Predicting perinatal outcomes with an obstructive sleep apnea screening tool

Abstract

The objective was to determine if a screening tool for obstructive sleep apnea could be used to predict adverse perinatal outcomes. This was a prospective observational study of patients receiving prenatal care and universally screened for obstructive sleep apnea with the STOP Questionnaire (four questions related to Snoring, Tiredness during daytime, Observed apnea, and high blood Pressure). Confounding variables were included in a backwards logistic regression model to predict adverse perinatal outcomes. The study population of 442 women had positive STOP screens (64; 14.5%) associated with preterm delivery and neonatal intensive care unit admissions. For preterm delivery, history of preterm delivery was the strongest predictor with odds ratios of 4.2 (95% confidence interval 2.0–8.8; p < 0.001), followed by STOP, odds ratios 2.8 (95% confidence interval 1.4–5.8; p = 0.004) and nulliparity, odds ratios 2.3 (95% confidence interval 1.2–4.4; p = 0.013). A positive STOP was the only significant predictor for neonatal intensive care unit admissions, odds ratios 2.5 (95% confidence interval 1.1–5.7; p = 0.036). STOP screening test performance indicated low sensitivity but high specificity: preterm delivery (28.3%, 87.4%), neonatal intensive care unit admissions (27.3%, 86.6%), low birth weight (25.0%, 86.9%), and preeclampsia (16.7%, 85.6%). As a stand-alone tool, the STOP Questionnaire has limited performance, but could be explored in combination with other factors that might increase sensitivity to predict preterm delivery and neonatal intensive care unit admission.

Introduction

Obstructive sleep apnea (OSA) is a condition characterized by recurrent upper airway obstruction, hypoventilation, and intermittent nocturnal hypoxia. Given the increasing incidence of obesity among reproductive-aged women, OSA presents unique complications during pregnancy including increased risk for gestational diabetes, preeclampsia, cesarean delivery, preterm delivery (PTD), fetal growth restriction (FGR), wound complications, neonatal intensive care unit (NICU) admission, and hyperbilrubinemia.^1–3 The most accurate method to diagnose OSA remains polysomnography, but its cost and prolonged time involvement presents clear limitations. ⁴ Screening questionnaires such as the Berlin, modified Berlin, Epworth Sleepiness Scale, OSA50 questionnaires, sleep apnea clinical scores, STOP-Bang, and STOP are tools for identifying potential OSA. ⁵ In a recent systemic review, the STOP Questionnaire (four questions related to Snoring, Tiredness during daytime, Observed apnea, and high blood Pressure) was compared to other screening tools in nine studies with varying populations using polysomnography validation and had the highest prediction sensitivity (100%), specificity (92%), and negative predictive value (100%) for moderate OSA. ⁶ The purpose of this study was to determine if the STOP screening tool could be used to predict adverse perinatal outcomes.

Patients/Methods

The research method employed was a prospective observational study of patients receiving prenatal care at a tertiary care clinic during a 1-year time period (June 2013 to May 2014). Patients were universally screened for OSA using the STOP Questionnaire, which consists of four concise questions: (1) Do you s nore loudly? (2) Do you often feel tired, fatigued, or sleepy during the daytime? (3) Has anyone o bserved you stop breathing during sleep or do you choke at night? (4) Do you have or have you been treated for high blood p ressure? Patients who answered “yes” to two or more of the questions or answered yes to number 3 were a positive screen and were referred to the hospital's Sleep Center for polysomnography.

Univariate analysis on the following adverse perinatal outcomes: cesarean delivery, preeclampsia, PTD (<37 weeks), low birth weight (LBW) (<2500 g), FGR, and NICU admission, compared positive versus negative screens. Significant outcome variables were included in a backwards logistic regression with the following confounding variables: maternal age, body mass index, parity, previous stillbirth, previous PTD, maternal ethnicity, hypertension disorder, diabetes, gastroesophageal reflux disease, asthma, chronic obstructive pulmonary disease, depression, thyroid disorder, polyhydramnios, oligohydramnios, history of stroke, history of heart attack, and history of heart failure.

Results

Included in analysis were 442 (84.8%) patients with a mean age of 25.3 ± 5.6 years and high tobacco (241; 54.5%) and illicit drug use (155; 35.1%). The mean body mass index of those who screened positive was statistically significantly higher than of those that screened negative (33.0 ± 9.5 vs 29.4 ± 7.9; p = 0.003). In addition, rates of hypertension (21.9% vs 4.2%; p < 0.001), asthma (25% vs 13.5%; p = 0.023), thyroid disease (10.9% vs 3.2%; p = 0.012), substance use (45.3% vs 33.3%; p = 0.045), and depression (37.5% vs 20.9%; p = 0.005), were statistically significantly higher in women screening positive versus negative. Positive STOP screens were assessed in 64 (14.5%) patients and were statistically significantly associated with more PTD (23.4% vs 10.1%; p = 0.006, significance was adjusted to 0.008 due to the number of tests), LBW infants (20.3% vs 10.3%; p = 0.023) and NICU admissions (14.1% vs 6.3%, p = 0.039). Backward logistic regression indicated that a positive STOP screen was associated with PTD and NICU admission. For PTD, history of PTD was the strongest predictor with an odds ratio (OR) of 4.2 (95% confidence interval (CI) 2.0–8.8), followed by a positive STOP screen with an OR 2.8 (95% CI 1.4–5.8) and nulliparity with an OR 2.3 (95% CI 1.2–4.4). For NICU admission, a positive STOP was the only significant predictor with 2.5 OR (95% CI 1.1–5.7; p = 0.036).

The positive STOP screening test performance (Table 1) had a specificity around 86% for all outcomes, but was low in sensitivity (preterm 28.3%, NICU 27.3%, and LBW 25%). False positive rate for all outcomes was less than 15%.

Table 1.

Positive screening test performance for perinatal outcomes.

Outcome	Sensitivity (%) (95% CI)	Specificity (%) (95% CI)	PPV (%) (95% CI)	NPV (%) (95% CI)
Preeclampsia	16.7 (13.2–20.1)	85.6 (82.3–88.9)	4.7 (0.2–9.2)	96.0 (91.8–100)
FGR	8.7 (6.1–11.3)	85.2 (81.9–88.5)	3.1(–0.6–6.8)	94.4 (89.5–99.4)
Cesarean	13.5 (10.4–16.7)	85.0 (81.7–88.3)	32.8 (22.8–42.8)	64.6 (52.4–76.7)
Preterm	28.3 (24.1–32.5)	87.4 (84.3–90.5)	23.4 (14.6–32.3)	89.9 (83.4–96.5)
LBW	25.0 (21.0–29.0)	86.9 (83.8–90.1)	20.3 (11.9–28.8)	89.7 (83.0–96.4)
NICU	27.3 (23.1–31.4)	86.6 (83.4–89.7)	14.1 (6.7–21.4)	93.7 (88.4–98.9)

PPV: positive predictive value; NPV: negative predictive value; CI: confidence interval; FGR: fetal growth restriction; LBW: low birth weight; NICU: neonatal intensive care unit.

Discussion

Our study corroborated others that found a positive STOP screen associated with a statistically significant increased risk of LBW, PTD, and NICU admission, and those that found that a positive STOP screen was predictive of PTD and NICU admission using logistic regression. However, a positive STOP screen was not associated with caesarean delivery, FGR, or cardiovascular outcome perhaps because of the high prevalence of obesity among the study population.

While self-assessments or screening tools for OSA have independently been shown to be predictive of PTD, they have not included the STOP questionnaire. Our finding of a positive STOP screen predicting NICU admissions, with an OR of 2.5, is similar to the results of a study that found women with sleep disorder breathing had a higher rate of infants admitted to the NICU (OR = 1.90; CI: 1.38–2.61) ⁷ and a systematic review reporting a significant association between sleep disorder breathing and NICU admission (OR = 2.43; 95% CI, 1.61–3.68; p < 0.001). ⁸ This study is an example of finding reasonable association between the screening tool and outcome, but screening performance is too low and therefore does not pass the test of being a worthwhile screening tool. ⁹

Our study had several strengths (a prospective design, moderate sample size and inclusion of confounding variables) and limitations (single center, conducted several years ago (2013–2014). There was potential inaccuracy in screening results as responses were self-reported and the clinic setting might not be representative of all obstetric clinics because of higher than usual rates of tobacco and substance use. With higher rates of overweight and obese patients, there could be an obesity-related effect that was not accounted for during our analyses. There may be other confounding variables we may not have considered that could impact birth outcome.

Conclusion

In summary, our study suggests that the STOP screening questionnaire as a stand-alone tool for predicting perinatal outcomes has limited performance. However, it could be explored in combination with other factors, especially if this can increase the sensitivity for preterm birth and NICU admission where the sensitivity for STOP alone was modest. OSA is likely to become more prevalent during pregnancy with the worldwide trend in increasing obesity rates. Thus, there will be an even greater need to screen for OSA.