure outcomes. The investigators further reported upregulation of gene methylation pathways associated with energy production in adipose tissue of IH-exposed offspring, suggesting a potential mechanism by which IH-exposure may result in increased risk of obesity. ^{Reference Khalyfa, Cortese and Qiao49}

Sleep fragmentation. Exposure to SF in utero may have deleterious effects on offspring metabolism which do not manifest until later in life. In mouse offspring exposed to gestational SF, BW was not different than controls. ^{Reference Cortese, Khalyfa, Bao, Andrade and Gozal54–Reference Khalyfa, Mutskov, Carreras, Khalyfa, Hakim and Gozal56} However, as the mice entered adulthood (16–18 weeks of life), offspring of SF-exposed mothers had significantly higher mean body weight, fat mass, blood glucose, and higher fasting total triglyceride and cholesterol concentrations than controls. ^{Reference Mutskov, Khalyfa, Wang, Carreras, Nobrega and Gozal55,Reference Khalyfa, Mutskov, Carreras, Khalyfa, Hakim and Gozal56} The negative metabolic effects of gestational SF on offspring were more pronounced in males, once again, supporting sex differences. ^{Reference Khalyfa, Carreras, Almendros, Hakim and Gozal57} Similar to IH, changes in gene methylation patterns for PPARα activation (critical for fatty acid oxidation, ketogenesis, ^{Reference Sander58} and hepatic gluconeogenesis) were seen in male offspring exposed to SF compared to controls. ^{Reference Khalyfa, Carreras, Almendros, Hakim and Gozal57} In a separate study, offspring exposed to gestational SF had epigenetic changes resulting in overexpression of the FoxO1 gene, a transcription factor that plays an important role in gluconeogenesis and glycogenolysis. ^{Reference Mutskov, Khalyfa, Wang, Carreras, Nobrega and Gozal55} Taken together, these findings suggest that SF exposure, especially in males, changes the offspring’s ability to regulate glucose and lipid metabolism, increasing risk of future metabolic disease. Alterations in offspring metabolism may, in part, be due to an integrated stress response involving cortisol and sympathetic activation. ^{Reference Trzepizur, Khalyfa, Qiao, Popko and Gozal59} Offspring of double mutant mice of integrated stress response genes CHP and GADD34 (unable to mount a stress response) exposed to SF did not have metabolic perturbations as seen in normal mice offspring exposed to SF. ^{Reference Trzepizur, Khalyfa, Qiao, Popko and Gozal59} Importantly, postnatal physical activity implemented early in SF-exposed offspring reduced FoxO1 acetylation and methylation to levels similar to controls. ^{Reference Mutskov, Khalyfa, Wang, Carreras, Nobrega and Gozal55}

The evidence from rodent models is compelling; however, it must be underscored that unlike humans, mice and rodents are born with very little fat and size at birth is in part due to the size of the litter and driven more by lean body mass. In summary, evidence from these rodent models suggests distinct mechanisms whereby maternal IH results in normal or low BW and subsequent accelerated postnatal growth, and maternal SF results in normal BW followed by later metabolic derangement.

Human studies: strong link between SDB and birthweight outcomes

Over 30 published reports described a strong association between maternal SDB and birth outcomes in humans, supporting that mechanisms and outcomes in rodent models may have translational relevance. In accordance with the DOHaD model, infants born with IUGR/SGA or LGA are both at risk for future metabolic diseases such as obesity and T2D. ^{Reference Hales and Barker60} Notably, while the purpose and study population were clearly defined in most reports, most lacked power for infant outcomes, made a single assessment of SDB during pregnancy (instead of multiple measures over gestation), did not measure severity of SDB, and body composition at birth was rarely measured (Table 3). In 32 of the 48 studies, investigators attempted to at least partially control for important confounding variables: maternal age, gestational age, BMI, smoking, and preexisting hypertension, which could highly influence growth outcomes (Table 3). It is also notable that all humans studies are descriptive in nature, so cause and effect cannot be determined. Further, even after attempting to control for known confounders, there may still be unmeasureable or unknown variables which can influence study outcomes.

IUGR/SGA. 27 studies were reviewed in which SGA or IUGR were reported as outcomes, three of which were meta-analyses. Ding et al. reported a significant association between SDB and IUGR across 8923 women in 11 studies (OR = 1.44 [95% CI: 1.22–1.71]). ^{Reference Ding, Wu and Xu61} Similarly, Warland et al. reported by meta-analysis that SGA was significantly associated with SDB across 9478 women (13 studies) in which SDB was measured subjectively (adjusted OR = 1.6 [1.1–2.2]) and across 56,423,715 women (7 studies) which included objective measures (adjusted OR = 1.4 [1.1–1.9]). ^{Reference Warland, Dorrian, Morrison and O’Brien62} In contrast, Brown et al. found no significant association between SDB and SGA by meta-analysis of 21 studies (adjusted OR 1.19, [0.94–1.51]). ^{Reference Brown, Turner and Kumar63} Findings were mixed across the 24 original (non-meta-analysis) studies reviewed for this manuscript. While investigators reported an association between SDB and growth restriction in nine studies, ^{Reference Franklin, Holmgren, Jonsson, Poromaa, Stenlund and Svanborg7,Reference Chen, Kang, Lin, Wang, Keller and Lin9,Reference O’Brien, Bullough and Owusu13,Reference Fung, Wilson and Lappas31,Reference Micheli, Komninos and Bagkeris64–Reference Miyagawa, Emori, Kawano, Sakurai and Tanigawa68} no significant association was detected across 15 studies. ^{Reference Bourjeily, Raker, Chalhoub and Miller4,Reference Louis, Auckley, Sokol and Mercer12,Reference Facco, Ouyang and Zee69–Reference Okun and O’Brien81} Mixed findings may be due to lack of consistent assessment of SDB, the severity of the SDB and provocation of IH, and a lack of adequate power (Table 3). Timing of maternal SDB assessment varied widely across studies with the majority of subjective SDB assessment done at time of delivery. Type of SDB assessment was also highly varied; the majority of studies included self-reported symptoms of snoring, which does not reflect severity of SDB. Of the 19 studies, only five ^{Reference O’Brien, Bullough and Owusu13,Reference Fung, Wilson and Lappas31,Reference Pamidi, Marc and Simoneau65,Reference Olivarez, Ferres and Antony71,Reference Yin, Williams and Burton78} were powered to detect an association between SDB and SGA as the primary outcome; a positive association was reported in three of the studies and no association was reported in two (Table 2). In the majority of the studies (n = 20), investigators controlled (at least in part) for important maternal variables (BMI, hypertension, diabetes, smoking) that could influence infant growth outcomes (Table 3). Studies in which severity of OSA is measured objectively throughout gestation (pre-pregnancy through 3rd trimester) in relation to SGA are lacking. In summary, the results from descriptive studies in humans are mixed in support of an association between SDB and SGA which may be due to a lack of consistency of measurement of SDB, a lack of power or other confounding variables that were not measured.

BW: Results from three meta-analyses supported a relationship between SDB and low BW with similar ORs (OR = 1.75 [95% CI: 1.33–2.32], ^{Reference Ding, Wu and Xu61} 1.67 [1.0–2.78], ^{Reference Brown, Turner and Kumar63} 1.39 [1.14–1.65] ^{Reference Pamidi, Pinto, Marc, Benedetti, Schwartzman and Kimoff82} ), while Li et al. concluded that there was no association between SDB and BW in a meta-analysis across 8749 women (15 studies). ^{Reference Li, Zhao, Hua and Li83} OSA diagnosis (documented by ICD-9) was associated with an increased risk of low BW (<2500 g; OR = 1.76 [1.28–2.40], n = 4746). ^{Reference Chen, Kang, Lin, Wang, Keller and Lin9} Further, self-reported frequent snoring in the 3rd trimester was associated with low BW (RR = 2.6 [1.2–5.4]); however, after adjustment for pre-maturity, snoring was no longer associated with low BW. ^{Reference Micheli, Komninos and Bagkeris64} SDB was not associated with increased risk of low BW in three smaller studies, ^{Reference Owusu, Anderson and Coleman84–Reference Sharma, Nehra and Sinha86} none of which were powered for BW as the outcome. Louis et al. reported that compared to women with obesity alone, women with OSA had infants with significantly lower BW (3288 ± 590 vs. 3013 ± 968 g). ^{Reference Louis, Auckley, Sokol and Mercer12} Diagnosis of OSA was made either before or during pregnancy and continuous positive airway pressure use/adherence was not reported, important because treatment of OSA could have affected BW in the OSA group. In three studies where BW was compared between women with OSA and those without, one group of investigators reported higher BW in women with OSA ^{Reference Olivarez, Ferres and Antony71} and two reported no significant difference. ^{Reference Sahin, Koken and Cosar74,Reference Louis, Auckley and Miladinovic87} Higgins et al. reported significantly higher BW in infants of women with a positive Berlin Score (high risk of OSA), ^{Reference Higgins, Leong, Park, Facco, McCarthy and Wong88} but Ugur et al. reported slightly lower BW in infants of women with a positive Berlin score. ^{Reference Ugur, Boynukalin, Atak, Ustuner, Atakan and Baykal89} Discrepant findings may be due to multiple factors. There was inconsistent timing of SDB measurement throughout the studies from pre-pregnancy to delivery, severity was not consistently assessed, and the degree of IH was usually unreported. BW was used differently across studies; five of the studies used low BW (<2500 g) as a categorical variable ^{Reference Chen, Kang, Lin, Wang, Keller and Lin9,Reference Micheli, Komninos and Bagkeris64,Reference Owusu, Anderson and Coleman84–Reference Sharma, Nehra and Sinha86} (requiring a larger sample to detect an association), while the other 10 studies used BW as a continuous variable. ^{Reference Ayrim, Keskin, Ozol, Onaran, Yiidirim and Kafali1,Reference Olivarez, Ferres and Antony71,Reference Sahin, Koken and Cosar74,Reference Tauman, Sivan, Katsav, Greenfeld and Many77,Reference Louis, Auckley and Miladinovic87,Reference Bassan, Uliel-Sibony, Katsav, Farber and Tauman90–Reference Leung, Hui, Leung, Yuen and Lau92} Further, none of the studies were powered on the BW outcome and body composition was not assessed. Importantly, low BW is different from IUGR as it is not corrected for gestational age (i.e., a preterm infant is expected to have a lower BW but may or may not have growth restriction). In seven of the studies found, investigators made an effort to at least partially adjust for confounding factors, while in nine of the studies, there was no control for confounding factors (Table 3). In summary, overall results from studies support an association between SDB and low BW, but, in over half of the studies, potential confounding variables were not controlled thus limiting the strength of evidence to support this association.

LGA: One meta-analysis supported an association between SDB and LGA across 678,310 women (7 studies) (OR = 1.6 [95% CI: 1.3–1.9]), ^{Reference Warland, Dorrian, Morrison and O’Brien62} while the additional four meta-analyses did not include LGA as an outcome. In eight studies, investigators reported an association between SDB and LGA. ^{Reference O’Brien, Bullough and Owusu13,Reference Facco, Ouyang and Zee69,Reference Telerant, Dunietz, Many and Tauman79–Reference Okun and O’Brien81,Reference Bin, Cistulli and Ford93,Reference Howe, Signal and Paine94} Seven groups found a significant association between SDB symptoms and LGA (BW > 90th percentile) and one group reported a significant association between SDB and macrosomia (defined as BW > 97th percentile). Maternal diagnosis of OSA in the year before or during pregnancy (across 636,227 women) was significantly associated with increased risk of LGA in a large, retrospective cohort (adjusted OR = 1.27 [1.04–1.55], p < 0.05), after adjustment for obesity. ^{Reference Bin, Cistulli and Ford93} Similarly, women who self-reported apnea had increased frequency of LGA compared to women who denied apnea (22.4% vs. 12.8%, p = 0.049). ^{Reference O’Brien, Bullough and Owusu13} Maternal report of pregnancy-onset breathing pauses during sleep was associated with LGA (OR = 3.5 [95% CI: 1.3–9.6], p = 0.01). ^{Reference Howe, Signal and Paine94} OSA+ women (identified via ICD-9 codes) had higher rates of macrosomia (BW > 97th percentile), but this group had a significantly higher rate of maternal diabetes and higher body mass index (BMI) which could also explain the macrosomia rate. ^{Reference Ravishankar, Bourjeily, Lambert-Messerlian, He, De Paepe and Gundogan73} Telerent et al. reported that the occurrence of LGA was significantly more frequent in infants born to mothers who had mild sleep apnea compared to those who did not. ^{Reference Telerant, Dunietz, Many and Tauman79} The investigators studied women in their 3rd trimester and excluded women with more severe sleep apnea (apnea hypopnea index [AHI] > 15), which might be expected to result in placental insufficiency and attenuated growth. ^{Reference Telerant, Dunietz, Many and Tauman79} Facco and colleagues, on the other hand, did not find an association between OSA (diagnosed using in-home objective measurement) and LGA (19.4% LGA in no SDB group vs. 12.5% in Moderate/Severe SDB group). ^{Reference Facco, Ouyang and Zee69} The lack of significant finding for this study may be due to the fact that it was designed to look at maternal and not fetal outcomes and a higher cutoff for definition of LGA (>95th percentile). Across studies, there are limited data on the classes of obesity, and no measures of newborn body composition. We recently completed a study in which we examined the relationship between severity of maternal OSA and infant % body fat in maternal obesity (BMI 30–40). There was no correlation between severity of OSA (by AHI) and infant % body fat (r = 0.34, p > 0.05) ^{Reference Farabi, Barbour, Heiss, Hirsch, Dunn and Hernandez95} ; however, overnight minimum oxygen saturation was correlated with infant % body fat, suggesting that increasing severity of maternal OSA (with greater reduction oxygen saturation) was related to higher newborn fat (r = −0.63, p = 0.02). ^{Reference Farabi, Barbour, Heiss, Hirsch, Dunn and Hernandez95} Importantly, none of the studies were powered on LGA, the assessment of SDB symptoms were variable and often did not include the degree of hypoxia, and the timing of assessment across studies was inconsistent. However, in seven of the eight studies, investigators attempted to control for confounding factors (Table 3). The available evidence supports an association between more severe maternal SDB and fetal overgrowth even after controlling for maternal BMI (in most studies).

Clinical studies investigating potential mechanisms

Some investigators studied the relationship between SDB in pregnancy and biomarkers that may indicate potential mechanisms for the link between SDB in pregnancy and fetal outcomes.

SDB and fetal stress. Maternal apneas were accompanied by fetal heart rate decelerations in three of four women with OSA, ^{Reference Sahin, Koken and Cosar74} suggesting increased fetal stress with maternal apnea. In contrast, Olivarez and colleagues found no association between degree of oxygen saturation and fetal heart rate in 20 women with OSA (severity of OSA not reported). ^{Reference Olivarez, Maheshwari and McCarthy96}

SDB, inflammation and oxidative stress. Koken and colleagues ^{Reference Koken, Sahin and Cosar97} reported that markers of oxidative stress (maldondialdehyde [MDA], Myeloperoxidase [MPO]) were higher and an anti-oxidant marker (Glutathione peroxidase [GSH-Px]) was lower in plasma of women who reported snoring compared to non-snorers. Of note, women who snored were older with higher BMI, which may confound the relationships reported, and the gestational week of blood collection was not controlled. In another study, women who reported snoring during pregnancy (n = 48) had higher levels of the inflammatory cytokine IL-6 in cord blood compared to women who denied snoring (n = 75). ^{Reference Tauman, Many and Deutsch98} Erythropoietin and nucleated (immature) red blood cell counts were higher in the cord blood of women who reported snoring, ^{Reference Tauman, Many and Deutsch98} supporting that exposure to SDB in utero may result in hypoxic changes in the fetus. In contrast, Khan and colleagues reported in a retrospective study that women with diagnosed OSA (vs. self-reported non-snorers) had lower oxidative and carbonyl stress markers (Advanced glycation end products [AGE] and advanced oxidation protein products [AOPP]), but higher anti-oxidative stress markers (total antioxidant capacity [TAC]). ^{Reference Khan, Lambert-Messerlian and Monteiro99} Taken together, these findings support that SDB, and especially OSA, may increase fetal exposure to oxidative stress, inflammation, and hypoxia-induced compensatory enhanced erythropoiesis, however further study is warranted as the findings are inconsistent. This is interesting as inflammation and oxidative stress may exacerbate IR. ^{Reference Chen, Chen, Wang and Liang100,Reference Khodabandehloo, Gorgani-Firuzjaee, Panahi and Meshkani101}

SDB and placental changes. Placental expression of carbonic anhydrase IX (CAIX), a marker of hypoxia, was higher in women who snored or with OSA compared to non-snoring women (n = 47). ^{Reference Ravishankar, Bourjeily, Lambert-Messerlian, He, De Paepe and Gundogan73} Fetal normoblastemia (immature red blood cells), a sign of exposure to hypoxia, was significantly higher in placentas of women who snored/had OSA. However, women who snored and women with OSA had significantly higher BMI and higher rates of chronic hypertension and diabetes (vs. non-snorers), which likely contributed to differences in biomarkers (preeclampsia rates were similar between groups). ^{Reference Ravishankar, Bourjeily, Lambert-Messerlian, He, De Paepe and Gundogan73} Women with OSA had lower levels of placenta-associated plasma protein-A (PAPP-A) ^{Reference Bourjeily, Curran, Butterfield, Maredia, Carpenter and Lambert-Messerlian102} and lower estriol levels ^{Reference Bourjeily, Butterfield, Curran and Lambert-Messerlian103} compared to women without OSA, suggesting poorer placental and fetal well-being, which could affect growth. Salameh et al., however, showed no differences in markers of fetal well being (PAPP-A, AFP, uE3, hCG, inhibin-A) between snorers and non-snorers. ^{Reference Salameh, Lee and Palomaki104} Women with mild OSA (mean AHI 7.8 ± 2.7), objectively measured in the 3rd trimester, had significantly higher placental weight and higher placental leptin mRNA expression compared to women without OSA, even after controlling for maternal BMI. ^{Reference Kidron, Bar-Lev, Tsarfaty, Ariel and Tauman105} Placental weight was positively correlated with infant adiposity (measured by skinfolds). ^{Reference Kidron, Bar-Lev, Tsarfaty, Ariel and Tauman105} This was a small study and a power analysis was not reported; however, the findings support that mild OSA may result in placental changes that contribute to increased infant adiposity. The effect of OSA on maternal IR was not explored in any of these studies as a possible mechanism for the increased in adiposity.

SDB and chronic disease risk. Salihu and colleagues reported that women at high risk of OSA (measured by Berlin Questionnaire) had neonates with shorter telomeres than women at low risk of OSA (n = 64 women). ^{Reference Salihu, King and Patel106} Telomere length is indicative of chromosomal aging and has been proposed as a mechanism contributing to chronic disease development, ^{Reference D’Mello, Ross, Briel, Anand, Gerstein and Pare107} suggesting that exposure to OSA could have long-term implications in accordance with DoHad theory. ^{Reference Hanson and Gluckman42}

Long-term implications of SDB exposure in humans. In a single longitudinal study, BW between infants born to mothers with SDB were similar to controls. ^{Reference Tauman, Zuk and Uliel-Sibony108} However, despite similar neurological development, at 1 year, mothers with SDB vs. control were more likely to report infant snoring (41.7% vs. 7.5%, p = 0.004). ^{Reference Tauman, Zuk and Uliel-Sibony108} Human studies in which body composition of the infant, and longer term metabolic phenotyping of offspring into childhood and young adulthood are needed to elucidate if SDB exposure that increases risk for future offspring metabolic disease as studies in rodents suggest.