Prevalence of obstructive sleep apnea (OSA) is high in obese subjects, many of whom may not be overtly symptomatic. Polysomnography (PSG) is a costly and time-consuming investigation. Since it is not feasible to subject all obese individuals to PSG, it is useful to define predictors of OSA among these subjects. One hundred and eighteen obese subjects [body mass index (BMI)> or =25 kg/m(2)] presenting to the hospital with non-sleep related complaints were included, of which 53 subjects with PSG evidence of OSA [apnea-hypopnea index (AHI)> or =15/h] were defined as cases and 65 subjects without any evidence of OSA (AHI<15/h) were defined as controls. Anthropometry, biochemical investigations, blood gas analysis, pulmonary function tests, and PSG were performed for all subjects. Waist hip ratio (WHR) (as percentage of a standard) [odds ratio (95% CI): 1.07 (1.00-1.14); P = 0.049] male gender [odds ratio (95% CI): 3.97 (0.99-15.81); P = 0.046] and neck circumference (NC) [odds ratio (95% CI): 1.23 (1.03-1.47); P = 0.023] were found to be independent predictors of OSA. Overnight oxygen desaturation data were evaluated in patients selected as having OSA on the basis of these clinical markers, and the best cut-off for level of desaturation (10%) was defined. The stepped approach had a specificity, sensitivity, positive and negative predictive value of 89.2, 88.5, 86.8 and 90.6%, respectively, for the diagnosis of OSA. Male gender, WHR and NC are independent predictors of OSA in overtly asymptomatic obese subjects. A stepped approach to diagnose OSA should be used, as it is accurate and cost-effective.