How should the multiple sleep latency test be analyzed?

Abstract

Results of the multiple sleep latency test (MSLT) are commonly recorded as 4-5 latencies to sleep onset, together with their mean. The suitability of these and other MSLT measures of daytime sleep tendency was examined in 111 non-narcoleptic patients of a sleep disorders center. Sleep and breathing patterns had been recorded for one night, followed by a five-nap MSLT. It was found that MSLT latencies to sleep stages 1 and 2 were distributed non-normally. There was typically an excessive frequency of long latencies and a secondary peak at 20 minutes, the longest allowed latency. Wake efficiencies (WE) (100% time asleep) were similarly distributed. Mean sleep latencies (SL) and mean WE were generally distributed more normally than the corresponding medians. Regression analysis showed that the MSLT variables best predicted by sleep-disordered breathing during the previous night were those based on WE. Effects were greatest for the first 1-2 naps. Whether the 4- or 5-nap MSLT can be shortened will depend on corroborating data from other labs. For now, mean WE appears to be the best measure of daytime sleep tendency. Regardless of which MSLT measure is used, sleep-disordered breathing, nocturnal sleep time, age, and gender together explain less than one-fourth of the variance found in the non-narcoleptic patients of a sleep disorders center. Much daytime sleepiness therefore remains unexplained.