Utility of overnight oximetry indices in the evaluation of children with snoring and suspected obstructive sleep apnea

Study objectives: Optimal cutoff values of oximetry indices that differentiate obstructive sleep apnea (OSA) from primary snoring (PS) is not well established. Our study aimed to assess the utility of overnight oximetry indices in differentiating PS from OSA and assessing OSA severity, compared to polysomnography (PSG), in children with suspected OSA.

Methods: This was a retrospective study of children (1-18 years) with snoring who underwent PSG. Patients with Down syndrome, craniofacial anomalies, known genetic syndromes, neuromuscular conditions and central apnea index ≥ 5 were excluded. Demographic data, PSG variables and oximetry indices (e.g. oxygen desaturation index [ODI_3, defined as number of ≥ 3% desaturation episodes/hour of artifact free recording time and SpO₂ nadir]) were collected.

Results: Of 1,203 children (mean age 9.1±3.9 years, 67.7% males), 91.8% (847/923) ≤ 12 years and 84.3% (236/280) > 12 years had OSA. Optimal cutoff of ODI₃ for differentiating PS from OSA was 2.4 [Se: 78.8% (75.9%-81.6%), Sp: 80.5% (69.9%-88.7%)] in ≤ 12 years and 3.6 [Se: 71.1% (64.8%-76.8%), Sp: 91.1% (78.8%-97.5%)] in > 12 years. The optimal cutoff of ODI₃ for differentiating PS from mild, moderate and severe OSA categories were 2.0 [Se: 70.1% (65.3%-74.5%), Sp: 70.1% (58.6%-80.0%)]; 3.7 [Se: 82.3% (76.6%-87.1%), Sp: 94.8% (87.2%-98.6%)] and 4.3 [Se: 99.1% (96.8%-99.9%), Sp: 98.7% (93.0%-100.0%)] in ≤ 12 years; and 1.9 [Se: 78.8% (75.9%-81.6%), Sp: 80.5% (69.9%-88.7%)]; 4.1 [Se: 85.4% (72.2%-93.9%), Sp: 91.1% (78.8%-97.5%)] and 6.9 [Se: 98.4% (91.2%-100.0%), Sp: 97.8% (88.2%-99.9%)] in > 12 years, respectively.

Conclusions: This study provides optimal cutoff values for ODI₃ in differentiating PS from OSA and assessing OSA severity in children. As oximetry is cheaper and widely available, ODI₃ has the potential to be incorporated into cost-effective clinical decision-making algorithms, especially in resource limited settings.