Multiple Testing, Cumulative Radiation Dose, and Clinical Indications in Patients Undergoing Myocardial Perfusion Imaging FREE

Utilization of medical imaging has grown rapidly in recent years.¹ Along with the benefits patients have received from medical imaging has come an increase in the burden of ionizing radiation associated with many such tests and the attendant potential risks of cancer. The National Council on Radiation Protection and Measurements has estimated that the per capita dose of medical radiation in the United States increased nearly 6-fold from the early 1980s to 2006.² This increased medical radiation burden has raised public health concerns, leading to a US Food and Drug Administration initiative to reduce unnecessary radiation exposure from medical imaging,³ with one of its focuses being nuclear imaging, and discussion in Congress of new legislation to regulate medical radiation.⁴

Although much attention has been paid to radiation from computed tomography (CT) scans,^5- 6 a recent study demonstrated that the single test with the highest radiation burden, accounting for 22% of cumulative effective dose from medical sources, is myocardial perfusion imaging (MPI).⁷ Volume of MPI increased from less than 3 million procedures in the United States in 1990 to 9.3 million in 2002,⁸ and it is now estimated to account for more than 10% of the entire cumulative effective dose to the US population from all sources, excluding radiotherapy.²

Estimates of the effective dose of radiation received by a patient undergoing a single standard MPI protocol have been previously reported, ranging from the equivalent of a few hundred to 2000 chest radiographs,^9- 11 although there are numerous sources of uncertainty in such estimates.¹¹ However, few data are available to characterize the total radiation burden received over an extended period by patients undergoing MPI, many of whom present with symptoms, such as chest pain or dyspnea, that would predispose them to receive multiple tests involving ionizing radiation. We analyzed procedures involving ionizing radiation received by a cohort of patients presenting for MPI to evaluate the total numbers of MPI examinations, other tests involving radiation, cumulative effective doses of radiation, and clinical indications for testing and repeat testing.

Our retrospective cohort study evaluated procedure counts, cumulative estimated effective doses of radiation, and clinical indications in a cohort of patients undergoing MPI within a single institutional system, Columbia University Medical Center/New York-Presbyterian Hospital (CUMC/NYPH), New York, New York, whose institutional review board approved the study and waived the requirement for informed consent.

All inpatients and outpatients undergoing single-photon emission computed tomography MPI at CUMC/NYPH during the first 100 days of 2006 (January 1-April 10) were included in the analysis. This MPI is identified herein as the index procedure for each patient; no patient had more than 1 index procedure.

Two electronic health record systems were manually queried by a single reader (A.B.) to identify all imaging and intervention procedures involving radiation at CUMC/NYPH and affiliated facilities that preceded the index examination, dating back to October 1988, and all subsequent tests through June 2008. This enabled identification of examinations performed over a nearly 2-decade period before the index examination, as well as providing more than 2 years for downstream testing resulting from the findings of the index MPI. Radiotherapy procedures were excluded. Procedures were classified using 328 procedure codes. Data were entered into a spreadsheet (Excel; Microsoft, Redmond, Washington) and checked for accuracy and edited by a second reader (A.J.E.).

Patient demographic data, including age at index procedure, sex, marital status, race/ethnicity, insurance coverage, and zip code, were obtained from the CUMC /NYPH Clinical Data Warehouse. Socioeconomic status was estimated by median annual household income in the patient's zip code, from 1999 US Census Bureau data.¹²

Effective dose was estimated for each procedure. For MPI and nuclear medicine tests, the radiopharmaceuticals used and corresponding administered activities (millicuries) were generally recorded; effective dose was estimated by multiplying administered activity by a radiopharmaceutical-specific conversion factor, as specified in International Commission on Radiological Protection Publications 80¹³ and 106,¹⁴ Radiation Dose Assessment Resource tables,¹⁵ and manufacturers' package inserts. In a few cases where activity was not recorded, a standard protocol was assumed. For cardiac fluoroscopic procedures such as invasive angiography and interventions, when available, recorded kerma-area product was multiplied by a standard conversion factor (0.2 mSv × mGy^-1 × cm^-2).¹⁶ For most other procedures, effective dose was estimated using a standard value for the procedure code. The values were determined for each code before the analysis by consensus of 2 investigators (A.J.E. and S.Balter) based on review of the literature relating to the procedure's dosimetry and standard imaging and intervention protocols at CUMC/NYPH. These values were in general agreement with estimates found in recent reviews,^9,17 although they are approximations that do not reflect secular trends in the dosimetry of individual procedures (eg, improvement in mammography equipment). Cumulative estimated effective dose was determined for each patient as the sum of the estimated effective doses for each procedure undergone by that patient.

A clinical cardiologist (S.D.W.) reviewed the electronic medical record of each patient, which contained the MPI report (including indications and history listed by the reporting physician), as well as in general, clinical, imaging, and procedural notes, and diagnostic codes. A single best reason for each test having been performed was identified.

Differences between groups were compared using Mann-Whitney, Kruskal-Wallis, or χ² tests, as appropriate. Correlation was measured using Spearman ρ. Two-tailed P < .05 was considered significant. A logistic regression model was developed for prediction of repeat MPI testing. Initially, all demographic variables with P < .20 on univariate analysis were included as independent variables in a preliminary model.¹⁸ Variables for which at least 1 level had P > .25 in the preliminary multivariate model were considered for removal, and likelihood ratio tests were performed to assess whether the variables significantly added to model fit, thereby warranting retention. Odds ratios (ORs) were estimated using the final logistic regression model and are reported as adjusted OR (95% confidence interval [CI]). Statistical analysis was performed by using Stata versions 10.1 and 11.1 (StataCorp, College Station, Texas).

The cohort included 1097 patients, including 565 women (51.5%). Mean age was 62.2 years (SD, 13.1; range, 11.6-96.8 years). A total of 424 patients (38.7%) were Hispanic, 314 (28.6%) were white, 228 (20.8%) were black, and 131 (11.9%) were other race (one of several other codes; eg, American Indian/Alaskan, Asian, Indian [India]). Mean (SD) income for zip code was $39.3 ($23.0) thousand (range, $14.3-$146.8 thousand). All but 78 patients (7.4%) had at least 1 health insurance plan.

Patients underwent a median of 15 (interquartile range [IQR], 6-32; mean, 23.9) procedures involving radiation exposure, of which 4 (IQR, 2-8; mean, 6.5) were high-dose procedures, defined as an effective dose of at least 3 mSv, the equivalent of 1 year's natural background radiation² (Table 1). These procedures included 1 (IQR, 1-2; mean, 1.8) MPI study per patient, of which 66% were dual-isotope, 28% were thallium, 4% were single-injection technetium Tc99m, 1% were dual-injection technetium, and less than 1% were positron emission tomographic perfusion studies. A total of 200 patients (18.2%) had at least 3 MPIs and 54 (4.9%) had at least 5 MPI examinations. Administered activity was available for 99% of MPI examinations. Previous procedures were identified for 996 patients (90.8%), dating back a median of 7.9 (IQR, 2.0-13.4; mean, 7.8) years before the index procedure.

Median cumulative estimated effective dose from MPI alone was 28.9 mSv (IQR, 27.9-55.6 mSv; mean [range], 44.6 [6.5-406.9] mSv). For all medical testing, median cumulative estimated effective dose was 64.0 mSv (IQR, 34.5-123.3 mSv; mean [range], 96.5 [6.5-917.9] mSv). A total of 71 patients (6.5%) received cumulative doses of more than 100 mSv due to MPI alone. A total of 344 patients (31.4%) received cumulative estimated effective dose from all medical sources of more than 100 mSv, including 120 patients (10.9%) who received cumulative dose of more than 200 mSv. The distributions of cumulative estimated effective doses for MPI and for all procedures are shown in the Figure.

Women underwent significantly more procedures involving exposure to ionizing radiation than men did, even excluding mammograms (Table 2). Nevertheless, cumulative effective dose was higher in men, reflecting primarily greater use of fluoroscopic procedures, including cardiac catheterization.

No significant differences in total number of procedures involving radiation were observed between black, Hispanic, and white patients; however, white patients underwent more MPI and fluoroscopy/catheterization procedures and correspondingly higher cumulative effective dose (Table 3). No strong correlation was observed between socioeconomic status (median income for zip code) and number of MPI examinations, number of ionizing radiation procedures, or cumulative effective dose (ρ<0.10 for each). Patients without health insurance underwent fewer tests involving radiation (median [IQR], 5.5 [3-19] vs 16 [7-34]; P < .001; mean, 13.1 vs 25.7) and correspondingly lower cumulative effective dose (median [IQR], 37.4 [28.8-76.8] vs 68.5 [36.7-131.8] mSv, P < .001; mean, 66.3 vs 101.2 mSv) than patients with any health insurance.

Reasons for testing are shown in Table 4. The primary reason for testing was chest pain, dyspnea, or both in 1296 examinations (66.8%). More than 80% of initial and 90% of repeat MPI examinations were performed in patients with known cardiac disease or symptoms consistent with it.

Of 1097 patients undergoing index MPI, 424 (38.6%) underwent additional MPI studies (Table 5). For this group of patients undergoing multiple MPIs, the median cumulative estimated effective dose was 121 mSv (IQR, 81-189; mean, 149 mSv). Median (IQR) time between consecutive MPI examinations was 23.7 (13.0-42.3) months. A total of 236 patients (56%) undergoing multiple MPI examinations had 2 examinations within 2 years of each other, and 117 (28%) had 2 MPI examinations within 1 year of each other. Repeat tests were more likely to demonstrate ischemia (36% vs 24%, P < .001) or scar (25% vs 14%, P < .001) than initial tests. Indications for repeat testing varied depending on interval between repeat examinations (Table 4). In multivariate analysis, patients undergoing multiple MPI examinations were significantly more likely to be older (OR, 1.31; 95% CI, 1.17-1.46 for 10 years increase in age) and insured (OR, 2.11; 95% CI, 1.15-3.87). There was a trend toward increased odds of undergoing multiple MPI examinations for male patients (OR, 1.29; 95% CI, 0.98-1.69) and patients of higher socioeconomic status (OR, 1.05; 95% CI, 0.99-1.12 for $10 000 increase in median household income). Race/ethnicity, religion, and marital status were not significantly associated.

Our study reveals very high cumulative estimated effective doses to many patients undergoing MPI, and especially to patients undergoing repeat MPI testing. More than 30% of patients received a cumulative estimated effective dose of more than 100 mSv, a level at which there is little controversy over the potential for increased cancer risks.¹⁹ The median cumulative estimated effective dose for the 39% of patients undergoing more than 1 MPI examination was 121 mSv, higher than that in the exposed (≥5 mSv) cohort in the Life Span Study of Japanese atomic bomb survivors.²⁰

Nevertheless, although effective dose reflects cancer risk from radiation, it is a population-averaged metric that does not account for individual characteristics such as age and health status. The population of patients undergoing MPI is fundamentally very different in several respects from both the Life Span Study cohort and, more importantly, the general US population. These differences favorably shift the balance of benefit vs risk of the ionizing radiation associated with MPI. Patients undergoing MPI are older than the general population, averaging 62 years herein, and based on primary indications for testing—more than 80% of initial MPIs and 90% of repeat MPI examinations were performed in patients with known cardiac disease or symptoms consistent with it—would be expected to have lower life expectancy than that predicted by age, noteworthy because solid tumors typically develop only after a lag period of at least 5 to 10 years following radiation exposure.²¹ The clear majority of MPI examinations were performed for reasons presently regarded as appropriate²² and with the potential to effect therapeutic management.

Although the high cumulative doses observed are certainly a matter of concern and an important target for improvement, these doses should not be viewed in isolation but rather within the clinical context where radiation risk for a specific patient is balanced against potential benefits. In particular, MPI plays a critical role in risk stratification of patients with established coronary artery disease.²³

Few studies have evaluated cumulative effective doses of radiation to patients undergoing multiple tests. In a series of 50 consecutive patients admitted to a cardiology service in Pisa, Italy, Bedetti et al²⁴ found median cumulative estimated effective dose from procedures performed during hospitalizations between 1970 and January 2006 of 61 mSv, similar to the median dose in our study. Two studies have examined cumulative dose from CT scans alone. One study observed patients who were receiving CT scans underwent a median of 3 scans during a 22-year period, with median cumulative effective dose of 24 mSv.²⁵ The second study, limited to patients with at least 3 emergency department visits in a year during which they received certain types of CT scans, found median cumulative estimated effective dose of 91 mSv during an 8-year period.²⁶ The largest study investigating cumulative estimated effective dose considered doses between 2005 and 2007 for nearly a million individuals covered by a single health care benefits organization. Median estimated effective dose was 0.1 mSv per enrollee per year, although 21% of individuals received cumulative estimated effective doses of more than 9 mSv over the 3 years, and 2% received more than 60 mSv.⁷ Nearly 10% of these patients underwent at least 1 cardiac imaging procedure using radiation. Among this subset, the mean cumulative estimated effective dose from cardiac procedures was 23.1 mSv, with 74% of this accounted for by MPI.²⁷ This lower cumulative dose noted in the health care benefits organization cohort vs in the CUMC/NYPH cohort can be accounted for by multiple factors, including the former's limitation to cardiac procedures, the 3- vs 20-year period during which procedures were observed, and the lower estimated mean dose for an MPI examination.

The findings of these studies, together with our findings, suggest that although most individuals receive little radiation from medical procedures, there exist certain groups of patients who receive high cumulative doses of radiation. Patients undergoing MPI, particularly those undergoing repeat MPI, are one such group. Efforts to reduce cumulative radiation dose should be especially targeted toward such groups.

Two cornerstones of radiological protection are justification (ensuring expected benefit exceeds harm for each exposure) and optimization (keeping the likelihood of incurring exposure and magnitude of individual exposures as low as reasonably achievable).²⁸ Even if an individual MPI study is justified, the cumulative radiation burden from all medical imaging and intervention procedures may not be optimized. Herein, the index MPI examination accounted for only 26% of cumulative radiation dose on average. While current appropriate use criteria provide detailed guidance for MPI utilization in terms of an individual test,²² they do not yet simultaneously consider the appropriateness of other modalities that involve no ionizing radiation exposure, and only superficially address longitudinal management strategies, which have clear implications for both radiation dose and health care costs.

For example, for patients with prior MPI presenting with new chest pain, current criteria view repeat MPI as “appropriate” if a prior MPI or angiogram was abnormal and “uncertain” if the prior test was normal. No distinction is made as to the number of prior tests performed, the interval between tests, the nature of any abnormal findings, and whether any were inconclusive. Several studies support existence of a “warranty period” for normal MPI, during which coronary events are unlikely to occur.^29- 31 Consistent with this, we observed normal myocardial perfusion for all 16 patients undergoing repeat MPI within 1 year for recurrent chest pain without prior evidence of ischemia or scarring. Future efforts need to focus not just on individual test justification but on optimizing and validating longitudinal imaging strategies to lower cumulative doses while ensuring performance of imaging needed for therapeutic decision making. Such strategies should consider tests without radiation, including stress echocardiography, stress magnetic resonance imaging, and exercise electrocardiography, and consider CT or invasive coronary angiography to “close the book” on cardiac sources of persistent, atypical symptoms.

Several approaches exist to decrease effective dose of an individual MPI examination. Two-thirds of MPI examinations in the 20-year period studied herein were performed using a dual-isotope protocol, with rest imaging using thallium chloride Tl 201 followed by stress with technetium Tc 99m. Dual-isotope imaging, while facilitating laboratory workflow, is associated with considerably higher radiation dose than other protocols,⁹ and our laboratory has subsequently moved away from its routine use, thereby decreasing the average effective dose of MPI, although avoidance of thallium chloride Tl 201 has been challenged by worldwide shortages of technetium Tc 99m.³² Use of protocols limited to technetium Tc 99m typically reduces radiation dose by more than 50% in comparison with dual-isotope imaging. This is especially the case for stress-only protocols, which often can be performed with low-administered activity.³³ Positron emission tomography MPI^9,34- 35 and low-dose protocols using newer camera³⁶ and image reconstruction³⁷ technologies offer the potential to decrease dose even further.

An interesting finding is the existence of several differences between groups. Men, white, and insured patients had higher odds of undergoing multiple MPI studies and received more MPI examinations, fluoroscopic procedures, and higher cumulative dose. These findings are consistent with previous studies demonstrating lower use of stress imaging tests, cardiac catheterization, and coronary revascularization in women and nonwhite patients.^38- 40 Whether this disparity in radiation doses represents an advantage or disadvantage is dependent on whether increase in utilization results in improved cardiovascular outcomes and requires further study.

One limitation of our study was that radiation dose records are limited to tests performed within a single hospital system. This restriction could potentially underestimate the problem of multiple testing by missing procedures and radiation doses received by patients at other facilities. Although CUMC/NYPH has a large pool of referring physicians, the findings of frequent repeat testing and high cumulative doses should be confirmed in regions of the country for which practice patterns differ.⁴¹ Another limitation was the use of median income from zip code as an area-based socioeconomic measure. Although zip-code level data can minimize incomplete matching during geocoding,⁴² in some settings it may be less accurate than block group– or census tract–level data as a surrogate for patient-level socioeconomic data.⁴³

In conclusion, in our single-center study, we observed multiple testing with MPI to be common and in many patients to be associated with very high cumulative estimated doses of radiation. Efforts are needed to decrease this high cumulative dose and its potential attendant risks.