Case-control studies of childhood cancer and foetal exposure to diagnostic x-rays suggest that doses as small as 10 mSv increase the risk of cancer to a detectable extent. A comparison of the risk coefficient derived from the largest such study with that obtained from the Japanese atomic bomb survivors irradiated in utero (average dose, approximately 300 mGy) shows that, once all sources of uncertainty are taken into account, these risk estimates are not incompatible. The absence of a discernible variation in the risk per unit dose over this dose range is consistent with a linear dose-response. However, uncertainties are such that definitive conclusions on the shape of the dose-response at low doses cannot be drawn from this epidemiological evidence alone. Nonetheless, the evidence does suggest that the risk is not zero at doses of the order of 10 mSv.