Three types of analysis were used.
1. Analysis of variance and Students t tests were used to compare groups. Separate t or pooled t was used, depending on the results of the Levene test for equality of variances. The Bonferroni criteria for significance levels were used when multiple comparisons were made. Nonparametric Mann-Whitney or Kruskall-Wallis rank sum H statistics were used to analyze group differences when variables were not normally distributed.
2. In covariate and multiple stepwise regression analyses, we used mean sleep latencies from the MSLT as dependent variables. When the distribution of an independent variable was skewed with a kurtosis >2, a logarithmic transformation was done and the logarithms were added to the model.
3. We also did K-means (K. M.) cluster analysis. Cluster analysis is a multivariate statistical procedure for detecting natural groups or clusters in the data and can be compared to a one-way analysis of variance where the groups are unknown. After a number of iterations, each case joins the cluster whose center (mean of cases in the cluster) is closest to the case. At this point, the clusters are best differentiated from each other.
The researcher requests the desired number of clusters in advance. With too few clusters, interpretation of the results is difficult because many types of patients will be in the same clusters. With too many clusters, it is difficult to find meaningful differences between the clusters. We requested five clusters and used the BMDP statistical program.
Descriptive data on our population of 90 men and ten women are presented in Table 1. Analysis of the MSLT indicated the overall presence of sleepiness: mean scores were 5.9 ±3.5, with a median of 5.4 and a range of 0 to 17. It also revealed an abnormal daytime sleep state distribution. The mean number of SOREMPs was 0.8±1.1. Of the patient population, 55 percent had no SOREMPs during the MSLT, while 21 percent had one period and 24 percent, two or more periods. This high frequency was unexpected, since REM sleep normally occurs about 90 min after sleep onset.
Table 1—Description of Population and Polygraphic Data
|Age (yr)||50.5 ±10.4||51||18-78|
|BMI (kg/sq m)||31.2±5.6||30.6||18.9-46.9|
|Lowest Sa02t (%)||73 ±16||77||32-92|
|% T<90% Sa02t||20.2 ±25||9.7||0-100|
|MSLT scores (min)||5.9±3.5||5.4||0-17|