Quantifying Cancer Risk in the Age of Routine CT: A National Projection of Diagnostic Radiation Harm

Computed tomography (CT) has revolutionized diagnostic medicine, providing fast, high-resolution imaging that supports the detection and management of countless diseases. From trauma evaluations in emergency departments to cancer staging in oncology clinics, CT is now a routine fixture in clinical practice. In 2023 alone, an estimated 93 million CT scans were performed in the United States, involving over 61 million patients. This widespread use has unquestionably enhanced clinical decision-making, enabling early diagnoses and improved outcomes across various medical specialties. However, the increasing reliance on CT has also raised concerns about the cumulative exposure of patients to ionizing radiation—a known carcinogen—especially given that many scans deliver radiation doses significantly higher than those associated with standard radiographs. The paradox of CT imaging lies in its dual nature: it is both a diagnostic marvel and a potential public health hazard. Ionizing radiation, particularly when used frequently or at high doses, can damage DNA and increase the risk of malignancies over a patient’s lifetime. While this risk is relatively small on a per-scan basis, the sheer volume of CT imaging performed nationally magnifies the public health implications. Notably, cancer risks are not uniformly distributed across populations. Children are particularly vulnerable due to their longer post-exposure life expectancy and greater sensitivity of developing tissues. Yet, adults account for the majority of cumulative radiation burden simply because they undergo the most imaging.

Historically, efforts to quantify CT-associated cancer risk have been limited by incomplete data, generalizations about patient anatomy, and coarse approximations of radiation doses. A pivotal 2009 study led by Berrington de González provided the first comprehensive projection, estimating that CT use in 2007 could result in 29,000 future cancers. That estimate, although groundbreaking at the time, was based on national imaging averages and dose assumptions derived from relatively outdated technology. In the years since, CT technology and utilization patterns have evolved substantially. Multiphasic imaging, for instance, which entails multiple scans in a single session, is now more common, and it significantly increases radiation exposure. Despite the availability of dose-reduction techniques, such as iterative reconstruction algorithms and automatic exposure control, their adoption remains inconsistent across health care systems.

These trends raised a critical question: how much cancer risk are we incurring under the current landscape of CT imaging? More specifically, how do age, sex, anatomical region imaged, and scan parameters contribute to that risk at the national scale today? The lack of up-to-date, granular data linking real-world CT practices to projected cancer outcomes created a knowledge gap with significant implications for both clinical practice and public health policy.

To this account, Professor Rebecca Smith-Bindman from the University of California, San Francisco and her team undertook a rigorous, methodologically sophisticated study to update and refine national projections of lifetime cancer risk associated with CT scans performed in 2023. Their work, published in JAMA Internal Medicine, leverages individual-level scan data from the UCSF International CT Dose Registry, which includes over 120,000 CT exams from diverse U.S. hospitals and imaging centers. Unlike previous assessments, this study does not rely on assumed dose distributions. Instead, it incorporates examination-specific acquisition parameters—such as scan length, tube current, and patient size—to simulate absorbed radiation doses in 18 organs using Monte Carlo radiation transport modeling. These doses are then paired with state-of-the-art risk modeling tools, including the National Cancer Institute’s RadRAT software, to generate lifetime cancer risk projections.

To investigate the cancer risk associated with modern CT imaging practices in the United States, the research team led by Professor Rebecca Smith-Bindman designed a study that integrated large-scale real-world imaging data with high-resolution dosimetry and advanced cancer risk modeling. Their approach was both methodologically rigorous and deeply grounded in clinical realities, reflecting a clear effort to bridge epidemiological insight with the granular details of radiological practice. Rather than relying on outdated assumptions or generalized scan categories, the team assembled a patient-level dataset drawn from the UCSF International CT Dose Registry, which includes scans performed across 143 hospitals and outpatient centers in 20 U.S. states. In total, this registry captured more than 120,000 CT exams, each tagged with detailed metadata such as patient age, sex, body region scanned, scan length, kilovoltage, milliamperage, and more.

What set this study apart was its precision in reconstructing the radiation dose absorbed by specific organs for each type of CT exam. Using Monte Carlo simulations (specifically, the MCNPX code), the team modeled radiation transport through anatomically realistic digital phantoms representing various ages and body types. This allowed them to compute absorbed doses for 18 distinct organs, taking into account real-world scan parameters rather than idealized ones. They found, for example, that abdomen and pelvis CT scans delivered some of the highest doses, particularly when performed with multiphase protocols, which are often unnecessary yet still commonly used. In contrast, extremity CT scans resulted in minimal organ exposure, highlighting how radiation risks vary significantly depending on the scan type.

Armed with these organ-specific dose estimates, the researchers moved to the next phase: estimating lifetime cancer risks. For this, they employed the National Cancer Institute’s RadRAT software, which uses risk models largely based on data from the Life Span Study of atomic bomb survivors, but updated and tailored to modern U.S. demographics. Importantly, the software factors in baseline cancer rates by age and sex, as well as mortality risks unrelated to cancer, providing a probabilistic output rather than a deterministic one. Applying RadRAT across the stratified dataset—418 unique combinations of patient age, sex, and CT exam type—the team generated cancer risk estimates for the entire U.S. population exposed to CT scans in 2023. The findings were sobering: roughly 103,000 future cancers were projected to arise from the 93 million CT scans performed that year. What made these projections especially compelling was the researchers’ ability to dissect them by subgroups. Although children had the highest per-scan risk due to their developing tissues and longer life expectancy, adults contributed over 90% of the projected cancers simply because they underwent the vast majority of CT scans. Lung cancer emerged as the most common outcome, with an estimated 22,400 cases attributed to radiation exposure, followed by colon cancer, leukemia, and bladder cancer. In women, breast and thyroid cancers also appeared prominently in the projections. Interestingly, head CTs were the leading cause of radiation-induced cancers in children, whereas abdomen and pelvis CTs were the dominant contributor in adults.

Moreover, the authors ran a battery of sensitivity analyses to test the robustness of their projections under different assumptions. For instance, when organ doses were varied by ±20%, the total number of projected cancers ranged from roughly 80,000 to 127,000—still substantial, even at the lower end. They also examined how results shifted if more or fewer CTs were performed on pediatric patients or if multiphase imaging was excluded. Notably, even under the most conservative modeling assumptions, the projected cancer burden remained high, reinforcing the urgency of the issue. Another revealing insight came from their analysis of CT use at the end of life. Since radiation-induced cancers take years to manifest, scans performed in the final year of life were excluded from risk calculations. By analyzing data from Kaiser Permanente Northern California, the team estimated that 10.6% of all CTs fell into this category and were unlikely to result in future cancers. Removing these from the risk pool gave a more accurate picture of long-term cancer incidence. It also suggested that a nontrivial portion of imaging might be used in contexts where the long-term risk is irrelevant, further emphasizing the need to differentiate between justified and potentially excessive use.

What became clear from these experiments and their resulting data is that radiation exposure from CT scans is not a hypothetical concern—it is a quantifiable, population-wide risk that demands greater attention. The researchers found that CT could account for up to 5% of all new cancer cases annually if current practices persist, putting it on par with other major modifiable risk factors like alcohol and obesity. And yet, unlike lifestyle choices, radiation exposure from imaging is externally administered and, in many cases, potentially avoidable or reducible through better scanning practices, protocol standardization, and stronger justification requirements.

The significance of this UCSF clinical study is in its ability to quantify, with unprecedented precision, the hidden long-term risks embedded in one of modern medicine’s most frequently used technologies. While  CT scans have become essential for diagnostic clarity, the research by Smith-Bindman and her team reveals that the cumulative consequence of this widespread imaging practice may be far more substantial than previously recognized. By using real-world data and detailed dose modeling for individual organs across diverse patient groups, the study not only updates prior estimates—it reframes the discussion entirely. What makes the findings so impactful is that they are not abstract predictions, but grounded projections based on actual scan parameters, patient demographics, and modern imaging frequencies. The estimate that CT imaging performed in just one year—2023—may result in over 100,000 future cancers gives new urgency to a conversation that has, until now, remained largely within the realm of theoretical concern. The implications are clear: the medical community must treat radiation exposure from diagnostic imaging with the same seriousness it affords other preventable health risks. One of the most striking implications is the sheer scale of risk created by standard clinical operations. These are not edge cases or experimental exposures but routine scans administered daily across the country. If nothing changes, CT use alone could account for approximately 5% of all new cancer diagnoses annually—a proportion comparable to other major modifiable risk factors. Unlike tobacco or diet, however, this risk is largely imposed rather than chosen, and therefore places an even greater ethical obligation on health systems to mitigate harm.

Clinically, the new study highlights the need for rigorous justification before ordering any CT scan, particularly when alternative imaging modalities like MRI or ultrasound could suffice. It also calls for immediate re-evaluation of multiphase scanning practices, which significantly increase dose but are often used reflexively, without clear added value. The research further highlights the importance of tailoring scan protocols to patient-specific factors—such as age, size, and clinical indication—to minimize unnecessary radiation exposure. On a policy level, the findings strengthen the argument for national radiation dose benchmarking and mandatory reporting, particularly for institutions that lag in adopting dose-reduction technologies. The study also supports broader implementation of decision-support tools that guide clinicians in selecting the most appropriate imaging study for each case. Insurance providers and regulatory agencies may find in this data a strong rationale for incentivizing low-dose practices and penalizing unjustified or excessive scanning.