|Home | About | Journals | Submit | Contact Us | Français|
The 600% increase in medical radiation exposure to the US population since 1980 has provided immense benefit, but potential future cancer risks to patients. Most of the increase is from diagnostic radiologic procedures. The objectives of this review are to summarize epidemiologic data on cancer risks associated with diagnostic procedures, describe how exposures from recent diagnostic procedures relate to radiation levels linked with cancer occurrence, and propose a framework of strategies to reduce radiation from diagnostic imaging in patients. We briefly review radiation dose definitions, mechanisms of radiation carcinogenesis, key epidemiologic studies of medical and other radiation sources and cancer risks, and dose trends from diagnostic procedures. We describe cancer risks from experimental studies, future projected risks from current imaging procedures, and the potential for higher risks in genetically susceptible populations. To reduce future projected cancers from diagnostic procedures, we advocate widespread use of evidence-based appropriateness criteria for decisions about imaging procedures, oversight of equipment to deliver reliably the minimum radiation required to attain clinical objectives, development of electronic lifetime records of imaging procedures for patients and their physicians, and commitment by medical training programs, professional societies, and radiation protection organizations to educate all stakeholders in reducing radiation from diagnostic procedures.
Since the discoveries of x-rays, radium, and radioactivity from uranium salts during the late 19th century, remarkable experimental, clinical and technological developments in radiologic imaging have continued to transform medicine as summarized in Table 1.1-2 A few years after x-rays were first used for radiologic imaging, physicians and other medical radiation workers developed skin carcinomas, leukemia, dermatitis, cataracts, and other adverse health effects.7-10 Despite early recommendations to decrease stray radiation to the patient and restrict the x-ray beam,8, 11 25 years passed before these recommendations were implemented1 and radiation protection committees were established.12 With the development and evolution of measures of radiation dose, film badge monitoring, and personal (eg, lead aprons) and general (eg, lead shields) radiation protection equipment,2 occupational doses declined dramatically 3, 13-14 and the excesses of leukemia, skin, and female breast cancer in medical radiation workers employed before 1950 were no longer apparent in subsequent medical radiation workers.3
From 1956 to the present, epidemiologic studies have also linked diagnostic x-rays with cancer increases in patients, including modest excesses of pediatric leukemia in offspring of mothers undergoing diagnostic x-rays during pregnancy,15-19 and increased breast cancer risks in women with tuberculosis monitored using fluoroscopy20-23 and in women with scoliosis evaluated with repeated x-rays.24 During the past 30 years, newer imaging modalities (such as computed tomography [CT], myocardial perfusion scans, positron-emission tomography [PET], and other radiologic procedures) dramatically increased. These procedures have provided immense clinical benefit but also higher ionizing radiation exposures to patients. Medical radiation now comprises almost 50% of the per capita radiation dose, compared with 15% in the early 1980s (Fig. 1).25 Although the individual risk for developing radiation-related cancer from any single medical imaging procedure is extremely small, the substantial increase in per capita effective dose during 1980 and 2006, as well as reports of a substantial fraction of patients undergoing repeated higher dose examinations, motivate this review.25-26
The objectives of this review are to summarize the key epidemiologic and experimental data on cancer risks associated with diagnostic radiologic procedures, to relate radiation exposures from recent and current imaging procedures to radiation levels statistically associated with cancer risks, and to propose a framework of strategies for reducing future cancer risks projected from current levels of diagnostic imaging procedures in patients.
The radiation dose is the amount of energy absorbed in the body from radiation interactions. Early non-quantitative measures of dose, based on skin erythema, were replaced by measures of exposure (eg, the ability of x-rays to ionize air, measured in roentgens [or R]) and measures of absorbed dose (eg, energy absorption, measured initially in radiation absorbed dose [or rad] and more recently in gray [Gy] or milligray [mGy] [1 Gy = 100 rad; 1 rad = 10 mGy or 0.01 Gy]).2 Shown in Table 2 are definitions of the key dose quantities and units. Different types of radiation may produce different biological effects and the magnitude of the effect can vary according to the rate at which radiation is received (dose rate). The dose rate is a primary factor in determining the biological effects of a given absorbed dose. For example, as the dose rate is reduced and the exposure time extended, the biologic effect of a given dose is generally reduced. Relative biological effectiveness, which denotes the ability of a given type of radiation to produce a specific biological outcome compared to x-rays or gamma rays, is taken into account by the sievert (Sv), a metric for biological equivalent dose which can be used to measure mixed types of radiation exposure. Effective dose is the sum of the equivalent doses to each tissue and organ exposed multiplied by the appropriate tissue weighting factor or, in other words, the whole body dose of x-rays that would have to be delivered to produce the same carcinogenic risk as the partial dose that was delivered. This quantity provides an easy assessment of overall risk and makes comparison of risks much simpler. Although effective dose is emphasized in many surveys because this metric is related to risk of carcinogenic effects, effective dose cannot be measured and cannot be used for individual risk assessment. Only absorbed dose to a given tissue or organ can be used for estimating cancer risks.30-31
Ionizing radiation is an established carcinogen, based on animal studies and studies of early radiologists; radium dial workers (who used radium-containing paint for glow-in-the-dark watch dials); uranium miners; the Japanese atomic bomb survivors; patients treated with radiotherapy and those undergoing repeated fluoroscopic or radiographic diagnostic examinations.13, 23, 32-34 Two types of cellular damage, deterministic and stochastic effects, are produced by radiation in the absence of adequate repair. Deterministic effects occur above a threshold dose and are characterized by a dose-related increasing risk and associated severity of outcome. A long-recognized adverse deterministic effect is radiation-induced dermatitis,35 which was initially described in 1902.7 After radiotherapy or fluoroscopically guided interventional procedures, generalized erythema may occur within hours then fade within hours to days, followed by a second phase of sustained erythema manifest 10 to 14 days after the exposure. The early erythema is considered to be an acute inflammatory reaction with an increase in vascular permeability, while the more sustained erythema, without other epidermal changes, is thought to be mediated by cytokines.36 Radiation cataractogenesis, particularly occurrence of posterior subcapsular opacities has been considered to be another classic example of a deterministic late effect. Formerly, the threshold was reported to be 2 Gy for acute radiation exposure, 4 Gy for fractionated doses and even higher levels for long-term exposure,31 but recent human and mechanistic studies suggest a lower (eg, around 0.5 Gy) or no threshold.37
Stochastic effects, including cancer and hereditary effects, are caused by a mutation or other permanent change in which the cell remains viable. The probability of a stochastic effect increases with dose (probably with no threshold, an assumption based on molecular knowledge of carcinogenesis: a very small x-ray dose can cause a base change in DNA), but the severity of the outcome is not related to the dose.2 For many years, radiation dose-related cancer risks at low doses were generally estimated from results of the follow-up studies of the atomic bomb survivors and of patients treated with moderate-to high-dose radiation. Major national and international radiation expert committees concluded in comprehensive reviews published during 2005 to 2008 that the available biological and biophysical data support a linear no-threshold risk model for cancer (eg, dose response at low levels occurs in a generally linear pattern without evidence of a threshold),31, 38-39 and that this combined with an uncertain dose and dose rate effectiveness factor for extrapolation from high doses continues to be considered a conservative basis for radiation protection at low doses and dose rates. Some recent reports, based mostly on findings from radiobiology, suggest that there is substantial greater complexity regarding low-dose and low-dose rate effects from nontargeted effects of low-dose radiation (eg, effects in nonirradiated cells near and at distant sites from irradiated cells).40-41
Epidemiologic literature on low-dose and low-dose rate effects is hampered by limited statistical power at cumulative lifetime radiation levels of less than 100 millisieverts (mSv), even for very large studies. Nevertheless, despite wide confidence limits, the results of individual large and pooled studies of radiation workers reveal modest exposure-related increases in risk of solid tumors at low-dose levels.42-43 More research is needed on radiobiologic effects along with continuing follow-up of existing and newer studies of radiation workers to clarify the shape of the dose-response relationship at low-dose and low-dose rate radiation levels.41
Epidemiologic studies have shown minimum latency periods of 2 to 5 years between radiation exposure and the onset of leukemias, with many of the excess leukemias occurring within the first 2 decades of exposure. There is variation in the temporal pattern of radiation-related leukemia risks between exposures in childhood and adulthood (with the decline in risk occurring sooner and more pronounced manner for the former than the latter) and for different major subtypes of leukemia (with the excess risk for chronic myeloid leukemia decreasing substantially about 10 years after exposure for this form of leukemia, the excess risk declining much more slowly for acute myeloid leukemia, and the excess risks of acute lymphocytic leukemia decreasing with attained age based on data from follow-up of the atomic bomb survivors).13, 44-45 Minimum latency periods are longer for solid tumors, ranging from 10 years to many years after the initial radiation exposure. Risks for most solid tumors continue to increase throughout the radiation-exposed person’s lifetime.46 Radiation-related cancers generally occur at the same ages as non-radiation-related cancers.
Much is known about cancer risks associated with a single high-dose rate external radiation exposure from studies of the Japanese atomic bomb survivors,44, 46-47 fractionated high-dose external radiation exposures in patients treated with radiotherapy for benign or malignant disorders 13, 22-23 and to a lesser extent chronic low-dose low dose rate exposures.42-43 The Life Span Study of more than 105,000 atomic bomb survivors (including 30,000 children), remains one of the richest sources of information because of the wide dose range (less than 0.005 Gy to 2-4 Gy [mean, 0.2 Gy]), wide range in age at exposure, and long-term follow-up. This study has demonstrated evidence of a linear dose response for all solid tumors combined, including a statistically significant dose response for survivors with estimated doses under 0.15 Gy (Table 3).44-47 For the 17,448 incident first primary cancers diagnosed during 1958 and 1998 (including 850 causes or 11% diagnosed in individuals with estimated doses greater than 0.005 Gy attributabl‘e to the atomic bomb radiation exposure), significant radiation-associated excess risks were observed for most, but not all, specific types of solid tumors.46 Excess relative risks (ERRs) per Gy (excess compared with baseline population risks) and excess absolute rates (EARs) varied according to organ or tissue and by age at exposure. ERRs per Sv for acute lymphoid, acute myeloid, and chronic myeloid leukemias were 9.1, 3.3, and 6.2, respectively, while excess absolute rates per 10,000 person-year Sv were 0.6, 1.1, and 0.9, respectively.44 Minimum latency periods of 2-5 years were apparent for the leukemias (excluding chronic lymphocytic leukemia), but were longer for solid tumors. Excess risk persisted throughout life for most malignancies.
Among approximately 2500 atomic bomb survivors who were in utero at the time of the bombings, there was no evidence of a radiation dose-related increase in cancer mortality among persons aged younger than 15 years at the time of follow-up.49 In a follow-up of cancer incidence in this population during 1958 through 199947 that compared solid cancer incidence risks among in utero cohort members (based on 94 incident cancers) with risks following postnatal exposures among survivors aged younger than 6 years at the time of the bombings (based on 649 incident cancers), the investigators found that the ERRs per Sv at the same attained age of 50 years were higher for the children exposed postnatally (1.7 per Sv, 95% confidence interval [95% CI] 1.1Sv-2.5 Sv) than for those exposed in utero (0.42 per Sv, 95% CI,0.0 Sv to 2.0 Sv). The EARs per 10,000 person-years per Sv increased markedly with attained age among those exposed in early childhood (EAR = 56, 95%CI = 36 – 79), but showed substantially lower increase with attained age among those exposed in utero (EAR, 6.8; 95% CI, 0.002-48). This landmark study demonstrated that in utero radiation exposure from the bombings was associated with increased adult-onset solid tumor risk,47 but could not provide detailed radiation-related childhood cancer incidence risk estimates in the absence of complete incidence during 1945 and 1957 (the period after the bombings but before establishment of population-based cancer registries in Hiroshima and Nagasaki).
The dose-response patterns for cancer risks associated with high-dose fractionated radiotherapy are generally similar to those of the atomic bomb survivors, but the ERRs per Gy are lower for patients treated with high-dose fractionated radiotherapy compared with those for atomic bomb survivors, likely due to cell killing (Table 3). At high doses radiation kills cancer cells by irrevocably damaging DNA so the cells are nonviable, whereas at lower doses cells may undergo DNA damage, but a large proportion of irradiated cells remain viable. In radiotherapy, extensive efforts are usually made to limit lower dose “radiation scatter” to surrounding tissue so only a small proportion of cells irradiated receive low doses. Nuclear workers have experienced radiation dose-related incidence and mortality risk increases for leukemias (excluding chronic lymphocytic leukemia). In the United Kingdom, incidence was slightly more elevated (ERR per Gy, 1.712; 90% CI, 0.06-4.29) than the dose-associated risks of the atomic bomb survivors (ERR per Gy = 1.4; 90% CI, 0.1-3.4). These workers also had statistically significant increases for all cancers combined other than leukemia. 42-43 Dose-associated increases were also apparent for lung cancer in the 15-country study, 42-43 although the associations with lung cancer may have been confounded by smoking (Table 3).
Prior to 1980, exposures to the U.S. general population from environmental sources of ionizing radiation (eg, radon, natural background gamma radiation, and cosmic rays) were estimated at about 2.8 mSv per capita versus 0.53 mSv from medical sources (the latter comprising about 15% of the estimated 3.6 mSv total).25 The estimated per capita dose from medical radiation in the United States increased approximately 600% from about 0.53 mSv in the early 1980s to about 3.0 mSv in 2006 (the latter including about 1.5 mSv per capita from CT scans, 0.8 mSv from nuclear medicine procedures, 0.4 mSv from interventional procedures and 0.3 mSv from standard radiographic procedures) (Fig. 1). Within the 25-year period, the proportion of per capita individual radiation exposure from medical sources increased from 15% to close to 50% (Fig. 1).25
Although US surveys for specific categories of radiologic procedures have been conducted periodically since the early 1950s, comprehensive assessment across different radiologic procedures has been relatively infrequent. Comparison of the estimated annual numbers and per capita doses for categories of procedures performed during 1980 to 1982 with the annual numbers performed in 2006 showed more than 2-fold increases in the total numbers of all radiographic examinations excluding dental procedures, a 20-fold increase in CT scans, a 5-fold increase in dental radiographic examinations, and a 1.5-fold increase in nuclear medicine procedures, accompanied by a notable change in the specific types of nuclear medicine procedures.25, 29 Compared with an estimated 3.3 million CT scans performed in 1980 and 1982, there were an estimated 80 million CT scans performed in 2010.50 The nearly 6-fold increase in annual estimated per capita effective dose from all sources of medical radiation between 1980 through 1982 and 2006 was due mostly to the nearly 100-fold increase in per capita dose from CT scans and the 5-fold and 2.5-fold increases from nuclear medicine and interventional procedures, respectively.25, 29 Although usage has also increased in other countries, average annual per capita exposure in the United States is 50% higher than in other high-income countries (3 mSv vs 2 mSv per year, respectively).29 Recently, however, there has been evidence of a decline in the percentage of annual increase in CT imaging among Medicare fee-for-service beneficiaries from a compound annual growth rate of 9.5% during 1998 to 2005 to 4.3% during 2005 to 2008.51 Among the Medicare beneficiaries, the decline in the compound annual growth rate for all noninvasive procedures was greater for tests ordered by radiologists (from 3.4% annual growth rate during 1998-2005 to 0.8% annually during 2005-2008) than for tests ordered by all other physicians (from 6.6% annual growth rate during 1998-2005 to 1.8% annually during 2005-2008).
Survey data from the United Kingdom and the United States demonstrate substantial variation in estimated effective doses for different radiologic procedures (Table 4).13, 52-55 For a given type of radiologic procedure, estimated effective doses differ by anatomic site examined (Table 4), by age at examination (particularly for children and adolescents) (Table 5), and by facility where the examination was performed (Fig. 2). Variation among hospitals in estimated effective doses associated with a specific radiologic procedure has been recognized for decades, 60-61 despite early recommendations to restrict the x-ray beam to anatomic sites under study, reduce the numbers of x-ray projections, incorporate standardized proto cols, and improve physician training.61Notable variation in estimated effective doses persists as was reported in 1999 for fetal doses from radiologic examinations62 and more recently for CT scans in adults (Fig. 2).63
The key studies examining the association of various diagnostic radiological procedures and subsequent cancer risk are reviewed below according to age at radiation exposure. Methodologic issues related to the quality and importance of the studies include the source of information about the radiologic procedures (self-reported vs those collected from medical records), the study design (case-control vs cohort studies), the method for estimating doses (dose reconstruction for individual patients vs other approach), the timing of exposure in relation to the cancer, and adequacy of the sample size.
During the late 1940s through the 1960s, obstetricians frequently evaluated pregnancy-related medical problems with whole-fetal imaging using abdominal radiographs and gauged the likelihood of successful vaginal delivery with radiographic imaging of the maternal pelvis and fetal structures within the pelvis (pelvimetry). More than 50 years ago, Stewart et al, in the large Oxford Survey of Childhood Cancers (OSCC) case-control study,15 described a 2-fold statistically significantly higher risk of total pediatric cancer mortality in offspring of women who underwent diagnostic x-ray procedures compared with risk in offspring of women who did not undergo radiographic procedures during pregnancy. Radiation doses to maternal and fetal gonads from pelvimetry based on nationwide UK surveys in the 1950s ranged from 1.4 mGy to 22 mGy per exposure, depending upon the projection and number of exposures.61 There was also notable variation within and among countries19 and over time64-65 in the proportion of pregnant women undergoing pelvimetry or abdominal x-rays. Although the interview-based 2-fold increase in risk reported by Stewart et al15 was initially received with skepticism, more notice was taken when the significant risk excess (RR: 1.39, 95% CI, 1.31-1.47) persisted after the accrual of more than 15,000 pediatric cancer cases in the OSCC during 1953 and 1981,66-67 maternal self-reports correlated well with radiologic reports,67 and a similar 1.4-fold significantly increased risk of total pediatric cancer based on medical records was reported in offspring of mothers undergoing prenatal radiographic examinations in the northeast United States.17 Subsequently, other studies from the United Kingdom, the United States, Finland and Sweden19, 68 replicated the findings.
A 2008 meta-analysis of 32 case-control studies of pediatric leukemia (excluding the hypothesis-generating OSCC study)18 revealed a similar (RR, 1.32; 95% CI, 1.19 – 1.46), albeit slightly lower, risk based on the 4052 pediatric leukemia cases in the OSCC (RR, 1.49; 95% CI, 1.33-1.67).66 Risk of pediatric leukemia from fetal diagnostic x-ray exposure in case-control studies of twins69-71 was comparable to the risks observed in singletons. In the OSCC, the estimated RR for all solid tumors (1.47; 95% CI,1.34-1.62) was similar to the risk of leukemia (RR, 1.49; 95% CI, 1.33-1.67). A few early studies reported modest 20 to 30% increased risks of pediatric central nervous system tumors in offspring of mothers undergoing diagnostic radiologic procedures with abdominal radiation,17, 66, 72 but more recent studies generally found no increase in risk.73-74 A limited number of case-control studies with small numbers of cases have assessed risks of other pediatric tumors associated with in utero diagnostic x-rays.19
OSCC data showed a dramatically declining risk of total pediatric cancer associated with fetal radiation exposure over time, from a 5.4-fold excess among offspring born during 1946 and 1947 to a 1.3-fold increase among children born during 1962 and 1963.64 Compared with the 1.5-to 2.2-fold increased risks of pediatric acute lymphoblastic leukemia in the offspring of mothers undergoing abdominal or pelvic diagnostic x-ray procedures reported in earlier studies,66, 75-76 risks were substantially lower or not increased in more recent studies,65, 77-79 possibly due to decreases in estimated radiation dose levels.
Cohort studies of pediatric cancer risks associated with in utero diagnostic x-rays have included a few hundred to 39,166 exposed children, but the findings were based on 13 or fewer total pediatric cancer cases and 9 or fewer pediatric leukemia cases in each cohort. Summary RRs were initially reported by Doll and Wakeford 68 (RR, 1.2; 95% CI, 0.7-2.0) and subsequently by the International Commission on Radiological Protection (ICRP) 2003 report80 for a larger number of studies (RR, 1.08; 95% CI, 0.78 – 1.50). The estimated RRs for the combined cohort studies were not significantly increased, although the confidence intervals were compatible with both the 40% increase from the case-control studies and with a decreased risk due to limited power and substantial uncertainty.68, 80 A recent record linkage study from Ontario that reported a nonsignificantly reduced risk of total pediatric cancer (based on 4 childhood cancer cases) in offspring of 5590 mothers exposed to major radiologic procedures in pregnancy compared with cancer occurrence in offspring of 1.83 million non-exposed mothers also had wide 95% CIs.81
Because the association between in utero diagnostic x-ray exposure and pediatric cancer risk could be confounded by maternal or fetal medical conditions prompting diagnostic x-ray examinations, epidemiologic studies of twins were recommended to clarify whether confounding could explain the association since a high proportion of twins underwent pelvimetry in early years to determine fetal positioning rather than for medical conditions.82 Cancer risks have been investigated in twin cohorts ranging in size from 13,000 to more than 125,000, with total pediatric cancers ranging from 14 to 166 and pediatric leukemias ranging from 3 to 55.83-89 RRs ranged from 0.70 to 0.96 for total cancer and from 0.7 to 1.14 for leukemia. Cancer risks in twins have not changed over time as pelvimetry has been replaced with ultrasonography,85 but lower pediatric leukemia risks in twins compared with singletons may reflect biologic or clinical characteristics of twin such as low birth weight, intrauterine growth restriction, 5-fold higher mortality in the first year of life, or genetic factors, which may outweigh potentially carcinogenic risks associated with in utero radiation exposure.87, 90
To address concerns that the observed associations between fetal diagnostic x-ray exposure and elevated pediatric cancer risk in offspring might be confounded by medical indication for the x-ray tests, additional analyses were undertaken that demonstrated that the associations were still apparent when the reasons for the diagnostic radiologic examinations were considered.67 In the medical record-based northeast US study, the associations were specific for childhood cancer and not other causes of death in children, and there was no evidence of confounding by many other factors.17 The studies of diagnostic x-rays in utero and risk of pediatric leukemia and other cancers are characterized by several uncertainties, the most important being lack of dose measurement data.18, 68
In utero diagnostic x-rays in earlier decades have been consistently linked with a small excess of pediatric leukemia in offspring. There continues to be debate about whether a radiation dose estimated as to be approximately 10 mGy could give rise to cancer.91 Doll and Wakeford had previously estimated that the lifetime excess risk of cancer for those exposed in utero was 6%,68 which is 2 fold to 3-fold higher than the ICRP lifetime excess risk estimate for exposure in childhood,80 but data from the recent follow-up of the atomic bomb survivors comparing ERRs and EARs of those children exposed in utero and those exposed in early childhood do not support a projection of higher lifetime risk for the former compared with the latter.47 Additional follow-up is needed to quantify lifetime risks in the atomic bomb survivors exposed early in life. Although ultrasound replaced abdominal x-rays and pelvimetry several decades ago, recently there have been reports of increasing levels of radiologic imaging in pregnant women in the US. Investigators leading a large survey at one institution reported that CT increased by 25% per year and nuclear medicine by 12% per year during 1997 through 2006.92 Understanding the cancer risks from in utero exposures, therefore, remains important.
The OSCC found no association between early life diagnostic exposure and risks of total pediatric cancer as reported in interviews of mothers.16 Postnatal diagnostic x-rays of children born between 1980 and 1983 in the United Kingdom were associated with a nonsignificant 2-fold increase (95% CI, 0.32-12.51) of childhood cancer risk based on interview data, but this association was largely attenuated (RR, 1.11; 95% CI, 0.32-3.63) when risks were recalculated for maternal reports of radiologic examinations that were confirmed in medical records.93 More recently, a non-significant modest increase in risk of all pediatric cancer (RR, 1.19; 95% CI, 0.82-1.74) was found in 2690 UK childhood cancer patients born between 1976 and 1996 based on evaluation of medical records.79 There was a slight excess of cancer in 4891 Canadian children with congenital heart disease who underwent cardiac catheterization during 1946 through 1968, and additional follow-up of a subset revealed a nonsignificant 60% excess of leukemia (90% CI, 0.43-4.14, based on 3 cases among 5 total pediatric cancer cases).94 Among 675 Israeli children who underwent cardiac catheterization for congenital anomalies during 1950 through 1970, there was a significant cancer excess (observed vs expected, 2.3; 95% CI, 1.2-4.1) due to increased risks of lymphomas and melanomas, based very small numbers of these malignancies.95
While 2 interview-based studies of early postnatal diagnostic x-rays found a significantly elevated risk of leukemia96-97 and a third observed a significant excess of acute lymphoblastic leukemia (but not acute myeloid leukemia)98 with exposure to diagnostic radiation, other investigations, including studies based on medical record assessment, have not found significant increases.17, 79 Few studies have investigated whether early postnatal exposure to diagnostic x-rays were linked with increased risk of specific subtypes of pediatric acute lymphocytic leukemia, but Shu et al65 found that risk was significantly elevated for pre-B-cell acute lymphoblastic leukemia, and Bartley et al98 reported that risk was significantly increased for B-cell acute lymphocytic leukemia Postnatal radiation exposure from diagnostic radiographs has generally not been linked to an increased risk of childhood brain tumors.19, 99 There have been relatively few studies of pediatric cancers following postnatal radiation other than leukemia and brain tumors and most have had small numbers of exposed cases, including 2 studies that found increased risk of lymphoma.79, 100
Epidemiologic studies of atomic bomb survivors exposed as young children47 and children treated with radiotherapy for benign conditions22 or cancer101 found that children exposed at young ages to ionizing radiation were at increased risk of developing radiation-related cancer later in life. Other evidence also indicates that exposure to diagnostic radiation in childhood or adolescence may have implications for lifetime cancer risk. Repeated diagnostic radiology examinations in adolescents and young women monitored for scoliosis102 and for tuberculosis20 have been associated with increased breast cancer risks later in life. The ERR per Gy for breast cancer incidence was 2.86 (P=0.058) in those monitored for scoliosis (mean dose to the breast was 120 mGy), and risks remained elevated for at least 5 decades following exposure. Risks of lung cancer and leukemia, however, were not elevated in either of these 2 groups of patients.103-104
Overall, studies of pediatric cancer risks in children undergoing radiographic examinations have produced ambivalent results,18-19, 105 perhaps due in part to methodologic limitations or differences (eg, insufficient age matching, recall bias, incorporation of varying latency periods, differing types of radiologic examinations evaluated, and reduction in radiation doses over time for standard radiologic procedures). In addition, if diagnostic radiation exposures are truly associated with very small risk increases, many epidemiologic studies may be too small to detect these increases. Few epidemiologic studies of diagnostic radiation exposures in young children have followed the population for sufficiently long periods to assess risks in adulthood.20, 47, 102 There are major initiatives currently underway around the world, however, to assess the cancer risks from CT scans received in childhood. These studies address many of the limitations described above.106
There have been several large retrospective cohort studies with tuberculosis patients who were monitored frequently using fluoroscopy.20-21 There was a wide range in the number of examinations. The mean dose to the most highly exposed organs (the breast and the lung) was close to 1 Gy. Significant dose-response relationships were found for breast cancer (RR, 1.29; 95% CI, 1.1-1.5), but there was no evidence of an increased risk of lung cancer. There have been no other epidemiologic studies assessing cancer risks in patients undergoing repeated fluoroscopic imaging procedures. Epidemiologic studies of adults undergoing non-fluoroscopic imaging procedures have provided more limited information due to the limited size of such studies, the lower sensitivity of adults to carcinogenic effects of ionizing radiation compared with children, the lack of individual patient dosimetry and the potential for recall bias. Findings from larger studies characterized by stronger methodology and efforts to minimize biases are summarized below.
In a large case-control study conducted in a health maintenance organization in which over 25,000 x-ray procedures were abstracted from medical records and each x-ray procedure was assigned a score based on estimated bone marrow dose, there was small, nonsignificant elevations in risk for leukemias other than chronic lymphocytic leukemia using different lag periods (3-month lag: RR, 1.17 [95% CI, 0.8-1.8]; 2-year lag: RR, 1.42 [95% CI, 0.9-2.2]; 5-year lag: RR, 1.0 [95% CI, 0.6-1.8]), but no evidence of dose-response relationships.109 Preston-Martin and Pagoda found that risks rose with increasing estimated doses to bone marrow to a 2.4-fold excess risk associated with estimated dose of 20 mGy in the 3 to 20 years prior to diagnosis in a medical record-based case-control study of adult-onset acute myeloid leukemia in Los Angeles that utilized a unique database of estimated doses and dose ranges based on review of the dosimetry literature and consultation with radiology experts.107 Radiographic procedures of the gastrointestinal tract and multiple spinal x-rays were linked with an increased risk of chronic myeloid leukemia in a case-control study in Los Angeles.108 Three of 4 earlier studies of chronic myeloid leukemia and diagnostic radiographic procedures (2 examining medical records) found evidence of small risks and one found a dose-response relationship with increasing number of x-ray films in the 20 years prior to diagnosis.108
From the large case-control study by Boice et al, small, nonsignificant increases were apparent for multiple myeloma for all lag periods, and dose-response trends approached statistical significance due to high RRs of patients in the highest exposure score category. There was no significant dose-response relationship for non-Hodgkin lymphoma.109 In Sweden, the cumulative number of x-ray examinations (derived from medical record review) was not linked with thyroid cancer risk.110 Meningiomas111-112 and parotid tumors in adults in Los Angeles113 were associated with full-mouth and substantial numbers of dental x-rays prior to age 20 or before 1945. Comparison of interview data with dental records showed similar levels of agreement for cases and controls, suggesting that the findings were not due to recall bias.114
Overall, the most compelling results are the significant dose-response associations with breast cancer, but not lung cancer, in the cohort studies of patients undergoing repeated fluoroscopic imaging examinations for tuberculosis. Inconsistent findings, limited numbers of epidemiologic studies, and relatively small numbers of substantially exposed leukemia cases other than chronic lymphocytic leukemia make it difficult to draw clear conclusions about diagnostic radiography and risk of leukemia other than chronic lymphocytic leukemia. Limited data suggest a possible risk for chronic myeloid leukemia. There are too few studies examining risks of non-Hodgkin lymphoma, multiple myeloma, thyroid cancer, parotid tumors, or meningiomas to draw conclusions. Recently, a statistical association was reported between chromosome translocation frequencies in cultures of peripheral blood lymphocytes and increasing radiation dose score based on numbers and types of diagnostic x-ray examinations in a cohort of US radiologic technologists.115-116 Mechanistic approaches in conjunction with epidemiologic and genetic studies in selected populations may provide insights about the role of low-dose radiation procedures and genetic susceptibility in breast, thyroid, and other radiogenic cancer risks.
Excess risks of liver, pituitary, and ovarian cancers have been reported in offspring of pregnant mice who were irradiated with a single whole-body dose of 0.3 to 2.7 Gy in utero on days 16 to 18 postcoitus.293117-119 In contrast, offspring of mice irradiated with 1.0 Gy on each day of gestation experienced no significant increase in their incidence of tumors as adults.120 Offspring of 1343 pregnant Beagle dogs irradiated with a single dose of 0.16 or 0.81 Gy on days 8, 28, or 55 after breeding and 2, 70, and 365 days postpartum (120 dogs in each dose and treatment day group) had a significant increase in their incidence of benign and malignant neoplasms, including fatal malignancies at young ages and during their lifetime.121 Statistically significant increases in risk for lymphoma were seen in the beagles irradiated at 55 days postcoitus and significant increases of hemangiosarcomas occurred at 8 and 55 days postcoitus, respectively, but a significantly increasing trend with increasing dose was seen only for hemangiosarcoma among dogs irradiated on day 8 postcoitus.121
Studies examining effects of radiation exposure of 0.5 to 3 Gy in mice during gestation have demonstrated various effects consistent with radiation-related genomic instability in fetal murine hematopoietic cells that are transferred though cell migration to postnatal bone marrow and seen subsequently as chromosomal abnormalities in adult bone marrow, but to date studies have not shown induction of leukemia from prenatal irradiation.122 Efforts to track explicit chromosomal aberrations from fetus to adult revealed that cells with these aberrations are eliminated during the early postnatal stage.123 Nakano at el124 showed that mean translocation frequencies in peripheral blood T cells, spleen cells, and bone marrow cells evaluated in mice at 20 weeks of age were very low when the mice had been exposed to 1 or 2 Gy of x-rays during the fetal or early postnatal stages, but translocation frequencies increased with increasing age at irradiation and then plateaued for mice irradiated at 6 weeks of age or older. These findings in mice were consistent with the absence of a radiation dose-related increase in the frequency of chromosome translocations in atomic bomb survivors exposed in utero (and studied at age 40 years) although the mothers of these offspring were found to have a radiation dose-associated increase in chromosomal translocations.125
Studies of laboratory animals have demonstrated the shape of radiation-associated dose-response curves for cancer over a broad range of doses; carcinogenic effects of acute, single-dose versus fractionated or protracted doses; the radiation-related dose response for cancer according to age at exposure, sex, organ irradiated, genetic background, physiological condition, and environment of the animals; and cellular and molecular mechanisms of carcinogenesis39 Unfortunately, few studies have exposed animals to radiation levels in the range of diagnostic radiologic procedures (less than 0.10 Gy). In more recent years, investigators have developed experimental models to study effects of radiation, cellular interactions, and mechanisms at the cancer progenitor cell level for studies of carcinogenic initiation. From these studies, accumulating data suggests that processes other than induction of specific locus mutations may be important. Such processes may include increased transcription of specific genes, altered DNA methylation, delayed genomic instability (eg, radiation-induced chromosomal alterations, changes in ploidy, mini- and microsatellite instabilities or other changes occurring at delayed times after irradiation and manifest in the progeny of exposed cells), and bystander effects (eg, nontargeted cellular effects usually associated with direct exposure to ionizing radiation but occurring in nonirradiated cells).39
As described above, because the risks to individuals from diagnostic radiation exposures are generally small, it is often difficult to study them directly. However, because of the large number of people exposed annually, even small risks could translate into a considerable number of future cancers. Risk projection models, which utilize the wealth of existing information on the long-term cancer risks after radiation exposure, can provide a more timely assessment of the magnitude of the potential risks. A number of expert committees have developed methodology to estimate the future cancer risks from low-dose radiation exposures. The National Academy of Science BEIR VII committee was the most recent to develop models for the US population,38 and the United Nations Scientific Committee on the Effects of Atomic Radiation13 has also published models for a number of different populations. These reports were used in most of the examples described below.
Based on the frequency of X-ray use in the United States in the early 1990s, Berrington de Gonzalez and Darby126 estimated that about 1% of cancers in the United States might be related to diagnostic x-rays and CT scans. At that time only very basic US survey data were available. Using newly available detailed estimates of the frequency of diagnostic medical radiation exposures in the United States25 and state-of-the-art risk projection models for cancer risks associated with low-dose radiation exposure to the US population,38 they recently published updated risk projections for current levels of diagnostic radiation exposures in the United States.127-128 The projected levels of risk and confidence limits assume a linear dose-response relationship for solid tumors, although there is uncertainty about the magnitude of the risk at low doses.41
These recent estimates suggest that the 70 million CT scans performed in the United States in 2007 could result in approximately 29,000 future cancers (95% uncertainty limits, 15,000-45,000).128 One-third of the projected cancers were from scans performed at age 35 to 54 years, compared with 15% from scans performed before age 18years; abdomen/pelvis scans in adults contributed almost one-half the total risk. If CT scan use remains at the current level, these results suggest that eventually about 2% (95% uncertainty limits, 1%-3%) of the 1.4 million cancers diagnosed annually in the United States129 could be related to CT scans. 128 The most common projected cancers in decreasing order were lung cancer, colon cancer, and leukemias.
Risk projection models have been used in a number of studies to estimate the potential radiation risks from repeated screening. The results of those studies, (eg, screening frequencies and age ranges) are shown in Table 6.130-134 The risks range from about 40 radiation-related cancers per 100,000 screened for annual coronary artery calcification screening from age 45 to 70 years131 to 1900 cancers per 100,000 for annual whole-body CT screening from age 45 to 70 years.133
The decision to expose large numbers of asymptomatic individuals to radiation exposure from screening tests such as CT colonography needs careful assessment since most of the persons screened will not develop the disease of interest. In general, the benefits, where established, should outweigh all risks, including the radiation risks from the radiologic screening test. For example, the mortality reduction from regular mammographic screening in women aged 50 or older is much greater than the estimated risk of radiation-related breast cancer.134 This may not be the case, however, for some screening tests or for screening at ages younger than the recommended ages because the radiation risks are higher but the absolute benefits from screening are typically lower.135 Whole-body CT screening is not currently recommended as a screening tool as no clear benefit has been established.
Evidence for an association between radiation and cancer in genetically susceptible populations with radiation sensitivity comes primarily from studies of individuals with chromosome instability disorders, such as ataxia telangiectasia (AT) and Nijmegen breakage syndrome (NBS).136-138 These rare, autosomal recessive diseases predispose to malignancies (leukemia and lymphoma for AT, and B-cell lymphoma prior to age 15 years for NBS), and in vitro studies indicate that individuals with these disorders are unusually sensitive to ionizing radiation.139-140 Clinical sensitivity to radiation has been observed following radiotherapy in these individuals,141 but it is not known whether these individuals are unusually sensitive to lower radiation doses typically received from diagnostic exposures. Defects in DNA repair genes may predispose individuals to radiogenic cancer or lower the threshold for the development of deterministic effects.34, 142 Patients with serious and unanticipated radiation injuries may be among the 1% of the population heterozygous for the AT mutated (ATM) gene, an autosomal recessive gene responsible for AT, or may harbor some other ATM abnormality. 34, 142 Other clinical disorders with a genetic component affecting DNA breakage or repair also increase radiation sensitivity, including Fanconi’s anemia, Bloom syndrome and xeroderma pigmentosum.34, 142-143 Patients with familial polyposis, Gardner syndrome, hereditary malignant melanoma, and dysplastic nevus syndrome may also be characterized by increased radiation sensitivity.142
Increased cancer risks associated with radiotherapy have been noted for individuals with hereditary cancer syndromes including retinoblastoma (Rb), neurofibromatosis type 1 (NF1), Li-Fraumeni syndrome (LFS), and nevoid basal cell carcinoma syndrome (NBCCS).144 Genetic predisposition has a substantial impact on cancer risk in these populations, which is further increased by radiotherapy. A study of hereditary Rb patients found a statistically significant radiation dose response for bone and soft tissue sarcomas.145 patients with NF1 who were irradiated for optic pathway gliomas are at increased risk for developing other cancers including gliomas, soft tissue sarcomas, leukemia, and malignant peripheral nerve sheath tumors.146 Elevated risks for developing second and third cancers were observed in a cohort of 200 LFS family members, especially children, possibly related to radiotherapy.147 Children with NBCCS are very sensitive to radiation and develop multiple basal cell cancers in irradiated areas.148 Due to improved survival, patients with these syndromes are at risk of second and third cancers, and they generally undergo periodic imaging to detect new tumors. Although the association of diagnostic radiation and cancer risk has not been evaluated in these populations, magnetic resonance imaging (MRI) scans have been recommended in place of imaging studies that produce ionizing radiation exposures to follow up symptoms, evaluate abnormal physical findings, or monitor effects of cancer treatment, particularly in Rb survivors149 and children with NBCCS, especially those who have been diagnosed with medulloblastoma.150 In contrast, [F-18]-fluorodeoxyglucose (18FDG)-PET scans have been recommended for detection of tumors in patients with LFS151 and NF1.152
Despite much interest in the possibility that common genetic variants confer an increased risk of radiation-induced cancer,142 there has been little empirical evidence to date, particularly in the context of diagnostic radiation. One study of childhood leukemia reported potential modification of the relationship between diagnostic x-rays and risk of leukemia by variants in the DNA mismatch repair genes human mutS homolog 3 (hMSH3) (exon23 variant) and human MutL homolog 1 (hMLH1) (exon8 variant), but results from the study were sex-specific and were not consistent between the first and second phase of the study.96, 153 A population-based study of breast cancer154 and a series of nested case-control studies in US radiologic technologists have suggested that common variants in genes involved in DNA damage repair,155-156 apoptosis, and proliferation157 may alter the risk of radiation-related breast cancer from diagnostic radiation procedures, but these results need to be replicated. Similarly, there is some indication that single nucleotide polymorphisms in the O 6-methylguanine DNA methyltransferase (MGMT) and poly (ADP-ribose) polymerase 1 (PARP1) DNA repair genes could modify the relationship between diagnostic radiation exposure and risk of glioma,158 but this has not been reported in other studies.
A few rare genetic variants associated with human cancer susceptibility syndromes appear to increase radiation susceptibility in individuals with chromosome instability disorders and certain hereditary cancer syndromes. Although these syndromes affect only a small proportion of the general population, it is important to identify such individuals and reduce medical radiation exposure to the extent possible. Genetic pathways including DNA damage repair, radiation fibrogenesis, oxidative stress, and endothelial cell damage have been implicated in cell, tissue, and gene studies of radiosensitivity,159 indicating that at least some part of the genetic contribution defining radiation susceptibility is likely to be polygenic, with elevated risk resulting from the inheritance of several low-penetrance risk alleles (the “common-variant-common-disease” model). While common genetic variation underlying this susceptibility is likely, identifying this variation is not straightforward. It is essential that future studies addressing this question be large in size, and have sufficient power to adequately address variation in demographic factors, and also include high-quality radiation exposure information.
Radiation dose levels associated with significantly increased cancer risks are shown in Table 7.18, 20, 42-44, 46, 66, 102, 160-162 These data are derived from epidemiologic studies assessing low-dose radiation and cancer risks. Based on epidemiological data, an international, multidisciplinary group of radiation science experts concluded that the lowest dose of x- or gamma radiation for which there is good evidence of increased cancer risks in humans is approximately 10 to 50 mSv for an acute exposure and approximately 50 to 100 mSv for a protracted exposure, but they recognized the uncertainties of these estimates and the difficulties of increasing precision in estimating radiation dose-response.91 Data from the most recent follow-up of solid cancer incidence in the atomic bomb survivors revealed a statistically significant dose response in the range of 0 to 150 mGy, and the pattern of the trend at low doses was consistent with the trend for the full dose range.46 Although a linear extrapolation of cancer risks from intermediate to low radiation doses appears to be the most reasonable hypothesis, it is acknowledged that there is uncertainty about the true relationship.41 From Table 4, the range of estimated effective doses from a single CT scan is 2 to 15 mSv. Mettler et al have reported that 30% of patients who undergo CT have at least 3 scans, 7% have at least 5 scans, and 4% have at least 9 scans.26 Patients who undergo multiple CT scans, as described in studies assessing use of CT among patients with a wide range of medical disorders,163-166 may be exposed to radiation doses associated with increased cancer risks. A single CT examination may comprise multiple CT scan sequences. Data from 2008 Medicare claims revealed that some hospitals were performing 2-scan sequences for a chest CT examination more than 80% of the time, even though the national average is 5.4%. Overall, 2009 Medicare data showed little change from the 2008 data.167
The referring medical practitioner is responsible for ensuring that a diagnostic procedure involving ionizing radiation is necessary for a patient’s care and that the radiation dose from the procedure is expected to do more good than harm, a concept designated as justification by the ICRP.31
The radiological medical practitioner (who is not always a radiologist) is responsible for ensuring that the radiologic procedure provides images adequate for diagnosis and treatment while keeping the radiation dose as low as reasonably achievable (ALARA), a concept designated as optimization by the ICRP.31 Optimization requires identifying imaging parameters and using procedures and protocols to produce the clinically required information while keeping radiation doses as low as possible.
In addition, the imaging equipment must be properly set up and maintained. To achieve optimization, radiological medical practitioners and radiologic technologists, with substantial input from manufacturers, must work closely with medical physicists to ensure rigorous oversight of radiation-producing imaging units. This includes accuracy of settings, safeguards, calibration, and maintenance, as highlighted in reports of excess radiation during CT brain perfusion scans.168-169 In the United States, there are 2 more avenues for optimization of the CT unit. One is the yearly state requirements for evaluation of dose by physicist and by inspections. For CT, accreditation of technologists is rapidly becoming mandatory, while accreditation of the CT unit is now voluntary but will be mandated for payment by Medicare in 2014.
Referring medical practitioners need guidance to determine whether an imaging study is needed and, if an imaging study is required, which type of imaging study will yield the necessary clinical information at the lowest achievable radiation dose. Unfortunately, it has been well documented that many physicians are often not conversant with the pros and cons of various imaging modalities, with the types of imaging modalities producing ionizing radiation exposure, or with the levels of radiation dose associated with specific imaging modalities. 170-172 Therefore one of the most important roles of the radiological medical practitioner is to provide advice to the referring medical practitioner about the appropriate test for the patient. The advice from the radiologic medical practitioner can be provided in several ways. An efficient method would be for the radiologic medical practitioner to screen requests for “high-dose” examinations, such as CT and, if the appropriate indication is not given or if the patient has had the same or similar radiologic procedures recently, to contact the referring medical practitioner and discuss the case.
Reducing radiation exposure from diagnostic procedures is a shared responsibility of the referring medical practitioner radiologic medical practitioner.173 To assist referring medical practitioners in decision-making about imaging in the management of patients, the American College of Radiology (ACR)174-175 and the American College of Cardiology (ACC) in collaboration with other professional societies176-177 in the United States and the Royal College of Radiologists178 in the United Kingdom have developed evidence- and/or consensus-based guidelines. These guidelines, produced by a panel of experts, generally take the form of identifying which modalities are most appropriate. Below we summarize key elements of the strategy to guide referring medical practitioners in selecting the optimal imaging tests needed for clinical diagnosis and treatment while limiting associated radiation exposures to levels as low as reasonably achievable. A few examples of the relevant literature base are provided, but the scope of this review precludes comprehensive assessment.
In general, only limited data provide strong evidence to conclusively indicate who needs an imaging examination involving ionizing radiation instead of an alternative that does not expose the patient to ionizing radiation. Clearly, it is inappropriate to utilize an imaging test in lieu of obtaining a detailed medical history and a carefully performed physical examination (absent major trauma or a patient in extremis). The concept of benefit/risk ratio should underlie justification decisions. If there is no difference in the expected benefit, the least invasive imaging tests (or those that do not require ionizing radiation) should be preferred over more invasive imaging tests (or those that do expose patients to ionizing radiation). An effort should also be made to avoid repeating the same examination for a given constellation or bout of symptoms and to consider the clinical urgency of the need for an imaging test (eg, ordering a test that can be performed immediately [often a CT]) versus another test, free of radiation-related risk, to be undertaken when an appointment is available (eg, ultrasound) or scheduled in a few days (eg, MRI, which does not involve ionizing radiation).
Because children and adolescents are at higher risk of developing radiation-associated cancers than older persons,46 there has been substantial debate about the optimal type of imaging tests for children and adolescents for certain indications (eg, CT scan vs ultrasound for suspected appendicitis).179 The recognition that children are at higher risk of developing cancer following exposure to radiation than adults has led to increasing reliance on clinical history and physical examination for children suspected of appendicitis and, only if necessary, the use of laboratory tests and imaging to confirm the diagnosis.180-182
Two examples illustrate important aspects of justification: 1) if higher-dose imaging examinations are needed at all (eg, certain pediatric head trauma patients) or 2) if 2 or more higher-dose imaging tests are needed at the same time (eg, posttreatment response in pediatric cancer patients). A third example, guidelines for breast cancer screening using mammography, illustrates some complexities associated with justification given knowledge gaps.
Head trauma is one of the most common reasons that a CT is ordered. While there is little argument that patients with a more severe head injury (eg, Glasgow coma score less than 13) will experience a greater benefit from a CT than any future radiation-related cancer risk, there is a substantial debate on routine CT for a child with a less severe injury (eg, Glasgow coma score greater than 14). In a prospective cohort study of 42,412 children presenting with Glasgow coma scale scores of 14 to 15, trained investigators recorded patient history, injury mechanism, and symptoms and signs before imaging results were known, and followed children to ascertain outcomes (including death, neurosurgery, intubation for more than 24 hours, or hospital admission of 2 nights or more.183 CT scans were obtained at the discretion of the emergency department clinician (n=14,969 patients) and interpreted onsite (780 patients had traumatic brain injuries on CT scan). The investigators derived and validated age-specific prediction rules for clinically important traumatic brain injury. The prediction rules identified children at very low risk for whom the investigators concluded that CT scans were not required.183
Patients with pediatric cancer are frequently treated with radiotherapy, depending upon the diagnosis and treatment protocol implemented. Regardless of the specific treatments, patients with pediatric cancer also undergo extensive imaging for diagnosis and clinical staging, treatment response assessment, and follow-up monitoring after treatment has ended. This assessment entails significant cumulative radiation doses.184 Developing an evidence-based approach to diagnosis and ongoing monitoring of pediatric oncology patients is critical to limit cumulative radiation dose, but there is extensive debate.184 Although it is clear that CT or PET/CT scans are valuable for diagnostic purposes and during early stages of treatment, it may not be necessary to obtain diagnostic contrast-enhanced CT at the same time as PET imaging.184 As noted earlier, it is particularly important to consider alternative imaging procedures for cancer patients who are at high risk for developing radiation-related second malignancies. The high incidence of radiation-related second tumors in patients with hereditary Rb has led pediatric ophthalmologists and pediatric radiologists to propose guidelines that call for the use of MRI rather than CT in such patients.149
Strong evidence from randomized trials has shown that screening mammography from ages 40 to 69 years reduces mortality from breast cancer.185 There are differing interpretations of the evidence and some differences among the guidelines with regard to screening intervals and ages at which to start and stop screening. Nevertheless, there is good agreement about screening for women aged 50 to 74 years.186-188 Reasons for the differences are mostly due to absence of data from multiple large randomized trials to address the following knowledge gaps: lack of accurate and reproducible measures of sensitivity of mammography screening for identification of breast cancer, particularly in those with dense breast tissue, and insufficient evidence about benefits versus harms of screening mammography in older women (age 75 years and older), of annual versus biennial screening, and overdiagnosis (eg, limited knowledge about which ductal carcinomas in situ will go on to become invasive and the rapidity of spread of invasive breast cancers). Given these gaps, the screening guidelines that have been proposed are based on based on expert consensus informed by critical assessment of the literature186 or on statistical modeling.187, 189 The estimated radiation dose associated with a single view in mammography is presently about 2 mGy.190 As indicated above, the risk of radiation-induced breast cancer from routine mammographic screening of women aged 50 to 74 years is small compared with the expected mortality reduction from screening in the general population,134-135 but the benefit may not outweigh the risk of screening female BRCA mutation carriers younger than age 35 years.191
Once the decision has been made that a CT is appropriate, the radiologic practitioner must tailor the CT parameters (milliamperes, kilovoltage peak, automatic exposure control, and others) and protocol (cover only the anatomic region necessary) to the patient’s size and age. There should be as few phases as possible (usually one) as each run (without contrast, with contrast, delayed) multiplies the dose. These considerations should be applied to all patients, but young children, pregnant women, and obese patients require further protocol modifications to optimize dose. 192 Technological improvements, including automatic tube current modulation (which modifies the dose depending on the thickness of the anatomic site to be examined) and noise reduction filters,193-195 will reduce further the doses from CT while continuing to improve images
It is important to include the dose report on all CT and other radiation-producing diagnostic procedures. As the dose cannot be determined by the appearance of the images, this is the only way to verify that the correct protocol was used. For CT, the current metric is the volume-weighted CT dose (CTDIvol). In the future, better metrics, such as size-specific dose estimates CTDIvol as advocated by the American Association of Physicists in Medicine,196 will hopefully become the norm.
A prospective controlled, nonrandomized study enrolled 4995 sequential patients undergoing cardiac computed tomography angiography (CCTA) at 15 hospital imaging centers during a 2-month control period, followed by an 8-month intervention period using a best-practice CCTA scan model (including minimized scan range, heart rate reduction, electrocardiographic-gated tube current modulation, and reduced tube voltage), and then a 2-month follow-up period. Compared with the initial control period, patients’ estimated effective dose was reduced from 21 mSv to 10 mSv, with the most notable reduction in dose occurring at low-volume sites.197
In 1990, the metric of normative values for patient radiation dose from a given procedure was introduced in the United Kingdom and was subsequently recommended by the ICRP.198 These normative measures, designated “diagnostic reference levels,” typically correspond to the 75th percentile of the distribution of measured dose values for particular imaging procedures.199 Diagnostic reference levels serve as benchmarks for comparing dose levels for imaging tests at a given facility with the broad range of dose levels from many other institutions. Such benchmarks should be regularly evaluated, and if exceeded, addressed by medical physicists and radiological medical practitioners, as part of a facility’s quality assurance program in radiation protection.200 These benchmarks should be periodically reevaluated and reduced, as current practices will certainly lower the 75th percentile dose.
The observation of striking regional (including small area) variation in use of medical procedures201 and debate about overuse, underuse, and the “right” level of use202 led to the concept of “appropriateness of medical procedures.” This concept was defined to mean that the expected health benefits from procedures should exceed by a sufficiently wide margin the expected negative con-sequences of performing the procedures.203 The RAND Corporation and the University of California at Los Angeles operationalized the concept of appropriateness of a specific medical procedure for specific indications by basing it on a quantitative score provided by expert panels (drawn from multiple medical specialties and including physicians who did and those who did not perform the procedure) that were guided by formal literature review. Each specific procedure/indication for use category was established for a homogeneous group of patients meeting the criteria for appropriateness; there could be many specific indications for a given procedure. A rigorous, reproducible statistical technique was used to obtain a consensus score on an ordinal scale. The approach has demonstrated good reliability, validity, and predictive power, and has confirmed the efficiency of the method for estimating the appropriateness of a variety of specific procedures for medical care.204 Randomized trials comparing general guidelines with specific appropriateness criteria in decisions about diagnostic testing have found that appropriateness criteria were effective in achieving more appropriate test ordering.205
In 1993, the ACR developed the scientific-based ACR Appropriateness Criteria to guide decisions about ordering imaging proce-dures. These guidelines are comprehensive, currently address more than 175 topics with over 850 variants, are produced through consensus of panels of recognized experts, are updated regularly, and incorporate medical practice guidelines used by the Agency for Healthcare Research and Quality as designed by the Institute of Medicine. The approach relies not only on evidence-based assessment of the scientific evidence but also on expert consensus when data from scientific outcome and technology assessment studies are insufficient.206
The ACR Appropriateness Criteria have been criticized for not utilizing the rigorous methodology of the evidence-based medicine approach for radiology.207 Although there is support for the development of a systematic evidence-based approach to evaluate each specific radiologic procedure/indication, it is acknowledged that there is a lack of even limited measures, such as sensitivity and specificity for certain procedures, let alone more rigorous types of evaluation as randomized trials. These major limitations, in conjunction with the rapid adoption and use of new imaging technologies, limit more comprehensive use of evidence-based approaches.208-209 Similar limitations apply to the Appropriate Use Criteria for Cardiac Computed Tomography developed by the ACC and other collaborating organizations. Studies have identified large proportions of clinical indications for which matching clinical fields or variants cannot be identified in the ACR or ACC Appropriateness Criteria.210-211 Another major problem is the low utilization of the ACR and perhaps the ACC appropriateness criteria, likely due to a lack of awareness of these resources.212
To evaluate a child with a first nonfebrile seizure (which occurs in 1%-2% of children and is generally idiopathic), unless a child is at high risk (eg, the presence of a predisposing condition) an emergent CT is not indicated and well-appearing children who meet low-risk criteria can be discharged if follow-up is assured.213 For low-risk children, an evidence-based assessment demonstrates that MRI is a sensitive neuroimaging modality that can detect neurodevelopmental lesions (eg, heterotopic gray matter, cortical dysplasia, and polymicrogyria, among others), some of which may be difficult to detect on CT214-215. Since many of the causes of seizures are not seen as well or at all on CT, the use of CT exposes children to risk without adequate benefit. That is, CT in these children is not justified. Similarly, for a child with new onset of headaches, the Quality Standards Subcommittee of the American Academy of Neurology and the Practice Committee of the Child Neurology Society has concluded that routine neuroimaging is not indicated for children with recurrent headaches and a normal neurologic examination.216 Neuroimaging evaluation is justified in children with an abnormal neurologic examination or other physical findings or symptoms that may suggest disease of the central nervous system. MRI in this situation will provide more information without radiation exposure compared with CT.
Evaluation of data from the National Hospital Ambulatory Medical Care Survey (1998-2007) provides indirect evidence of inappropriate ordering of CT or MRI examinations in emergency department visits. These data revealed that there was no change during the period in the prevalence of patients admitted to the hospital or intensive care unit from emergency departments, whereas the prevalence of CT or MRI use in the emergency departments increased from 6% to 15%.217 Review of data to assess the use of screening cervical CT examinations performed after trauma revealed that close to 24% of the CT scans of patients that were negative for an acute injury had no written documentation of any of the 5 criteria established by the National Emergency X-Radiography Utilization Study to identify patients with a low probability of cervical spine injury who do not require cervical spine imaging.218 Retrospective review of medical records from 459 outpatient CT and MRI examinations from primary care physicians in the state of Washington using appropriateness criteria from a radiology benefit management company similar to the ACR Appropriateness Criteria revealed that 74% of the imaging examinations were considered appropriate, while 26% were not considered appropriate (examples of the latter included brain CT for chronic headache, lumbar spine MRI for acute back pain, knee or shoulder MRI in patients with osteoarthritis, and CT for hematuria during a urinary tract infection).219 The investigators followed up the results of the examinations and found that 58% of the appropriate studies but only 24% of the inappropriate studies had positive results and affected subsequent management.
For some patients with chronic remitting and relapsing disorders, such as Crohn disease, who may require multiple imaging examinations, evaluation of appropriateness criteria may be less important than consideration of alternate imaging procedures that provide the data for clinical decision-making while reducing radiation-related risks.220 Despite the ACR Appropriateness Criteria, the continuing increase in imaging has led to consideration of preauthorization programs based on Appropriateness Criteria. Utilization patterns of CT and MRI before and after implementation of an Israeli managed care preauthorization program, based on the ACR Appropriateness Criteria and the UK Royal College of Radiology guidelines, demonstrated that annual performance rates of CT and MRI decreased from 25.9 and 7 examinations, respectively, per 1000 in the year 2000 to 17.3 and 5.6 examinations, respectively, per 1000 in 2003, representing reductions of 33% for CT and 9% for MRI.221. Decision support software that uses the ACR Appropriateness Criteria has been built into a computerized radiology examination ordering system, making it available at the time the imaging study is requested.222-223 This method has been shown to be effective in decreasing the rate of imaging utilization.223 It is also essential for reports of all CT and other radiologic examinations to be incorporated into medical records immediately to reduce the frequency of repetition of the same or similar diagnostic radiologic procedures.
The Society for Pediatric Radiology sponsored the first ALARA conference on CT dose reduction in 2001, bringing together physicists; radiation biologists; manufacturers; members of the US Food and Drug Administration (FDA), the National Cancer Institute, and the National Council on Radiation Protection and Measurements with referring and radiologic practitioners. The Society has continued to sponsor biennial conferences focusing on various topics to limit unnecessary procedures and decrease radiation doses from CT.224-227
A crucial offshoot of these efforts was the formation of the Alliance for Radiation Safety in Pediatric Imaging in 2007. By 2008, this advocacy group was formalized with the founding organizations including the Society for Pediatric Radiology, the American Society of Radiologic Technologists, the ACR, and the American Association of Physicists in Medicine. This coalition of professional health care organizations joined with manufacturers of imaging equipment to work together for both appropriate imaging and for reducing the radiation dose from imaging procedures. The organization has continued to grow and now includes more than 65 organizations committed to reducing radiation dose.228-229 The Image Gently campaign is an initiative of this organization (available at: www.imagegently.org).
The Society for Pediatric Radiology has a program to expose second and third year medical students to information about imaging and radiation-producing tests. The Society is also working with the nationwide Children’s Oncology Group to devise dose-reducing protocols for diagnosis, treatment, and surveillance of patients with pediatric cancers.
The ACR, the Radiological Society of North America, the American Association of Physicists in Medicine, and the American Society of Radiologic Technologists have collaborated with the Image Gently campaign of the Alliance for Radiation Safety in Pediatric Imaging to create the Image Wisely campaign, whose objectives are to apply the same principles of appropriate and lower radiation doses to diagnostic procedures undertaken in adults.
A 2009 summit cosponsored by the American Board of Radiology Foundation, the National Institute of Biomedical Imaging and Bioengineering, and the American Board of Radiology identified several contributors to overutilization, including the payment system and reimbursement of procedures on a procedure basis; little control over the number of imaging devices available in populations of patients; high reimbursement for imaging procedures encouraging nonradiologists to add imaging to services provided to patients; little legislative or regulatory action to control inappropriate, financially motivated self-referral practices that have led to higher utilization;230 defensive medicine practices (43% of 824 physicians completing a survey on defensive medicine reported using imaging technology in clinically unnecessary circumstances231 and 28% of CT scans were ordered primarily for defensive purposes in one state);232 lack of education of referring medical practitioners from medical school through residency training, practice, and continuing medical education at meetings; failure to educate referring medical practitioners when inappropriate tests are ordered; failure of radiologists to review imaging requests for appropriateness; failure to educate patients who demand imaging tests about benefits and risks; and inadvertent or deliberate duplication of imaging studies (20% of all patients surveyed in 2007 had duplicate imaging examinations).233-234 Areas for improvement identified by summit participants included better education and training of referring medical practitioners, a national collaborative effort to develop comprehensive evidence-based appropriateness criteria for imaging; greater use of practice guidelines in requesting and conducting imaging studies, decision support at the point of care, education of patients and the public,235 accreditation of imaging facilities, management of self-referral and defensive medicine by the physician community acting in concert or by legislative action to place restrictions on self-referral, and payment reform.234
In February 2010 the FDA launched an Initiative to Reduce Unnecessary Radiation Exposure. The overarching goals are to promote safe use of medical imaging devices, support informed clinical decision-making, and increase patient awareness. To promote safe use of medical imaging devices, the FDA will establish requirements for manufacturers of CT and fluoroscopic devices to incorporate additional safeguards into equipment design, labeling, and user training; partner with the Centers for Medicare and Medicaid Services to incorporate key quality assurance practices into accreditation and participation criteria for imaging facilities and hospitals; and recommend that the health care professional community, in collaboration with the FDA, continue efforts to develop diagnostic reference levels for CT, fluoroscopy, and nuclear medicine procedures locally and also through a national radiation dose registry. To support informed clinical decision-making, the FDA will establish requirements for manufacturers of CT and fluoroscopic devices to record radiation dose information for use in patient medical records or a radiation dose registry and will recommend that the health care community continue to develop and adopt criteria for appropriate use of CT, fluoroscopy, and nuclear medicine procedures that use these techniques. To increase patient awareness, the FDA will provide patients with tools to track their personal medical imaging history.
Professionals and professional organizations that play a key role in appropriate utilization of medical imaging are the referring medical practitioners who are responsible for ensuring that a diagnostic procedure involving ionizing radiation is necessary for a patient’s care and should be expected to do more good than harm (designated as justification) and the radiological medical practitioner who, together with qualified medical physicists and manufacturers of x-ray equipment, provide images adequate for diagnosis and treatment while keeping the radiation dose as low as reasonably achievable (designated as optimization). Only limited data provide strong evidence about which categories of patients should be evaluated with an imaging examination involving ionizing radiation instead of an alternative. Approaches for optimizing doses from imaging procedures have undergone limited assessment. Diagnostic reference levels (corresponding to the 75th percentile of the distribution of doses from all such examinations) provide normative values and serve as benchmarks for comparing dose levels and for investigating imaging practices if these levels are exceeded. The history, methodology, and limitations of the ACR Appropriateness Criteria program to guide decisions about ordering imaging procedures are described. Growing evidence provides documentation that a substantial proportion of imaging examinations are inappropriately ordered and performed. Imaging examinations that do not require ionizing radiation should be preferred, when appropriate, for patients with chronic disorders who require repeated imaging for diagnostic and treatment purposes. Strategies that can reduce unnecessary imaging examinations include pre-authorization and the use of decision support software. Finally, efforts to reduce radiation doses from diagnostic procedures include those by radiation safety alliances of radiologists, physicists, radiobiologists, clinicians, and manufacturers; a summit of 60 organizations to discuss the causes and effects of overutilization of imaging and to identify areas for improvement; and the FDA Center for Devices and Radiological Health Initiative to promote safe use of medical imaging devices, support informed clinical decision-making, and increase patient awareness of radiation exposures from medical imaging.
We are grateful to Annelie Landgren, MPH, and Stephanie Glagola, BA, for technical support.
DISCLOSURE: This review was supported by the Intramural Research Program of the National Institute of Health and the National Cancer Institute. Published 2012 American Cancer Society, Inc.
†This article is a US Government work and, as such, is in the public domain in the United States of America. doi: 10.3322/ caac.21132. Available online at http://cacancerjournal.com