Mother's age at menarche showed a normal distribution (mean ± SD = 12.8 ± 1.5 y), and values were divided into approximate quintiles for analysis and display. BMI, fat mass index, and truncal fat index showed skewed distributions; analyses were therefore based on log-transformed data, and the data in Table 1 represent geometric means. Multiple linear regression models were performed to test the associations between quintiles of mother's age at menarche (adjusted for sex, age, and mother's educational achievement) and various outcome measures of mother's body size and children's body size, body composition, and growth. Models with outcome measures of infant (0–2 y) growth were further adjusted for gestational age, parity, and breast-feeding status at 4 mo. Statistical comparisons of obesity risk between the earliest and the oldest quintiles of mother's menarche were performed by multivariate logistic regression, resulting in the estimation of adjusted odds ratios (OR) and 95 percent confidence intervals (CIs). For childhood obesity risk, mother's BMI was entered into the model as a covariable, because mother's prepregnancy BMI was positively related to her child's BMI at 9 y (r = 0.31, p < 0.001) [16].
