Sir,

The recent paper by Journy et al (2015) addresses an important issue regarding the interpretation of epidemiological studies of CT scans and cancer risk. It has been suggested that raised risks reported in the studies in Northern England (Pearce et al, 2012) and Australia (Mathews et al, 2013) might reflect the early symptoms of undetected cancer, or of factors that predispose to cancer and which are the indications for the CT scans, rather than an effect of the CT scans per se (Walsh et al, 2014). The study of Journy et al–based on a cohort of children who received CT scans at 23 radiology departments in France–benefits from the availability of information on predisposing factors for cancer. However, I have concerns that their findings could be misinterpreted.

Table 1 here combines the results from Table 5 and Supplementary Table 6 from the study by Journy et al. The authors have highlighted that – for each cancer type – the estimate of the excess relative risk (ERR) per 1 mGy cumulative organ dose is lower with adjustment for predisposing factors than without such an adjustment. At face value, this might suggest confounding by indication, reflecting higher cancer risk and potentially higher radiation doses from CT scanning among children with predisposing factors compared with children without such factors. However, Table 1 here also shows that – for each cancer type – the ERR among children without predisposing factors is at least as large as the unadjusted value for the cohort overall, whereas the ERR among children with predisposing factors is close to zero. This suggests that the difference between the unadjusted and adjusted values principally reflects modification of the ERR by predisposing factors, rather than confounding.

Table 1 Number of cases and associated risks of primary tumours of the CNS, leukaemia, and lymphoma
Full size table

It is unclear from the study by Journy et al to what population the adjusted ERR estimates apply. Looking at Table 1, the adjusted estimates appear to be similar to a weighted average of the ERR estimates for those either with or without a predisposing factor, with weighting based on the numbers of cancer cases in each group. This would suggest that the adjusted estimates reflect the prevalence of predisposing factors among those children who developed cancer. However, from a public health perspective, it is more relevant to consider the prevalence of predisposing factors in the general population, rather than in the selected population of cancer patients. Fewer than 4% of the children in the cohort of Journy et al had a predisposing factor and the correspondence percentage for the general population is likely to be lower still, given that children with a predisposing factor may be more likely to receive CT scans than other children. On that basis, the ERR estimates specific to children without a predisposing factor would seem to be much more relevant to the general population than the adjusted estimates of Journy et al.

In view of the small number of cases in this study, inferences are limited. Further follow-up of this cohort and results from other studies that collect information on predisposing factors (e.g., Meulepas et al, 2014) would be valuable in providing further insights. Nevertheless, the findings of Journy et al do not indicate that the association between cancer risk and radiation exposure from CT scans has been confounded by predisposing factors for cancer.