Table 4.
Investigating mediatory effect of birth weight between continuous CRP levels and child brain morphology at 9–12 (model 4)
| Total effect | Natural direct effect | Natural indirect effect | Proportion mediated (%)** | ||||
|---|---|---|---|---|---|---|---|
| β (95% CI) | P-value | β (95% CI) | P-value | β (95% CI) | P-value | ||
| Total brain volume | 0.002 (−0.032, 0.030) | 0.9 | 0.005 (−0.027, 0.040) | 0.8 | −0.003 (−0.012, 0.010) | 0.5 | 0 |
| Cortical gray matter volume | 0.002 (−0.010, 0.010) | 0.8 | 0.002 (−0.010, 0.010) | 0.8 | 0.000 (0.000, 0.000) | 0.7 | 0 |
| Total white matter volume | 0.011 (−0.002, 0.020) | 0.1 | 0.011 (−0.002, 0.020) | 0.1 | 0.000 (0.000, 0.000) | 0.7 | 0 |
| Subcortical gray matter volume | −0.010 (−0.033, 0.010) | 0.4 | −0.00 (−0.032, 0.010) | 0.4 | 0.000 (−0.002, 0.000) | 0.5 | 0 |
| Cerebellum cortex volume | −0.055 (−0.087, −0.020) | <0.001 * | −0.055 (−0.087, −0.020) | <0.001 * | 0.000 (−0.002, 0.000) | 0.7 | 0 |
| Cerebrospinal fluid volume | −0.029 (−0.069, 0.010) | 0.2 | −0.029 (−0.069, 0.010) | 0.2 | 0.000 (−0.001, 0.000) | 1.0 | 0 |
| Hippocampus volume | 0.020 (−0.012, 0.050) | 0.2 | 0.020 (0.013, 0.040) | 0.2 | 0.000 (−0.001, 0.000) | 0.9 | 0 |
| Amygdala volume | 0.021 (−0.009, 0.050) | 0.2 | 0.021 (−0.009, 0.050) | 0.2 | 0.001 (−0.002, 0.000) | 0.5 | 0 |
Significant results are presented in bold.
Model 1 was corrected for sex, child age at assessment, gestational age at which CRP was measured, model 2 was additionally corrected for maternal national background, maternal education, household income and maternal age, and model 3 was additionally corrected for psychoactive substances, smoking, alcohol, and maternal psychopathology. Model 4 was further adjusted for total brain volume for all the regional brain volume outcomes.
For easier interpretation of multiple outcomes with different ranges, the exposure, mediator and the outcome were all standardized to enable comparison.
The TE, NDE and NIE (with corresponding p-values and 95% confidence intervals) were calculated with quasi-Bayesian methods (sampling n=10,000). To compute the confidence intervals the bias-corrected and accelerated (BCa) intervals were chosen.
Represents significant p-values after Bonferroni correction.
The proportion mediated is only displayed when it could be calculated, i.e., when the NDE and NIE were in the same direction.