Triple network model of brain connectivity changes related to adverse mood effects in an oral contraceptive placebo-controlled trial

Abstract

Combined oral contraceptives (COC) are among the most commonly used contraceptive methods worldwide, and mood side effects are the major reason for discontinuation of treatment. We here investigate the directed connectivity patterns associated with the mood side effects of an androgenic COC in a double-blind randomized, placebo-controlled trial in women with a history of affective COC side effects (n = 34). We used spectral dynamic causal modeling on a triple network model consisting of the default mode network (DMN), salience network (SN) and executive control network (ECN). Within this framework, we assessed the treatment-related changes in directed connectivity associated with adverse mood side effects. Overall, during COC use, we found a pattern of enhanced connectivity within the DMN and decreased connectivity within the ECN. The dorsal anterior cingulate cortex (SN) mediates an increased recruitment of the DMN by the ECN during treatment. Mood lability was the most prominent COC-induced symptom and also arose as the side effect most consistently related to connectivity changes. Connections that were related to increased mood lability showed increased connectivity during COC treatment, while connections that were related to decreased mood lability showed decreased connectivity during COC treatment. Among these, the connections with the highest effect size could also predict the participants' treatment group above chance.

Fig. 1. Experimental design.

Each participant had two sessions, before and during treatment (day 1–10 and day 15–21 respectively after onset of menses). Therefore, for the placebo group endogenous hormone levels increased from the first to the second appointment, while in the COC group synthetic hormone levels were stable and endogenous hormones low. Expected hormonal variation in a physiological menstrual cycle and during COC treatment is represented by green (estradiol), and yellow (progesterone) lines. Actual values for each of the MRI-session are displayed in Table 1.

Fig. 2. Interactive effects of group by treatment and relation to mood lability differences.

a Interactive effects of group by treatment. These connections correspond to changes in the COC group during treatment compared to the placebo group if they surpassed a posterior probability of 95% and an estimated value of 0.10. The results are organized by changes observed in the placebo group, in the COC group, or both COC vs. placebo group in opposite direction. b Connections showing relationship to mood lability differences from DMN, from the SN, and from ECN. These connections surpassed a posterior probability of 95% and an estimated value of 0.10 and related during treatment to the mood-lability changes from the previous cycle. Only those changes in the same direction as the interactive effect of group by treatment are displayed. Dashed connections were those driven by the change only in the placebo group from follicular to luteal phase. For (a) and (b): The differential connectivity strengths (Ep) are depicted by the width of the arrow. Black arrows reflect positive values and red arrows reflect negative values.

Fig. 3. Leave-one-out cross-validation analysis.

Left: scatter plot displaying the correlation between the actual treatment group in the left-out-subject’s design matrix and the predicted treatment group based on the left-out-subject’s connectivity. Centre: the resulting posterior probability for each treatment group for each subject. Right: Differential connectivity strength for the interactive effect of group by treatment. The differential connectivity strengths (Ep) are depicted by the width of the arrow. Black arrows reflect positive values and red arrows reflect negative values.