Statistical Analysis
Statistical analysis was performed with Microsoft Excel (Microsoft 365
MSO, Version 2405 Build 16.0.17628.20006, 64-bit). If means and standard
deviations were not reported in the study of interest, they were
manually calculated from the data provided if possible. For combining
data, means were weighted based on study population size and standard
deviations were pooled based on unequal study population sizes, and a
two-tailed paired t-test was performed with P < 0.05 as the
cutoff for significance. Reported data was converted to common units of
measurement if possible, however some data was incomplete and unable to
be converted without additional information. Review Manager (Version:
8.0.0, REVMAN) was used for the metanalysis. A random effects model was
used if heterogeneity existed. When pooling the data in the studies, the
means, standard deviations, p-values, and 95% confidence intervals (CI)
were calculated in either Microsoft Excel or REVMAN. Heterogeneity was
assessed with the I2 statistic (inconsistency levels:
low = 25%, moderate = 50%, high = 75%) and the Cochran Q statistic,
with significant heterogeneity considered when P < 0.1 was
obtained. If heterogeneity existed, a sensitivity analysis was performed
by removing each of the studies to identify the source(s). The null
hypothesis for this study was that no difference exists between pre- and
post- DISE-directed tongue surgery PSG outcomes for our study
population.