Determinants of Daytime Sleepiness in Obstructive Sleep Apnea: Stepwise Multiple Regression Analysis

Determinants of Daytime Sleepiness in Obstructive Sleep Apnea: Stepwise Multiple Regression AnalysisStepwise Multiple Regression Analysis
Clinical insight suggests that poor nocturnal sleep, repetitive apneas, and significant hypoxia during sleep, which are influenced by obesity, could be responsible for daytime sleepiness. We thus asked if the respiratory variables monitored during nocturnal sleep (apnea, drops in SaOg) could be significant determinants of mean MSLT results. We selected age, BMI, percentage of total time spent in Sl-4 and REM sleep, RDI and 02-80-I or percent TOO percent Sa02 as independent variables and the mean sleep latency score as the dependent variable. The distribution of percent TOO percent SaOz and 02-80-I was skewed with a kurtosis >2, and a logarithmic transformation was done. After covariate analysis, significant variables were selected for stepwise multiple regression analysis. We could not derive a model, as none of the selected variables was shown to be statistically significant.

Cluster Analysis. We next questioned whether we could define, in our population, “sleepy” and “alert” OSAS groups by cluster analysis of MSLT results, and evaluate if each group had statistically different determinants. Cluster analysis identified two groups: a “sleepy” group with a mean MSLT of 4.5 (or 4’30”)±3.4 minutes, and a group defined as “alert,” with a mean MSLT of 7.9 (or 7’55”)±4.3 minutes (Table 3). Although the “sleepy” group had more disturbed nocturnal sleep and longer TST, it had a mean BMI very similar to the “alert” group, and paradoxically, a mildly lower mean RDI. The significantly higher percentage of SOREMPs at MSLT in the “sleepy” group indicated a more serious disturbance of the circadian distribution of REM sleep, and sleep deprivation. The “sleepy” group had a mean lower TOO percent SaOz than the “alert” one.
The Role of Obesity. Cluster analysis recognized two other groups: an “obese” group with mean BMI of 33.3±7.3 vs a “non-obese” group with mean BMI of 28.9 ±2. The results of the analysis are presented in Table 4. Non-obese patients had a higher mean RDI than obese patients, although their mean MSLT scores were moderately better. The mean percent T<90 percent Sa02 was higher in the non-obese group, but these mean differences were not statistically significant. When the results of the cluster analysis were submitted to mean squares analysis, the only significant difference involved TST The BMI was not statistically different, despite a trend (F ratio = 2.134, p<0.07).
Table 3—Cluster Analysis of “Sleepy” vs “Alert Subjects”

Variablet “Sleepy” “Alert” MeanSquares

Analysis

X MSLT 4.5±3.4 7.9±4.3 p<0.005
X BMI (kg/sq m) 29.8±4.9 28.9±4.1 NS
XRDI 48 ±20 53 ±27 NS
X % T<90% Sa02 21 ±25 26 ±31 NS
X 02-80-I 7.1 ± 14 8.1 ± 17 NS
X nocturnal TST (min) 434 ±28.6 238 ±27.1 p<0.0001
X % nocturnal S3-4 NREM 3.7±5 13 ±14 NS
XSOREM 1 ± 1.4 0.4±0.5 p<0.05

Table 4—Cluster Analysis of “Obese” vs uNon-obese” Subjects”

Variablet “Obese” “Non-obese” MeanSquares

Analysis

X BMI (kg/sq m) 33.3±7.3 28.9±2.2 NS (p<0.07)
X MSLT (min) 6.4±3.7 6.0±2.9 NS
XRDI 48 ±24.1 55±33 NS
X nocturnal TST (min) 353 ±28.3 382 ±37.5 p<0.05
X % nocturnal S3-4 NREM 5.2±9.0 6.2±9.5 NS
X % T<90% Sa02 28 ±31 38 ±33.7 NS
X 02-80-I 8.3± 18 7.4± 13 NS
This entry was posted in Sleep Apnea and tagged daytime sleepiness, obstructive sleep apnea, respiratory.