Potential for adult-based epidemiological studies to characterize overall cancer risks associated with a lifetime of CT scans

Abstract

Recent epidemiological studies have suggested that radiation exposure from pediatric CT scanning is associated with small excess cancer risks. However, the majority of CT scans are performed on adults, and most radiation-induced cancers appear during middle or old age, in the same age range as background cancers. Consequently, a logical next step is to investigate the effects of CT scanning in adulthood on lifetime cancer risks by conducting adult-based, appropriately designed epidemiological studies. Here we estimate the sample size required for such studies to detect CT-associated risks. This was achieved by incorporating different age-, sex-, time- and cancer type-dependent models of radiation carcinogenesis into an in silico simulation of a population-based cohort study. This approach simulated individual histories of chest and abdominal CT exposures, deaths and cancer diagnoses. The resultant sample sizes suggest that epidemiological studies of realistically sized cohorts can detect excess lifetime cancer risks from adult CT exposures. For example, retrospective analysis of CT exposure and cancer incidence data from a population-based cohort of 0.4 to 1.3 million (depending on the carcinogenic model) CT-exposed UK adults, aged 25-65 in 1980 and followed until 2015, provides 80% power for detecting cancer risks from chest and abdominal CT scans.

FIG. 1

Predicted CT-induced cancer incidence rates peak during the same “cancer-prone” age range as background cancer (roughly 50–80), even if CT exposure occurred much earlier. Here the sex-averaged background cancer incidence rate for the UK population (http://www.cancerresearchuk.org/) is compared with the incidence rates of excess cancers induced by a single chest CT scan administered at ages 10 or 40 [predicted by the BEIR-VII model (5)]. Both CT-induced and background cancer incidence rates were corrected for age-dependent mortality (http://www.statistics.gov.uk/hub/index.html). Qualitatively similar curves were produced if other radiation carcinogenesis models were used instead of the BEIR VII model.

FIG. 2

The age at exposure dependences of radiation carcinogenesis, predicted by three different models: Model 1, BEIR VII (5); Model 2, Preston et al. (14, 18); and Model 3, Shuryak et al. (15). The y-axis represents sex-averaged excess relative risk per unit dose (ERR/Gy) at an attained age of 70 for all cancers combined.

FIG. 3

A schematic diagram of the process used here to estimate the sample size of a population-based cohort study designed to detect CT-induced cancer risks. Details are described in this article.

FIG. 4

Effects of cohort age and study duration on total numbers of CT-exposed cohort members needed to satisfy the statistical criteria of 95% significance and 80% power for detecting CT-induced cancer risks. The curves were produced from calculations using the Preston et al. model (14, 18) to estimate CT-induced risks from ten cancer types accumulated within a UK cohort beginning in 1980. Qualitatively similar curves were produced if other radiation carcinogenesis models were used.

FIG. 5

The relative contributions of CT-exposed cohort members of different ages to the power of a hypothetical study designed to measure CT-induced cancer risks. This calculation was based on estimation of combined risks from ten cancer types, which would accumulate (over 10 million person-years of follow-up between 1980 and 2015) within simulated UK cohorts composed of individuals of a specific age in 1980. CT-induced cancer risks were calculated using the Preston et al. model (14, 18), and these risk estimates were used to calculate study power. Qualitatively similar curves were produced if other radiation carcinogenesis models were used.

FIG. 6

Estimated cohort sizes required for population-based cohort studies of CT-induced cancer risks. The cohort sizes represent the numbers of CT-exposed individuals needed to satisfy the statistical criteria of 95% significance and 80% power for detecting CT-induced cancer risks, while assessing combined risks from ten cancer types accumulated within the cohort from 1980 to 2015 in the UK, calculated using three different radiation carcinogenesis models. Model 1, BEIR-VII (5); Model 2, Preston et al. (14, 18); and Model 3, Shuryak et al. (15). Each cohort was assumed to be composed of individuals of a specific age in 1980.