Table 4.

Significant adolescent predictors for class III obesity at age 24 by logistic regression models.

Dependent variable	Significant explanatory variables	Sign, p	Odds ratio, 95% CI
Class III obesity at age 24 Yes (BMI ≥40, n=33) vs No (BMI <40,n=300)	SHBG at age 14 (nmol/l)	-, <.0001	0.82, 0.76-0.88
333 observations used (140 W, 193 B) AUC=0.881	Race (W=1, B=2)	+, .0002	11.4, 3.14-41.7
	Oligomenorrhea at age 14 (Yes as 1, No as 0)	+, .037	3.64, 1.08-12.3
Class III obesity at age 24 Yes (BMI ≥40, n=33) vs No (BMI <40,n=300)	Metabolic syndrome at age 14 (Yes as 1, no as 0)	+, .018	4.12, 1.28-13.2
	Race (W=1, B=2)	+, .0002	13.8, 3.45-55.3
333 observations used (140 W, 193 B) AUC=0.806	SHBG at age 14 Race-spec. bottom decile as 2, Upper 9 deciles as 1	+, .012	3.85, 1.35-10.9
	Oligomenorrhea-hyperandrogenism (PCOS category)	+, .016	2.41, 1.18-4.94
	Childhood insulin Race-spec. top decile as 2, Lower 9 deciles as 1	+, .048	2.93, 1.01-8.48

Stepwise selection from race, age 14 variables (FT, E2, DHEAS, SHBG, metabolic syndrome, oligomenorrhea [yes, no], PCOS category [3 levels—no oligomenorrhea as 0, oligomenorrhea as 1, oligomenorrhea plus hyperandrogenism, i.e. PCOS as 2]), and childhood insulin.

Stepwise selection from race, age 14 categorical variables (race-specific top decile vs lower 9 deciles of FT, DHEAS, race-specific bottom decile vs upper 9 deciles of E2 and SHBG, metabolic syndrome, oligomenorrhea and PCOS category), and race-specific top decile vs lower 9 deciles of childhood insulin.