Value of clinical, functional, and oximetric data for the prediction of obstructive sleep apnea in obese patients.

OBJECTIVE: To evaluate the diagnostic value of clinical features, pulmonary function testing, blood gas tensions, and oximetric data for case finding of obstructive sleep apnea (OSA) before polysomnography (PSG) in a series of consecutive overweight patients.

METHODS: We studied a population of 102 consecutive patients referred by an obesity clinic for suspected OSA, in whom body mass index was > or = 25 kg/m(2). The following tests were performed: clinical score (CS), pulmonary function tests (PFTs), measurement of arterial blood gas tensions, nocturnal oximetry, and full-night PSG.

RESULTS: Six of 34 women and 34 of 68 men had OSA, defined by an apnea-hypopnea index > or = 15. CS and the cumulative time spent below 80% arterial oxygen saturation (SaO(2)) were higher, and PaO(2), minimal SaO(2), and mean nocturnal SaO(2) (mSaO(2)) were lower in OSA patients than in non-OSA patients. Logistic regression showed that sex, CS, and the ratio of FEV(1) over forced expiratory volume in 0.5 s (an index of upper airway obstruction on flow-volume curves) and mSaO(2), expressed as categorical variables, were independent predictors of OSA. None of these individual variables had a satisfactory diagnostic value for the diagnosis of OSA. A logistic regression model including sex and all continuous variables would have allowed us to predict the presence or absence of OSA confidently in 72.5% of the population, in whom the positive predictive value of the model was 94% and the negative predictive value was 90%.

CONCLUSION: In obese patients referred to a respiratory sleep laboratory and evaluated by CS, PFTs, arterial blood gases, and oximetry, no individual sign or symptom may accurately predict the presence or absence of OSA. Provided that it is validated in prospective studies, a logistic regression model using these variables may be useful for the prediction of OSA.