Today, I review, link to, and excerpt from JAMA Internal Medicine‘s “Projected Lifetime Cancer Risks From Current Computed Tomography Imaging”. [PubMed Abstract] [Full-Text HTML] [Full-Text PDF]. JAMA Intern Med. 2025 Jun 1;185(6):710-719. doi: 10.1001/jamainternmed.2025.0505.
All that follows is from the above resource.
Key Points
Question How many future cancers could result from radiation exposure from annual computed tomography (CT) examinations in the United States?
Findings In this risk model, the 93 million CT examinations performed in 62 million patients in 2023 were projected to result in approximately 103 000 future cancers. Although the per-examination cancer risk was higher in children, higher CT utilization among adults accounted for the majority of the projected cancers.
Meaning These findings suggest that if current radiation dosing and utilization practices continue, CT-associated cancers could eventually account for 5% of all new cancer diagnoses annually.
AbstractImportance Approximately 93 million computed tomography (CT) examinations are performed on 62 million patients annually in the United States, and ionizing radiation from CT is a known carcinogen.
Objective To project the number of future lifetime cancers in the US population associated with CT imaging in 2023.
Design, Setting, and Participants This risk model used a multicenter sample of CT examinations prospectively assembled between January 2018 and December 2020 from the University of California San Francisco International CT Dose Registry. Data analysis was conducted from October 2023 to October 2024.
Main Outcomes and Measures Distributions of CT examinations and associated organ-specific radiation doses were estimated by patient age, sex, and CT category and scaled to the US population based on the number of examinations in 2023, quantified by the IMV national survey. Lifetime radiation-induced cancer incidence and 90% uncertainty limits (UL) were estimated by age, sex, and CT category using National Cancer Institute software based on the National Research Council’s Biological Effects of Ionizing Radiation VII models and projected to the US population using scaled examination counts.
Results An estimated 61 510 000 patients underwent 93 000 000 CT examinations in 2023, including 2 570 000 (4.2%) children, 58 940 000 (95.8%) adults, 32 600 000 (53.0%) female patients, and 28 910 000 (47.0%) male patients. Approximately 103 000 (90% UL, 96 400-109 500) radiation-induced cancers were projected to result from these examinations. Estimated radiation-induced cancer risks were higher in children and adolescents, yet higher CT utilization in adults accounted for most (93 000; 90% UL, 86 900-99 600 [91%]) radiation-induced cancers. The most common cancers were lung cancer (22 400 cases; 90% UL, 20 200-25 000 cases), colon cancer (8700 cases; 90% UL, 7800-9700 cases), leukemia (7900 cases; 90% UL, 6700-9500 cases), and bladder cancer (7100 cases, 90% UL, 6000-8500 cases) overall, while in female patients, breast was second most common (5700 cases; 90% UL, 5000-6500 cases). The largest number of cancers was projected to result from abdomen and pelvis CT in adults, reflecting 37 500 of 103 000 cancers (37%) and 30 million of 93 million CT examinations (32%), followed by chest CT (21 500 cancers [21%]; 20 million examinations [21%]). Estimates remained large over a variety of sensitivity analyses, which resulted in a range of 80 000 to 127 000 projected cancers across analyses.
Conclusions and Relevance This study found that at current utilization and radiation dose levels, CT examinations in 2023 were projected to result in approximately 103 000 future cancers over the course of the lifetime of exposed patients. If current practices persist, CT-associated cancer could eventually account for 5% of all new cancer diagnoses annually.
Sensitivity AnalysesThe sensitivity analyses generated a range in estimated future cancers from 79 900 to 126 600 by reducing and increasing organ doses by 20%, respectively, reflecting 22.2% fewer cancers to as many as 23.3% more cancers than the primary analysis (eTable 6 in Supplement 1). Using the IMV-estimated proportion of pediatric examinations resulted in an 11.0% increase in projected cancers overall (to 114 000) and an increase in the proportion of cancers from childhood imaging from 9.4% to 23.2%.
DiscussionCT is frequently lifesaving, yet its potential harms are often overlooked, and even very small cancer risks will lead to a significant number of future cancers given the tremendous volume of CT use in the United States. For current utilization and radiation dosing practices, we projected approximately 103 000 future cancers could result from CT use in the United States in 2023 (with sensitivity analyses projecting a range of 80 000 to 127 000) among the 62 million people who underwent CT. To provide context, if the number of new cancer diagnoses in the United States remains stable (1.95 million in 2023) and both the utilization and radiation doses from CT remain unchanged in future decades, CT could be responsible for approximately 5% of cancers diagnosed each year. This would place CT on par with other significant risk factors, such as alcohol consumption (5.4%) and excess body weight (7.6%).22
The projected number of radiation-induced cancers in this analysis is 3 to 4 times higher than the earlier assessment of CT exposure for several reasons.10 First, while growth in utilization has slowed over the intervening years,20 CT use is 30% higher today than in 2007, due to growth in low-value, potentially unnecessary imaging23–27 as well as population aging. Second, dose modeling in this study accounted for multiphase scanning, which occurs in 28.5% of examinations but was not modeled in the prior study, as multiphase frequency was unknown. Third, the substantially higher organ doses in this study were reconstructed using newer dosimetry methods with examination-level data from more than 120 000 actual examinations, while the prior study modeled doses from national survey data or imaging protocols and assumed examinations in children were performed using pediatric-specific settings. Lastly, we included more granular CT categories reflecting imaging indications that have important dose differences. Both studies used the same BEIR VII risk models; thus, this would not explain the large observed differences.
Lung cancer was projected to be the most common radiation-induced cancer, with 22 400 cases (eTable 4 in Supplement 1). Approximately 70% of these were in female patients, reflecting the higher BEIR VII risk coefficients in female patients. However, even when we applied male risk coefficients for female examinations, lung cancers were still the most common radiation-induced cancer in female patients. Colon cancer was the next most common, with 8700 cases (58.6% in males). It is unclear whether the current, unexplained increase in these 2 cancers as well as others at unexpectedly younger ages28 may be partly due to CT. Thyroid cancer also revealed notable sex differences, likely due to risk coefficients. For example, 1400 vs 320 thyroid cancers were projected to result from CT exposure in female and male patients, respectively, performed when the patient was younger than 1 year, despite equal thyroid organ doses (74.4 and 75.2 mGy) (eTables 2 and 4 in Supplement 1) and more examinations in male patients (Table 1). Our estimates from childhood CT exposure are higher than those in the large EPI-CT cohort study of pediatric cancer outcomes2,3 because we estimated lifetime risk of all cancer types, while EPI-CT examined brain and hematologic cancers within 15 years after exposure.
Abdomen and pelvis CT were projected to cause the greatest number of cancers. These and other examination types, such as high-dose abdomen and pelvis, full body, and spine CT, incur greater risks on average per examination because they frequently use multiple scan phases that result in higher doses.29,30 Often these examinations could use single-phase scanning, which would lower doses without impacting diagnostic accuracy.
This study estimated future cancers using the National Academy of Science BEIR VII–based modeling approach, which is widely accepted in the field of radiation epidemiology.
Many of the model assumptions were conservative. For example, we used the ACR’s percentage of estimated examinations in children, which is lower than the percentage from IMV. We did not include CT-guided procedures, such as biopsies, which often use higher doses. Furthermore, it is possible that the low-energy x-rays emitted by CT scans cause more cellular damage compared with gamma rays, which were the primary source of radiation released from the atomic bombs.33 Lastly, RadRAT applied the DDREF of 1.5 (90% uncertainty interval, 1.1-2.3) recommended by BEIR VII to account for differences between low-dose exposure and the higher doses for which the models were developed. This assumes lower radiation doses are less harmful (per unit) than higher doses, based on the BEIR VII estimate that the risk of solid cancer per unit of radiation dose may be 1.5 times lower for doses of 100 mGy or less.6 However, several systematic reviews of low dose (<100 mGy) and low dose rate exposures support a DDREF of 1.34,35 If accurate, the true number of projected cancers would be closer to the upper end of the sensitivity estimates than the primary analysis projects.
Strengths and LimitationsThis study has several strengths, including detailed data on CT utilization and associated radiation dose, detailed calculation of risks with uncertainty limits, and sensitivity analyses that provide a range of estimates under widely varying assumptions. There are several limitations: first, the BEIR VII risk estimated model parameters are based primarily on the Japanese survivor outcomes, and questions remain about the transfer of radiation risks from the mid-20th century Japanese population to the current US population. The use of a weighted average of the excess relative and excess absolute risk models aims to partly account for this, but these weights are subjective.36 Second, our risk calculations factored in average life expectancies, and the degree to which patients who undergo CT have shorter life expectancy due to underlying illness may overestimate future cancer risk. However, we excluded on average 10.6% of CTs that were likely performed during the last year of life, given these patients are not at risk of a radiation-induced cancer. A recent analysis found that 9.6% of patients who undergo CT died within 1 year,37 similar to our estimate. Third, while the CT categorization algorithm was 90% accurate compared with expert review,13 some examinations in the registry may have been miscategorized; however, this is unlikely to significantly impact our results.
ConclusionsIn this study, approximately 5% of annual cancer diagnoses or 100 000 cancers were projected to result from CT utilization in 2023. Despite public attention to the potential adverse effects, CT use has grown significantly in the United States since 2009. In 2023, 93 million CT examinations were performed in the United States; in 2007, the number was 68.7 million—a 35% increase incompletely explained by population growth.38 Justification of use and optimization of dose, including consideration of the need for multiphase examinations, are the tenets of CT imaging and must be applied uncompromisingly to mitigate potential harm.
