Table 3.

The relationships between the decline of psychosomatic symptoms after treatment and predictors for the predictive component as evaluated by the OPLS model and multiple regression (for details see Statistical analysis).

Data Type	Variable	OPLS (Predictive Component)
		Component Loading	t-Statistic	R c	Regression Coefficient	t-Statistic
Relevant predictors (matrix X)	16α-Hydroxypregnenolone	0.208	2.88	0.341 *	−0.098	−1.35
	Dehydroepiandrosterone	0.205	2.28	0.337 *	−0.002	−0.02
	7α-Hydroxy-DHEA	0.194	2.24	0.318 *	−0.132	−1.01
	Androstenediol	0.225	2.68	0.368 *	0.023	0.24
	Androstenedione	0.252	3.20	0.412 **	0.086	1.15
	Allopregnanolone sulfate	0.192	8.04	0.314 **	0.028	0.33
	Isopregnanolone sulfate	0.189	3.74	0.309 **	−0.059	−1.08
	Conjugated pregnanolone	0.244	2.34	0.399 *	0.101	1.50
	Conjugated 5α-pregnane-3α,20α-diol	0.147	2.94	0.241 *	−0.047	−0.68
	Conjugated 5β,20α-tetrahydroprogesterone	0.219	3.23	0.359 **	0.157	3.19 **
	Conjugated 5β-pregnane-3α,20α-diol	0.191	2.50	0.312 *	−0.033	−0.56
	17-Hydroxyallopregnanolone sulfate	0.233	3.64	0.382 **	0.144	2.10 *
	Androsterone	0.188	3.40	0.309 **	0.027	0.67
	Epietiocholanolone sulfate	0.150	2.26	0.246 *	0.018	0.19
	Conjugated 5β-androstane-3α,17β-diol	0.128	1.34	0.210	−0.015	−0.19
	Psychosomatic symptoms	0.573	5.33	0.939 **	0.804	8.97 **
	Δr11β-Hydroxyandrostenedione a	−0.269	−2.66	−0.440 *	−0.153	−1.91 *
(matrix Y)	ΔPsychosomatic symptoms b	−1.000	−9.07	−0.920 **
Explained variability of dependent variable		84.7% (72.9% after cross-validation)

^a Δ_r represents the relative change calculated as (value after intervention − basal value)/basal value, ^b Δ represents the absolute change calculated as (value after intervention − basal value); ^c R represents component loadings expressed as correlation coefficients with a predictive component, * p < 0.05, ** p < 0.01.