Methods: : We utilized data among 1020 infants from a mother-offspring cohort, who were Singapore citizens or permanent residents of Chinese, Malay or Indian ethnicity with homogeneous parental ethnic backgrounds, and did not receive chemotherapy, psychotropic drugs or have diabetes mellitus. Ethnicity was self-reported at recruitment and later confirmed using genotype analysis. Subject-specific BMI curves were fitted to infant BMI data using natural cubic splines with random coefficients to account for repeated measures in each child. We estimated characteristics of the child's BMI peak [age and magnitude at peak, average pre-peak velocity (aPPV)]. Systolic (SBP) and diastolic blood pressure (DBP), BMI, sum of skinfolds (SSF) and fat-mass index (FMI) were measured during a follow-up visit at age 48 months. Weighted multivariable linear regression was used to assess the predictors (maternal BMI, gestational weight gain, ethnicity, infant sex, gestational age, birthweight-for-gestational age and breastfeeding duration) of infant BMI peak and its associations with outcomes at 48 months. Comparisons between ethnicities were tested using Bonferroni post-hoc correction.

Results: : Of 1020 infants, 80.5% were followed up at the 48-month visit. Mean (SD) BMI, SSF and FMI at 48 months were 15.6 (1.8) kg/m 2 , 16.5 (5.3) mm and 3.8 (1.3) kg/m 2 , respectively. Mean (SD) age at peak BMI was 6.0 (1.6) months, with a magnitude of 17.2 (1.4) kg/m 2 and pre-peak velocity of 0.7 (0.3) kg/m 2 /month. Compared with Chinese infants, the peak occurred later in Malay {B [95% confidence interval (CI): 0.64 mo (0.36, 0.92)]} and Indian infants [1.11 mo (0.76, 1.46)] and was lower in magnitude in Indian infants [-0.45 kg/m 2 (-0.69, -0.20)]. Adjusting for maternal education, BMI, gestational weight gain, ethnicity, infant sex, gestational age, birthweight-for-gestational-age and breastfeeding duration, higher peak and aPPV were associated with greater BMI, SSF and FMI at 48 months. Age at peak was positively associated with BMI at 48 months [0.15 units (0.09, 0.22)], whereas peak magnitude was associated with SBP [0.17 units (0.05, 0.30)] and DBP at 48 months [0.10 units (0.01, 0.22)]. Older age and higher magnitude at peak were associated with increased risk of overweight at 48 months [Relative Risk (95% CI): 1.35 (1.12-1.62) for age; 1.89 (1.60-2.24) for magnitude]. The associations of BMI peak with BMI and SSF at 48 months were stronger in Malay and Indian children than in Chinese children.

STUDY DESIGN: A wide range of socio-demographic characteristics of Chinese, Malay and Indian women attending routine gynecologic care in Singapore were prospectively collected. Physical performance was objectively measured by hand grip strength and the Short Physical Performance Battery. Percent VAT was determined by dual-energy X-ray absorptiometry. Fasting serum concentrations of glucose, insulin, IL-6, TNF- α, and hs-CRP were measured.

MAIN OUTCOME MEASURE: was insulin resistance, expressed as the homeostatic model assessment of insulin resistance (HOMA-IR).

RESULTS: 1159 women were analyzed, mean age 56.3 (range 45-69) years, comprising women of Chinese (84.0%), Indian (10.2%), and Malay (5.7%) ethnic origins. The adjusted mean differences for obesity (0.66, 95% CI 0.32-1.00), VAT area in the highest vs lowest tertile (1.03, 95% CI 0.73-1.34), low physical performance (0.63, 95% CI 0.05-1.24), and highest vs lowest tertile of TNF- α (0.35, 95% CI 0.13-0.57) were independently associated with HOMA-IR. Women of Malay and Indian ethnicity had higher crude HOMA-IR than Chinese women. However, after adjustment for obesity, VAT, physical performance, and TNF- α, no differences in mean HOMA-IR remained, when comparing Chinese women with those of Malay ethnicity (0.27, 95% CI -0.12 to 0.66) and with those of Indian ethnicity (0.30, 95% CI -0.01 to 0.66).