The patient population that is evaluated for bariatric surgery is characterized by a very high body mass index (BMI). Since obesity is the most important risk factor for obstructive sleep apnea (OSA), sleep disordered breathing is highly prevalent in this population. If undiagnosed before bariatric surgery, untreated OSA can lead to perioperative and postoperative complications. Debate exists whether all patients that are considered for bariatric surgery should undergo polysomnography (PSG) evaluation and screening for OSA as opposed to only those patients with clinical history or examination concerning sleep disordered breathing. We examined the prevalence and severity of OSA in all patients that were considered for bariatric surgery. We hypothesized that, by utilizing preoperative questionnaires (regarding sleepiness and OSA respiratory symptoms) in combination with menopausal status and BMI data, we would be able to predict which subjects did not have sleep apnea without the use of polysomnography. In addition, we hypothesized that we would be able to predict which subjects had severe OSA (apnea-hypopnea index (AHI) > 30).
Three hundred forty-two consecutive subjects, evaluated for bariatric surgery from November 1, 2005 to January 31, 2007 underwent overnight polysomnography and completed questionnaires regarding sleepiness, menopausal status, and respiratory symptoms related to OSA. Apneas and hypopneas were classified as follows: mild apnea 5 ≤ AHI ≤ 15, moderate apnea 15 < AHI ≤ 30, and severe apnea AHI > 30.
The overall sample prevalence of OSA was 77.2%. Of these, 30.7% had mild OSA; 19.3% had moderate OSA, and 27.2% had severe OSA. Among men, the prevalence of OSA was 93.6% and 73.5% among women. The mean AHI (events per hour) for men with OSA was 49.2 ± 35.5 and 26.3 ± 28.3 for women with OSA. Separate logistic regression models were developed for the following three outcomes: AHI ≥ 5 events per hour, AHI > 15 events per hour, and AHI > 30 events per hour. When predicting these three levels of OSA severity, the area under the curve (AUC) values were: 0.8, 0.72, and 0.8, respectively. The negative predictive value for the presence of sleep apnea (AHI ≥ 5) was 75% when using the most stringent possible cutoff for the prediction model.
The prevalence of OSA in all patients considered for bariatric surgery was greater than 77%, irrespective of OSA symptoms, gender, menopausal status, age, or BMI. The prediction model that we developed for the presence of OSA (AHI ≥ 5 events per hour) has excellent discriminative ability (evidenced by an AUC value of 0.8). However, the negative prediction values for the presence of OSA were too low to be clinically useful due to the high prevalence of OSA in this high-risk group. We demonstrated that, by utilizing even the most stringent possible cutoff values for the prediction model, OSA cannot be predicted with enough certainty. Therefore, we advocate routine PSG testing for all patients that are considered for bariatric surgery.