Determination of Oxygen Saturation Indices
We calculated mean nocturnal sleep Sa02 using the formula of Bradley et al, recording the highest and lowest Sa02 of each polygraphically recorded epoch. Because the pattern of desaturation and resaturation in OSAS approximates a sine wave, the mean Sa02 of each epoch was estimated by averaging the high and low values. Mean nocturnal Sa02 for TST was then calculated, using the mean values of all epochs. To further focus on events leading to significant 02 desaturation, even if short-lived, we calculated the number of apnea/hypopnea-related Sa02 drops below 80 percent per hour of sleep to obtain the index 02-80-I. The percentage of TST spent <90 percent Sa02 is self-explanatory.
Measurement of Sleep Variables
We scored sleep epochs according to the international criteria of Rechtschaffen and Kales, using percentage of nocturnal REM (NREM) sleep (SI) as an index of nocturnal sleep disruption. We also analyzed the number of awakenings. “Awakening” is polygraphically defined as a sleep disturbance of >15 s, as indicated by EEG, EOG and EMG. Awakenings and percentage of SI were highly correlated (0.78 percent Pearson correlation coefficient).
Analysis of the Multiple Sleep Latency Test
The daytime and nighttime recording montages were similar: for the MSLT, we used EEG, chin EMG, EOG, ECG (modified V^, respiration, airflow and movements, but not leg movements or ear oximetry. We scored the MSLT epochs by criteria of Rechtschaffen and Kales. Sleep latency was defined as the period from lights-out to the first epoch of scored sleep, as discussed later. Rapid eye movement sleep latency was defined as the time from the first epoch of scored sleep until the beginning of the first epoch of REM sleep. An epoch had to contain at least 50 percent of a sleep stage or state to be scored as such. Sleep onset was confirmed if three successive epochs were scored as SI, or if one epoch was scored as any other sleep stage or state. The REM sleep appearing within 15 min of sleep onset was called a sleep onset REM period (SOREMP). Each sleep latency test (SLT) lasted for 20 min if no sleep, as defined previously, occurred. If sleep occurred, the subject was allowed to sleep for 15 min so that we could note any early, abnormal appearance of REM sleep.
We collected the following variables: sex, age, (BMI) (weight/ height,2 expressed in kg/sq m) computed by the method of Khosla and Lowe, RDI, mean Sa02, percent T< 90 percent Sa02, 02-90-I, 02-80-I, percentage of REM sleep and other sleep stages, number of awakenings, and amount of wake time after onset of sleep (WASO). Sleep latencies and REM sleep latencies were extracted from the MSLT.