Childhood overweight and obesity at the start of primary school: external validation of pregnancy and early-life prediction models

Publication authors

Nida Ziauddeen, Paul Roderick, Gillian Santorelli, John Wright, Nisreen Alwan.


Background Tackling the childhood obesity epidemic can potentially be facilitated by risk-stratifying families at an early-stage to receive prevention interventions and extra support. As part of the Studying Lifecourse Obesity Predictors (SLOPE) study, we developed prediction models for childhood overweight and obesity using routinely-collected antenatal and early-life healthcare data in Hampshire, South of England.1 As a model usually performs better in the data used for its development, external validation is needed to check its performance in similar but new target population. This analysis aimed to externally validate these models using data from the Born in Bradford (BiB) cohort in the North of England.

Methods BiB is a longitudinal multi-ethnic birth cohort study which recruited 12,453 women (13,776 pregnancies) at around 28 weeks gestation between 2007 and 2010 in Bradford. Data from the routine National Child Measurement Programme measurement at 4–5 years in school was linked to the maternal and early-life BiB cohort data. The outcome was defined as body mass index (BMI) ≥91st centile at 4–5 years based on the UK clinical cut-off used to develop the SLOPE models. Maternal predictors included BMI, highest educational attainment, partnership and smoking status at booking, ethnicity and intake of folic acid supplements. Early life predictors included birthweight, gestational age, sex and weight at 1 or 2 years. Discrimination was assessed using the area under the receiver operating curve (AUC) and calibration using calibration slope (equal to one in well-calibrated model).

Results Data were available for 6292 women for the early pregnancy models and 3801 women and children for the early-life models. The AUC was comparable to the development model at all stages (early pregnancy, birth, ~1 year and ~2 years). The AUC at development was 0.66 (95% confidence intervals (CI) 0.65 to 0.67) compared to 0.64 (95% CI 0.62 to 0.66) on external validation. Similarly, the AUC was 0.83 (95% CI 0.82 to 0.84) at ~2 years at development and 0.81 (95% CI 0.79 to 0.83) on external validation. Models were less well-calibrated on external validation ranging from 0.87 (standard error (SE) 0.04) to 0.91 (SE 0.06) across the stages compared to 0.98 (SE 0.03) to 0.99 (SE 0.01) at development.

Conclusion The SLOPE models developed for predicting childhood overweight and obesity risk performed reasonably well on external validation in a birth cohort with a different geographical location and ethnic composition. However, recalibration by updating the model intercept may be required to improve calibration in other populations.