Growth trajectory during the first 1000 days and later overweight in very preterm infants

Objective: To identify the characteristics of early life growth associated with later overweight or obesity (OWO) in very preterm population.

Design: Length, weight and body mass index (BMI) were prospectively recorded from three prospective, population-based cohorts with 5 years (Loire Infant Follow-up Team (LIFT), EPIPAGE2 (Etude EPIdémiologique sur les Petits Ages GEstationnels 2)) and 15 years (EPIPAGEADO, Etude EPIdémiologique sur les Petits Ages GEstationnels-Adolescents) of follow-up. Missing data were imputed.

Setting: Regional (LIFT), national (EPIPAGE2) and multiregional (EPIPAGEADO) cohorts in France.

Patients: Eligible infants born before 33 weeks of gestation in 1997 (EPIPAGEADO), between 2003 and 2014 (LIFT), and in 2011 (EPIPAGE2).

Main outcome measures: OWO was determined as BMI Z-score >85th percentile of the WHO reference curves at 5 years (LIFT, EPIPAGE2) and 15 years (EPIPAGEADO).

Results: In EPIPAGEADO, LIFT and EPIPAGE2, BMI Z-scores were known for 302 adolescents, 1016 children and 2022 children, respectively. In EPIPAGEADO, OWO was observed in 42 (13.9%, 95% CI 10.5 to 18.3) adolescents. In multivariable models, birthweight Z-score, increase in weight Z-score during neonatal hospital stay and increase in BMI between discharge and at 2 years of corrected age were positively associated with OWO at 15 years (adjusted OR (aOR)=3.65, 95% CI 1.36 to 9.76; aOR=3.82, 95% CI 1.42 to 10.3; and aOR=2.55, 95% CI 1.72 to 3.78, respectively, by Z-score), but change in length Z-score during neonatal hospital stay was negatively associated (aOR=0.41, 95% CI 0.21 to 0.78, p=0.007). These four associations with OWO assessed at 5 years were confirmed in the LIFT and EPIPAGE2 cohorts.

Conclusions: Change in length Z-score during hospitalisation, a putative proxy of quality of neonatal growth, was negatively associated with risk of later OWO when change in BMI between discharge and at 2 years was included in the multivariable model.