Child-to-adult body size change and risk of type 2 diabetes and cardiovascular disease

Aims/hypothesis Childhood overweight increases the risk of type 2 diabetes and cardiovascular disease in adulthood. However, the impact of childhood leanness on adult obesity and disease risk has been overlooked. We examined the independent and combined influences of child and adult body size on the risk of type 2 diabetes and cardiovascular disease. Methods Data from the UK Biobank on 364,695 individuals of European ancestry and free of type 2 diabetes and cardiovascular disease were divided into nine categories based on their self-reported body size at age 10 and measured BMI in adulthood. After a median follow-up of 12.8 years, 33,460 individuals had developed type 2 diabetes and/or cardiovascular disease. We used Cox regression models to assess the associations of body size categories with disease incidence. Results Individuals with low body size in childhood and high body size in adulthood had the highest risk of type 2 diabetes (HR 4.73; 95% CI 4.50, 4.99), compared to those with average body size in both childhood and adulthood. This was significantly higher than the risk in those with high body size in both childhood and adulthood (HR 4.05; 95% CI 3.84, 4.26). By contrast, cardiovascular disease risk was determined by adult body size, irrespective of childhood body size. Conclusions/interpretation Low body size in childhood exacerbates the risk of type 2 diabetes associated with adult obesity but not the risk of cardiovascular disease. Thus, promoting healthy weight management from childhood to adulthood, among lean children, is crucial. Graphical Abstract Supplementary Information The online version contains peer-reviewed but unedited supplementary material available at 10.1007/s00125-023-06058-4.


Electronic supplementary material (ESM)
Title: Child-to-adult body size change and risk of type 2 diabetes and cardiovascular disease Authors: Carrasquilla et al

Lifestyle factors, family history, and prevalent disease definitions
We defined healthy diet scores based on shared features from the World Health Organization recommendations [1] and traditional healthy Western and Oriental diets [2].Such diets predominantly emphasize an increased intake of vegetables, fresh fruits, whole grains, and healthy fats while recommending reduced consumption of fatty and processed meats and salt.To gather data, we used a touchscreen questionnaire, which captured participants' reported weekly intake of various food and drink items as part of a food frequency questionnaire.The healthy diet score was calculated by summing daily servings of cooked and fresh vegetables and fresh fruits.It further considered factors such as consumption of whole grain bread (1 for yes, 0 for no), bran/oatmeal cereal (1 for yes, 0 for no), one healthy fish serving per week (1 for yes, 0 for no), ≤ 1 processed meat serving per week (0 for yes, 1 for no), ≤ 1 red meat serving per week (0 for yes, 1 for no), and the addition of extra salt to food (0 for yes, 1 for no).
Sedentary time was assessed considering data from three activities: driving, computer use, and television watching time.If an individual reported any of these activities, that value was taken into account.When two or three activities were reported, the mean value was used.This provided a single measure of sedentary behaviour for each individual.
Physical activity time (minutes per day) was defined by combining three distinct types of physical activities: walking, moderate physical activity, and vigorous physical activity.To consider the diverse intensities and contributions of these activities, we applied distinct weightings: 1 for walking (lighter intensity), 2 for moderate activity (moderate intensity), and 3 for vigorous activity (higher intensity).To ensure that our measure of physical activity remained robust, we accommodated missing data by implementing an imputation strategy.Specifically, we calculated the mean value for situations with available data and replaced missing values with this calculated mean.
Smoking status was defined into three categories: current smokers, former smokers, or individuals who have never smoked tobacco products.
Family history of diabetes encompassed the presence or absence of a medical history of diabetes mellitus within an individual's immediate family, including their mother, father, or siblings, where one or more of these relatives had received a diagnosis of diabetes mellitus.
Prevalent cases were defined as medical conditions that have been previously diagnosed by a doctor, including diseases such as cancer and severe respiratory illnesses (i.e., chronic obstructive pulmonary disease).

Genome-wide association analysis and polygenic scores for adult body mass index
We conducted a genome-wide association study (GWAS) to elucidate the genetic determinants of adult body mass index (BMI) utilizing the REGENIE software [3] that accounts for relatedness, including the following covariates: age, sex, genotype chip, assessment centre, and the first 15 genetic principal components.
A total of 418,133 European individuals were enrolled in this analysis.Details about the genotyping and single nucleotide polymorphism (SNP) imputation procedure have been previously outlined [4].This study focused on 776 variants that reached genome-wide significance and were independent (p<5x10-9, R2<0.1 and ±500 kilobases window) to be considered further for constructing polygenic risk scores.
We obtained a weighted polygenic risk score (PRS) using recently validated polygenic score methods in the UK Biobank [5].Each SNP was categorized as having 0, 1, or 2 riskincreasing alleles based on their genetic makeup.These values were then multiplied by their respective effect sizes.These scores were then split into tertiles to define individuals with low, medium, and high genetic risk for adult obesity.We then rerun Cox models for the groups of interest.

ESM Fig. 4 .
Cardiovascular risk across child and adult body size groups separately.The top panels are the Kaplan-Meier estimates of incidence disease (A-B), and the bottom panels are adjusted hazard ratios for disease risk (C-D).Cox regressions were adjusted for age, sex, Townsend deprivation index, and assessment centre.ESM Fig. 8. Risk of myocardial infarction and stroke by child-adult body size categories.Shown are adjusted hazard ratios for disease risk across childhood-adulthood body size categories for combined and sex-stratified analysis.In these comparisons, individuals in the ChildAverage-AdultAverage served as the reference group.Cox regression was adjusted for age, sex, educational attainment, Townsend deprivation index, and assessment centre.Bars indicate 95% confidence intervals.ESM Fig. 9. Odds ratios for type 2 diabetes and cardiovascular risk, including prevalent cases across child to adult body size categories.Panels A, B, and C show odd ratios for type 2 diabetes, and Panels D, E, and F show odds ratios for cardiovascular disease.Logistic regression models were adjusted for age, sex, educational attainment, Townsend deprivation index, and assessment centre.Bars indicate 95% confidence intervals.ESM Fig.10.Risk of type 2 diabetes and cardiovascular disease across child to adult body size categories with and without accounting for birth weight.Hazard ratios for disease risk across child-adult body size categories for combined and sex-stratified analysis (women/men).Type 2 diabetes is shown in the panels A and C, and cardiovascular disease in panels B and D. Models were adjusted for the same covariates as the main model, including age, sex, Townsend deprivation index, and assessment centre (A and B).Birth weight was added to the model on top of the covariates above (C and D).In these comparisons, individuals in the ChildAverage-AdultAverage served as the reference group.Bars indicate 95% confidence intervals.ESM Fig.13.Disease risk across genetic risk tertiles for adult obesity for ChildLow-AdultHigh and ChildHigh-AdultHigh categories.Hazard ratios for disease risk are reported across tertiles of polygenic scores for adult obesity (high, medium, and low).Type 2 diabetes is shown on the left panel, and cardiovascular disease on the right.Models were adjusted for the same covariates as the main model, including age, sex, Townsend deprivation index, and assessment centre.In these comparisons, individuals in the ChildAverage-AdultAverage served as the reference group.Bars indicate 95% confidence intervals.

Table 1 . Education qualifications based on the International Standard Classification
1. World Health Organization."Action framework for developing and implementing public food procurement and service policies for a healthy diet."(2021).2. Cena, Hellas, and Philip C. Calder."Defining a healthy diet: evidence for the role of