The behavioural results
Table 1 shows the behavioural results. The results indicated that CR scores were significantly negatively correlated with PSQI scores. However, the ES scores were significantly positively correlated with PSQI scores. No significant correlations were found between CR scores and ES.
INSERT TABLE 1. ABOUT HERE
Functional connectivity linking ER to PSQI
To begin with, in order to examine the association between the brain functional connectivity within the DMN and ER, we selected the DMN as a network-of-interest and conducted the resting-state functional connectivity (rsFC) analysis of brain regions in the DMN. Results showed that the four rsFC including the left superior temporal gyrus (lSTG)-left lateral occipital cortex (lLOC), lSTG-left anterior cingulate gyrus (lACG), right lateral occipital cortex (rLOC)-left middle frontal gyrus (lMFG), and rLOC-lSTG were positively correlated with CR (see Fig. 1a), and two rsFC including the left precuneus cortex (lPrcu)-right Temporal Pole (rTP) and lPrcu-lSTG were positively correlated with ES (see Fig. 1b).
INSERT FIG. 1. ABOUT HERE
In addition, to further confirm the hypothesis that ER mediates the association between the brain functional connectivity and sleep quality, we also performed the Pearson’s correlation analysis to examine the association between these rsFC have been identified and sleep quality. Results showed that the lSTG-lLOC, lSTG-lACG, rLOC-lMFG, and rLOC-lSTG were negatively associated with PSQI scores. Similarly, results also showed that lPrcu-rTP and lPrcu-lSTG were negatively correlated with PSQI scores. Table 2 shows the results of the correlation analysis.
INSERT TABLE 2. ABOUT HERE
Mediation analysis
To further explore the relationship among brain functional coupling, ER, and sleep quality, we constructed the mediation model to test whether ER mediated the effect of brain functional connectivity on sleep quality. We also took sex and age as covariates in the mediation model.
To begin with, we tested the mediating effect of CR on the relationship between the lSTG-lLOC, lSTG-lACG, rLOC- lMFG, and rLOC-lSTG rsFC and PSQI scores (see Fig. 2). The mediation analysis further revealed that CR mediated the association between the lSTG-lLOC and sleep quality (a = 0.17, p = .019, bootstrapped 95% CI = [0.74, 7.99]; b = -0.16, p = .030, bootstrapped 95% CI = [-0.14, -0.01]; direct effect: c’ = -0.14, p = .044, bootstrapped 95% CI = [-3.33, -0.04]; indirect effect: ab = -0.03, bootstrapped 95% CI = [-0.06, -0.01]), the association between the lSTG-lACG and sleep quality (a = 0.20, p = .006, bootstrapped 95% CI = [1.72, 10.01]; b = -0.16, p = .030, bootstrapped 95% CI = [-0.14, -0.01]; direct effect: c’ = -0.14, p = .059, bootstrapped 95% CI = [-3.73, 0.07]; indirect effect: ab = -0.03, bootstrapped 95% CI = [-0.07, -0.01]), the association between the rLOC-lMFG and sleep quality (a = 0.17, p = .024, bootstrapped 95% CI = [0.62, 9.10]; b = -0.15, p = .035, bootstrapped 95% CI = [-0.13, -0.01]; direct effect: c’ = -0.20, p = .007, bootstrapped 95% CI = [-4.51, -0.72]; indirect effect: ab = -0.02, bootstrapped 95% CI = [-0.06, -0.01]), and the association between the rLOC-lSTG and sleep quality (a = 0.23, p = .002, bootstrapped 95% CI = [2.44, 10.0]; b = -0.15, p = .048, bootstrapped 95% CI = [-0.13, -0.01]; direct effect: c’ = -0.17, p = .021, bootstrapped 95% CI = [-3.91, -0.32]; indirect effect: ab = -0.04, bootstrapped 95% CI = [-0.08, -0.01]).
INSERT FIG. 2 ABOUT HERE
Besides, we tested the mediating effect of ES on the relationship between the lPrcu-rTP and lPrcu-lSTG rsFC and PSQI scores (see Fig. 3). The mediation analysis further revealed that ES mediated the association between the lPrcu-rTP and sleep quality (a = 0.16, p = .026, bootstrapped 95% CI = [0.43, 6.70]; b = 0.27, p < .001, bootstrapped 95% CI = [0.08, 0.28]; direct effect: c’ = -0.19, p = .001, bootstrapped 95% CI = [-4.91, -0.76]; indirect effect: ab = 0.04, bootstrapped 95% CI = [0.01, 0.09]), and the association between lPrcu-lSTG and sleep quality (a = 0.16, p = .040, bootstrapped 95% CI = [0.12, 5.29]; b = 0.27, p < .001, bootstrapped 95% CI = [0.08, 0.27]; direct effect: c’ = -0.19, p = .006, bootstrapped 95% CI = [-4.08, -0.69]; indirect effect: ab = 0.04, bootstrapped 95% CI = [0.01, 0.09]).
INSERT FIG. 3. ABOUT HERE
It is worth noting that, as suggested by methodologists, the direct effect and indirect effect were of the opposite direction in this mediation model, which is recognized as inconsistent mediation (Li et al., 2015; MacKinnon et al., 2000; Shrout & Bolger, 2002). More specifically, as the indirect effect is of an opposite sign with the direct effect, these two opposite effects counteracted with each other (Li et al., 2015). In this model, the positive effect of the two rsFC on ES (path a) was significant, and the positive effects of ES on sleep quality (path b) was also significant. Bootstrapping mediation analyses suggested that the mediating effect (path ab) of ES between the rsFC and sleep quality was significant, and indicated that the partial mediating effect as the relationship between the two rsFC and sleep quality was weakened but still significant upon addition of ES to this mediation model.