For these analyses, we collapsed breast stage 2 and breast stage 3 to contrast girls who were prepubertal (breast stage 1) from those who had breast development (breast stage ≥2), and contrasted pubic hair stage 1 from those with pubic hair stage ≥2. Maturation status was described at age 7 and 8 years, as described already. A given participant could have contributed to analyses at both ages 7 and 8 for age-specific maturation status.
Logistic regression models were used to examine factors that are potentially related to breast stage (1 vs ≥2) as the outcome. The initial models included BMI percentile, race, age, and site, as well as all interaction terms of main effects with site. A final logistic regression model, derived from backward elimination, included only variables that were significant at a probability level of .05. Linear regression was used to estimate the strength of the relationship between height velocity and potential predictors, including age, race, breast stage, site, and all sites by main effect interactions. A backward elimination procedure was used to derive a final model with only significant main effects and site interactions.
To assess interrater agreement for pubertal staging, 1 of the experts (Dr Galvez) visited each clinical site to perform maturation assessment in tandem with clinical staff. The degree of agreement between the master trainers and the research staff members who were conducting the examinations was measured in 127 dual examinations by using the κ statistic, which evaluates observed agreement contrasted to agreement that is attributable chance.
Logistic regression models were used to examine factors that are potentially related to breast stage (1 vs ≥2) as the outcome. The initial models included BMI percentile, race, age, and site, as well as all interaction terms of main effects with site. A final logistic regression model, derived from backward elimination, included only variables that were significant at a probability level of .05. Linear regression was used to estimate the strength of the relationship between height velocity and potential predictors, including age, race, breast stage, site, and all sites by main effect interactions. A backward elimination procedure was used to derive a final model with only significant main effects and site interactions.
To assess interrater agreement for pubertal staging, 1 of the experts (Dr Galvez) visited each clinical site to perform maturation assessment in tandem with clinical staff. The degree of agreement between the master trainers and the research staff members who were conducting the examinations was measured in 127 dual examinations by using the κ statistic, which evaluates observed agreement contrasted to agreement that is attributable chance.
