Cortisol dynamics and sleep quality: the role of sex and oral contraceptive use 117 Sleep disruptions and CAR: Moderation analyses NC women, OC users and men CAR values stratified by level of reported sleep disruptions, as measured using the PSQI 5B, and by group are shown in the lower panel of Figure 4.3. The association between CAR and sleep disruptions was dependent on sex and OC use: in NC women, there was no significant association between CAR and sleep disruptions, whereas a lower CAR was associated with a higher probability of sleep disruptions in OC users and men. As shown in Table 4.3, the association between CAR and sleep disruptions was stronger in OC users compared to NC women in both the marginal and covariate-corrected models (-0.129 per 100 nmol/L*minutes increase in CAR, 95% CI [-0.239 to -0.019], p=0.021 and -0.123 per 100 nmol/L*minutes increase in CAR, 95% CI [-0.241 to -0.005], p=0.042, padjusted=0.084b.), although on trend-level after adjustment for repeated testing. The association between the CAR and sleep disruptions was not significantly different in men compared to NC women (-0.047 per nmol/L*minutes increase in CAR, 95% CI [-0.117 to 0.022], p=0.18). The raw estimates of all PIM analyses are shown in Table 4.3, and as shown in Table 4.3, the covariate-controlled models show similar directions and effect size magnitudes. Sleep quality and CAR: Mediation analysis To assess whether the difference in sleep qualities between NC women and OC users was mediated by the difference in the CAR, we conducted an additional mediation analysis with group (OC vs. NC) as independent variable, CAR as mediator and sleep quality score as dependent variable, as shown in Figure 4.4. For this analysis, we excluded 6 participants with missing CAR measurements, resulting in a sample of 84 NC women and 28 OC users. Results from this mediation analysis reveal a significant total effect, indicating that OC users report poorer sleep than NC women (B=1.07, 95% CI: 0.21 to 1.94, p=0.016). The direct effect (i.e., when adjusting for the CAR) was estimated to be B=1.11 (95% CI: 0.18 to 2.01, p=0.019, padjusted=0.076). The indirect effect was estimated to be -0.036 (95% CI: 0.035 to -0.10, p=0.82), based on the product of the group difference in the CAR (B=-2.34, 95% CI: -3.56 to -1.12, p<0.0005) and the effect of the CAR on the sleep quality score (B=-0.015, 95% CI: -0.12 to 0.15, p=0.82). This means that the estimated proportion mediated was -0.034 (CI: -0.32 to 0.26), i.e., there was no clear evidence to support that the CAR was mediating the group difference in sleep quality.
RkJQdWJsaXNoZXIy MTk4NDMw