Table 1.
Descriptive characteristics of study participants.
Figure 1.
Stress experience in subjects suffering from current insomnia.
Correlation between insomnia severity score (indicated by the Insomnia Severity Index (ISI)) and stress perception (indicated by the Perceived Stress Scale (PSS)). Analysis showed a significant correlation between scores on the ISI and the PSS across the whole sample (rp = 0.548, p<0.001). * Denotes statistical significance at p<0.05.
Figure 2.
Correlation between serum BDNF levels and scores on the Perceived Stress Scale (PSS) by insomnia severity groups according to the Insomnia Severity Index (ISI).
Analyses showed a significant correlation (partial correlation controlled for smoking) between BDNF and stress only in subjects with no insomnia (rp = −0.511, p = 0.013) compared to subjects with sub threshold (rp = 0.069, p = 0.814) or clinical insomnia (rp = 0.199, p = 0.608). White squares represent subjects with no insomnia, black circles represent subjects with sub threshold and black triangles represent subjects with clinical insomnia. (Inset) Mean serum BDNF levels of the insomnia severity groups. Plotted means and standard errors estimated by ANCOVA with serum BDNF as dependent variable, insomnia severity group as independent variable and smoking as covariate. For all three insomnia severity groups the overall effect on serum BDNF was not significant (F(2) = 2.67; p = 0.080). Contrasts showed that serum BDNF levels in the group with no insomnia were significantly higher compared to the groups reporting sub threshold and clinical insomnia (F(1) = 5.33; p = 0.026); (no insomnia n = 24; sub threshold insomnia n = 16, clinical insomnia n = 10). * Denotes statistical significance at p<0.05
Figure 3.
Mediation models for the interplay between stress, sleep and BDNF.
(A) Stress as a mediator in the relationship between sleep and BDNF. Analyses revealed a significant ‘a’ path (t = 4.36; p<0.001) between sleep and stress. The ‘b’ path between stress and BDNF was not significant (t = −0.92; p = 0.365). The ‘c’ path between sleep and BDNF was significant after inclusion of stress as mediator (t = −2.05; p = 0.046). The indirect ‘ab’ path was not significant (bootstrap 95% confidence intervals: lower = −0.385, upper = 0.169, p>0.05). (B) Sleep as mediator in the relationship between stress and BDNF. The ‘a’ path between stress and sleep was significant (t = 4.36; p<0.001), as was the ‘b’ path between sleep and BDNF (t = −2.05; p = 0.046). The ‘c’ path between stress and BDNF was not significant when the mediator sleep was included (t = −0.92; p = 0.365). In this model the indirect ‘ab’ path was significant (bootstrap 95% confidence intervals: lower = −0.522, upper = −0.046; p<0.05). Both models explain 27.3% of the variation (adjusted R square) in serum BDNF levels (F(7.45) = 7.013; p<0.001). Smoking was included as covariate in both models. * Denotes statistical significance at p<0.05 and *** p<0.001.