Introduction Residual apnoea hypopnoea index (AHI) derived from positive airway pressure (PAP) device downloads is frequently used to assess treatment effectiveness in obstructive sleep apnoea (OSA). However, the accuracy of device reported AHI in real‑world settings remains uncertain. This study evaluated the agreement between PAP device derived AHI and AHI measured using home based cardiorespiratory polygraphy. Methods This secondary analysis of the 3DPiPPIn randomised controlled trial included adults with OSA (AHI ≥15 events/hr) naïve to domiciliary PAP therapy. Participants underwent two‑night home cardiorespiratory polygraphy at three and six months. Residual AHI was obtained from Löwenstein PAP devices via data download. Agreement between methods was assessed using Bland–Altman analysis, intraclass correlation coefficients (ICC), and weighted Kappa. Sensitivity and specificity for detecting inadequate control (AHI ≥7 events/hr) were calculated. The association between average leak and AHI discrepancy was examined using linear regression. Results Ninety‑eight participants were included. PAP derived AHI values consistently underestimated polygraphy derived AHI across all time points. Agreement was poor, with ICCs of 0.02 at three months and 0.16 at six months, and low diagnostic agreement across severity categories at both three months (K0.10, 95%CI: 0 to 0.26) and six months (K0.09, 95%CI: 0 to 0.28). Bland–Altman plots demonstrated systematic underestimation with wide limits of agreement. Sensitivity for detecting inadequate control was very low at both time points (3 months 12% and 6 months 21%), despite high specificity (97% at 3 months and 95% at 6 months). Higher interface leak was significantly associated with greater underestimation of AHI (three months: slope = 0.45, R² = 0.18, p < 0.0001; six months: slope = 0.38, R² = 0.14, p < 0.0001). Conclusion PAP device downloads substantially underestimated residual AHI and demonstrated poor agreement with home cardiorespiratory polygraphy. These findings highlight important limitations of device derived AHI in real‑world practice, particularly in the presence of interface leak. Where treatment adequacy is uncertain, confirmatory assessment with polygraphy remains necessary.
Mansell et al. (Thu,) studied this question.