Cumulative effective dose among 5.6 million CT exams in a multinational radiation dose registry

Abstract

Objectives

To understand the percentage of patients who exceed a cumulative effective dose of 50 or 100 mSv within 1-5 years of first CT scan, while accounting for data censoring (i.e. considering patients who are no longer being followed in the study and whose unobserved exams may influence the estimates)—a major limitation of prior reports.

Methods

A retrospective, multi-national study of 5.6 million CT exams in 2.8 million patients from 161 hospitals between 2010 and 2021. We quantified the percentage of exams with effective doses >50 and >100 mSv, and using survival analysis, projected the percentage of patients with cumulative effective doses above these levels at 1 and 5 years and identified relevant drivers (eg, patient size, exam type, country).

Results

A total of 189 981 (3.4%) examinations were >50 mSv and 14 906 (0.3%) >100 mSv, varying by patient age, size, and exam type. Overall, 13% of average-sized adults and 5% of average-sized children were estimated to accumulate 50 mSv over 5 years (56% and 27% of large-sized adults and children, respectively), whereas 4% of average-sized adults and 1% of average-sized children were estimated to accumulate 100 mSv over 5 years (31% and 13% of large-sized adults and children, respectively). England had a 1.7- to 5-times lower than average percentage of patients exceeding levels.

Conclusions

High cumulative doses were not rare, particularly high in larger patients, and repeat imaging was a strong driver of these exposures.

Advances in knowledge

When accounting for data censoring, a considerably larger share of patients accumulated a high effective dose from CT than has been previously reported, exceeding radiation dose ranges associated with radiation-induced cancer.

Introduction

CT plays an invaluable role in disease diagnosis and monitoring but is associated with a small but significant increased risk of cancer.^1–5 Furthermore, cancer risk increases with cumulative exposure from repeated imaging over a patient’s lifetime.^1,6,7 Thus, quantifying how many and which patients receive high cumulative doses may help clinicians fully assess the potential benefits and harms of imaging decisions, particularly in patients who undergo repeat imaging.

Cumulative dose is commonly measured by effective dose (in units of millisieverts, mSv), which reflects a whole body, risk-weighted sum of absorbed doses to individual organs and tissues. Effective dose is helpful for tracking cumulative exposure in a single patient, as it allows for the summation of doses from different exam types and over time. Although it is widely accepted that cancer risk increases with dose, there is no known threshold below which radiation is risk-free, reflected in the linear-no-threshold risk model of radiation carcinogenesis.¹ Indeed, a recent cohort study that assessed medical imaging use and hematologic cancers in nearly 4 million children provides directly observed evidence of the dose-response relationship, with elevated risk observed even at low doses.⁵ Nevertheless, there is ample evidence that doses above 50 and 100 mSv raise lifetime cancer risk, with the National Academies’ Biologic Effects of Ionizing Radiation (BEIR) VII report estimating 1 cancer in 100 people exposed to 100 mSv.^1,8–10 Thus, levels of 50 and 100 mSv are commonly used in research to identify elevated radiation exposures from medical imaging. It is understood that these are markers of patients with elevated exposures and not discrete thresholds where risk abruptly increases.

A growing body of literature has found that a small proportion of patients exceed a cumulative effective dose of these levels, and it has been estimated that nearly 1 million patients globally will accumulate >100 mSv each year.^7,11–15 However, a shortcoming of prior studies is their lack of accounting for incomplete or unobserved data (or censoring). For example, patients may have obtained CT at one hospital but gone elsewhere for subsequent care. Patients enter and leave the defined study populations at different times, and previous studies that have not accounted for this have undercounted repeated exams. Herein, we performed survival analysis—a useful technique for handling censoring—using data from a large CT radiation dose registry to assess cumulative effective dose and how it varies by demographic and anthropometric patient factors.

Methods

Data were derived from the UCSF International CT Dose Registry (“registry”), created to pool CT data from consecutive exams performed at hospitals and outpatient facilities associated with 27 healthcare organizations across 20 US states and 7 countries. Registry details and methods have been previously reported.^16,17 Radiation and imaging data stored in Digital Imaging and Communications in Medicine (DICOM) format are exported to the registry through Radimetrics dose management software. The registry was approved with a waiver of informed consent by the UCSF institutional review board, and participating institutions either relied on the UCSF approval or obtained local ethics approval to contribute.

Data sources

Our cohort consists of patients who underwent diagnostic CT exams between January 1, 2010, and March 11, 2021, at a facility that contributed to the registry for at least 3 months and where the patient’s first exam occurred at least 7 days prior to the facility’s last day of participation. CT exams performed for image-guided procedures, cardiac imaging, research, or missing key variables (medical record number, sex, age, CT category, effective diameter, or radiation dose) were excluded. Patients’ CT use over time was tracked through individual medical record numbers.

CT category

CT exams were classified into 1 of 14 categories, which reflect a combination of anatomic area and radiation dose required (high, routine, low) based on the image quality needs of the clinical indication.¹⁸ Two additional categories, whole body CT and PET-CT, which were not included in the initial publication of CT categories, were added to this analysis for a total of 16. For PET-CT, we included the CT dose only, not the radiopharmaceutical component. Exams were sorted into CT categories by applying a validated natural language processing algorithm on protocol name and study description.¹⁸

Radiation dose metric

The effective dose was calculated using 2 methods. The first method (computationally intensive) accounted for 90% of exams in children and 33% in adults. Patient age, sex, diameter, and body region, in combination with exam-level CT acquisition parameters, were used to reconstruct patient-dependent organ-absorbed doses through Monte Carlo radiation transport simulations.^19–21 The resulting organ doses were multiplied by ICRP Publication 103 tissue weighting factors and summed to calculate effective dose.²² The second method was used in the remaining exams, where the machine-reported dose length product was multiplied by published adult- and pediatric-specific effective dose coefficients developed using summary results from the first method.^23,24 A patient’s cumulative effective dose was defined as the sum of effective doses from all CT exams up to a given time.

Accounting for patient size

As radiation dose varies by patient size, we stratified the analyses by patient size into 2 strata. Large-sized patients were defined as those with any exam-level effective diameter higher than the 80th percentile for the corresponding CT category, sex, and age stratum (<1, 1-4, 5-11, 12-17, and 18+ years of age). If a patient was defined as large on any CT exam, they were considered large in all analyses. Remaining patients were classified as average-sized.

Statistical analyses

We first assessed the percent of individual CT exams with an effective dose exceeding 50 or 100 mSv and described variation by patient age (at first exam), sex, size, and CT category.

To account for censoring, we used Cox regression (survival analysis) to compute the percentage of patients with a cumulative effective dose exceeding 50 or 100 mSv, categorized by age, sex, size, CT category, and country, at 1 and 5 years after their first observed CT exam. These computations were done separately for average- and large-sized patients, using 3 different approaches to censor patients from the cohort. In method A, patients were censored at the time of their last observed exam. This tends to overestimate imaging, as it assumes patients continue to undergo imaging outside of the observed facility through the remainder of the observation period. In method B, patients were censored at the date of their associated facility’s last observed CT exam. This tends to underestimate, as it assumes patients receive no imaging outside of the given facility throughout the observation period, thus missing exams obtained outside observed facilities. In method C, patients were assumed to remain in the cohort for 3 years after their last observed exam, or until their facility exited the cohort, whichever came first. This method assumed, in the absence of observed exams, patients received no further imaging for up to 3 years, after which their absence is attributed to having left the cohort.

The hazard ratio (HR) and 95% CI for each predictor were extracted from the Cox regression model using method C, as we believe it represented an intermediate approach to dealing with unobserved data that neither underestimated (by assuming no further outside imaging occurred) nor overestimated (by assuming an inaccurately high amount of outside imaging). For patient sex, age, and country, the HR reflects the ratio of hazard functions between the stratum of interest and the referent stratum shown in the tables, whereas for CT category, it reflects the hazard function between patients with at least 1 exam in the CT category vs patients without any exams in the category.

We estimated the percentages of patients exceeding the 50 and 100 mSv levels, and the mean time to meet these levels, for the average patient overall (defined as a patient who is statistically average in all studied predictors) and for the typical patient within each predictor stratum (defined as a patient who presents the trait denominated by the stratum, but who is otherwise statistically average).

Results

The study included 2 841 051 patients (52.4% female, 6.2% children <21 years, and 17.1% large-sized patients) who underwent 5 575 682 CT exams (Table 1). Most exams (84.7%) were performed in the United States. Abdomen and Pelvis Routine Dose (1 597 896 exams, 28.7%), Head Routine Dose (1 246 551 exams, 22.4%), and Chest Routine Dose (1 179 820 exams, 21.2%) accounted for 72.3% of all exams. The combined chest, abdomen, and pelvis had the highest exam-to-patient ratio, making it the exam type most repeated. The mean number of exams per patient was 1.96, and the median was 1. Facilities contributed to the registry for a mean of 5 years (SD of 2.5 years, range of 103 days to 11 years, Figure S1).

Table 1.

Characteristics of patients and CT exams included in the study cohort.

	Patients		Exams
	n	Column, %	n	Column, %
Total	2 841 051		5 575 682
Sex
Female	1 489 842	52.4	2 901 994	52.0
Male	1 351 209	47.6	2 673 688	48.0
Age group (years) a
<1	9214	0.3	10 723	0.2
1 to <5	17 657	0.6	22 882	0.4
5 to <12	32 204	1.1	41 221	0.7
12 to <18	59 865	2.1	78 354	1.4
18 to <21	59 620	2.1	84 616	1.5
21 to <35	359 220	12.6	566 875	10.2
35 to <65	1 298 791	45.7	2 486 083	44.6
65+	1 004 480	35.4	2 284 928	41.0
Size group
Average-sized	2 356 315	82.9	4 286 146	76.9
Large-sized	484 736	17.1	1 289 536	23.1
Country
England	82 902	2.9	136 033	2.4
Germany	46 523	1.6	84 578	1.5
Netherlands	54 175	1.9	104 198	1.9
Switzerland	62 359	2.2	119 176	2.1
Israel	224 026	7.9	291 735	5.2
Japan	62 653	2.2	117 336	2.1
United States	2 308 413	81.3	4 722 626	84.7
Exams per patient
1	1 795 595	63.2
2	510 086	18.0
3	216 302	7.6
4+	319 068	11.2
CT category b , c
Head
Low dose	200 456	5.0	217 145	3.9
Routine dose	863 674	21.7	1 246 551	22.4
High dose	16 749	0.4	19 499	0.3
Abdomen and pelvis
Low dose	76 547	1.9	86 912	1.6
Routine dose	1 095 385	27.6	1 597 896	28.7
High dose	121 228	3.1	143 450	2.6
Chest
Low dose	10 218	0.3	12 410	0.2
Routine dose	754 278	19.0	1 179 820	21.2
Spine
Neck and cervical	280 074	7.1	319 169	5.7
Thoracic and/or lumbar	57 816	1.5	62 107	1.1
Extremities	55 123	1.4	59 917	1.1
Head and neck
Routine dose	135 507	3.4	150 872	2.7
High dose	13 567	0.3	14 938	0.3
Combined chest, abdomen, and pelvis	239 117	6.0	387 613	7.0
Whole body	5733	0.1	6839	0.1
PET-CT	47 088	1.2	70 544	1.3

^aThe Patient column reflects the patient’s age upon entry to the registry at the time of their first CT; the exam column reflects the patient’s age at the time of the exam.

^bThe CT category in an examination-level variable. The patient column shows the number of patients who have at least 1 exam in the category, but since patients may have multiple scans of different CT categories, the sum of the patient-level column will be greater than the number of patients.

^cCardiac exams (2% of total) were excluded as they cannot be size-adjusted due to the method in which Radimetrics determines the effective diameter.

Individual exams exceeding 50 or 100 mSv

In average-sized patients, 90 054 exams (2.1%) exceeded an effective dose of 50 mSv (median 61.7 mSv), and 5483 (0.13%) exceeded 100 mSv (median 115 mSv), with rates varying by CT category (Tables S1 and S2). In large-sized patients, 99 927 exams (7.7%) exceeded 50 mSv, and 9423 (0.73%) exceeded 100 mSv. CT categories with a higher proportion of exams exceeding levels included Abdomen and Pelvis High Dose (2.6% of all exams, 15% exceeded 50 mSv), Neck and Cervical Spine (5.7% of exams, 7.8% exceeded 50 mSv), Thoracic and/or Lumbar Spine (1.1% of exams, 13% exceeded 50 mSv), and exams that cover multiple body regions such as combined head and neck high dose (0.3% of exams, 9.6% exceeded 50 mSv), chest, abdomen, and pelvis (7.0% of exams, 7.0% exceeded 50 mSv), whole body (0.1% of exams, 3.7% exceeded 50 mSv), and PET-CT (1.3% of exams, 16% exceeded 50 mSv).

Cox regression results

The 3 censoring assumptions yielded a wide range of estimates of how many patients exceed dose levels, where methods A and B bracket the method C estimate (Figure S2). For average-sized patients overall at 5 years, the proportion exceeding 50 mSv ranged from 8% (method B, the most conservative assumption) to 65% (method A, the most permissive), with our intermediate method C estimate of 12%; the corresponding range for 100 mSv was 2%–28% (method C estimate: 3%). For large-sized patients, the proportion exceeding 50 mSv at 5 years ranged from approximately 40% to 85% (method C: 56%), and the proportion exceeding 100 mSv ranged from approximately 22% to 78% (method C: 31%). Subgroup-specific method A and method B estimates followed a comparable pattern. The mean time to reaching 50 mSv was estimated at 9.5 years (Table 2), and mean time to 100 mSv was 10.6 years (Table 2).

Table 2.

Percent of patients who exceed 50 and 100 mSv dose levels.

	Average-sized patients						Large-sized patients
	Exceed 50 mSv			Exceed 100 mSv			Exceed 50 mSv			Exceed 100 mSv
	Percent		Mean time to dose (years)	Percent		Mean time to dose (years)	Percent		Mean time to dose (years)	Percent		Mean time to dose (years)
	at 1 year, %	at 5 years, %		at 1 year, %	at 5 years, %		at 1 year, %	at 5 years, %		at 1 year, %	at 5 years, %
All patients	4	12	9.5	1	3	10.6	22	55	4.8	7	31	7.1
Female	4	10	9.7	1	3	10.7	21	52	5.1	6	27	7.4
Male	5	15	9.2	1	4	10.4	24	58	4.5	8	34	6.7
Adults	5	13	9.3	1	4	10.5	23	56	4.7	7	31	7.0
Children < 21 years	2	5	10.4	0	1	10.9	10	27	7.7	3	13	9.1
Age group (years)
<1	2	7	10.2	0	1	11	8	24	8	2	9	9.7
1 to <5	1	4	10.6	0	1	10.9	8	23	8.1	2	11	9.4
5 to <12	1	3	10.7	0	1	11	5	14	9.2	1	6	10.1
12 to <18	1	3	10.7	0	1	11	7	20	8.4	2	10	9.6
18 to <21	3	8	10	0	2	10.8	19	48	5.5	5	25	7.6
21 to <35	3	10	9.8	1	2	10.7	21	52	5.1	6	27	7.4
35 to <65	5	13	9.3	1	4	10.5	23	56	4.7	7	32	6.9
65+	5	15	9.2	1	4	10.5	23	57	4.7	7	31	7
Country
England	1	4	10.6	0	1	11	12	33	7	2	12	9.4
Germany	2	7	10.2	0	2	10.8	13	37	6.6	3	17	8.7
Netherlands	4	11	9.6	1	4	10.5	18	48	5.5	6	27	7.5
Switzerland	5	13	9.4	1	3	10.7	15	41	6.1	3	17	8.7
Israel	5	13	9.3	1	3	10.7	26	62	4.2	7	32	6.9
Japan	3	9	9.9	0	2	10.8	15	40	6.3	4	21	8.1
United States	5	13	9.4	1	4	10.5	23	56	4.7	7	32	6.9
CT Category
Head
Low dose	2	7	10.2	0	2	10.8	15	41	6.1	5	22	8
Routine dose	3	9	9.9	1	3	10.7	19	48	5.5	5	26	7.6
High dosea	6	15	9.1	1	5	10.4	26	61	4.3	8	35	6.7
Abdomen and pelvis
Low dosea	4	12	9.5	1	3	10.6	20	51	5.2	7	31	7.1
Routine dose	7	19	8.6	1	6	10.1	25	60	4.4	9	38	6.3
High dose	13	33	7.1	2	9	9.7	44	84	2.4	15	58	4.6
Chest
Low dose	2	5	10.4	0	1	10.9	8	24	7.9	2	11	9.5
Routine dose	8	22	8.3	2	8	9.9	25	61	4.3	9	41	6
Spine
Neck and cervical	10	25	7.9	1	7	10.1	28	65	4	9	41	6
Thoracic and/or lumbar	14	34	6.9	2	9	9.7	35	74	3.2	12	51	5.2
Extremities	2	6	10.2	0	2	10.8	15	41	6.2	5	23	7.9
Head and neck
Routine dose	8	22	8.3	1	6	10.2	30	68	3.7	8	38	6.4
High dose	16	39	6.4	2	9	9.7	37	77	2.9	11	48	5.4
Chest, abdomen, and pelvis	16	40	6.2	4	16	8.8	41	81	2.6	17	63	4.2
Whole body a	5	13	9.3	1	5	10.4	16	43	5.9	6	28	7.3
PET-CT	14	35	6.8	2	11	9.5	47	87	2.2	18	65	4.1

^aPredictor does not meet statistical significance at the 95% CI.

Results varied by patient factors, CT category, and country (Table 2). Males exceeded the levels at a higher percentage than females and did so more quickly (likely due to their larger physical size), as did adults compared to children: 13% of average-sized adults vs 5% of average-sized children accumulated 50 mSv at 5 years, and 4% vs 1% accumulated 100 mSv. With the exception of the youngest children (<1 year), the overall likelihood of exceeding levels increased with age, with 15% and 4% of adults aged 65+ exceeding 50 mSv and 100 mSv, respectively, in 5 years (Figure 1 and Table 2).

Figure 1.

Expected proportion of patients exceeding 50 and 100 mSv effective dose levels over time, by patient age group.

Compared with average-sized patients, a far higher percentage of large-sized patients exceeded dose levels. 56% of large-sized adults exceeded 50 mSv (range approximately 40%–97% across methods A-C) and 31% exceeded 100 mSv (range approximately 19%–78%) by 5 years. Among large-sized children, 27% exceeded 50 mSv and 13% exceeded 100 mSv by 5 years (Table 2). Large-size patients exceeded these levels more quickly than average-size patients: 4.8 vs 9.5 years to reach 50 mSv and 7.1 vs 10.6 years to reach 100 mSv.

CT categories associated with high single-exam effective doses (eg, Abdomen and Pelvis High Dose) were also strongly associated with a high probability of exceeding these dose levels (Table 2 and Figure S3). The chest, abdomen, and pelvis showed the highest accumulation, with 40% of average-sized patients exceeding 50 mSv (81% in large patients) and 16% exceeding 100 mSv (63% in large patients) at 5 years. However, even CT categories not associated with particularly high doses per exam were associated with high cumulative doses. For example, average-sized patients who underwent Abdomen and Pelvis Routine Dose also frequently exceeded 50 mSv (19%) and 100 mSv (6%) at 5 years.

The United States had among the highest percentages of patients exceeding levels, while England had the lowest. Compared to the average adult, England had a 1.7- to 5-times lower percentage of patients exceeding levels (Figure 2).

Figure 2.

Expected proportion of patients exceeding 50 and 100 mSv effective dose levels over time, by country.

The multivariate results are provided in Table 3. These confirm the strongest predictors of high cumulative dose are CT category (eg, the HR for exceeding 50 mSv in an average-sized patient who undergoes Abdomen and Pelvis Routine Dose vs a patient who did not is 2.12) and country; eg, the HR for exceeding 50 mSv in average-sized patients in the United States is 3.46, or approximately 3 times higher than in England; also see Figure 2.

Table 3.

Hazard ratios (95% CI) for the likelihood of reaching a cumulative effective dose of 50 mSv or 100 mSv by patient sex and age, CT category, and country.

	Average-sized patients				Large-sized patients
	Exceed 50 mSv		Exceed 100 mSv		Exceed 50 mSv		Exceed 100 mSv
	Hazard ratio	P-value	Hazard ratio	P-value	Hazard ratio	P-value	Hazard ratio	P-value
	[95% CI]		[95% CI]		[95% CI]		[95% CI]
Female	Ref.		Ref.		Ref.		Ref.
Male	1.42 [1.37, 1.47]	<.001	1.57 [1.49, 1.65]	<.001	1.20 [1.16, 1.24]	<.001	1.32 [1.25, 1.39]	<.001
Age group (years)
<1	0.68 [0.38, 1.21]	<.001	0.32 [0.14, 0.75]	<.001	0.37 [0.24, 0.58]	<.001	0.30 [0.16, 0.59]	<.001
1 to <5	0.40 [0.27, 0.60]		0.40 [0.19, 0.84]		0.35 [0.24, 0.52]		0.38 [0.26, 0.55]
5 to <12	0.27 [0.18, 0.40]		0.26 [0.16, 0.45]		0.21 [0.14, 0.30]		0.20 [0.12, 0.32]
12 to <18	0.30 [0.24, 0.38]		0.28 [0.19, 0.41]		0.31 [0.22, 0.44]		0.32 [0.22, 0.46]
18 to <21	0.81 [0.67, 1.00]		0.76 [0.57, 1.02]		0.90 [0.79, 1.03]		0.92 [0.76, 1.11]
21 to <35	Ref.		Ref.		Ref.		Ref.
35 to <65	1.40 [1.27, 1.56]		1.59 [1.40, 1.80]		1.14 [1.05, 1.23]		1.23 [1.10, 1.38]
Age 65+	1.58 [1.39, 1.80]		1.65 [1.41, 1.94]		1.15 [1.04, 1.27]		1.17 [1.01, 1.37]
Country
England	Ref.	<.001	Ref.	<.001	Ref.	<.001	Ref.	<.001
Germany	1.80 [1.06, 3.06]		3.21 [1.91, 5.40]		1.16 [0.62, 2.18]		1.51 [0.79, 2.89]
Netherlands	3.07 [2.58, 3.64]		7.17 [6.14, 8.37]		1.64 [1.21, 2.23]		2.56 [1.92, 3.41]
Switzerland	3.37 [2.65, 4.29]		5.34 [4.20, 6.79]		1.35 [0.97, 1.87]		1.48 [1.06, 2.06]
Israel	3.62 [1.92, 6.83]		5.27 [2.80, 9.92]		2.47 [1.41, 4.32]		3.19 [1.66, 6.12]
Japan	2.42 [0.75, 7.80]		4.28 [1.18, 15.49]		1.29 [0.55, 3.02]		1.96 [0.70, 5.45]
United States	3.46 [2.79, 4.29]		7.60 [5.91, 9.76]		2.09 [1.52, 2.89]		3.16 [2.27, 4.39]
CT Category
Head
Low dose	0.50 [0.42, 0.60]	<.001	0.59 [0.53, 0.66]	<.001	0.66 [0.60, 0.73]	<.001	0.69 [0.62, 0.75]	<.001
Routine dose	0.59 [0.53, 0.65]	<.001	0.64 [0.57, 0.72]	<.001	0.80 [0.74, 0.86]	<.001	0.80 [0.73, 0.88]	<.001
High dose a	1.27 [0.80, 2.02]	.315	1.38 [0.82, 2.34]	.229	1.18 [0.84, 1.67]	.342	1.16 [0.74, 1.84]	.519
Abdomen and pelvis
Low dose a	1.00 [0.82, 1.22]	.99	1.01 [0.81, 1.26]	.947	0.88 [0.71, 1.10]	.269	1.00 [0.73, 1.38]	.995
Routine dose	2.12 [1.89, 2.38]	<.001	2.65 [2.24, 3.14]	<.001	1.29 [1.18, 1.41]	<.001	1.80 [1.60, 2.03]	<.001
High dose	3.15 [2.73, 3.63]	<.001	2.90 [2.50, 3.37]	<.001	2.42 [2.14, 2.74]	<.001	2.51 [2.21, 2.84]	<.001
Chest
Low dose	0.40 [0.32, 0.51]	<.001	0.38 [0.29, 0.49]	<0.001	0.35 [0.29, 0.43]	<.001	0.30 [0.25, 0.37]	<.001
Routine dose	2.32 [1.98, 2.71]	<.001	3.15 [2.61, 3.81]	<.001	1.28 [1.12, 1.46]	<.001	1.84 [1.59, 2.13]	<.001
Spine
Neck and cervical	2.37 [1.97, 2.86]	<.001	2.08 [1.68, 2.58]	<.001	1.36 [1.16, 1.59]	<.001	1.55 [1.29, 1.87]	<.001
Thoracic and/or lumbar	3.30 [2.40, 4.53]	<.001	2.83 [2.21, 3.64]	<.001	1.72 [1.40, 2.11]	<.001	1.97 [1.59, 2.45]	<.001
Extremities	0.49 [0.42, 0.57]	<.001	0.63 [0.56, 0.71]	<.001	0.65 [0.60, 0.71]	<.001	0.71 [0.63, 0.78]	<.001
Head and neck
Routine dose	1.94 [1.39, 2.71]	<.001	1.70 [1.06, 2.72]	.027	1.44 [1.12, 1.86]	.004	1.31 [0.94, 1.82]	.116
High dose	3.77 [2.34, 6.08]	<.001	2.78 [1.70, 4.54]	<.001	1.87 [1.19, 2.94]	.007	1.79 [1.06, 3.01]	.029
Chest, abdomen, and pelvis	4.37 [3.80, 5.02]	<.001	5.84 [4.91, 6.95]	<.001	2.29 [2.07, 2.53]	<.001	3.16 [2.80, 3.56]	<.001
Whole body a	1.09 [0.30, 4.04]	.894	1.45 [0.45, 4.64]	.534	0.71 [0.25, 2.04]	.521	0.91 [0.32, 2.59]	.856
PET-CT	3.34 [2.23, 5.00]	<.001	3.37 [2.10, 5.43]	<.001	2.56 [1.79, 3.68]	<.001	2.92 [1.99, 4.29]	<.001

^aPredictor does not meet statistical significance at the 95% CI.

Discussion

While effective doses exceeding 50 or 100 mSv in a single exam were uncommon for most CT categories, we estimated that 13% of average-sized adults in this analysis would accumulate a dose exceeding 50 mSv in 5 years, and that 4% would exceed 100 mSv (56% and 31% in large-sized adults, respectively). While we do not have detailed information on why the patients received these doses or whether they received necessary repeated imaging (eg, was imaging for oncologic or trauma imaging or surveillance of incidental findings), doses in this range have the potential to cause cancer in a small number of patients, underscoring the importance of balancing this risk against the potential benefits of CT imaging. As awareness grows regarding the rise in low-value imaging and the overuse of CT scans, it is essential to prioritize efforts to minimize patient exposure at these dose levels. Strategies to achieve this include avoiding unnecessary exams and optimizing protocols to reduce radiation exposure per exam whenever feasible.

In lieu of universal patient effective dose tracking, there is no exhaustive way to measure cumulative dose from medical imaging. Prior studies failed to account for unobserved imaging, yet survival analysis relies on assumptions about such incomplete data. Our 3 censoring methods reflect varying assumptions about periods without observed CT exams, significantly impacting results. Method C, used in our primary analysis, provides a middle-ground estimate (notably similar to the most conservative method B, as shown in Figure S2) and still indicates a meaningful proportion of patients accumulated high exposures. Our primary results are similar to a 2025 report by Ferretti et al., which found 6.6% of adult patients in a single hospital had a cumulative dose >100 mSv over 11 years without accounting for censoring, compared to our 4% in 5 years.¹⁴ Universal patient effective dose tracking could be a valuable approach to monitoring patient exposures. While several hospitals already track cumulative doses, adopting a universal standard for dose monitoring could improve awareness of this issue and encourage more judicious use of medical imaging.

While previous studies have not examined cumulative dose by patient size in quite the same way, large-sized patients in this analysis reached dose levels in far higher proportions than average-sized patients. Larger patients require some increase in radiation output to maintain adequate image quality, and a portion of the difference we observed is attributable to that physical requirement. However, the magnitude of the increase—with approximately one-third of larger patients exceeding 100 mSv over 5 years (method C estimate)—is greater than would be expected from size-driven dose adjustment alone, consistent with prior reports that size-specific dose estimates rise rather than remain constant across the patient-size spectrum. This pattern suggests that protocol optimization for body habitus remains an important opportunity for improvement. Where feasible, efforts to reduce these doses include avoiding multiphase protocols, repeat surveillance, and choosing the lowest dose protocols that will answer the clinical questions. These findings are driven both by high-dose individual CT exams in some categories (particularly chest, abdomen, spine, and PET-CT), but more importantly, by the fact that patients undergo repeated imaging: 37% of patients in the cohort received more than 1 observed exam, despite the limited observation period, a figure that almost certainly understates the true rate of repeat imaging because patients may receive follow-up imaging at facilities outside the registry.

For average-sized children <1 year, 7% will exceed 50 mSv of cumulative dose in 5 years. It could be assumed that these patients are generally sicker than older children in the cohort, and CT utilization may reflect essential imaging, such as frequent surveillance of a congenital condition. Similarly, patients undergoing chest, abdomen, and pelvis exams may be cancer patients getting both high-dose and frequent exams for surveillance. However, with the growing awareness in the use of low-value exams, these exams may also reflect CT that can and should be avoided.

We did not have clinical data to explain why individual patients received high cumulative doses, and therefore, cannot distinguish appropriate from potentially inappropriate exposures. A high cumulative dose may reflect clinically necessary repeated imaging (such as oncologic surveillance) or examinations that inherently require high-dose protocols, but may also reflect protocols that deliver more dose than needed to answer the clinical question or imaging that could have been avoided. Given the scale of exposure observed, these findings reinforce the need to ensure imaging is performed only when clearly indicated and that doses are uniformly optimized.

The analyses were completed by CT category, reflecting a combination of body region and the underlying dose requirement based on the clinical indication for scanning. For example, colonography is assigned to abdomen and pelvis low dose, enterography goes into Abdomen and Pelvis Routine Dose, while assessment of cancer or acute gastrointestinal bleeding goes into Abdomen and Pelvis High Dose.¹⁸ While these category names may not use nomenclature from everyday practice, they provide a useful framework for grouping clinical indications with similar radiation requirements for comparison across facilities. These demonstrate that higher dose categories contribute to high cumulative doses both because of the higher needed dose per exam, but also because patients frequently undergo repeated studies (eg, for combined chest, abdomen, and pelvis).

There were significant differences across countries in the proportion of patients exceeding 50 and 100 mSv, driven by both per-exam doses and the frequency of repeated imaging. Differences of this magnitude (the United States average vs England differing by approximately 1.7- to 5-fold) are unlikely to be fully explained by case-mix or differing scopes of practice. The healthcare-quality literature has long recognized unwarranted in-service delivery, that is, variation not explained by patient need, preference, or evidence, as a primary marker of opportunities for systematic improvement.^25,26 Patterns observed here, therefore, motivate further evaluation of the systemic factors influencing CT utilization and dose, including the emphasis placed on justification, the availability and preference for alternative modalities (ultrasound and MRI), and reimbursement structures. We cannot, from these data alone, attribute these differences to any specific quality factor at any country or facility, and the relatively smaller number of non-US facilities means we may not fully represent imaging practice in those countries.

A primary strength of this study is the large, diverse cohort of 5 575 682 exams (including a large sample in children, N = 237 796), making this the largest known study of cumulative effective dose. The exclusion of cardiac and procedure-related CT (because of challenges related to measuring patient size and dose) will have resulted in an underestimate of exposures; however, because these examination types are relatively uncommon, the impact will have been small. The use of survival analysis accounts for variation in the opportunity to observe whether patients attain these dose levels, as patients who enter the records near the end of a facility’s observation period, or whose facility was observed only for a short time, or who left an observed facility and underwent imaging at outside facilities, have some probability of exceeding these dose levels within or outside the observation period. We examined 3 analytical models that provided minimum and maximum estimates of reaching these dose levels and presented results from the intermediate model, which was closer to the conservative estimate. One limitation of the study was the 2 different methods used to calculate effective dose. This was done due to constraints on running the computationally intensive Monte Carlo simulations on the full 5.5 million exams. However, the methods are closely aligned, as the effective dose coefficients (method 2) were generated based on results of method 1, and the simpler method results in a close approximation of the more computationally complex method.^23,24 Additionally, it is important to note that while effective dose is widely used for population-level comparisons as done here, its application to individual patient risk estimation has limitations and should be interpreted cautiously.

Conclusion

The decision to image must always weigh potential benefits such as reliable and timely diagnosis with risks, which include not only ionizing radiation exposure but also harms such as patient pain or complications from intravenous contrast, stress, financial cost, and invasive or unnecessary downstream care. The likelihood that the benefits of CT outweigh the harms is greater in high-value exams, such as severe trauma, vs low-value exams, such as for chronic headache, pulmonary embolism in low-risk patients, or ongoing surveillance of very small lung nodules that have not changed on 1 year imaging—all of which are on the rise.^27–31 Cumulative dose concerns should never prevent a medically necessary CT exam, but ordering clinicians should be aware that certain exam types in certain patients routinely deliver effective doses in ranges known to be carcinogenic, and repeat imaging may compound the risk considerably.

Among 5.6 million CT exams in 2.8 million patients, individual examinations exceeding 50 or 100 mSv were uncommon, but a sizable proportion of patients accumulated effective doses above these levels over time, primarily through repeated imaging. The likelihood of accumulating a high cumulative dose was substantially greater for larger‑bodied patients and varied widely by CT category and country. Across the 3 censoring assumptions, the estimated proportions varied widely but were of meaningful magnitude under all 3, and the relative rankings across patient subgroups, exam types, and countries were consistent. These findings support continued attention to 2 longstanding principles of radiation protection in medical imaging: that each CT examination be justified and that radiation doses be optimized to the lowest level consistent with diagnostic image quality.

Supplementary material

Supplementary material is available at BJR online.

Conflicts of interest

Dr R.S.B. is a co-founder of Alara Imaging, Inc., a company focused on improving the clinical and operational aspects of health systems, including collecting and reporting CT radiation dose and image quality as part of payer-led quality programs. Alara Imaging had no role in the study. The other authors of this manuscript have no disclosures.

Funding

This research was supported by awards from the National Cancer Institute (1R01CA181191-01A1), by the Patient-Centered Outcomes Research Institute (CD-1304-7043), and by residual class settlement funds in the matter of April Krueger v. Wyeth Inc., case No. 03-cv-2496 (US District Court, SD of Calif). Funders were not involved in the study design, conduct, interpretation, reporting of data, or the publication. The views in this article are solely the authors’ and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute, its Board of Governors or Methodology Committee, nor other funders.