Table 4.

Mediation effects of prenatal cortisol (Cortisol) on the relationship between prenatal symptoms of depression (CES-D) and gestational age at birth (GA) and fetal growth rates (FG).

Gestational Age at Birth (R² =.35 , Adjusted R² = .33, F(2, 76) =20.09 ,p < .001)
Regression Path	b	s.e.	t	p
Total effect c (CES-D on GA; Cortisol not included)^*	-.032	.015	2.15	.035
Mediation path a (CES-D on Cortisol)^*	3.39	.979	3.46	<.001
Mediation path b (Cortisol on GA)^*	-.008	.001	5.79	<.001
Direct effect c’ (CES-D on GA; Cortisol is included)^*	-.003	.013	0.25	.801
Indirect effect tests			95% Confidence Intervals
Sobel	-.029	.009	(-.047 to -.009) Z = 2.94, p = .003
Bootstrap (z= 1000)			(-.050 to -.010)^b

Fetal Growth Rate (R² = .17, Adjusted R² = .15, F(2, 76) = 7.78, p < .001)
Regression Path	b	s.e.	t	p

Total effect c (CES-D on FG; Cortisol not included)^*	-.472	.195	2.42	.018
Mediation path a (CES-D on Cortisol)^*	3.39	.979	3.46	<.001
Mediation path b (Cortisol on FG)^*	-.065	.022	3.01	.035
Direct effect c’ (CES-D on FG; Cortisol is included)^*	-.252	.199	1.26	.210
Indirect effect tests c - c’ ^a			95% Confidence Intervals
Sobel	-.214	.093	(-.414 to -.026) Z = 2.22, p = .026
Bootstrap (z= 1000)			(-.421 to -.055)^b

b = unstandardised coefficient; s.e. = standard error.

Effects that need to be significant to meet the criteria for mediation according to Baron and Kenny, 1986.

If this coefficient is reduced to zero then we have perfect mediation, if it is reduced significantly but not to zero then we have partial mediation, (Baron & Kenny, 1986). The effect of CES-D on both GA and FG is significantly reduced when Cortisol is introduced as a mediator.

If the 95% confidence interval produced by the bootstrap does not include zero then criteria for mediation has been meet (Preacher & Hayes, 2004).