Agreement between apnea-hypopnea index and hypoxic burden for severity classification of obstructive sleep apnea.
Level 4 - case-series / case-control
Retrospective single-center cross-sectional diagnostic/comparative study
PubMed 42657971 · doi:10.4081/monaldi.2026.4016
What was done
A retrospective single-center study assessed 95 adults (aged 18–79 years, mean age 58.4 ± 18.3 years) undergoing full-night in-laboratory polysomnography. Obstructive sleep apnea (OSA) severity classifications based on the apnea-hypopnea index (AHI: mild 5 to <15, moderate 15 to <30, severe ≥30) and hypoxic burden (HB: mild <3.34, moderate 3.34 to 6.66, severe ≥6.67) were compared among the 73 patients diagnosed with OSA (AHI ≥ 5) using Cohen's kappa and Pearson correlation.
What was found
Continuous AHI and HB were moderately correlated (r = 0.663, p < 0.001). Categorical agreement between AHI and HB severity classifications showed only slight concordance beyond chance (κ = 0.168, p < 0.001; reported observed agreement 30.1%). Among the 73 patients with OSA, HB severity exceeded AHI severity in 27.4%, was concordant in 61.6%, and was lower in 11.0%. Patients with HB severity greater than AHI severity demonstrated significantly lower nadir oxygen saturation (p = 0.011).
Why it matters
Standard AHI measures only respiratory event frequency, whereas hypoxic burden reflects the depth and duration of desaturation; the limited agreement suggests AHI alone may misclassify physiological severity.
Limits
The study is limited by a small single-center sample (n = 95 total, n = 73 with OSA) and a retrospective design. Clinical outcomes, cardiovascular endpoints, and patient symptoms were not assessed. The abstract also notes an apparent internal numerical tension between reported concordant cases (61.6%) and observed agreement (30.1%).
Cited by
- supports Having up to 4 or 5 breathing pauses per hour during sleep is clinically normal, with mild sleep apnea defined starting at 5 events per hour and moderate sleep apnea starting at 15 events per hour.