Table 2.

Association between neuroendocrine levels following a cold pressor test and indices of water intake to structural findings

	Neuroendocrine responses			Water intake
	Vasopressin (Z, P)	ACTH (Z, P)	Oxytocin (Z,P)	UOsmAM (t,P)	UOsmPM (t,P)
Linear group contrast w/o covariate #	1.85, .06	2.23, .03	2.95, .004	4.9, .0001	6.7, .0001
Covariate effects:
3rd Ventricle volume	1.40, .15	3.74,.0005 *	.66, .51	.76, .45	1.63, .11
Resulting linear group contrast	1.76, .07	2.09, .03	.74, .45	3.8, .001	5.0, .0001
Left hippocampal 5th eigenvector	1.98, .05	.15,.87	3.13, .001 *	.52, .86	.24, .81
Resulting linear group contrast	1.78, .07	2.35, .01	3.48, .0005	3.8, .001	5.0, .0001
Right hippocampal 5th eigenvector	1.37, .16	.26,.79	1.60,.11	.30, .76	.51, .60
Resulting linear group contrast	1.80, .07	2.36, .018	3.43, .0006	4.0, .001	4.8,.0001
Left amygdala 5th eigenvector	3.65, .0002 *	.30,.76	2.28, .02	.03, .97	.10, .91
Resulting linear group contrast	2.34, .019	3.24, .001	4.83, .0000	3.3, .003	5.34, .0001
Right amygdala 5th eigenvector	1.95, .05	.76, .44	2.36, .02	.87, .39	.82,.41
Resulting linear group contrast	2.33, .02	2.84, .004	5.65, .0000	3.6, .002	5.34, .0001

The table shows whether the structural measures defined in the text were significantly associated with neuroendocrine responses to a cold pressor test and to the indices of water intake in the psychiatric patients. The first row shows the significance of the linear group contrasts (PHS>PNS>HC>NNS) in the regression models for each of the dependent measures listed in the five vertical columns (see Supplemental Figure 2 for vasopressin and ACTH). The subsequent rows show the significance of each of the structural measures when added as covariates to these models, as well as the impact of their addition to the linear group contrast. A reduction in statistical significance of the group contrast indicates the covariate effect is not independent of the group effect, and thus is more likely an epiphenomenon, while an increase suggests an independent contribution.

^#AM and PM urine osmolality were analyzed by ANCOVA. Neuroendocrine responses were analyzed by mixed effects linear regression as described in Goldman et al. 2007a, 2008. This approach better accounts for the correlation between repeated measures and the variability in individual responses. For vasopressin and ACTH the statistic is for the interaction between the linear group contrast and the quadratic time trend (i.e. rise and fall). Other statistics are mean levels (i.e. there was no oxytocin response)

^*P<.05 after Bonferroni correction (i.e. 15 comparisons).