Skip to main content
. 2018 Sep 20;8:14134. doi: 10.1038/s41598-018-32142-6

Table 2.

Results of GEE models on OSA individuals and non-OSA matched controls.

Variable Model 1: All psychotropic drugs Model 2: Antidepressants Model 3: Anxiolytics Model 4: Hypnotics
Effect (p) 95% CI Effect (p) 95% CI Effect (p) 95% CI Effect (p) 95% CI
Group
Matched control ref. ref. ref. ref.
OSA  + 4% (NS) [−2%, + 11%]  + 19% (***) [ + 12%, + 27%] −14% (***) [−21%,−5%] −9% (**) [−17%,−1%]
Year
0 ref. ref. ref. ref.
1 −9% (***) [−11%,−7%] −8% (***) [−11%,−5%] −8% (***) [−11%,−5%] −11% (***) [−14%,−7%]
2 −14% (***) [−16%,−11%] −14% (***) [−17%,−10%] −15% (***) [−18%,−11%] −12% (***) [−16%,−8%]
3 −17% (***) [−20%,−15%] −16% (***) [−20%,−12%] −21% (***) [−24%,−17%] −16% (***) [−20%,−11%]
Group × Year
OSA × 1 −1% (NS) [−6%, + 4%] 0% (***) [−6%, + 6%] −4% (NS) [−11%, + 4%] −1% (NS) [−10%, + 9%]
OSA × 2  + 1% (NS) [−5%, + 7%]  + 4% (***) [−4%, + 13%] −4% (NS) [−13%, + 6%] −5% (NS) [−14%, + 6%]
OSA × 3  + 1% (NS) [−5%, + 8%]  + 3% (***) [−6%, + 12%]  + 4% (NS) [−7%, + 15%] −8% (NS) [−19%, + 4%]

GEE: generalized estimating equations, CI: confidence interval, p: P-value, OSA: obstructive sleep apnoea. NS: not significant, *p-value < 10%, **p-value < 5%, ***p-value < 1%. A logarithm function was used as link function. The estimated intercepts are not shown in the table. The multiplicative effect was computed from a transformation of the estimated coefficients β: Effect = (1 - β) × 100.