Article Text

Download PDFPDF

Radiation risk from coronary artery disease imaging: how do different diagnostic tests compare?
  1. Andrew J Einstein
  1. Dr A J Einstein, Columbia University Medical Center, 622 West 168th Street, PH 10-408, New York, NY 10032, USA; andrew.einstein{at}

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Invasive coronary angiography (ICA) has served as the “gold standard” for coronary artery disease diagnosis over the past two decades. The recent British Cardiovascular Society Working Group report on the role of non-invasive imaging, published in Heart, proposes that developments in non-invasive testing now challenge the primacy of ICA.1 Non-invasive testing modalities, such as exercise tolerance testing (ETT), myocardial perfusion scintigraphy (MPS), stress echocardiography (SE), cardiac computed tomography (CT) and cardiovascular magnetic resonance (CMR), offer alternative approaches to evaluate aspects of cardiac anatomy and physiology, while avoiding the morbidity, costs and risks of major complications from ICA.

Since selection of the optimal diagnostic test for a particular patient requires an assessment of the benefits and risks to the patient of each test under consideration, as well as the benefits and risks of not performing any testing, an understanding of the risks from ionising radiation can play an important role in patient management. In this editorial, I examine how radiation risk is quantified and how such estimates differ between different cardiac diagnostic tests.


The most common approach to characterising risks associated with ionising radiation, such as malignancies or genetic mutations in a patient’s progeny, is to describe it in terms of dose. The absorbed dose to an organ reflects the amount of energy imparted to the organ, divided by its mass. Absorbed dose does not reflect either the type of radiation or the sensitivity of the organ to radiation and as such is not itself a measure of risk. An equivalent dose to an organ normalises the absorbed dose, using a factor that reflects the relative biological effectiveness of the type(s) of radiation. In the current system of radiological protection that factor is 1 for the x rays and γ rays commonly used in cardiac imaging, and thus in corresponding units, the absorbed dose is equal to the equivalent dose.

The equivalent dose still does not reflect the chance of developing adverse outcomes in the organ of interest. A related quantity, the weighted equivalent dose, normalises the equivalent dose using a tissue weighting factor that reflects the relative detriment associated with radiation to different organs, serving as a composite measure of the risks of malignancies, their lethalities, years of life lost, effects on quality of life and, for the gonads, heritable effects as well. The organs currently with the greatest tissue weighting factors are the lungs, bone marrow, colon, stomach and female breasts.2 When the weighted equivalent doses of all organs are added up, one obtains a quantity called the effective dose (E). Since it is a single number, reflecting a radiation-associated risk from a study, E (measured in millisieverts; mSv) has become a popular quantity to compare the radiation risks of different diagnostic tests.

The effective dose, E, needs to be applied with some caution. Any estimate of E incorporates several sources of uncertainty in its determination.3 Moreover, E is defined to be applied to populations, and not for specific individual subjects. Thus, while it may be appropriate to assign an E of 10 mSv to a typical technetium-99m MPS study in a population, it is “off-label” to associate a dose of 9.3 mSv to Mr X’s study or 10.4 mSv to Mrs Y’s test.4

Describing E and/or organ doses of a typical study is one way to characterise radiation risk in cardiac imaging. An alternative approach is to estimate the risk of a particular adverse outcome—for example, cancer incidence—which is attributable to radiation exposure. Since radiation is a weak carcinogen with a typically long lag time between exposure and attributable cancer diagnosis, very large sample sizes and long follow-up are needed to establish a statistically significant relationship between low-dose radiation exposure and cancer risk in epidemiological studies. For example, the 15-Country Collaborative Study of Cancer Risk among Radiation Workers in the Nuclear Industry included more than 400 000 workers receiving an average cumulative dose of 19 mSv, with over 5 million person-years of follow-up to observe an excess relative risk of solid-tumour mortality of 0.97 (95% confidence interval 0.14 to 1.97) per Sv.5

To date, no comparable cohorts of patients exposed to cardiac imaging procedures have been studied. Thus cancer risks have been estimated by transporting risk from other populations exposed to radiation, most notably the Life Span Study cohort of about 120 000 Japanese atomic bomb survivors, but also medically and occupationally exposed cohorts.6 Point estimates of such cancer risks are associated with considerable uncertainty. One difference between such cancer risk estimates and E is that the former can reflect patient-specific factors that modulate cancer risk, such as age and gender, whereas E does not, by virtue of using tissue weighting factors derived for the whole spectrum of patient ages and genders. Thus, while E from a typical helical CT coronary angiogram (CTCA) will be identical for a population of young women and a population of older men, the respective estimated cancer risks attributable to the angiogram will be markedly different. Even so, E may be useful in comparing risk between two different tests—for example, dual isotope MPS versus prospectively gated CTCA.


E from ICA has been reported in at least 15 studies.7 This has been most commonly estimated by multiplying the kerma-area product derived from an ionisation chamber built into the fluoroscopy unit, by a conversion factor. E varies markedly from study to study, ranging from 2.3 to 22.7 mSv. As a typical value, the United Nations Scientific Committee on the Effects of Atomic Radiation cites a dose of ∼7 mSv,8 but at least four authors report attaining average doses below 4 mSv. Dose has been shown to be highly dependent on factors such as the fluoroscopy unit, procedural complexity, use of radiation-reducing techniques and angiographer experience and workload. Radial access is associated with higher dose than is femoral access.9 Left anterior oblique views, which tend to be more steeply angulated and therefore associated with a longer x-ray course through the thorax, are associated with higher dose than right anterior oblique or posteroanterior views. Lowering the fluoroscopy and acquisition frame rates, using retrospective storing of fluoroscopy rather than “cine” acquisition and decreasing detector dosing are among the radiation-reducing techniques that can be employed. An important radiation-related consideration in fluoroscopy is avoidance of skin injury, which should not occur from a single ICA but may occur from ICA in a patient with a history of multiple studies or interventional procedures. Careful intraprocedural dose monitoring and postprocedural follow-up of patients receiving significant doses can minimise morbidity from skin injury.10


E from MPS varies markedly, depending on the choice of radiopharmaceutical agent(s) and protocol used. E is typically estimated from compilations of dose coefficients, such as those of the International Commission on Radiological Protection or those of the Society of Nuclear Medicine’s Medical Internal Radiation Dose committee. Such dose coefficients are determined using mathematical biokinetic models that assume a typical anatomical model and incorporate organ- and radionuclide-specific time–activity data derived from human and/or animal studies, data on the absorption of energy in target organs deriving from Monte Carlo simulations, and data on the frequency and energy of nuclear transitions. As such, while such dose coefficients reflect the dosimetry to a “typical” patient, there are numerous potential sources of uncertainty.

Figure 1 summarises the effective doses to a typical patient from standard MPS protocols. For a typical patient, E for a standard 1-day low-dose/high-dose technetium-99m-sestamibi or tetrofosmin study is of the order of 10 mSv. A stress-only protocol, omitting the low-dose rest injection, results in a dose that is about 25% less. A typical single-injection thallium scan has an E of about 17 mSv, while a dual isotope thallium-sestamibi scan tops the charts with an E of about 24 mSv.

Whereas in the UK the vast majority of nuclear stress tests are performed with technetium-based agents, in the USA over a quarter of MPS studies are still performed with thallium, including 24% of studies that are dual isotope.13 Although in some patients, such as those with a history of technetium images obfuscated by gastrointestinal uptake, thallium may be the preferred radionuclide, from a dosimetric point of view patients are better off receiving technetium-99m. Even if a viability assessment might be desired, the total dose from rest–stress sestamibi MPS and [18F]fluorodeoxyglucose metabolic imaging is less than that of dual isotope imaging, while [18F]fluorodeoxyglucose provides better assessment of functional recovery.14 Dual isotope imaging, the protocol associated with the highest patient dose, may have advantages in its improved laboratory workflow; if it is used then it should at least be avoided in patients at higher risk of radiation-attributable cancer, such as younger patients without significant known comorbidities.

In the past couple of years, several vendors have attempted to develop more efficient gamma cameras—for example, incorporating multiple CdZnTe detectors as opposed to the now-standard two (NaI(Tl)) crystal detectors.15 While initial investigations of such cameras have focused on evaluating diagnostic performance using decreased imaging time, they also offer the possibility of decreasing the administered activity of radiopharmaceutical agents and thus correspondingly decreasing the dose of MPS.

Figure 2 Comparison of estimated effective doses for standard cardiac imaging protocols. For methodological details see Einstein et al.7 Administered activities of myocardial perfusion scintigraphy procedures here are average activities specified in American Society of Nuclear Cardiology guidelines (10 July 2006 version)11: 99mTc rest 10 mCi, 99mTc stress 27.5 mCi, 201Tl single dose 3.5 mCi, dual isotope study 201Tl 3.5 mCi and 99mTc 25 mCi, 82Rb 50 mCi rest and stress doses, 13NH3 15 mCi rest and stress doses. Standard administered activities vary from country to country12 and effective doses vary proportionally. Here 1 year of background radiation equals 2.4 mSv, reflecting the worldwide average.8 CTCA, computed tomography coronary angiography; ECTCM, electrocardiographically controlled tube current modulation; ICA, invasive coronary angiography; MPS, myocardial perfusion scintigraphy; PTG, prospective triggered gating.


E from a CT study can be estimated using either calculations based on dose measurements in a plexiglass phantom, computerised Monte Carlo scanner simulations, or by multiplying dosimetric information provided on the scanner console by an appropriate conversion factor. The current standard for CTCA is to perform a helical scan using a 64-slice multidetector-row scanner and retrospectively reconstruct images corresponding to a desired portion of the cardiac cycle.

Most commonly, x-ray tube current and voltage remain constant throughout the cardiac cycle. Each of the four major vendors of 64-slice scanners now offers the option to perform electrocardiographically controlled tube current modulation (ECTCM), decreasing the tube current outside of the prespecified desired portion of the cardiac cycle. This decreases E, typically by about one-third, but limits opportunities to reconstruct diagnostic-quality images from other phases. ECTCM should be applied when the heart rhythm is regular and rate is well controlled.

Effective dose estimates have been provided in at least 10 studies of 64-slice CTCA. These published estimates, from experienced centres largely employing Siemens scanners, have ranged from 8 to 21 mSv without ECTCM and from 7 to 14 mSv with ECTCM.7 Higher doses have been observed from less experienced laboratories. Studies incorporating aortic scanning as well as coronary scanning—for example, coronary artery bypass graft angiography, can be associated with higher doses, up to nearly 30 mSv. Calcium scoring is associated with a far lower dose, typically around 2 mSv.

For CTCA, estimates of the lifetime attributable risk of cancer incidence have been reported in addition to estimates of effective dose. These risk estimates vary markedly, depending on age, gender and scan type (use of ECTCM, aortic scanning). The estimated risk to a 60-year-old woman was about 1 in 700 and to a 60-year-old man was about 1 in 1900. In general, women had higher risks than men and younger patients had higher risks than older patients. Most risk was due to lung or breast cancer.6

Several new technologies offer the possibility of substantially lowering the dose from CTCA. Each major vendor now offers the possibility of axial scanning with prospective triggered gating, either using step-and-shoot scanning or whole-organ single-heartbeat imaging (eg, 320-slice CTCA). In this approach, the x-ray tube is off except during a specified portion of the cardiac cycle and the high degree of overlap in the areas irradiated in consecutive gantry rotations, characteristic of contemporary helical CTCA, is eliminated. Two early studies of prospective triggered gating report an average E of <3 mSv.16 17 Another method of lowering dose is to decrease x-ray tube voltage below the standard of 120 kVp.18 19 Other possibilities are currently more vendor-specific, including multiple-source scanning, more efficient detectors and improved ECTCM. While such approaches offer great promise for dose reduction, their diagnostic performance needs to be evaluated in comparison with the “gold standard” of ICA and compared with that of 120 kVp 64-slice helical scanning, which has demonstrated high accuracy in numerous studies.20


Ultrasound, and SE in particular, does not involve the use of ionising radiation. While ultrasonic energy has the ability to heat tissue, the consensus opinion of numerous organisations, such as the British Medical Ultrasound Society and the United States Food and Drug Administration, is that there is no current evidence that diagnostic ultrasound has resulted in harmful effects. Nevertheless, these organisations recommend using ultrasonography only for valid clinical purposes. Similarly, CMR involves no ionising radiation and current data do not support a relationship between exposure to radiofrequency radiation and cancer in humans. While radiofrequency pulses from CMR also produce heat, regulatory limits in the rate at which radiofrequency energy is absorbed in the patient, measured as the specific absorption rate, limit tissue heating to safe levels.


Radiation risk to patients from cardiac imaging is incompletely understood. The most commonly used measure of radiation risk has been the effective dose. Doses comparable to those typical for some cardiac imaging procedures have been associated with cancer risk. Point estimates of effective dose vary widely between procedures. Dual isotope MPS and helical 64-slice CTCA without ECTCM are associated with the highest E, which may be greater than 20 mSv in many populations. The lowest E are associated with [13N]ammonia PET MPS and calcium scoring. Newer CTCA techniques offer promising approaches for low-dose cardiac imaging but require further validation. ETT, SE and CMR do not involve the use of ionising radiation. Radiation risk is one factor among many that should be considered in selecting the optimal diagnostic test for a patient with known or suspected cardiac disease.


View Abstract


  • Funding: Supported in part by an NIH/NCRR Clinical and Translational Science Award (1 UL1 RR-24156-01) and by a Nuclear Cardiology Foundation Research Award.

  • Competing interests: AJE has received research support from Covidien, served as a consultant to GE Healthcare and received travel funding from GE Healthcare, INVIA, Philips Medical Systems and Toshiba America Medical Systems.