OBJECTIVE:To evaluate the performance of clinical criteria (CC) for diagnosis and initiation of empirical treatment with continuous positive airway pressure (CPAP) in patients with suspected obstructive sleep apnea (OSA) compared with the treatment decision based on sleep studies (polysomnography or respiratory polygraphy), guidelines, and experience of participating physicians. METHODS:This was a simulated intention-to-treat study in a retrospective (G1) and prospective (G2) cohort. Four observers (two per group) called CC1 and CC2 reviewed the sleep questionnaires and indicated CPAP if the patients presented snoring, frequent apneas (≥ 3-4/week), body mass index (BMI) > 25 kg/m2, sleepiness (Epworth > 11), or tiredness (at least 3-4 times per week) and some comorbidity (hypertension, coronary/cerebrovascular event, diabetes). Ten independent observers formed two groups of five (FD1 and FD2) and were blinded to each other's opinion. These observers in FD1 and FD2 decided CPAP treatment based on guidelines of the Spanish Society of Pulmonology and Thoracic Surgery (SEPAR) or guidelines of the American Academy of Sleep Medicine (AASM) and factored in their own opinion. Sensitivity (S), specificity (Sp), and positive/negative likelihood ratios (LR+/-) were calculated with the test method: CC1/2, and the reference method: majority decision of FD1/2. RESULTS:A total of 653 patients (264 women, 40%) were studied. Median age was 54 years, BMI 28 kg/m2, and apnea hypopnea index (AHI) 16.5 events/h. S ranged from 21 to 25% (p 0.60), Sp 96.1 to 97.6% (p 0.39), and LR+ of clinical criteria 6.4 to 8.9 (p 0.52). CONCLUSION:CPAP indication without a previous sleep study showed a low sensitivity (≅ 22%) but a specificity greater than 95% in patients with high pretest probability for OSA (snoring, report of frequent apneas, BMI > 25 kg/m2 and sleepiness or tiredness plus comorbidity).