Intrauterine growth, birthweight and childhood cancer risk in the UK and USA

Prenatal as well as postnatal factors influence risk of a wide range of childhood tumours, including both normal and abnormal intrauterine (IU) growth patterns. We investigated this in datasets larger than any used previously, extending to tumours never previously considered in detail. We metaanalysed data on the reduced occurrence of childhood cancer in twins compared to singletons. We analysed data from a UK study of head circumference at birth and CNS tumour risk. We put these in the context of existing metaanalyses of CNS tumours and the acute leukaemias. Our main results derive from two similar population-based cancer registry studies in the UK and USA. The National Registry of Childhood Tumours includes birthweight for most children born and registered with cancer between 1980-2007 in England and Wales, and matched controls, but limited other IU growth information. The pooled USA dataset derives from statewide linkage of birth records to cancer registries for cases and matched controls in 5 states, born and registered between 1970-2004. Extensive information on factors which may confound the birthweight/risk relationship, such as gestation at birth is available. Twins experience a significant 20% risk reduction in total childhood cancer. A 1 cm increase in head circumference is associated with a significant 30% increase in brain tumour risk. From the cancer registry studies we had birthweight for 24,402 childhood cancer cases (all types) and 34,095 individually matched controls in the UK. For the USA we had data for 17,434 cases (all types) and 57,591 matched controls. The Odds Ratios (OR) and 95% Confidence Intervals (95% CI) for birthweight as a continuous risk factor, after adjusting for matching variables only, were nearly identical in each: 1.06 (1.04-1.08) (UK), 1.06 (1.05-1.08) (USA) per 500 g increase in birthweight, for all childhood cancers combined. In the USA further adjustments made little difference. There were stronger relationships with birthweight for some tumours than others, and for some risk was not related to increasing birthweight. Hepatoblastoma risk was conspicuous for its strong inverse relationship with birthweight: 0.81 (0.72- 0.91) (UK) and 0.65 (0.60-0.70) (USA) per 500 g increase. There was generally substantial agreement in the patterns observed by tumour type for both the UK and USA, strengthening confidence in the individual observations. Modelling suggested the birthweight risk relationship might be of linear or higher order term form, with particularly strong effects at the highest birthweights, for several tumour types including acute lymphoblastic leukaemia and some brain tumours. Normal IU growth, resulting in high birthweight, is associated with risk for about 60% of all childhood tumours. Factors associated with high birthweight may be responsible for an attributable fraction of 4% of all childhood tumours. Investigation of the underlying mechanisms is a priority.