Very preterm infants experience poor postnatal growth relative to intra-uterine growth rates but have increased percentage body fat (%fat). The aim of the present study was to identify nutritional and other clinical predictors of infant %fat, fat mass (FM) (g) and lean mass (LM) (g) in very preterm infants during their hospital stay. Daily intakes of protein, carbohydrate, lipids and energy were recorded from birth to 34 weeks postmenstrual age (PMA) in fifty infants born <32 weeks. Clinical illness variables and anthropometric data were also collected. Body composition was assessed at 34–37 weeks PMA using the PEA POD Infant Body Composition System. Multiple regression analysis was used to identify independent predictors of body composition (%fat, FM or LM). Birth weight, birth weight z-score and PMA were strong positive predictors of infant LM. After adjustment for these factors, the strongest nutrient predictors of LM were protein:carbohydrate ratios (102–318 g LM/0·1 increase in ratio, P = 0·006–0·015). Postnatal age (PNA) and PMA were the strongest predictors of infant FM or %fat. When PNA and PMA were accounted for a higher intake of energy (–1·41 to –1·61 g FM/kJ per kg per d, P = 0·001–0·012), protein (–75·5 to –81·0 g FM/g per kg per d, P = 0·019–0·038) and carbohydrate (–27·2 to –30·0 g FM/g per kg per d, P = 0·012–0·019) were associated with a lower FM at 34–37 weeks PMA. Higher intakes of energy, protein and carbohydrate may reduce fat accumulation in very preterm infants until at least 34–37 weeks PMA.