Impact of Reverse Causation on Estimates of Cancer Risk Associated With Radiation Exposure From Computerized Tomography: A Simulation Study Modeled on Brain Cancer. Issue 1 (13th October 2021)

• Record Type: Journal Article
• Authors: Little, Mark P; Patel, Ankur; Lee, Choonsik; Hauptmann, Michael; Berrington de Gonzalez, Amy; Albert, Paul
• Is Part Of: American journal of epidemiology. Volume 191:Issue 1(2022)
• Page Start: 173
• Page End: 181
• Publication Date: 2021-10-13
• Subjects: brain tumors -- computed tomography -- ionizing radiation -- reverse causation -- simulation studies; Epidemiology -- Periodicals; Public health -- Periodicals; 614.4
• DOI: 10.1093/aje/kwab247
• Languages: English
• ISSNs: 0002-9262

Abstract

Abstract: Use of computed tomography (CT) scanning has increased substantially since its introduction in the 1990s. Several authors have reported increased risk of leukemia and brain tumors associated with radiation exposure from CT scans. However, reverse causation is a concern, particularly for brain cancer; in other words, the CT scan may have been taken because of preexisting cancer and therefore not have been a cause. We assessed the possibility of reverse causation via a simulation study focused on brain tumors, using a simplified version of the data structure for recent CT studies. Five-year–lagged and unlagged analyses implied an observed excess risk per scan up to 70% lower than the true excess risk per scan, particularly when more than 10% of persons with latent cancer had increased numbers of scans or the extra scanning rate after development of latent cancer was greater than 2 scans/year; less extreme values of these parameters imply little risk attenuation. Without a lag and when more than 20% of persons with latent cancer had increased scans—an arguably implausible scenario—the excess risk per scan was increased over the true excess risk per scan by up to 35%–40%. This study suggests that with a realistic lag, reverse causation results in downwardly biased risk, a result of induced classical measurement error, and is therefore unlikely to produce a spurious positive association between cancer and radiation dose from CT scans.