Varying optimal power for height-standardisation of childhood weight, fat mass and fat-free mass across the obesity epidemic

Childhood adiposity markers can be standardised for height in the form of indices (marker/height ) to make meaningful comparisons of adiposity patterns within and between individuals of differing heights. The optimal value of p has been shown to differ by birth year, sex, age, and ethnicity. We investigated whether height powers for childhood weight and fat mass (FM) differed by birth year, sex, or age over the period before and during the child obesity epidemic in Copenhagen. Population-based cross-sectional study of 391,801 schoolchildren aged 7 years, 10 years and 13 years, born between 1930 and 1996, from the Copenhagen School Health Records Register. Sex- and age-specific estimates of the height powers for weight and FM were obtained using log-log regression, stratified by a decade of birth. For weight, amongst children born 1930-39, optimal height powers at 7 years were 2.20 (95% CI: 2.19-2.22) for boys and 2.28 (95% CI: 2.26-2.30) for girls. These increased with birth year to 2.82 (95% CI: 2.76-2.87) and 2.92 (95% CI: 2.87-2.97) for boys and girls born in 1990-96, respectively. For FM, amongst those born 1930-39, powers at 7 years were 2.46 (95% CI: 2.42-2.51) and 2.58 (95% CI: 2.53-2.63) for boys and girls, respectively, and increased with birth year reaching 3.89 (95% CI: 3.75-4.02) and 3.93 (95% CI: 3.80-4.06) for boys and girls born 1990-96, respectively. Powers within birth cohort groups for weight and FM were higher at 10 years than at 7 years, though similar increases across groups were observed at both ages. At 13 years, height powers for weight and FM initially increased with the birth year before declining from the 1970s/80s. Due to increases in the standard deviation of weight and FM during the obesity epidemic, optimal height powers needed to standardise childhood weight and FM varied by birth year, sex, and age. Adiposity indices using a uniform height power mean different things for different birth cohort groups, sexes, and ages thus should be interpreted with caution. Alternative methods to account for height in epidemiological analyses are needed.