Radiation Protection of Patients Undergoing Cardiac Computed Tomographic Angiography

Interest in the use of computed tomography (CT) for cardiac evaluation has increased rapidly since the introduction of 64-slice scanners. Reflecting this, the installation base of CT scanners in US cardiology practices has tripled in the past 2 years.¹ Reports of the high diagnostic performance of coronary CT angiography (CTA), and especially its high negative predictive value in populations with low-to-intermediate prevalence of coronary disease,^{2 - 3} have been tempered by a concern about its high radiation dose to patients and the attendant risk of cancer.⁴

Despite a number of single-center studies that have reported a wide range of effective doses for coronary CTA,⁵ the existing literature does not adequately answer the questions of what radiation doses patients actually receive in clinical practice, and what factors are associated with higher radiation dose. Such information should help practitioners develop protocols that are in accordance with the goal of maintaining radiation exposure to patients as low as reasonably achievable (the ALARA principle). The Prospective Multicenter Study On Radiation Dose Estimates Of Cardiac CT Angiography In Daily Practice I (PROTECTION I), an observational study of worldwide cardiac CTA practice in 2007 reported by Hausleiter and colleagues in this issue of JAMA,⁶ represents an effort to fill this gap.

The primary outcome measure used to quantify radiation dose in PROTECTION I is the dose-length product (DLP),⁷ a CT-specific term unfamiliar to many physicians. The DLP is a reflection of the total amount of radiation deposited over the entire set of images comprising a patient's CT series, reported in mGy × cm. Better known is the effective dose, a measure applicable beyond the confines of CT and reported in millisieverts (mSv). Effective dose weights the concentrations of energy deposited in each organ from a radiation exposure using factors reflecting the type of radiation and the relative detriment to each organ of potential radiation-associated mutagenic changes. Although effective dose can be compared between different types of exposures, the factors used in its determination to weight each organ are approximate population averages, and therefore it is imprecise to report the effective dose of an individual patient's study.⁸ Thus, in characterizing the amount of radiation from a single CTA study, DLP is a more objective physical metric than effective dose, and the reason PROTECTION I is replete with DLP data. However, effective dose is appropriate to refer to in a population of patients and is especially useful for comparing between different types of exposures in populations with similar age and sex distributions.⁹ As in numerous previous studies, effective dose of CTA is estimated in PROTECTION I by multiplying a median DLP by a conversion factor suggested by the European Commission.^{7 ,10 - 11}

Hausleiter et al⁶ present a number of interesting and surprising findings about radiation dose from cardiac CTA. The estimated overall median effective dose for CTA, excluding calcium scoring when performed as part of the same study, was 12 mSv, somewhat less than the value reported in several earlier studies using 64-slice scanners.^{12 - 14} A few factors may account for this. Although the conversion factor used in PROTECTION I to estimate effective dose from DLP was derived using single-slice scanners and chest rather than cardiac scan sequences, this factor is the most current available.¹⁰ The value is approximately 20% lower than that used in previous studies, and thus would be expected to result in effective doses that are correspondingly lower. Additionally, as pointed out by Hausleiter et al, invitation of study sites on the basis of previous publications may have introduced a bias favoring more expert centers, which would be expected to be more proficient in managing radiation dose. Indeed, dose-reduction techniques were used in at least 80% of patients undergoing 64-slice CTA for coronary assessment.

Consistent with the suggestion of lower dose in centers with greater expertise, Raff et al¹⁵ recently reported a reduction in median effective dose of CTA from 25 mSv in the initial month to 13 mSv in the ninth month of a statewide quality improvement program in Michigan, with no change in image quality. This program used a collaborative process, involving initial counseling of high-dose sites, comparative reports, and scanner-specific training. In contrast, longer experience and increased frequency of performing cardiac CTA were not associated with sizable changes in DLP in PROTECTION I. The reason for this discrepancy is unclear, but raises the possibility that increased procedure volume alone does not translate to dose optimization in the absence of a concerted and educated effort to reduce patient radiation dose.

The variability in DLP between sites and scanners observed in PROTECTION I is striking. Median DLP at the highest-dose site was more than 6 times that at the lowest-dose site, and doses ran the gamut in between these extremes. Thus, cardiac CTA may be associated with significantly higher or lower effective dose than standard nuclear stress testing protocols,⁵ depending on how CTA is practiced at a site. Median DLP at the Latin American and South American sites was 3 times that at US and Canadian sites. Two factors evaluated by Hausleiter et al appear to contribute significantly to this variability: differential utilization of dose-reduction strategies and differences in dosimetry between scanners.

Four dose-reduction strategies were considered. Only 1 method, automated exposure control, was not associated with reduction in DLP. The electrocardiographically controlled tube current modulation (ECTCM), which decreases dose by adjusting the rate of delivery of x-rays as a function of the phase of the cardiac cycle, was used in 73% of studies and associated in multivariate analysis with a dose reduction of 25%. An alternative is sequential scanning, in which the CT scanner images the heart in discrete pieces and the x-ray beam is only on during a brief portion of the cardiac cycle for each piece, rather than being on continuously as in the conventional helical mode. Sequential scanning was used in 6% of studies and reduced dose by 78%. Low-voltage (100 kV) scanning, which reduces dose by decreasing the average energy of x-rays and can be used in conjunction with any of the previous methods, was used in 5% of studies, with a dose reduction of 46%.

Despite the greater dose reductions with sequential and low-voltage scanning, the strength of evidence supporting use of these methods is currently not as great as for ECTCM. The assessment of imaging tests has been described in terms of a hierarchical model of efficacy studies, beginning with image quality evaluations, then studies evaluating diagnostic accuracy efficacy, and progressing to higher levels in which the effect on patient management and outcomes are evaluated.¹⁶ For standard-voltage (120 kV) 64-slice helical coronary CTA, there are now 3 multicenter studies evaluating diagnostic accuracy efficacy in comparison with a gold standard diagnosis by invasive angiography,^{2 - 3 ,17} as well as 45 such single-center studies in the largest systematic review.¹⁸ Fifteen of these studies used ECTCM, including 1 of the multicenter studies.² For sequential scanning, several studies in the past year have observed image quality comparable with conventional helical scanning,¹⁹ and at least 3 single-center studies have demonstrated sensitivity and specificity comparable with helical CTA.^{20 - 22} For low-voltage coronary CTA, current evidence is limited to several studies evaluating image quality in selected patients.^{23 - 25} The implications for diagnostic accuracy efficacy of low-voltage CTA's increased image noise, reported previously by Hausleiter and colleagues,²⁴ remain to be demonstrated, as does the effect of body habitus on this efficacy.

An intriguing finding of PROTECTION I is the difference in DLP between scanner models. Median DLP for the highest-dose scanner was twice that for the lowest-dose scanner. This finding is sure to generate controversy; however, it should be interpreted with caution. Although there may be a significant difference between scanner models in the amount of radiation to which patients need to be exposed to obtain a given level of noise and image quality, due to its observational study design and limited binary assessment of image quality PROTECTION I cannot prove this and should thus be regarded as hypothesis-generating.

What implications should PROTECTION I have for patient care? First, the study results reinforce the observation that cardiac CTA is still a potentially high-dose procedure, and like all procedures involving the use of ionizing radiation, a patient-specific benefit-risk analysis should always be performed to justify the imaging study. Second, the findings suggest that dose-reduction methods can be used in the majority of patients, which should serve as a wake-up call to cardiac CT laboratories that do not routinely use these methods. Given the strength of evidence supporting it, ECTCM should be widely applied. The evidence for sequential scanning is rapidly accumulating, and it should also be given serious consideration for appropriate patients (ie, those with heart rates <65/min and regular cardiac rhythm). Low-voltage scanning should also be considered, perhaps especially for patients who are nonobese and at higher risk of radiation-associated cancer, such as children and young women,⁴ although advocating its routine usage awaits further data. Third, PROTECTION I reveals a degree of variability in radiation dose between sites that had not been previously appreciated, but which offers the potential to decrease radiation burden from cardiac CTA while maintaining diagnostic image quality by instituting quality improvement programs to close the gap. Fourth, the lack of clinically significant association between procedure volume and dose suggests that despite the general association between case volumes and quality of care,²⁶ even many high-volume centers can benefit from such quality improvement programs.

The International Basic Safety Standards recommend retaining and making available necessary information to allow retrospective radiation dose assessment,²⁷ and the 2008 multisocietal statement on key data elements for cardiac imaging includes DLP as a recommended data element for the description of cardiac CT studies.²⁸ The DLP should be recorded for each study and serve as the cornerstone of quality assurance efforts, with outliers investigated. A “diagnostic reference level” is a prespecified level for a test, commonly set at the 75th to 80th percentile based on survey data,²⁹ which if consistently exceeded calls for local review.³⁰ Hausleiter et al⁶ suggest a diagnostic reference level of 1200 mGy × cm for cardiac CTA. However, given the variability in DLP between scanners, it appears that use of a single diagnostic reference level may be problematic, as the median DLP for 1 scanner studied in PROTECTION I exceeded this level. Serious consideration should be given to routinely including DLP in the reports of cardiac CTA studies. This would facilitate quality assurance efforts and tracking of patients' cumulative radiation exposures, as well as serve as an additional motivation to technologists and physicians to control radiation dose. The European Union³¹ and Intersocietal Commission for the Accreditation of Computed Tomography Laboratories³² currently mandate quality assurance programs for CT including patient dose assessment. Other appropriate regulatory and accreditation bodies, working in conjunction with professional societies, might also consider whether this approach would be beneficial for their constituencies.

The international system of radiological protection stands on 3 principles: justification, optimization, and diagnostic reference levels.⁸ PROTECTION I provides valuable information pertaining to each of these in the context of cardiac CTA, and as such makes an important addition to the evidence base.