Erratum for: Bin YS, Cistulli PA, Ford JB. Population-based study of sleep apnea in pregnancy and maternal and infant outcomes. J Clin Sleep Med. 2016;12(6):871–877.
The authors have discovered an error which impacts on the results of the above published study. The study used linked population data in the form of hospital and birth records to compare birth outcomes for pregnant women with and without sleep apnea. The authors discovered a typographical error in the programming code used to select the relevant hospital records for analysis. This had the impact of including excess records and inflating the number of pregnant women with sleep apnea. On correcting the error, the authors found that although the absolute numbers have changed, the main findings and interpretation remain the same. The authors apologize for any inconvenience caused. A PDF of the original paper marked up with the corrections is available from the corresponding author (yusun.bin@sydney.edu.au). What follows are the corrected portions of the manuscript.
Abstract:
Last sentence of the Methods section revised to: “Modified Poisson regression models were used to examine associations between sleep apnea and each outcome taking into account maternal age, country of birth, socioeconomic disadvantage, smoking, obesity, parity, diabetes, hypertension, and plurality.”
Results section revised to: “Sleep apnea was significantly associated with pregnancy hypertension (adjusted RR 1.38; 95% CI 1.03–1.86), planned delivery (1.06; 1.02–1.11), preterm birth (1.64; 1.20–2.22), admission to neonatal intensive care/special care nursery (1.37; 1.13–1.66), large-for-gestational-age infants (1.47; 1.11–1.96) but not with gestational diabetes (1.41; 0.97–2.05), caesarean section (1.13; 0.98–1.32), perinatal death (1.16; 0.37–3.63), low Apgar at 5 minutes (1.71; 0.96–3.03), or small-forgestational-age infants (0.67; 0.41–1.09).”
Main Manuscript - Methods section:
Statistical Analysis sub-section, second paragraph, second sentence revised to: “For the outcome of gestational diabetes, only maternal age, country of birth, socioeconomic disadvantage, smoking, obesity, plurality (singleton/multiple pregnancy), and parity were included in the adjusted model.”
Statistical Analysis sub-section, third paragraph revised to: “For the outcome of pregnancy hypertension, we controlled for chronic hypertension and pre-existing diabetes in addition to maternal age, country of birth, socioeconomic disadvantage, smoking, obesity, plurality, and parity. For all other outcomes, the covariates of maternal age, country of birth, socioeconomic disadvantage, smoking, obesity, parity, chronic hypertension, pregnancy hypertension, preexisting diabetes, plurality, and gestational diabetes were included in the adjusted models.”
Main Manuscript - Results section revised to:
“Of the 636,227 women who delivered, 216 (0.03%) had a hospital admission with diagnosis of sleep apnea in the year before or during pregnancy. The characteristics of women with and without a hospital record for sleep apnea are presented in Table 1. Women with sleep apnea were older and more likely Australian-born than those without sleep apnea. Women with sleep apnea had higher rates of obesity, pre-existing diabetes, and chronic hypertension than without a diagnosis of sleep apnea. There was a tendency for more women with sleep apnea to be in their first pregnancy than those without sleep apnea, and more women with sleep apnea had twin and higher-order pregnancies.
Table 1.
Demographic, health, and pregnancy characteristics of 636,227 pregnant women with and without sleep apnea.
The associations between sleep apnea and the outcomes from the crude and fully adjusted models are presented in Table 2. In the crude models, women with sleep apnea were not more likely to have SGA infants or perinatal death. Sleep apnea was significantly associated with the other outcomes, although these associations were attenuated when maternal demographics and health risk factors were included in the models. Sleep apnea remained significantly predictive of pregnancy hypertension, planned delivery, preterm birth, NICU/SCN admission, and LGA infants. The association between preterm birth and sleep apnea was driven by planned deliveries, while risk of NICU/SCN admissions was significantly increased only in term infants.
Table 2.
Comparison of maternal and infant outcomes in 636,227 pregnant women with and without sleep apnea.
In the sensitivity analysis for gestational diabetes, which excluded women with pre-existing diabetes, the main result was unchanged, i.e. sleep apnea was not significantly associated with risk for gestational diabetes compared to no apnea (RR 1.48; 1.00–2.13).”
Main Manuscript - Discussion section:
Third paragraph, third sentence revised to: “It is unclear what might account for this disparity.”
Fifth paragraph revised to: “A novel finding in the present study was that sleep apnea was not significantly associated with perinatal death, which includes stillbirths and neonatal deaths. Clinical studies have been precluded from recruiting mothers with stillbirths due to ethical considerations. Louis and colleagues examined the association between OSA and stillbirth in maternal health records and found no significant association.33 That study was able to take into account more maternal comorbidities than the present study. However, that study did not include neonatal deaths, which are included in perinatal mortality here.”
Sixth paragraph, first sentence removed.
Seventh paragraph, third sentence revised to: “In the current study, this is consistent with the lower Apgar scores for these infants, reflecting poorer condition of the baby soon after birth.”
Strengths and Limitations sub-section, third paragraph revised to: “Previous studies have relied heavily on self-reported snoring as an indicator of sleep-disordered breathing, although only a small subset of those who snore will have significant sleep apnea.41 The use of objective sleep apnea diagnoses from hospital records improves on previous studies. The hospital records have high specificity (over 99%) for a range of medical conditions when compared to medical record review,23 meaning that the sleep apnea group likely contains true cases of clinically significant sleep-disordered breathing. 52.8% of those with sleep apnea also had a procedure code in the hospital records indicating they underwent a sleep test in the study period.
Strengths and Limitations sub-section, fourth paragraph, second sentence revised to: “Obstructive sleep apnea accounted for 63.8% of all cases in the current study, 32.8% had sleep apnea that was unspecified, and only 2.3% were coded as having central sleep apnea (see Table S1 in the supplemental material for details).”
Strengths and Limitations sub-section, fifth paragraph, first sentence revised to: “The rate of sleep apnea determined through hospital records was very low at 0.03%, although studies based on health records in the Taiwan and the United States report similarly low rates (0.03% and 0.01%, respectively).32,33”
Table S1.
ICD-10-AM codes used in the identification of sleep apnea.
Table 1 - revised below. Highlighted are those results where the significance of the between-group comparison has changed from the original:
Table 2 - revised below. Due to changes in Table 1, the covariates in the adjusted models have changed, namely the addition of parity and multiple pregnancy (plurality) as covariates in all models. Highlighted are the results where the significance of the between-group comparison has changed from the original.
Supplemental Material - revised below:
Table S2.
ICD-10-AM codes used in the identification of obesity.
