A total of 27 women were enrolled, 12 HG cases and 15 NV controls. Venous access could not be maintained for two women, one in each group, and they were dismissed from the protocol. Face validity of the case-control assignment was verified by their PUQE score (possible range 3–15) on the day of recruitment. The HG cases had scores from 8 to 15 (M = 11.4, SD=2.4). The NV controls had scores from 3 to 10 (M = 5.9, SD=2.5; t(23) = −5.6, p <.001). Ten women met PTSD diagnostic criteria. Half were HG cases, half were NV controls. The following two-by-two cells were formed: NV only (n=9), HG only (n=6), NV and PTSD (n=5), HG and PTSD (n=5).
Characteristics of the sample overall and the four groups are presented in . The sample was demographically diverse. Half were expecting their first infant. Their mean gestational age was 12 weeks. Seven women were using antiemetic medication, five in the HG only group and two in the HG and PTSD group (χ2 = 14.9, p = .002). The two PTSD groups had more exposure to childhood abuse (χ2 = 8.7, p = .034) and adult abuse (χ2 = 11.9, p = .008), but the sum of lifetime exposures did not differ significantly across groups. The PTSD symptom count rose from a low of 1.8 in the NV only group, 3.3 in the HG only group, 8.2 in the NV and PTSD group, and 12.0 in the HG and PTSD group (F(3, 21) = 19.8, p <.001), with post hoc Scheffe indicating significant contrasts between both PTSD-affected groups and the PTSD-negative groups (p values ranging from= .045 to <.001). Dissociation symptoms (M = 10.0, SD = 2.9) were more closely associated with HG status than with PTSD status (F(3, 21)= 2.6, p = .079), but the difference did not reach statistical significance. The HG-only group had the highest mean levels of all three pregnancy hormones, though only the human chorionic gonadotropin level difference reached statistical significance (F(3, 21)= 4.2, p = .019), with post hoc Scheffe indicating a significant difference in the NV and PTSD group versus HG only group contrast (p = .032).
Descriptive profile for the overall sample with tests for differences among the four groups.
We examined a correlation matrix with all background characteristics, pregnancy hormones, and PUQE, PTSD, and dissociation scores in relation to oxytocin concentration. Only the nausea and vomiting score (r = .464, p = .019) and dissociation score (r = .570, p = .003) were correlated with oxytocin. The nausea and vomiting and dissociation scores also were correlated with each other (r = .521, p = .008).
We compared the four groups on PUQE scores, dissociation scores, and noon oxytocin levels (). There was no difference in PUQE scores between the HG only and HG and PTSD groups. However, dissociation scores and oxytocin levels were significantly higher in the HG and PTSD group than in the HG only group. Therefore, we additionally tested the hypothesis that it is dissociation, rather than PTSD that is associated with oxytocin levels. We created a new four-group variable representing psychiatric subtypes from the two-by-two cross tabulation of PTSD diagnosis and dissociation median split, resulting in the categorical variable: neither PTSD nor dissociation (n = 10), PTSD only (n = 6), dissociation only (n = 5), both PTSD and dissociation (n = 4). The cases with both PTSD and dissociation had significantly higher oxytocin, a difference that had a large effect size (eta2 = .610).
Tests by ANOVA with effect size estimations, followed by additional tests of the post hoc hypothesis that it is dissociation, rather than PTSD per se that is predictive of oxytocin concentration.*
To test the hypothesis that the HG and PTSD group is distinct in its association with oxytocin, we entered into a linear regression model the three dummy group variables, using the NV only control group as the reference category. This hypothesis was affirmed; only the HG and PTSD group was associated with the noon, log-transformed oxytocin level (β= .558, p = .014, adjusted R2 = .172).
To test the hypothesis that dissociation would mediate the association of trauma-related HG with oxytocin level, we used a linear regression. We used only the trauma-related HG dummy variable in the first step, which was significant (β = .486, p = .014, adjusted R2 = .203). When we entered the dissociation score in a second step, the trauma-related HG subtype was no longer independently predictive (β = .282, p = .148), but the dissociation score was predictive (β = .439, p = .029), and this second model’s explained variance increased to 33% by adjusted R2. Thus, in this small pilot sample, the hypothesis that dissociation mediates the association of a trauma-related HG subtype with oxytocin level was supported.
To estimate effect size and observed power, we used a general linear model with the four-group variable and the dissociation median split variable as fixed factor main effects. We also specified an interaction due to the mediating effect of the dissociation level. and depict this model. The overall model is statistically significant (p = .005) with 66% of variance explained and with an observed power of 95.1%. Each of the three predictor variables is associated with oxytocin, at observed power levels that are nearly adequate despite the small sample size (76% for the dissociation median-split factor, 77% for the HG grouping factor, and 81% for the interaction term).
Results of general linear model of the four-group variable, dissociative median split variable, and their interaction to estimate effect sizes and observed power in relation to noon oxytocin level.
Estimated marginal means of oxytocin level across case and control groups by dissociation level.