Mean versus median for the multiple sleep latency test

Since its introduction in the mid-1970s, the multiple sleep latency test (MSLT) has become the standard method for evaluating hypersomnolence. The mean sleep latency is usually calculated and constitutes the traditional basis for interpretation. Mean and median are both measures of the central tendency of a distribution, but because the trials of the MSLT are limited to 20 minutes, the median may be more appropriate. The objective of this study was to compare the value of the mean versus the median sleep latency in the interpretation of the MSLT. We retrospectively analyzed 100 MSLTs performed for evaluation of excessive daytime sleepiness. Patients' ages ranged from 6 to 84 years (mean 43). Mean and median sleep latencies were calculated according to standard formulas. We classified each record into one of three categories, using both the mean and the median sleep latencies: normal (> 10 minutes), moderate (> or = 5 and < or = 10 minutes), and severe sleepiness (< 5 minutes). Of the 100 MSLTs, 89 remained in the same category (normal, moderate, severe) whether mean or median was used. In 11 cases, the category changed. All shifts were by one category, that is, no shift occurred between normal and severe. This study suggests that, despite valid theoretical arguments for the use of the median, both measures are equally acceptable for clinical purposes.