Which derivative from overnight oximetry best predicts symptomatic response to nasal continuous positive airway pressure in patients with obstructive sleep apnoea?

The cost and inconvenience of polysomnography for the investigation of obstructive sleep apnoea (OSA) has led to the search for simpler and cheaper alternatives. These are usually compared to an apnoea/hypopnoea index (AHI) derived from ore-nasal airflow and ribcage/abdominal effort sensors. However, AHI is not a proven gold standard for the identification of clinically important OSA. Recent work has shown that correctly analysed oximetry indices not only mimic AHI with considerable accuracy, but also predict the symptomatic response to nasal continuous positive airway pressure (nCPAP) at least as well as conventional polyisomnographic indices. This current study looks at 10 different derivatives of all-night oximeter tracings in 81 patients with OSA, and compares their ability to predict the improvement in subjective sleepiness after 6 months of nCPAP therapy. Sleep apnoea in this study was defined as a >4% SaO(2) dip rate of > 10 h(-1), and an Epworth sleepiness score (ESS) of >10 on presentation. Subjects took part in a placebo-controlled trial of nCPAP for 1 month and thereafter were all supplied with nCPAP set at an effective pressure. All night SaO(2) was sampled every 2 sec and later analysed for mean, median, mean nadir of SaO(2) dip, time below 90%, 2 and 12 sec delta index (an average measure of SaO(2) change across these two time periods), rapid resaturation index (rises in SaO(2) > 3% within 10 sec per hour), and >4, >3, and >2% dips per hour. These indices were then correlated with the change in ESS seen after 6 months on nCPAP. Median compliance for the group was 5.8 (5/95% range, 1.5-8.0) h night(-1). The two best correlates of improvement in ESS were the >4% SaO(2) dips h(-1) (Spearman's r = -0.33 P = 0.002) and the delta 12 sec index (r = -0.33, P = 0.003). Cumulative time below 90% SaO(2) was almost as good (r = -0.29, P = 0.009). The worst correlate was the >2% SaO(2) dips h(-1) (r = 0.01, not significant). This study has identified which of several analyses of overnight oximeter recordings best identify nCPAP responsive OSA. Both >4% SaO(2) dips and 12 sec delta index are equally predictive and presumably the most appropriate derivatives to calculate from overnight records of SaO(2).