Cancer risks among studies of medical diagnostic radiation exposure in early life without quantitative estimates of dose

Abstract

Background: There is accumulating evidence of excess risk of cancer in various populations exposed at acute doses below several tens of mSv or doses received over a protracted period. There is also evidence that relative risks are generally higher after radiation exposures in utero or in childhood.

Methods and findings: We reviewed and summarised evidence from 89 studies of cancer following medical diagnostic exposure in utero or in childhood, in which no direct estimates of radiation dose are available. In all of the populations studied exposure was to sparsely ionizing radiation (X-rays). Several of the early studies of in utero exposure exhibit modest but statistically significant excess risks of several types of childhood cancer. There is a highly significant (p < 0.0005) negative trend of odds ratio with calendar period of study, so that more recent studies tend to exhibit reduced excess risk. There is no significant inter-study heterogeneity (p > 0.3). In relation to postnatal exposure there are significant excess risks of leukaemia, brain and solid cancers, with indications of variations in risk by cancer type (p = 0.07) and type of exposure (p = 0.02), with fluoroscopy and computed tomography scans associated with the highest excess risk. However, there is highly significant inter-study heterogeneity (p < 0.01) for all cancer endpoints and all but one type of exposure, although no significant risk trend with calendar period of study.

Conclusions: Overall, this large body of data relating to medical diagnostic radiation exposure in utero provides support for an associated excess risk of childhood cancer. However, the pronounced heterogeneity in studies of postnatal diagnostic exposure, the implied uncertainty as to the meaning of summary measures, and the distinct possibilities of bias, substantially reduce the strength of the evidence from the associations we observe between radiation imaging in childhood and the subsequent risk of cancer being causally related to radiation exposure.

Keywords: Childhood; In utero; Radiation; cancer risk.

Figure 1.. In utero exposure, odds ratio/relative risk (+95% CI) by midpoint year of study data ascertainment for (a) any cancer, (b) leukaemia, (c) brain/CNS tumour and (d) lymphoma.

Each point corresponds to a single cancer endpoint (generally one per study), using all studies and endpoints in Table 1 (see Supplementary Table S1). Dashed red line is odds ratio/relative risk = 1

Figure 2.. Postnatal exposure odds ratio/relative risk (+95% CI) for (a) leukaemia, (b) brain/CNS tumour and (c) lymphoma by midpoint year of study data ascertainment.

Each point corresponds to a single study and relevant endpoint in Table 2 (see Supplementary Table S2). Dashed red line is odds ratio/relative risk = 1

Figure 3.. Meta-regression for studies of in utero exposure. Restricted maximum likelihood (REML) fits to odds ratio or relative risk by calendar year midpoint of study data ascertainment range (for <1950, 1950–1959, 1960–1969, 1970–1979, 1980–1989, 1990+).

Plots are shown for (a) the four cancer endpoints analysis (leukaemia, lymphoma, brain/CNS cancer, other cancer) and (b) the any cancer endpoint analysis, for each in utero exposure study in Table 1 (see Supplementary Table S1). Dashed red line is odds ratio/relative risk = 1.

Figure 4.. Meta-regression for studies of postnatal exposure. Restricted maximum likelihood (REML) fits to odds ratio or relative risk by calendar year midpoint of study data ascertainment range (for <1960, 1960–1969, 1970–1979, 1980–1989, 1990–1999, 2000+).

Plots are shown for (a) the four cancer endpoints analysis (leukaemia, lymphoma, brain/CNS cancer, other cancer) and (b) the any cancer endpoint analysis, for each postnatal exposure study in Table 2 (see Supplementary Table S2). Dashed red line is odds ratio/relative risk = 1.

Figure 5.

Funnel plot of in utero exposure study odds ratios/relative risks, using any cancer endpoints within all studies, as in Table 1.

Figure 6.

Funnel plot of postnatal exposure study odds ratios/relative risks, using any cancer endpoints within all studies, as in Table 2.