BACKGROUND Arterial blood gas (ABG) analysis is not a routine test in sleep laboratories due to its invasive nature. Therefore, the diagnosis of obesity hypoventilation syndrome (OHS) is underestimated. We aimed to evaluate the differences in subjects with OHS and pure obstructive sleep apnea (OSA) and to determine clinical predictors of OHS in obese subjects. METHODS Demographics, body mass index (BMI), Epworth Sleepiness Scale score, polysomnographic data, ABG, spirometric measurements, and serum bicarbonate levels were recorded. RESULTS Of 152 obese subjects with OSA (79 females/73 males, mean age of 50.3 ± 10.6 y, BMI of 40.1 ± 5.6 kg/m(2), 51.9% with severe OSA), 42.1% (n = 64) had OHS. Subjects with OHS had higher BMI (P = .02), neck circumference (P < .001), waist circumference (P < .001), waist/hip ratio (P = .02), Epworth Sleepiness Scale scores (P = .036), ABG and serum bicarbonate levels (P < .001), apnea-hypopnea index (P = .01), oxygen desaturation index (P < .001), and total sleep time with S(pO2) < 90% (P < .001) compared with subjects with pure OSA (n = 88). They also had lower daytime PaO2 (P < .001), sleep efficiency (P = .032), mean S(pO2) (P < .001), and nadir S(pO2) (P < .001). Serum bicarbonate levels and nadir S(pO2) were the only independent predictive factors for OHS. A serum bicarbonate level of ≥ 27 mmol/L as the cutoff gives a satisfactory discrimination for the diagnosis of OHS (sensitivity of 76.6%, specificity of 74.6%, positive predictive value of 54.5%, negative predictive value of 88.9%). A nadir S(pO2) of < 80% as the cutoff gives a satisfactory discrimination for the diagnosis of OHS (sensitivity of 82.8%, specificity of 54.5%, positive predictive value of 56.9%, negative predictive value of 81.4%). When we used a serum bicarbonate level of ≥ 27 mmol/L and/or a nadir S(pO2) of < 80% as a screening measure, only 3 of 64 subjects with OHS were missed. CONCLUSIONS Serum bicarbonate level and nadir saturation were independent predictive factors for the diagnosis of OHS.