CT Radiation Concerns Rise With Patients' Exposure
Download the PDF Here |
Download the PDF Here
By ALASTAIR GEE
Published: June 12, 2012
Even in health care systems in which doctors do not bill for each test they administer, the use of diagnostic imaging like CT and PET scans has soared, as has patients' radiation exposure, a new study has found.
The study, published online on Tuesday in The Journal of the American Medical Association, says that while advanced medical imaging has undoubted benefits, allowing problems to be diagnosed earlier and more accurately, its value needs to be weighed against potential harms, which include a small cancer risk from the radiation.
"The doses are not at a level that people should really be concerned," said Dr. Rebecca Smith-Bindman, the study's lead author and a radiologist and epidemiologist at the University of California, San Francisco. "It's rather that we need to minimize unnecessary exposures wherever possible."
Debate has grown louder over the role of advanced imaging, which many doctors say is overused. In April, a group of nine medical specialty boards recommended that doctors perform 45 common tests and procedures less often, with imaging prominent among them.
One board, the American Academy of Allergy, Asthma and Immunology, said a CT scan was not needed for cases of uncomplicated acute rhinosinusitis, or sinus infection, while another board, the American Society of Clinical Oncology, said doctors should cut back on CT and PET scans for early prostate and breast cancers that are unlikely to metastasize.
Harms that may result from radiation exposure during scans are also being examined. A recent study linked the use of CT scans in children to small but significant increases in the risk of leukemia and brain cancer.
Dr. Smith-Bindman's study looked at data on one million to two million patients a year from 1996 to 2010 in six health maintenance organizations across the United States, only some of whom had imaging. The number of CT scans tripled over the study period, to 149 per 1,000 patients in 2010, while the number of M.R.I.'s quadrupled, to 65 per 1,000 patients in 2010.
Financial incentives did not seem to drive the increase, the authors said, as doctors in H.M.O.'s do not charge for each service they provide. Rather, they said, the changes seemed to derive from improvements in scanning technologies that made them more widely applicable, along with the fact that patients more often requested the scans. Also, the authors said, some doctors practiced "defensive medicine," ordering tests to guard against malpractice lawsuits.
The radiation exposure of patients also jumped over time, driven by the increase in the number of scans ordered. CT and PET scans use ionizing radiation; M.R.I.'s do not.
"Getting one exam, one would not be particularly concerned - the risks are exceptionally tiny," said John D. Boice Jr., president of the National Council on Radiation Protection and Measurements. "Multiple exams are a different issue."
Some doctors said that although imaging is overused, aspects of the new study are to be welcomed.
"In many ways it's very good news," said William R. Hendee, a professor of radiology at the Medical College of Wisconsin. "A lot more patients are benefiting from advanced imaging procedures than were able to in the late 1990s."
An editorial accompanying the study suggests that doctors need to discuss the risks of radiation exposure with patients.
For now, it seems that imaging rates have stopped climbing, at least at H.M.O.'s. The study found that the number of CT scans and M.R.I.'s peaked in 2007.
JAMA. Editorial | June 13, 2012
Lung Cancer Screening, Radiation, Risks, Benefits, and Uncertainty
George T. O'Connor, MD, MS; Hiroto Hatabu, MD, PhD
Computed tomography (CT) scanning, which was introduced for imaging the head in 1972 and became widely available for imaging the rest of the body by the early 1980s, has revolutionized the practice of medicine and surgery. This technology, for which the Nobel Prize was awarded in 1979, has been used to diagnose and guide the management of diseases affecting every part of the body, improving quality of life and saving countless lives. Two articles in this issue of JAMA,1 - 2 however, point out the complexities involved in deciding whether to extend the use of CT scanning from diagnosis to screening and in determining whether the current use of CT scanning is appropriate or excessive.
As the value of CT scanning as a diagnostic tool became clear, it was natural to consider a potential role for this technology to screen for subclinical disease amenable to early intervention. The potential benefits of such screening must, of course, be weighed against the risks and costs. The risk that the ionizing radiation exposure from medical diagnostic tests will cause cancer appears to be small but not zero. Ionizing radiation causes DNA double-strand breaks that are repaired imperfectly, potentially leading to mutations and consequent cancers. An analysis of data from 15 countries has led to the estimate that from 0.6% to 3.2% of cancer diagnosed to age 75 years may be attributable to diagnostic x-rays, including CT scans, although these calculations involved assumptions subject to considerable uncertainty.3 Another risk of screening is the occurrence of false-positive findings that may lead to adverse psychological effects on patients as well as physical harm caused by diagnostic procedures undertaken to investigate the findings. Moreover, CT scans are expensive, as are the diagnostic procedures performed to evaluate abnormalities detected.
In this issue of JAMA, Bach and colleagues1 report the results of their systematic review of randomized clinical trials (RCTs) and cohort studies addressing the benefits and risks of screening for early-stage lung cancer using low-dose CT (LDCT) scans. The authors focus on lung cancer-specific and all-cause mortality outcomes in RCTs, avoiding the mistaken inferences that can result from lead-time bias, length-biased sampling, and overdiagnosis with other outcomes and designs.4 Their review yielded only 3 RCTs from which valid inferences can be drawn concerning the effect of LDCT screening for lung cancer among current or former smokers aged 50 years or older. Of these 3 studies, the National Cancer Institute's National Lung Screening Trial (NLST)5 was by far the largest and most persuasive, driving the authors' conclusion that lung cancer mortality is reduced by LDCT screening of adults meeting the NLST entry criteria: age 55 to 74 years, current or former smokers, 30 or more pack-years, and still smoking or having done so within the past 15 years.
The good news of a mortality benefit is tempered by some of the specifics. In the NLST, the number needed to screen to prevent 1 lung cancer death was 320 persons undergoing 3 annual LDCTs. Across all studies reviewed, the average rate of detecting nodules per round of screening was 20%, and more than 90% of these nodules turned out to be benign, leading to substantial follow-up testing including invasive procedures. Combining screening and follow-up diagnostic scans, the estimated mean 3-year radiation exposure of NLST participants in the screening group was 8 mSv, which Bach et al1 estimate would cause 1 cancer death per 2500 persons screened, although this death would likely occur many years later. The heterogeneity in nodule detection rate both among NLST sites and among the other studies reviewed by Bach et al,1 and the inconsistent mortality results of the 2 smaller RCTs, add a measure of uncertainty to the estimated benefit that would be obtained from broad application of LDCT screening. Nevertheless, the estimates of the benefits and risks of LDCT screening for lung cancer derived from the NLST are the best information currently available.
The American College of Chest Physicians, the American Society of Clinical Oncology, and the American Thoracic Society have endorsed an evidence-based practice guideline, included as an online appendix to the article by Bach et al,1 recommending that LDCT screening "should be offered" to persons meeting the NLST entry criteria, assuming this offer is made with counseling about risks and benefits and assuming the screening and follow-up are done at an institution with the resources for managing the findings of screening. The National Comprehensive Cancer Network has also recently issued a clinical practice guideline recommending LDCT screening for lung cancer in persons meeting NLST entry criteria as well as smokers older than 50 years with other lung cancer risk factors, including a history of chronic obstructive pulmonary disease or family history of lung cancer.6 These recommendations have been made with recognition that the cost-effectiveness of such screening has not been assessed and that the psychological effects on screened patients found to have a nodule7 are not well understood. Rigorous evaluation of these aspects of lung cancer screening-and the finding of a reasonable degree of cost-effectiveness-may be needed before the Centers for Medicare & Medicaid Services and other payers are willing to cover LDCT lung cancer screening.
Even without a new application of CT scans for lung cancer screening, the use of diagnostic CT and other advanced imaging modalities involving radiation exposure is frequent and increasing in the United States, as revealed by the report of Smith-Bindman and colleagues2 in this issue of JAMA. In their analysis of data from 6 large health maintenance organizations (HMOs), the use of CT scans increased from 52 per 1000 enrollees in 1996 to 149 per 1000 enrollees in 2010, an average annual increase of 7.8%, although the increase appears to have flattened after 2007. During this 15-year interval, there was an approximate doubling of the mean per capita radiation dose and of the percentages of enrollees who received a high or very high dose of radiation in a given year.
This report of HMO data and another recent report in a fee-for-service population8 both indicate that a nontrivial number of patients in the United States receive a high (20-50 mSv) or very high (>50 mSv) annual exposure to ionizing radiation from imaging studies in a given year. However, these data are not linked to clinical outcomes and do not reveal whether the radiation risks from these imaging studies are outweighed by the health benefits provided by the diagnostic information obtained. The data also cannot address how much of this testing is driven by defensive practice styles due to concerns about malpractice. They do, however, suggest that clinicians need to consider-and discuss with their patients-radiation risks when ordering diagnostic tests, possibly taking into account the cumulative radiation exposure a patient has received in recent months or years. Furthermore, the radiation risks and financial costs of advanced diagnostic imaging clearly warrant more research, including studies using informatics infrastructures such as that used by Smith-Bindman et al,2 to enhance decision support to guide the use of these technologies.
It is encouraging that advancing CT technology has permitted the reduction of ionizing radiation exposure,9 - 10 and in the near future, it may be possible to further decrease radiation exposure by an order of magnitude by combining modern scintillation materials for x-ray detectors, iterative physical model-based reconstruction algorithms, and more personalized image-acquisition protocols. Diagnostic modalities without radiation exposure, such as magnetic resonance imaging and ultrasonography, may be able to be substituted for some CT scans. For lung cancer screening, more selective patient targeting on the basis of genotype,11 gene expression profile,12 or plasma biomarkers13 may in the future reduce the number needed to screen and thereby reduce risk relative to benefit.
One of the authors of this Editorial recently had an office visit with a patient in her late 50s regarding obstructive lung disease. She reported difficulty quitting smoking in part due to stress related to her sibling's recent diagnosis of lung cancer, and she asked whether there was a test available to see whether she might have lung cancer herself. After a brief discussion of some of the major findings of the NLST-including the likelihood of discovering 1 or more small nodules that would need to be followed up over time, perhaps adding to her anxiety-the patient and physician together decided to pursue an LDCT scan. This seems like a reasonable decision based on available information in 2012, but it is important to recognize, as do Bach et al1 in the final sentence of their abstract, that "uncertainty exists."
Benefits and Harms of CT Screening for Lung CancerA Systematic Review
Peter B. Bach, MD, MAPP; Joshua N. Mirkin, BA; Thomas K. Oliver, BA; Christopher G. Azzoli, MD; Donald A. Berry, PhD; Otis W. Brawley, MD; Tim Byers, MD, MPH; Graham A. Colditz, MD, DrPH; Michael K. Gould, MD, MS; James R. Jett, MD; Anita L. Sabichi, MD; Rebecca Smith-Bindman, MD; Douglas E. Wood, MD; Amir Qaseem, MD, PhD, MHA; Frank C. Detterbeck, MD
Context Lung cancer is the leading cause of cancer death. Most patients are diagnosed with advanced disease, resulting in a very low 5-year survival. Screening may reduce the risk of death from lung cancer.
Objective To conduct a systematic review of the evidence regarding the benefits and harms of lung cancer screening using low-dose computed tomography (LDCT). A multisociety collaborative initiative (involving the American Cancer Society, American College of Chest Physicians, American Society of Clinical Oncology, and National Comprehensive Cancer Network) was undertaken to create the foundation for development of an evidence-based clinical guideline.
Data Sources MEDLINE (Ovid: January 1996 to April 2012), EMBASE (Ovid: January 1996 to April 2012), and the Cochrane Library (April 2012).
Study Selection Of 591 citations identified and reviewed, 8 randomized trials and 13 cohort studies of LDCT screening met criteria for inclusion. Primary outcomes were lung cancer mortality and all-cause mortality, and secondary outcomes included nodule detection, invasive procedures, follow-up tests, and smoking cessation.
Data Extraction Critical appraisal using predefined criteria was conducted on individual studies and the overall body of evidence. Differences in data extracted by reviewers were adjudicated by consensus.
Results Three randomized studies provided evidence on the effect of LDCT screening on lung cancer mortality, of which the National Lung Screening Trial was the most informative, demonstrating that among 53 454 participants enrolled, screening resulted in significantly fewer lung cancer deaths (356 vs 443 deaths; lung cancer-specific mortality, 274 vs 309 events per 100 000 person-years for LDCT and control groups, respectively; relative risk, 0.80; 95% CI, 0.73-0.93; absolute risk reduction, 0.33%; P = .004). The other 2 smaller studies showed no such benefit. In terms of potential harms of LDCT screening, across all trials and cohorts, approximately 20% of individuals in each round of screening had positive results requiring some degree of follow-up, while approximately 1% had lung cancer. There was marked heterogeneity in this finding and in the frequency of follow-up investigations, biopsies, and percentage of surgical procedures performed in patients with benign lesions. Major complications in those with benign conditions were rare.
Conclusion Low-dose computed tomography screening may benefit individuals at an increased risk for lung cancer, but uncertainty exists about the potential harms of screening and the generalizability of results.
Lung cancer is the leading cause of cancer death in the United States (and worldwide), causing as many deaths as the next 4 most deadly cancers combined (breast, prostate, colon, and pancreas).1 Despite a slight decline in US lung cancer mortality rates since 1990, lung cancer will account for more than 160 000 deaths in the United States in 2012.2 Most patients diagnosed with lung cancer today already have advanced disease (40% are stage IV, 30% are stage III), and the current 5-year survival rate is only 16%.3
Earlier randomized controlled trials (RCTs) involving chest radiographs and sputum cytology for lung cancer screening found that these strategies detected slightly more lung cancers, smaller tumors, and more stage I tumors, but the detection of a larger number of early-stage cancers was not accompanied by a reduction in the number of advanced lung cancers or a reduction in lung cancer deaths.4 - 14 Renewed enthusiasm for lung screening arose with the advent of low-dose computed tomography (LDCT) imaging, which is able to identify smaller nodules than can chest radiographs.
This systematic review focuses on the evidence regarding the benefits and harms of LDCT screening for lung cancer. Other potential screening methods (eg, chest radiographs, sputum cytology or biomarkers, exhaled breath) are not addressed. This review is a collaborative initiative of the American Cancer Society (ACS), the American College of Chest Physicians (ACCP), the American Society of Clinical Oncology (ASCO), and the National Comprehensive Cancer Network (NCCN) and forms the basis for the clinical practice guideline of the ACCP and ASCO. This work will be reassessed when pertinent new data become available, consistent with the Institute of Medicine recommendations for guideline development.15
This report summarizes the systematic review conducted by a multisociety collaborative effort examining the risks and benefits of LDCT screening for lung cancer and forms the basis of the American College of Chest Physicians and the American Society of Clinical Oncology clinical practice guideline ( Article and eAppendix 4). The guideline is based on the finding that a reasonable amount of data has been reported regarding the outcomes for LDCT screening for lung cancer and that some conclusions can be drawn regarding its risks and benefits despite many areas of uncertainty.
Box. Role of Computed Tomography Screening for Lung Cancer: Recommendations From the American College of Chest Physicians and the American Society of Clinical Oncology
For smokers and former smokers aged 55 to 74 years who have smoked for 30 pack-years or more and either continue to smoke or have quit within the past 15 years, we suggest that annual screening with low-dose computed tomography (LDCT) should be offered over both annual screening with chest radiograph or no screening, but only in settings that can deliver the comprehensive care provided to National Lung Screening Trial (NLST) participants. (Grade of recommendation: 2B.)
Counseling should include a complete description of potential benefits and harms (as outlined in the full guideline text) so the individual can decide whether to undergo LDCT screening.
Screening should be conducted in a center similar to those where the NLST was conducted, with multidisciplinary coordinated care and a comprehensive process for screening, image interpretation, management of findings, and evaluation and treatment of potential cancers.
A number of important questions about screening could be addressed if individuals who are screened for lung cancer are entered into a registry that captures data on follow-up testing, radiation exposure, patient experience, and smoking behavior.
Quality metrics should be developed such as those in use for mammography screening, which could help enhance the benefits and minimize the harms for individuals who undergo screening.
Screening for lung cancer is not a substitute for stopping smoking. The most important thing patients can do to prevent lung cancer is not smoke.
The most effective duration or frequency of screening is not known.
For individuals who have accumulated fewer than 30 pack-years of smoking or are either younger than 55 years or older than 74 years, or individuals who quit smoking more than 15 years ago, and for individuals with severe comorbidities that would preclude potentially curative treatment, limit life expectancy, or both, we suggest that CT screening should not be performed. (Grade of recommendation: 2C.)
Full text of the American College of Chest Physicians and the American Society of Clinical Oncology evidence-based practice guideline on the role of CT screening for lung cancer is available in eAppendix 4. This guideline has been endorsed by the American Thoracic Society.
A recent large, high-quality RCT (the NLST) found that annual LDCT screening reduced the relative risk of death from lung cancer by 20% and the absolute risk by 0.33% in a population with a substantially elevated risk for lung cancer. Two smaller RCTs (DANTE and DLCST) comparing LDCT with usual care found no benefit of LDCT screening but are best interpreted as neither confirming nor contradicting the NLST findings. Because a recent large RCT (N = 154 901) demonstrated no lung cancer mortality difference between chest radiographs screening and usual care, the interventions in these 3 studies are reasonably comparable.65
The literature supports the conclusion that LDCT screening can lead to harm. It identifies a relatively high percentage of subjects with nodules (average ~20%), the vast majority of which are benign. The additional imaging that these nodules trigger increases radiation exposure. The rates of surgical biopsy are variable (<1%-4%) as are the percentages of surgical procedures performed for benign disease. The rate of major, and sometimes fatal, complications among persons with benign lesions is low.
The unexplained heterogeneous rates of nodule detection, additional imaging, and invasive procedures that occurred within the structured settings of the controlled trials of LDCT raise concerns about how easily LDCT can be more broadly implemented. There is already substantial variability in the United States in the rates and complications of pulmonary needle biopsy66 and outcomes of lung cancer surgery, which are considerably better in dedicated centers (such as those conducting LDCT trials).67 - 68 Furthermore, adherence with screening is consistently lower in cohort studies than in the NLST and could be worse with unstructured implementation, with resulting diminished benefits. Analogous concerns in breast cancer screening led to the Mammography Quality Standards Act.69 The position statement by the International Association for the Study of Lung Cancer recommends demonstration projects to evaluate implementation of LDCT screening, establishment of quality metrics, and multiple task forces to address the many critical areas of uncertainty.70 Given all of these issues, performing an LDCT scan outside of a structured organized process appears to be beyond the current evidence base for LDCT lung cancer screening.
The fear and anxiety that patients can experience once there is even a slight suspicion of lung cancer highlights the need for careful education of LDCT participants and the need for carefully worded scan interpretations. In addition, even a small negative effect of screening on smoking behavior (either lower quit rates or higher recidivism) could easily offset the potential gains in a population.71 Smoking cessation should be considered a valuable component of any screening program.
In the setting of increasing health care costs, the relative cost-effectiveness of LDCT screening compared with other interventions will be a topic of discussion and concern in policy spheres. Medicare is allowed to contemplate a preventive service's cost-effectiveness before adding it to the package of preventive benefits (Medicare Improvements for Patients and Providers Act of 200872 ). Now that an estimate is available of effectiveness, an estimate of cost-effectiveness could be generated.
Some critical elements of such an analysis will include determining what the price of the component services will be and how frequently follow-up procedures will be required.
The ACCP and ASCO recommendations for LDCT screening should be interpreted in light of several limitations. We did not conduct a formal cost-effectiveness evaluation. LDCT would be expected to be less cost-effective when applied to individuals at lower risk of lung cancer, because more individuals will need to be screened to prevent each death from the disease. Making screening available in settings without an organized approach to the evaluation and management of LDCT findings may also lower cost-effectiveness, if the frequency of unnecessary interventions and procedures is higher in these settings.73 - 75
Second, the data on which to base these recommendations are relatively limited. Although LDCT screening appears promising, it is also a clinical intervention in its infancy. Many questions that clinicians and patients might reasonably ask when considering whether or not to pursue screening remain unanswered. How large are the risks from radiation? How does an individual's smoking history interplay with the size of the expected benefit or the risk of harms? How serious a problem is overdiagnosis?
Third, it is not clear whether individuals across the range of risk in the NLST are all sufficiently likely to benefit that all such individuals should be considered for screening, or alternatively if a narrower or broader group should be targeted to achieve an appropriate balance between benefit, costs, and harms. How often to perform screening, or over what period of time it should continue, are also important questions that have not been answered by the available data. It is possible to speculate that benefits of screening could be enhanced if screening were continued for longer periods, but the risks could be amplified as well.
A substantial amount of data on LDCT screening should be reported in the near future, including numerous planned analyses of the NLST data both by its investigators and by the Cancer Intervention and Surveillance Modeling Network (CISNET) investigators. The ongoing RCTs in Europe will also be reporting estimates of both the magnitude of LDCT's mortality benefit and the extent of its harms soon. These data may help inform some of the important questions that still linger regarding LDCT screening.
Screening a population of individuals at a substantially elevated risk of lung cancer most likely could be performed in a manner such that the benefits that accrue to a few individuals outweigh the harms that many will experience. However, there are substantial uncertainties regarding how to translate that conclusion into clinical practice.
Eight RCTs (Table 1)16 ,18 - 27 and 13 cohort studies of LDCT screening (Table 2)28 - 47 were selected from 591 citations identified by the literature search (eAppendix 3). Two smaller RCTs are related to 2 larger RCTs: the Lung Screening Study (LSS) was a pilot study of the National Lung Screening Trial (NLST) and there is a prespecified plan to combine data from the Danish Lung Cancer Screening Trial (DLCST) with the Dutch Belgian Randomised Lung Cancer Screening Trial (NELSON). Several trials were ongoing with only preliminary data available. Two RCTs were excluded because they lacked data on key end points; 1 RCT and several cohort studies were excluded because they involved populations at risk because of factors other than smoking or were for general population screening. For studies reported in multiple publications, all reports were reviewed, but earlier reports superseded by additional data in later reports are not referenced.
The NLST and DLCST had a low risk of bias (eTable 1). Other studies had variable risks of bias, in part because only preliminary reports of ongoing studies are available. The risk of bias in the cohort studies was variable and often high (usually because they lacked justification for the sample size, a definition of the primary end point, or description of funding sources).
Across the RCTs, the minimum smoking history required for enrollment ranged from 15 to 30 pack-years (ie, cigarette packs smoked per day multiplied by years of smoking), with a maximum time since quitting smoking ranging from 10 years to an unlimited number of years (Table 1). The lower age limit ranged from 47 to 60 years, and the upper limit from 69 to 80 years. There was greater variation in entry criteria in the cohort studies (Table 2).
The underlying risk for lung cancer varied substantially between the studies.The NLST,23 LSS,25 and study by Garg et al16 generally focused on higher risk; DLCST,19 ITALUNG,21 and DANTE22 on both higher and intermediate risk; and NELSON18 and Dˇpiscan27 on a broad range of risk among participants.
Although estimating the average risk of all participants in any of these studies is difficult because of a lack of granular data, the minimum risk level in each study was approximated using established formulas.48 - 49 Over 10 years, the risks of being diagnosed with lung cancer for participants meeting minimum entry criteria of each study, assuming they had quit smoking at time of study entry, were approximately 2% for NLST, 1% for DLSCT, and considerably less than 1% for NELSON. The nodule size deemed large enough to investigate further ranged from "any size" to greater than 5 mm; the size that triggered an invasive intervention (when specified) ranged from 6 to 15 mm.
Potential Benefits of LDCT Screening
Effect on Mortality. Three RCTs reported the effect of LDCT screening on lung cancer-specific mortality (Table 3). The NLST found that 3 annual rounds of screening (baseline and 1 and 2 years later) with LDCT resulted in a 20% relative decrease in deaths from lung cancer vs chest radiographs over a median of 6.5 years of follow-up (P = .004).23 In absolute terms, the chance of dying from lung cancer was 0.33% less over the study period in the LDCT group (87 avoided deaths over 26 722 screened participants), meaning 310 individuals must participate in screening for typically 3 rounds to prevent 1 lung cancer death. Based on a slightly different denominator, the NLST authors reported the number needed to screen with LDCT was 320 to prevent 1 lung cancer death.
The considerably smaller ongoing DANTE and DLCST studies each compared 5 annual rounds of LDCT screening to usual care; after a median of 34 and 58 months of follow-up, respectively, no statistically significant difference in lung cancer mortality was observed in either study (DANTE: relative risk [RR], 0.97; 95% CI, 0.71-1.32; P = .84; DLCST: RR, 1.15; 95% CI, 0.83-1.61; P = .43).19 ,22
All 3 studies reported on the risk of death from any cause (Table 4) between study groups and directly or indirectly on the risk of death from any cause other than lung cancer. Only the NLST found a difference in this end point, with fewer deaths overall in the LDCT vs the chest radiograph group (1303 vs 1395 deaths per 100 000 person-years, respectively). Analyses focusing exclusively on deaths not due to lung cancer found no significant differences in any of the 3 studies.23
Effect on Smoking Behavior. Speculation exists that undergoing LDCT screening may result in justification of continued smoking or, alternatively, may represent an opportunity for smoking cessation interventions. But studies examining the smoking behavior of LDCT-screened individuals have not found evidence that cessation or reinitiation rates are meaningfully altered by participation in screening.
Potential Harms of LDCT Screening
Detection of Abnormalities. Low-dose computed tomography identifies both cancerous and benign noncalcified nodules; the latter are often called "false positives." Although most LDCT screening studies have reported on nodules detected, the categorization and manner of reporting are inconsistent (eg, it is sometimes unclear if newly identified nodules are assigned to that round or to an earlier round if they can be retrospectively seen on an earlier LDCT). Likewise, size thresholds that would trigger an invasive workup are variously and inconsistently reported, as are the potential denominators for estimating false-positive rates, such as per screening round or per person-year.
Based on the study's own size cutoffs, the average nodule detection rate per round of screening was 20% (Table 5, eFigure 1) but varied from 3% to 30% in RCTs and 5% to 51% in cohort studies. Most studies reported that more than 90% of nodules were benign. In general, there is a tendency toward lower nodule detection rates in repeat screening rounds, but the data and reporting are inconsistent (Table 5, eFigure 2). In the NLST, the rate of detection did not decrease until the third round. In that round, the study protocol allowed for ignoring nodules that had been present in the prior rounds. We were unable to find any statistically significant relation between study parameters, such as smoking history of study enrollees, CT scan settings, nodule size cutoffs, and reported nodule detection rates.
Most often, a detected nodule triggered further imaging, but the underlying management protocols were inconsistently reported in the studies. Whether all additional imaging tests were captured in the studies was also uncertain: reported follow-up imaging rates may be underestimated.
The frequency of further CT imaging among screened individuals ranged from 1% in the study by Veronesi et al27 to 44.6% in the study by Sobue et al.32 The frequency of further positron emission tomography (PET) imaging among screened individuals exhibited much less variation, ranging from 2.5% in the study by Bastarrika et al39 to 5.5% in the NLST.23 The frequency of invasive evaluation of detected nodules was generally low but varied considerably (Table 6, eFigure 3). No patterns were apparent that explained this heterogeneity. In the NLST, 1.2% of patients who were not found to have lung cancer underwent an invasive procedure such as needle biopsy or bronchoscopy, while 0.7% of patients who were not found to have lung cancer had a thoracoscopy, mediastinoscopy, or thoracotomy.23 In the NELSON study, these numbers were 1.2% and 0.6%, respectively.18 Invasive nonsurgical procedures in patients with benign lesions were common (eg, 73% in NLST).
Complications of Diagnostic Procedures. The only study reporting on complications resulting from LDCT screening is the NLST. Overall, the frequency of death occurring within 2 months of a diagnostic evaluation of a detected finding was 8 per 10 000 individuals screened by LDCT and 5 per 10 000 individuals screened by chest radiographs. Some of the deaths that occurred after a diagnostic evaluation were presumably unrelated to follow-up procedures, as 1.9 and 1.5 per 10 000 occurred within 2 months when the diagnostic evaluation involved only an imaging study.
Deaths most clearly related to follow-up procedures were those occurring within 2 months when the most recent procedure was a bronchoscopy or needle biopsy (3.4 per 10 000 screened by LDCT and 2.2 per 10 000 screened by chest radiographs). Approximately one-third of the deaths occurred within 2 months of a surgical procedure in both study groups, and the vast majority of these were in the patients with cancer, suggesting perhaps that the surgical procedures in those with cancer were more extensive (ie, resection rather than biopsy; such details were not reported). The 60-day perioperative mortality for patients with lung cancer who underwent a surgical procedure was 1% for the LDCT group and 0.2% for the chest radiographs group.
Overall, the frequency of major complications occurring during a diagnostic evaluation of a detected finding was 33 per 10 000 individuals screened by LDCT and 10 per 10 000 individuals screened by chest radiographs. The rate of (presumably unrelated) complications following imaging alone was similar and low (1.1 and 1.5 per 10 000 screened, respectively); the complication rate after a bronchoscopy or needle biopsy was also low (1.5 and 0.7 per 10 000 for LDCT and chest radiographs, respectively). The vast majority of major complications occurred after surgical procedures and in those patients with lung cancer. The rate of major complications in those patients with lung cancer who underwent surgery was 14%.
Focusing only on patients who had nodules detected by LDCT that were determined to be benign, death occurred within 60 days among 0.06% and major complications occurred among 0.36%. Approximately half of the deaths occurred after imaging alone, whereas the majority of major complications occurred after a surgical procedure (details unknown). Calculating these numbers for an entire screened population, the risks of death and major complications following diagnostic events (including imaging) for nodules that were determined to be benign is 4.1 and 4.5 per 10 000, respectively. These rates are higher than those in the chest radiographs group (1.1 and 1.5 per 10 000 for risks of death and major complications, respectively).
Overdiagnosis. Overdiagnosis refers to histologically confirmed lung cancers identified through screening that would not affect the patient's lifetime if left untreated. This includes patients who are destined to die of another cause (eg, a comorbidity or an unexpected event).53 Earlier studies suggested that chest radiographs screening may have an overdiagnosis rate of roughly 25%.54 - 55 The overdiagnosis rate for LDCT screening cannot yet be estimated; NLST data show a persistent gap of about 120 excess lung cancers in the LDCT group vs the chest radiographs group, but further follow-up is needed.
Radiation Exposure. The effective dose of radiation of LDCT is estimated to be 1.5 mSv per examination, but there is substantial variation in actual clinical practice. However, diagnostic chest CT (~8 mSv)56 or PET CT (~14 mSv)56 - 58 to further investigate detected lesions rapidly increases the exposure and accounts for most of the radiation exposure in screening studies. We estimate that NLST participants received approximately 8 mSv per participant over 3 years, including both screening and diagnostic examinations (averaged over the entire screened population). Estimates of harms from such radiation come from several official bodies and commissioned studies,59 - 60 based on dose extrapolations from atomic bombings and also many studies of medical imaging.61 - 62 Using the NLST data, these models predict that approximately 1 cancer death may be caused by radiation from imaging per 2500 persons screened.
Therefore, the benefit in preventing lung cancer deaths in NLST is greater than the radiation risk-which only becomes manifest 10 to 20 years later. However, for younger individuals or those with lower risk of developing lung cancer, the trade-off would be less favorable. Preliminary modeling studies suggest that potential risks may vastly outweigh benefits in nonsmokers or those aged 42 years or younger.63 Further study is needed, including evaluation of the effects of ongoing annual LDCT beyond 3 successive years.
Quality of Life. The effect of LDCT screening on quality of life is uncertain. We found only 1 study, in which 88% to 99% of 351 participants reported no discomfort, but 46% reported psychological distress while awaiting results.64 Although there may be quality-of-life benefits due to lower morbidity from advanced lung cancer, there are also potential detriments due to anxiety, costs, and harms from the evaluation of both false-positive scans and overdiagnosed cancers.
Patients Likely to Benefit
The NLST population is the only one for whom a lung cancer mortality benefit from LDCT has been demonstrated (those aged 55-74 years with ³30 pack-years of smoking who quit ²15 years prior to entry [if they have stopped smoking]). Other studies are too small, too preliminary, or too poorly designed to support meaningful conclusions. The value of LDCT screening is likely determined primarily by the risk of lung cancer vs competing causes of death. Little information exists regarding comorbidities, but presumably the NLST participants were deemed healthy. We estimate an average risk of developing lung cancer within 10 years of approximately 10% for the NLST population in the absence of screening (estimated median age 62 years and ~50 pack-years of smoking). However, the calculable risk for individual NLST participants most likely varied by more than 10-fold across the participants, from less than 2% to greater than 20%, and it is unclear which groups experienced benefit.48 - 49 But there is no evidence base for determining how selection criteria for screening should be refined. Incorporating other well-known risk factors has not been studied.
Effective Screening Setting
A summary of the setting of the NLST (the only positive study) demonstrates that most (76%) of the NLST sites were National Cancer Institute-designated cancer centers, and 82% were large academic medical centers with more than 400 hospital beds, although screening may have taken place at satellite facilities in some cases (eTable 3). All of these centers are likely to have specialized thoracic radiologists and board-certified thoracic surgeons on staff. The CT scanners used in the NLST underwent ongoing extensive quality control, and the scans were interpreted by chest radiologists who underwent specific training and quality control in the interpretation of images and wording of screening LDCT findings.57 A nodule management algorithm was included in the NLST, but adherence or the setting in which nodules were managed was not mandated or tracked by the study.57
Most other RCTs and cohort studies of LDCT screening were conducted in facilities similar to the NLST sites: academic medical centers, large hospitals, with the involvement of relevant subspecialist services and a defined nodule management algorithm. The association between the setting of LDCT screening and outcome has not been tested, but variability in rates of false-positive LDCT scans, further imaging, and procedures suggests these may be important.
June 13, 2012
Use of Diagnostic Imaging Studies and Associated Radiation Exposure for Patients Enrolled in Large Integrated Health Care Systems, 1996-2010
Rebecca Smith-Bindman, MD; Diana L. Miglioretti, PhD; Eric Johnson, MS; Choonsik Lee, PhD; Heather Spencer Feigelson, PhD, MPH ; Michael Flynn, PhD; Robert T. Greenlee, PhD, MPH; Randell L. Kruger, PhD; Mark C. Hornbrook, PhD; Douglas Roblin, PhD; Leif I. Solberg, MD; Nicholas Vanneman, MA; Sheila Weinmann, PhD; Andrew E. Williams, PhD
Context Use of diagnostic imaging has increased significantly within fee-for-service models of care. Little is known about patterns of imaging among members of integrated health care systems.
Objective To estimate trends in imaging utilization and associated radiation exposure among members of integrated health care systems.
Design, Setting, and Participants Retrospective analysis of electronic records of members of 6 large integrated health systems from different regions of the United States. Review of medical records allowed direct estimation of radiation exposure from selected tests. Between 1 million and 2 million member-patients were included each year from 1996 to 2010.
Main Outcome Measure Advanced diagnostic imaging rates and cumulative annual radiation exposure from medical imaging.
Results During the 15-year study period, enrollees underwent a total of 30.9 million imaging examinations (25.8 million person-years), reflecting 1.18 tests (95% CI, 1.17-1.19) per person per year, of which 35% were for advanced diagnostic imaging (computed tomography [CT], magnetic resonance imaging [MRI], nuclear medicine, and ultrasound). Use of advanced diagnostic imaging increased from 1996 to 2010; CT examinations increased from 52 per 1000 enrollees in 1996 to 149 per 1000 in 2010, 7.8% annual increase (95% CI, 5.8%-9.8%); MRI use increased from 17 to 65 per 1000 enrollees, 10% annual growth (95% CI, 3.3%-16.5%); and ultrasound rates increased from 134 to 230 per 1000 enrollees, 3.9% annual growth (95% CI, 3.0%-4.9%). Although nuclear medicine use decreased from 32 to 21 per 1000 enrollees, 3% annual decline (95% CI, 7.7% decline to 1.3% increase), PET imaging rates increased after 2004 from 0.24 to 3.6 per 1000 enrollees, 57% annual growth. Although imaging use increased within all health systems, the adoption of different modalities for anatomic area assessment varied. Increased use of CT between 1996 and 2010 resulted in increased radiation exposure for enrollees, with a doubling in the mean per capita effective dose (1.2 mSv vs 2.3 mSv) and the proportion of enrollees who received high (>20-50 mSv) exposure (1.2% vs 2.5%) and very high (>50 mSv) annual radiation exposure (0.6% vs 1.4%). By 2010, 6.8% of enrollees who underwent imaging received high annual radiation exposure (>20-50 mSv) and 3.9% received very high annual exposure (>50 mSv).
Conclusion Within integrated health care systems, there was a large increase in the rate of advanced diagnostic imaging and associated radiation exposure between 1996 and 2010.
The use of diagnostic imaging in the Medicare population has increased significantly over the last 2 decades, particularly using expensive new technologies such as computed tomography (CT), magnetic resonance imaging (MRI), and nuclear medicine positron emission tomography (PET).1 - 2 The development and improvement in these advanced diagnostic imaging technologies is widely credited with leading to earlier and more accurate diagnoses of disease using noninvasive techniques. However, utilization and costs of advanced diagnostic imaging in the United States are high and rapidly growing,3 - 4 and payments to physicians for diagnostic imaging have had the highest rate of growth among all physician services over the last decade.3 - 4 Computed tomography and nuclear medicine examinations deliver much higher doses of ionizing radiation than conventional radiographs, and extensive epidemiological evidence has linked exposure to radiation levels in this range with the development of radiation-induced cancers.5 - 6 It is estimated that 2% of future cancers will result from current imaging use, if imaging continues at current rates.7 - 8
Most studies that have evaluated patterns of diagnostic imaging have assessed insurance claims for fee-for-service insured populations1 ,9 - 11 where financial incentives encourage imaging.12 - 13 No large, multisite studies have assessed imaging trends in integrated health care delivery systems that are clinically and fiscally accountable for the outcomes and health status of the population served.13 - 14 Understanding imaging utilization and associated radiation exposure in these settings could help us determine how much of the increase in imaging may be independent of direct financial incentives.
We conducted a population-based study of diagnostic imaging trends between 1996 and 2010 among members of 6 geographically diverse integrated health care delivery systems that have both care delivery and insurance relationships with their member-patients. The availability of administrative and electronic medical record data on all health care received-including diagnostic imaging-allowed us to assess patterns of imaging over time as they varied by health system and patient demographics.
We found the increase in imaging studies among 6 large integrated HMOs was substantial over the last 15 years, paralleling the increase reported among fee-for-service insured populations. For example, among the HMO enrollees 65 years and older, rates of imaging with CT increased an average of 10.2% annually between 1998 and 2005, and slowed to 4.2% annual growth from 2005 to 2008, similar to the respective 10.1% and 5.1% growth rates recently reported among Medicare fee-for-service beneficiaries during the same time periods.2 ,24 Use of MRI also increased rapidly among the HMO enrollees, with 14.5% and 6.5% average annual increase in these 2 periods-similar to that reported for Medicare fee-for-service beneficiaries (13.5% and 2.2%, respectively).2
The increase in imaging use over this period was likely driven by many factors, including improvements in the technology that have led to expansion of clinical applications, patient- and physician-generated demand, defensive medical practices,25 and medical uncertainty26 - 27 -all factors that would be expected to influence utilization across all systems of medical care. However, strategies that have been adopted by most private commercial payers to control imaging costs, such as use of radiology benefit management firms that require preapproval or prenotification28 - 29 and member copayments, have not been widely adopted within these settings. Only 2 of these health plans have recently adopted copayments for advanced imaging (one at $10 and one at $50). Although several plans have recently adopted decision support software, it is too early to assess whether greater adherence to appropriateness criteria included with the software products may influence utilization rates.
Although the increase in imaging studies was similar between HMO members and Medicare fee-for-service insured beneficiaries, the rates of imaging seem to be modestly lower among HMO enrollees. For example, in 2006, HMO enrollees 65 years and older underwent 474 CTs and 123 MRIs per 1000 enrollees, whereas Medicare fee-for-service enrolled adults underwent 550 CTs and 192 MRIs, 15% and 35% lower rates, respectively.2 While some of this difference may be due to underlying geographic variation in imaging rates (we included 6 HMOs where the cited numbers were based on a larger national sample)9 ,30 and due to possible differences in age and health status among HMO enrollees (who may be healthier than non-HMO enrollees), imaging rates do seem to be lower within the included HMO settings.
We found that imaging use increased steeply with age, particularly for CT and nuclear medicine examinations, resulting in high radiation exposures received by the oldest enrollees. Among enrollees 45 years and older who underwent imaging, nearly 20% received high or very high radiation exposure annually. Although cancer risks from radiation are often considered to decline with age, recent models suggest that cancer risks decline with age until middle age, when cancer risks may then increase in a U-shaped distribution.6 ,31 Thus, radiation-related cancer risks after exposure in middle and older ages may be higher than previously believed.26 Because the utilization of imaging is higher in older adults, and because the potential harm from these tests may also be higher in these patients, it is particularly important to quantify the benefits of imaging in these patients.
We found the per capita exposure to radiation from diagnostic imaging was 2.7 mSv in 2006, similar to the annual per capita exposure reported by the National Council on Radiation Protection (NCRP), describing the entire US population32 (3.0 mSv), and Fazel and colleagues,11 describing a fee-for-service insured population (2.4 mSv). A notable difference is that we found a significantly larger number of patients received very high radiation exposures annually. For example, Fazel et al reported that 0.2% of insured individuals incurred a very high (>50 mSv) annual radiation dose, in contrast to our finding that 1.4% of enrollees received such high exposures-a 7-fold increase. There are several possible reasons for this difference, including that we did not assume that every test delivered the same radiation exposure but rather modeled the actual distribution in dose when estimating the proportion of patients with high exposures. Our work adds to the work of the NCRP and Fazel et al by assessing doses limited to enrollees who underwent imaging, as it is only these individuals who are at risk of radiation-related carcinogenesis. By 2010, 10.8% of enrollees who underwent imaging received an annual exposure greater than 20 mSv. It is notable that even when limited to patients who were imaged during both time periods, the average dose per person nearly doubled, suggesting more intensive medical imaging among those who undergo any imaging.
Considering governmental limits on radiation exposure can provide context for these typical patient doses: 20 mSv is the annual allowable occupational exposure to radiation in Europe,33 - 34 and 50 mSv is the annual allowable occupational exposure in the United States.35 While it is not appropriate to set exposure limits when radiation is required for health benefit, the number of patients exposed to such levels highlights the need to consider this potential harm when ordering imaging tests and to track radiation exposures for individual patients so that this information is available to physicians who are ordering tests. The National Academy of Sciences' National Research Council concluded, after a comprehensive review of the published literature, that patients who receive radiation exposures in the same range as a single CT-10 mSv-may be at increased risk of developing cancer6 ,36 - 37 ; 16.5% of patients who underwent imaging in 2010 received a dose at least this high.
We did not assess costs for imaging within these integrated settings. Costs for imaging among fee-for-service insured elderly adults have declined since 2005, despite increasing utilization.10 ,28 As part of the Deficit Reduction Act of 2005, Congress enacted a provision to equalize the reimbursement rate for imaging examinations regardless of where they were performed; among fee-for-service Medicare enrollees, a 12.7% reduction in imaging costs followed enactment.28 Because of bundled payments for imaging within our integrated settings, these types of per-examination reductions in payment would not be expected to have had the same effect on utilization as they have in the fee-for-service environment.13 ,28
The HMO Research Network that we relied on provides a unique opportunity to conduct analyses of patterns of imaging because of the complete capture of health care utilization by their members, including all diagnostic testing, standardization of how these data are collected, and storage of detailed imaging records so that actual radiation exposures could be measured. However, there are several limitations of our work. We focused on individuals enrolled in comprehensive health care plans and excluded data from fee-for-service enrollees because of incompleteness of the available data. For inpatient imaging examinations, only the admission date was available to us; thus, collapsing claims could lead to undercounting of multiple examinations performed during the same hospital stay. Similarly, patients who underwent multiple examinations with the same procedure code on a single day were only counted once, and for these patients we have likely underestimated their exposures. We limited assessment to beneficiaries enrolled throughout a given year, and imaging may differ for patients who leave the HMO program.
To assess medical radiation dose, we used an estimate of the dose for each patient based on a sampling of high-dose studies, but we did not use actual dose information for each individual patient examination, as these data are not routinely stored in an easily accessible format. To account for underlying variations in dose, we included an estimate of the variation in dose when estimating the number of individuals above dose thresholds, and we believe our estimates to be conservative, as they do not allow for any within-person correlation in the random deviation. We only evaluated cumulative exposure within each year and not cumulative exposure over time. We did not study cumulative exposures because of the fluidity of enrollment over time. We used effective dose as the measure to summarize multiple exposures and this measure is imprecise, particularly when trying to sum across different anatomic areas that might be imaged. No alternative measure exists, and this measure will almost certainly capture those patients receiving high exposures.38 We only assessed radiation exposure as a potential harm of testing, but there are several other potential harms associated with imaging, such as false-positive test results that may begin a cascade of unnecessary testing, and over-diagnosis of otherwise indolent disease that leads to unnecessary treatment.
The increase in use of advanced diagnostic imaging has almost certainly contributed to both improved patient care processes and outcomes, but there are remarkably few data to quantify the benefits of imaging. Given the high costs of imaging39 - 41 -estimated at $100 billion annually-and the potential risks of cancer and other harms, these benefits should be quantified and evidence-based guidelines for using imaging should be developed that clearly balance benefits against financial costs and health risk.
Between 933 897 and 1 998 650 enrollees were included during each year of the study. The age distribution of health plan members roughly paralleled that of the states in which the members were enrolled, and 52.5% of enrollees were female (see the eTable). Enrollees underwent a total of 30.9 million imaging examinations during the 15-year study (25.8 million person-years), reflecting an average of 1.18 tests per person per year (95% CI, 1.17-1.19), of which 35% were advanced diagnostic imaging (ie, CT, MRI, nuclear medicine, and ultrasound).
The rates of imaging examinations per 1000 enrollees in 2008 by modality, site, and anatomic area are provided in Table 1. The total rate of imaging was 1420 examinations per 1000 enrollees per year-the most common being radiography (783/1000, 55.1% of all examinations), followed by ultrasound (271/1000, 19.1%), CT (177/1000, 12.5%), MRI (72/1000, 5.1%), angiography/fluoroscopy (64/1000, 4.5%), and nuclear medicine (53/1000, 3.7%).
Although the percentage of examinations for each modality was relatively similar across health systems (51.8%-57.6% of examinations involved radiography across sites and 10.8%-13.9% of examinations used CT across sites), there were significant differences across health systems in the rates of imaging for each examination type. For example, the utilization of MRI ranged from 55 to 88 examinations per 1000 enrollees (relative risk [RR] between highest and lowest site, 1.6; P < .001). Angiography/fluoroscopy utilization varied the most across sites (RR, 4.4; P < .001) and radiography had the least variation in use across sites (RR, 1.3; P < .001).
Utilization Over Time
Radiography and angiography/fluoroscopy rates were relatively stable over time: radiography increased 1.2% per year, and angiography/fluoroscopy decreased 1.3% per year. In contrast, the utilization of advanced diagnostic imaging changed markedly (Figure 1). Computed tomography examinations tripled (52/1000 enrollees in 1996 to 149/1000 in 2010, 7.8% annual growth; 95% CI, 5.8%-9.8%); MRIs quadrupled (17/1000 to 65/1000, 10% annual growth; 95% CI, 3.3%-16.5%); ultrasounds approximately doubled over the same period (134/1000 to 230/1000, 3.9% annual growth; 95% CI, 3.0%-4.9%). Nuclear medicine rates decreased (32/1000 to 21/1000, 3% annual decline; 95% CI, 7.7% decline to 1.3% increase), although after 2004, PET imaging rates increased from 0.24 per 1000 enrollees to 3.6 per 1000 enrollees, 57% annual growth.
The increase in advanced diagnostic imaging with both age and year is shown in Figure 2. Diagnostic imaging increased with age, and within each age group, advanced diagnostic imaging rates increased rapidly for many years and then flattened or minimally declined in the more recent years. For CT, growth in imaging tended to flatten around 2007. For MRI, rates peaked around 2007, with slight declines in subsequent years. For nuclear medicine, a marked reduction in imaging rates occurred from 2006 onward; however, PET imaging rates increased steadily through 2010.
Radiation Exposure and Changes Over Time
The increase in the utilization of CT resulted in an increase in enrollee exposure to radiation, with the mean per capita effective dose rising from 1.2 mSv in 1996 to 2.3 mSv in 2010. The percent of enrollees who received high (>20-50 mSv) or very high (>50 mSv) radiation exposure during a given year also approximately doubled across study years. By 2010, 2.5% of enrollees received a high annual dose of greater than 20 to 50 mSv, and 1.4% received a very high annual dose of greater than 50 mSv (Table 2).
The average effective dose to those individuals who were exposed to any radiation from medical imaging increased from approximately 4.8 mSv in 1996 to 7.8 in 2010 (3.2% annual growth; 95% CI, 3.1%-3.3%). By 2010, 6.8% of patients who underwent imaging received a high dose of more than 20 to 50 mSv and 3.9% of patients received a very high dose above 50 mSv during this single year. The distribution in dose over time is shown in Figure 3. There was a gradual increase in the radiation dose received by individuals in the top 1% and 10% of those exposed. By 2010, the highest 1% of exposed individuals received around 100 mSv, and the highest 10% of exposed individuals received around 20 mSv.
The increase in CT use accounted for the increase in the number of enrollees exposed to high (>20-50 mSv) and very high (>50 mSv) radiation exposures. In 1996, CT accounted for 5.7% of examinations and 30.3% of enrollees' exposure to ionizing radiation while contributing to a per capita exposure of 0.38 mSv per enrollee (Table 3). By 2010, CT accounted for 12.0% of examinations and 67.8% of radiation exposure and contributed 1.58 mSv per enrollee (a 4-fold increase in per capita radiation exposure from CT). Angiography/fluoroscopy accounted for a declining proportion of examinations (reduced from 7.4% to 4.6%) and a reduced contribution to absolute radiation exposure (reduced from 0.52 to 0.34 mSv per year), reflecting a reduction from 42.3% to 14.6% of enrollees' total radiation exposure.