Barrett’s esophagus and the risk of obstructive sleep apnea: a case–control study

Esophageal adenocarcinoma (EAC) has dramatically increased in incidence in the United States over the last 3 decades [1, 2] and is strongly linked to its precursor metaplastic lesion, Barrett’s esophagus (BE), and gastroesophageal reflux disease (GERD) [3]. Current guidelines recommend considering endoscopic screening of patients with multiple risk factors for Barrett’s esophagus including chronic GERD symptoms [4]. Studies suggest that 8%-13% of patients with GERD have BE [5, 6], compared with a rate of 1%-6% in patients without heartburn symptoms [6, 7]. The risk of BE increases with increasing duration of reflux symptoms, particularly nocturnal symptoms [3, 8]. Moreover, GERD patients with BE typically have a greater duration of esophageal acid exposure than those without BE.

BE and esophageal adenocarcinoma are also associated with obesity [9–11]. The increase in obesity in the United States over the past several decades has been proposed as an explanation for the rising incidence of esophageal adenocarcinoma. Obesity and its associated health conditions represent a growing public health concern in the United States. Obesity-associated obstructive sleep apnea (OSA), for example, is an important albeit underdiagnosed syndrome with detrimental cardiovascular and neurocognitive effects. Like those at risk for BE, patients at risk for OSA are typically male, obese, and middle-aged.

Prior theory and evidence suggest that OSA is associated with gastroesophageal reflux disease. It has been theorized that negative intrapleural pressures generated during nocturnal airway obstruction in OSA could contribute to nocturnal gastroesophageal reflux [12, 13], although studies have not demonstrated a direct relationship between obstructive events and nocturnal GERD events [12, 14]. Alternatively, it is possible that obstructive events could lead to weakening at the lower esophageal sphincter over time [14]. A population-based study reported increased nocturnal GERD symptoms in patients with sleep-disordered breathing symptoms [15]. A large cohort study suggested that nocturnal GERD may be associated with increased body mass index and sleep complaints [16]. A study of 1093 OSA patients found that frequent nocturnal GERD was associated with more severe OSA, independent of body mass index [13]. Other studies have demonstrated that treatment of OSA with continuous positive airway pressure effectively treats nocturnal gastroesophageal reflux symptoms [17, 18]. Continuous positive airway pressure treatment in OSA patients with GERD significantly reduced nocturnal acid contact time as measured by esophageal pH monitoring [18]. Metabolic changes associated with OSA including increased oxidative stress [19] and alteration in serum levels of insulin and insulin-like growth factors [20] could predispose to BE. The role of nocturnal reflux in Barrett’s esophagus needs further investigation. Although some studies have suggested that nocturnal reflux symptoms are not associated with Barrett’s esophagus among GERD patients [21, 22], other studies have measured increased nocturnal esophageal acid exposure [23] or nocturnal non-acid reflux [24] in BE patients compared with GERD patients without BE.

Given the growing body of evidence associating OSA with gastroesophageal reflux disease, this pilot study aimed to determine whether patients with BE are more likely to be at risk for OSA. We hypothesized that BE patients would be more likely to score at high risk for OSA on a survey instrument for OSA screening compared with controls. We also aimed to determine if nocturnal gastroesophageal reflux symptoms affected the relationship between BE and high risk for OSA.

The study was approved by the University Hospitals Case Medical Center Institutional Review Board and was performed in accordance with the Declaration of Helsinki. We performed a pilot case–control study comparing risk for OSA between subjects with Barrett’s esophagus and 2 control groups. Risk for OSA was determined based on an addended version of the Berlin Questionnaire, a validated survey instrument used for OSA screening that identifies subjects at high risk for OSA [25, 26]. Two control groups identified from patients undergoing outpatient endoscopy were recruited including patients from another study as previously described [27]: 1) Subjects matched to BE cases by age ± 5 years, race, and gender with upper endoscopy negative for BE (hereafter referred to as the EGD group); and 2) Patients presenting for outpatient colonoscopy, including open-access cases, who had no prior upper endoscopy at our facility (hereafter referred to as the COL group). If possible, we matched up to 3 EGD control subjects per BE case subject by age (at time of survey) ± 5 years, race, and gender. The survey was distributed via mailings to BE patients who had previously undergone upper endoscopy and in person at the time of endoscopy. The vast majority of control subjects were recruited at the time of endoscopy, with 2 EGD control subjects completing their surveys < 14 days after an upper endoscopy negative for BE. Informed consent was obtained from all subjects.

Demographic characteristics, body mass indices, and GERD symptoms of the three comparison groups are presented in Table 1. Patients in the COL group were somewhat younger than the BE group (mean age, 57 years versus 63 years, p = 0.02). Although EGD patients had a significantly lower mean self-reported BMI than the BE group (27 kg/m² versus 30 kg/m²), that in the COL group was not significantly different from the BE group. Not surprisingly, patients in the COL group were less likely to be male and white. BE patients were significantly more likely to report nocturnal regurgitation or nocturnal heartburn symptoms than either of the control groups (Table 1). In addition, BE patients reported more severe typical heartburn and regurgitation symptoms than either the EGD or COL groups (Figure 1A and 1B).

	BE (n = 54)	EGD (n = 62)	P value (BE vs. EGD)	COL (n = 171)	P value (BE vs. COL)
Mean age, years (SD)	63 (11.1)	60 (11.7)	NS	57 (13.6)	0.001*
White (%)	53 (98.2)	59 (95.2)	NS	70 (41.9)†	<0.0001¥
Male (%)	38 (70.4)	36 (58.1)	NS	75 (43.9)	0.001°
Mean BMI, kg/m2 (SD)	30 (5.9)	27 (5.5)	0.03*	28 (5.6)	0.06*
Heartburn symptoms (%)	43 (79.6)	33 (53.2)	0.003°	53 (31.0)	< 0.0001°
Nocturnal heartburn symptoms (%)	27 (50.0)	16 (25.8)	0.01°	21 (12.3)	<0.0001°
Regurgitation symptoms (%)	42 (77.7)	38 (61.3)	0.06°	41 (24.0)	<0.0001°
Nocturnal regurgitation symptoms (%)	31 (57.4)	18 (29.0)	0.002°	18 (10.5)	<0.0001°

SD, Standard deviation

NS, Not significant

BMI, Body mass index

* Independent samples t test with equal variances

° Chi-square test

†Among respondents; 4 subjects did not respond to the question about race.

^¥ Fisher’s exact test

Although the proportion of BQ+ patients was not significantly different between the BE and EGD groups, BE patients were more likely to score positively for 2 of the 3 symptom categories on the Berlin Questionnaire: snoring symptoms (56% for BE versus 32% for EGD, p = 0.01) and the presence of obesity/HTN (74% for BE versus 48% for EGD, p = 0.005). Figure 2 displays the proportion of subjects scoring positively for each symptom category of the Berlin Questionnaire among the 3 groups. However, within self-reported BMI strata (normal, overweight, or obese), there was no significant difference in snoring symptoms between the BE and EGD groups, suggesting that differences in BMI could explain the initial differences in snoring symptoms between the 2 groups. When BE patients were compared with the COL group, the proportion of subjects scoring positively for snoring symptoms or daytime somnolence did not differ significantly (56% for BE versus 45% for COL, p = 0.18 for snoring symptoms; 29% for BE versus 20% for COL, p = 0.13 for daytime somnolence/drowsiness while driving). BE patients were more likely to score positively for the 3^rd BQ symptom category, presence of obesity/HTN, although this did not reach statistical significance (74% for BE versus 59% for COL, p = 0.05).

Because COL patients who were not getting upper endoscopy served as controls and could potentially have been misclassified, we performed a sensitivity analysis to evaluate the extent to which misclassification could have affected our results. A previous study of 961 colonoscopy patients undergoing upper endoscopy revealed that 6.8% had Barrett’s esophagus; among colonoscopy patients with history of heartburn symptoms, 8.3% had Barrett’s esophagus, compared with 5.6% among those with no history of heartburn symptoms [6]. We therefore assumed that 7%, or 12, patients in the COL group in the current study could have been misclassified and actually had Barrett’s esophagus. Assuming that the BQ+ rate among these misclassified patients was 1) the same as that in the original COL group (a conservative estimate) versus 2) the BQ+ rate in the original BE group, then the odds ratio would be 1.75 (95% CI 0.98, 3.13) for the former scenario, versus 2.28 (95% CI 1.27, 4.10) for the latter scenario. In all likelihood, the true odds ratio probably lies somewhere in between.

The majority of BE patients in our pilot study scored at high risk for OSA on the Berlin Questionnaire. After adjusting for age and BMI, the association between Barrett’s esophagus and scoring at high risk for OSA compared with the COL group was no longer statistically significant. Because older age and higher BMI are risk factors for OSA [30], these factors likely had a large effect on the study results because the BE group was older than the COL group and had a higher BMI than the EGD group. Patients with known OSA were excluded from our study; however, almost all of these patients came from the BE group which could have affected our results. In our study, 37% of COL patients scored at high risk for OSA, which is slightly greater than the 32%-36% rate previously described in primary care settings and a cardiology referral clinic [26, 29]. Patients presenting for outpatient endoscopy at our hospital may be a higher risk population than patients in primary care settings, who in turn likely have a greater prevalence of OSA than the general population [31]. Nonetheless, the high rate of BQ+ status among BE patients in our study is striking. Barrett’s esophagus patients may be more likely to have obstructive sleep apnea than GERD patients without BE due to insulin/insulin growth factor pathways which may have a role in the development of BE [32]; obstructive sleep apnea is associated with insulin resistance [20]. Since our study was conducted, two other recent studies presented in abstract form have suggested that BE is associated with obstructive sleep apnea [33] or obstructive sleep apnea characteristics [34]. Taken together, these findings hold particular relevance because BE patients who undergo endoscopic surveillance or lengthier endoscopic ablative procedures may thereby be at increased risk for sedation-related complications. Benzodiazepines and opioids can depress respiratory drive, decrease upper airway reflexes, and decrease pharyngeal muscle tone [35, 36]. Propofol can also lead to upper airway collapse [35, 37]. These effects are particularly important for patients with obstructive sleep apnea due to redundant pharyngeal tissue and generally narrower airways [38].

In the current study, 44% of EGD patients scored at high risk for OSA. This high rate is not surprising since many of these patients were likely undergoing EGD for GERD symptoms, and previous studies suggest that GERD is associated with sleep complaints [15, 16]. In addition, Siupsinskiene and colleagues reported in abstract form results from a prospective Lithuanian study in which 42 OSA patients underwent upper endoscopy regardless of whether they had upper GI symptoms. This study revealed that 83.3% had pathologic GI findings: 64.3% had hiatal hernia, 45.2% had erosive esophagitis, and 21.4% had histological esophagitis [39]. There was no significant correlation between OSA severity and frequency of endoscopic findings. Although there was no mention of Barrett’s esophagus as an endoscopic finding, BE presumably could have been masked by erosive esophagitis.

In our study, BE patients were more likely than either of the control groups to report nocturnal heartburn or nocturnal regurgitation. BE patients also reported more severe typical GERD symptoms than patients in either control group. These findings are consistent with previous studies demonstrating higher esophageal acid exposure in BE patients [40, 41]. Within the BE and COL groups, nocturnal GERD was associated with BQ+ status, although the rate of nocturnal GERD was low (23%) among BQ+ subjects in the COL group. These findings raise questions of whether BE could be associated with obstructive sleep apnea through a higher incidence of nocturnal GERD or whether nocturnal GERD itself increases the risk of obstructive sleep apnea. Previous studies have been inconsistent in demonstrating a link between nocturnal GERD and BE [21–24, 42]. The relationship between GERD and OSA is incompletely understood. Nocturnal GERD was associated with sleep complaints [16] or OSA symptoms [43] in large cohort studies, and continuous positive airway pressure improved GERD symptoms in OSA patients [17, 18]. However, a recent small study suggested no difference in the number of nocturnal gastroesophageal reflux events as measured by pH monitoring and impedance among OSA patients without GERD, OSA patients with GERD, GERD patients without OSA, and healthy controls [44]. In addition, high-resolution manometry measurements in that study suggested that during OSA events, gastroesophageal junction and upper esophageal sphincter pressures actually increased—changes which would be expected to prevent rather than induce reflux events.

Our pilot study had several limitations. Our study was limited by the use of open-access colonoscopy patients who were not getting EGD as controls, since we assumed that those patients did not have Barrett’s esophagus. The decision to include the COL control group in our study was partially due to slow accrual of EGD controls. Based on our sensitivity analysis, misclassification could have potentially strengthened or weakened an association between BE and BQ+ status. In addition, the study was limited due to exclusion of patients with known OSA; since most of these patients had Barrett’s esophagus, this could have affected the study results. Our study was also limited because results of the Berlin Questionnaire were not confirmed with polysomnography. Previous validation of the Berlin Questionnaire in a surgical patient population demonstrated a sensitivity of 68.9%-87.2% at various apnea-hypopnea index cutoffs [28]. Our assessment of nocturnal GERD was based on patient self-assessment of symptoms rather than an objective measure such as esophageal pH monitoring. The Berlin Questionnaire is partially based on self-reported BMI, which can vary significantly and unpredictably from the true BMI value and can potentially be misleading in very muscular patients. Finally, a type II error might explain why the association between BE and BQ+ status disappeared after adjusting for age.

In summary, the findings from this pilot study demonstrate that a high proportion of Barrett’s esophagus patients score at high risk for obstructive sleep apnea. This finding is particularly important because many BE patients undergo endoscopic surveillance, and a minority undergo lengthier endoscopic procedures for ablation of dysplasia. BE patients may therefore be periodically exposed to sedation with its inherent risks. Given the high proportion of BE subjects at high risk for OSA in our study, screening for OSA should be considered in BE patients. At the very least, raising awareness that BE patients are likely to score at high risk for OSA may facilitate the judicious use of opioids and sedatives, consideration for use of capnography during sedation for lengthier endoscopic ablative procedures, and more vigilant post-procedure monitoring in BE patients undergoing surveillance endoscopy. Our results also demonstrated that patients scoring at high risk for obstructive sleep apnea were more likely to report nocturnal heartburn or acid regurgitation symptoms. Our findings support a growing body of literature linking nocturnal GERD to sleep disturbances. Future studies could explore mechanistic links among Barrett’s esophagus, nocturnal GERD, and obstructive sleep apnea.