Uncertainty of Calculated Risk Estimates for Secondary Malignancies After Radiotherapy

Int. J. Radiation Oncology Biol. Phys., Vol. 68, No. 4, pp. 1265–1271, 2007 Copyright Ó 2007 Elsevier Inc. Printed in the USA. All rights reserved 0360-3016/07/$–see front matter

doi:10.1016/j.ijrobp.2007.04.014

PHYSICS CONTRIBUTION

UNCERTAINTY OF CALCULATED RISK ESTIMATES FOR SECONDARY MALIGNANCIES AFTER RADIOTHERAPY STEPHEN F. KRY, PH.D.,* DAVID FOLLOWILL, PH.D.,* R. ALLEN WHITE, PH.D.,y MARILYN STOVALL, PH.D.,* DEBORAH A. KUBAN, M.D.,z AND MOHAMMAD SALEHPOUR, PH.D.* Departments of *Radiation Physics, y Biostatistics and Applied Mathematics, and z Radiation Oncology, The University of Texas M. D. Anderson Cancer Center, Houston, TX Purpose: The significance of risk estimates for fatal secondary malignancies caused by out-of-field radiation exposure remains unresolved because the uncertainty in calculated risk estimates has not been established. This work examines the uncertainty in absolute risk estimates and in the ratio of risk estimates between different treatment modalities. Methods and Materials: Clinically reasonable out-of-field doses and calculated risk estimates were taken from the literature for several prostate treatment modalities, including intensity-modulated radiotherapy (IMRT), and were recalculated using the most recent risk model. The uncertainties in this risk model and uncertainties in the linearity of the dose–response model were considered in generating 90% confidence intervals for the uncertainty in the absolute risk estimates and in the ratio of the risk estimates. Results: The absolute risk estimates of fatal secondary malignancy were associated with very large uncertainties, which precluded distinctions between the risks associated with the different treatment modalities considered. However, a much smaller confidence interval exists for the ratio of risk estimates, and this ratio between different treatment modalities may be statistically significant when there is an effective dose equivalent difference of at least 50%. Such a difference may exist between clinically reasonable treatment options, including 6-MV IMRT versus 18-MV IMRT for prostate therapy. Conclusion: The ratio of the risk between different treatment modalities may be significantly different. Consequently risk models and associated risk estimates may be useful and meaningful for evaluating different treatment options. The calculated risk of secondary malignancy should be considered in the selection of an optimal treatment plan. Ó 2007 Elsevier Inc. Second cancers, IMRT, Risk, Uncertainty.

for estimating carcinogenic risk after low-dose radiation exposure. Carcinogenic risk has received increased attention with the advent of intensity-modulated radiotherapy (IMRT), which has been associated with higher out-of-field doses than conventional radiotherapy (11–13) and estimated to have correspondingly higher risks (14–17). For example, Followill et al. estimated that the risk of induced cancer after highenergy IMRT could be as great as 8.4% and could be much greater with high-energy tomotherapy (14). Such risk estimates of induced malignancy are commonly calculated using the NCRP/ICRP risk model (13–15, 18–20). However, no uncertainty has been cited in calculated risk estimates based on this risk model. Although it has been suggested that the risk of fatal secondary malignancies is higher with IMRT than with conventional radiotherapy, and that the risk varies

INTRODUCTION Low doses of radiation have been shown to induce cancer in a number of populations, most notably survivors of the atomic bombing (1, 2), but also in patients who undergo radiotherapy (1, 3–7). Although epidemiologic studies are the ideal method for estimating the risk of secondary cancers after low-dose exposures, such studies are challenging and require decades of follow-up. Consequently this risk of fatal secondary malignancies is often estimated using risk models. A common risk model has been agreed upon by the National Council on Radiation Protection and Measurement (NCRP) and the International Commission on Radiation Protection (ICRP) (8, 9), whereas the Environmental Protection Agency (EPA) has a risk model that is slightly different (10) (Table 1). These two risk models are the best currently available

Supported by the Rosalie B. Hite foundation fellowship for graduate research in cancer. Conflict of interest: none. Received July 27, 2006, and in revised form April 4, 2007. Accepted for publication April 10, 2007.

Reprint requests to: Stephen F. Kry, Ph.D., Department of Radiation Physics, University of Texas M.D. Anderson Cancer Center, 1515 Holcombe Blvd., Box 94, Houston, TX, 77030; Tel: (713) 563-2594; Fax: (713) 563-2482; E-mail: [email protected] 1265

1266

I. J. Radiation Oncology d Biology d Physics

Table 1. Risk models for induced fatal malignancies from the NCRP (8), ICRP (9), and EPA (25) Risk of fatal secondary malignancy (%/Sv) Organ site

NCRP/ICRP

EPA

Bladder Bone marrow Bone surface Breast Esophagus Colon Kidney Liver Lung Ovary Skin Stomach Thyroid Remainder Total

0.30 0.50 0.05 0.20 0.30 0.85 – 0.15 0.85 0.10 0.02 1.10 0.08 0.50 5.00

0.24 0.56 0.01 0.51 0.12 1.04 0.05 0.15 0.99 0.15 0.01 0.41 0.03 1.50 5.75

Abbreviations: EPA = Environmental Protection Agency; ICRP = International Commission on Radiation Protection; NCRP = National Council on Radiation Protection and Measurement. Numbers in parentheses after abbreviations are reference numbers.

depending on the beam energy, it is not clear whether these differences in risk are significant. The uncertainty in the risk estimates involves many different contributions, each of which may introduce large uncertainties. The two risk models mentioned above were based primarily on the incidence of radiation-induced malignancies in Japanese individuals who were exposed to an acute dose of radiation during the World War II atomic bombing. It would be worthwhile to know how well these risk models can be applied to a population of cancer patients exposed to a lowdose-rate as a by-product of radiotherapy. Studies comparing the risk associated with the atomic bomb exposure to epidemiologic data regarding exposure to a low-dose rate of medically administered radiation have produced mixed results depending on the site of exposure (7, 21–23). The issue is further complicated by the fact that during radiotherapy, areas near to the treatment field receive intermediate- to highdose exposures, i.e., from a few Gray up to the target-site prescription dose. Normal tissues irradiated to these doses are susceptible to secondary malignancies (4, 5); however, the standard linear dose–response model for secondary cancer induction does not apply in such cases because cell killing becomes an important factor (16). Although recent work has been done in this area (24), no validated risk model exists for calculating the risk in this region of higher doses. In the current work, we analyze the uncertainty in absolute risk estimates and the uncertainty in the ratio of the risk estimates between different treatment modalities. This analysis is limited to second-cancer induction in low-dose areas, that is, doses less than a few sieverts where there is a linear nothreshold dose response. The goal of the current study is not to determine the risk of induced cancer from any partic-

Volume 68, Number 4, 2007

ular therapy, but rather to evaluate the impact of the uncertainty of the risk models on which such risk estimates are based. If the uncertainty in the risk model is extremely large, the risk model may not have any value because it may be impossible to give a meaningful estimate of the risk or to distinguish between the risks associated with different treatment options. In contrast, if the uncertainty in the risk model is small enough, significant differences in the risk may exist between different treatment options, which may offer clinical guidance on the optimal treatment selection. The analysis in this study is applicable to the calculated risk of secondary malignancy from any arbitrary radiation therapy; however we apply it to previously published dose and risk data for various prostate cancer treatment modalities (15). We analyzed uncertainties associated with conventional radiotherapy and different energies and equipment for IMRT as an evaluation of the current risk models. METHODS AND MATERIALS To place the uncertainty analysis of the risk model in a clinical perspective, out-of-field doses and associated calculated risk estimates for fatal secondary malignancies were obtained from a previous study for a variety of prostate therapies (15). The treatment modalities assessed included 10-field conformal radiotherapy delivered at 18 MV and 8-field step-and-shoot IMRT delivered at 6 MV, 10 MV, 15 MV, and 18 MV via a Varian accelerator (Varian Medical Systems, Palo Alto, CA); and at 6 MV and 15 MV via a Siemens Primus accelerator (Siemens Medical Solutions, Concord, CA). The MU requirements were based on treatment plans for 10 clinical patients. The conventional plan was developed in Pinnacle v6.2 (Pinnacle3, Philips Medical Systems, Andover, MA) with 78 Gy prescribed to isocenter. The IMRT plans were developed with Corvus v4r6 (NOMOS Corp., Cranberry, PA) with 75.6 Gy prescribed to 98% of the prostate and base of the seminal vessicles in 42 fractions. The risk estimates originally produced by Kry et al. (15) were calculated based on the dose equivalent received by the colon, stomach, liver, esophagus, lung, thyroid, and active bone marrow. The average dose equivalent to these organs was used as the dose equivalent to the ‘‘remainder of body.’’ The risk coefficients for the kidneys, bone surface, skin, and bladder were added to the remainder of body coefficient. The dose to the ovaries and breast were not considered because prostate treatments were being examined. These original risk estimates were recalculated for the current study for two reasons. First, the original risk estimates were calculated using the NCRP/ICRP risk model; however, the EPA risk model is a more recent reevaluation of the NCRP/ICRP risk model and the risks were therefore recalculated according to this model (25) (Table 1). Second, the original risk estimates were intended to be conservative maximal estimates of the risk, representing the worst-case scenario by considering the upper limit of the number of monitor units (MUs) required to deliver each treatment and considering the highest rather than the average dose to each organ. The risk estimates were therefore recalculated on the basis of the average number of MUs for each treatment and the average dose to each organ. Along with the recalculated risk estimates, the second-cancer low-dose effective-dose equivalent was determined for each treatment modality. That is, the effective-dose equivalent considering only the risk of secondary malignancies (not genetic effects from gonad dose) and considering only organs that received a low dose

Uncertainty of calculated risk estimates d S. F. KRY et al.

of radiation (<2 Sv) so the traditional linear risk model can be applied. In practice, this effective dose equivalent is the calculated risk estimate (based on organ-specific doses) divided by 5.75%/ Sv, giving a form of whole-body dose equivalent. The uncertainty in the absolute risk estimates was based on the results of the addendum to the EPA report, which included a confidence interval on the whole-body absolute risk estimate of 5.75%/ Sv. The 90% confidence interval was 2.0% to 11.0% (25). This confidence interval is similar to that of the NCRP Report 126 (26) and was generated on the basis of an uncertainty analysis that included sampling variability, diagnostic misclassification, temporal dependence, population mapping, dosimetric uncertainties, and doserate effects. We next examined the ratio of the risk from each therapy as compared with the risk associated with conventional therapy. The second-cancer induction rate reported from low-dose exposures is consistent with a linear no-threshold dose–response model (except at a few sites and for leukemia) (1, 6, 27–29). This means that if the dose is doubled, the risk is also doubled, independent of the absolute risk estimate. The uncertainty in the ratio of risk between two treatment modalities hinges on the uncertainty in the dose–response model (i.e., the accuracy of the linear no-threshold model) and is independent of the uncertainty in the absolute risk estimate. An estimate of the uncertainty in the dose–response relationship was determined from examination of Fig. 2 in Preston et al. (2), which shows the excess relative risk as a function of dose and includes one standard-error bound on the data. Based on the error bounds on the data, the uncertainty in the linearity was taken to be 21% of the risk (90% confidence interval). For an absolute risk rate of 5.75%/Sv, a 21% range corresponds to 4.5%/Sv to 7.0%/Sv.

RESULTS Table 2 lists the average organ doses and average MU requirements for each of the seven prostate treatment approaches examined. Table 3 lists the original risk estimates for fatal secondary malignancy calculated by Kry et al. (15), the recalculated risk estimates produced using the EPA risk model and the dose data from Table 2 (5.75%/ Sv), and the second-cancer low-dose effective dose equivalent. Despite the increase in the per-unit-dose risk estimate

1267

from using the EPA risk model, consideration of the average MU requirements and average organ dose rather than the maximum MU requirements and maximum organ dose lead to a reduction in the recalculated risk estimates. In all cases, the IMRT risk estimates were greater than the risk estimates for conventional radiotherapy. Figure 1 illustrates the absolute risk of fatal secondary malignancy as a function of effective dose equivalent (solid line), along with the 90% confidence interval (dashed lines). The points along the solid line represent the effective dose equivalent and associated absolute risk estimates for each treatment modality shown in Table 3. The point at the lowest effective dose equivalent represents the conventional therapy, whereas the point at the highest effective dose equivalent represents the 18 MV IMRT treatment. The absolute risk estimate and confidence interval for each treatment modality is presented in Table 4. The confidence interval for the risk estimates shown in Fig. 1 is much larger than the differences in risk estimates between any two of the treatment modalities, i.e., the confidence intervals overlapped for all points considered. The ratio of risks of fatal secondary malignancy as a function of effective dose (Fig. 2) was normalized to 0.23 Sv, which is the effective dose of the conventional therapy. The ratio of risks is represented by the solid line, and the confidence interval for the ratio of the risks is represented by the dashed lines. The points along the black line represent the ratio of the risk for each treatment modality shown in Table 3. These ratios, along with the corresponding confidence intervals, are presented in Table 4. Significant differences were observed between the ratio of risks of the different treatment modalities. As illustrated in Fig. 2, a dose difference of more than 50% corresponded to a significant difference in risk, i.e., nonoverlapping confidence intervals. For the treatment modalities considered in this study, Fig. 2 and Table 4 show that there was a significant difference between the risks associated with the conventional therapy and that associated with most of the IMRT therapies.

Table 2. Average dose equivalent to sensitive organs from different prostate therapies (entire treatment, in mSv) and average monitor unit (MU) requirements for each treatment approach Treatment type, energy, and accelerator Conventional

IMRT

18 MV Organ site

10 MV

6 MV

18 MV

15 MV

Varian

Siemens

Varian

Varian

Siemens

Varian

Colon Liver Stomach Lung Esophagus Thyroid Bone marrow

482 242 232 126 96 127 329

786 340 341 154 131 106 379

891 429 414 284 270 289 496

529 278 270 152 134 108 293

684 422 428 245 181 244 597

866 504 479 365 274 351 637

1014 693 687 447 351 546 968

MU/fraction

233

986

1147

808

830

1049

864

Abbreviation: IMRT = intensity-modulated radiotherapy.

1268

I. J. Radiation Oncology d Biology d Physics

Volume 68, Number 4, 2007

Table 3. Original risk estimates of fatal secondary malignancy (based on Reference 15), recalculated risk of fatal secondary malignancy for different therapy modalities, and the effective dose equivalent considering only the risk of secondary malignancy in the low-dose region Treatment type, energy, and accelerator Conventional

IMRT

18 MV

Original risk (%) Recalculated risk (%) Effective dose equivalent (Sv)

10 MV

6 MV

18 MV

15 MV

Varian

Siemens

Varian

Varian

Siemens

Varian

2.9 1.9 0.34

3.7 2.5 0.44

2.1 1.5 0.26

3.4 2.2 0.39

4.0 2.8 0.48

5.1 3.6 0.63

1.7 1.3 0.23

Abbreviation: IMRT = intensity-modulated radiotherapy.

Furthermore, although the confidence intervals for the Siemens IMRT treatments overlapped, this was not the case for the Varian IMRT treatment modalities. The most striking difference was that the confidence interval for the Varian 18 MV IMRT treatment did not overlap with that of the Varian 6 MV or the Varian 10 MV IMRT modalities. Although this analysis has been done based on previously published prostate therapy data, this result is general to the risk model and could be applied to a comparison of arbitrary radiation treatments assuming certain conditions are met (as outlined in the Discussion). That is, a difference of 50% in the out-of-field dose would correspond to a significantly different risk of induced cancer.

DISCUSSION There are very large uncertainties in the absolute risk model. No statistically significant differences were found in the absolute risk estimate between any of the treatment modalities examined in this study. More generally, it is unlikely that a statistically significant difference would be found

Fig. 1. Calculated risk of fatal secondary malignancy as a function of second-cancer low-dose effective dose equivalent of secondary fatal malignancies. The solid line is the EPA risk model for second cancer (10), the dashed lines are the 90% confidence intervals, and the points along the line are the data from Table 3.

between any reasonable radiation therapies. Therefore, the absolute risk as calculated by current risk models is not a particularly meaningful number, and is not particularly useful for evaluating potential treatment approaches for associated risks. In contrast, the ratio of the risk was associated with a relatively narrow confidence interval that produced statistically significant differences in the risk with a difference in the effective dose equivalent of at least 50%. This difference in dose is reasonable among different treatment approaches, and was seen among several of the clinical prostate therapies examined. This suggests that the ratio of the risks may be a clinically useful criterion for evaluating different treatment options. The analysis and confidence intervals of Fig. 1 include many sources of uncertainty in the current risk models, including potential dose rate effects and applying the risk model to a particular population (i.e., present-day Westernhemisphere cancer patients). However, they do not include other substantial sources of uncertainty. For example, although the risk model is based on a range of all ages, radiotherapy patients tend to be older and are consequently less sensitive to induced cancer (1, 6, 28). Not only are radiotherapy patients less sensitive to induced malignancies, they also generally have a shorter lifespan than a population of all ages, and they may not live long enough for second cancers to develop (1). The risk model also neglects the incidence of second cancers near to or in the treatment field. In addition, dosimetric uncertainties are introduced from the relative biologic effectiveness of neutrons relevant during high-energy therapy. The failure for the current analysis to include these sources of uncertainty emphasizes the implications of Fig. 1. Even without these additional sources of uncertainty, the confidence interval on the absolute risk estimates is much larger than any trend in the data. If the additional sources of uncertainty were included in the confidence interval, the true uncertainty in the risk model would be even larger. In short, the ability of the current risk model to estimate the absolute risk of fatal secondary malignancy induced by exposure to a low dose of ionizing radiation is extremely limited. It is unlikely that a significant difference could ever be detected between different treatment approaches, and the absolute risk is therefore a very limited measure for evaluating out-of-field dose and associated risks.

Uncertainty of calculated risk estimates d S. F. KRY et al.

1269

Table 4. Absolute risk of fatal secondary malignancy and 90% confidence interval on the absolute risk estimate, the ratio of the risk of fatal secondary malignancy normalized to the conventional therapy, and 90% confidence interval for the ratio of the risk Treatment type, energy, and accelerator Conventional

IMRT

18 MV

Risk (%) Confidence interval Risk ratio Confidence interval

1.3 (0.5, 2.6) 1.0 (0.8, 1.2)

10 MV

6 MV

18 MV

15 MV

Varian

Siemens

Varian

Varian

Siemens

Varian

1.9 (0.7, 3.7) 1.4 (1.1, 1.8)

2.5 (0.9, 4.8) 1.9 (1.5, 2.3)

1.5 (0.5, 2.8) 1.1 (0.9, 1.3)

2.2 (0.8, 4.2) 1.6 (1.3, 2.0)

2.8 (1.0, 5.3) 2.1 (1.6, 2.5)

3.6 (1.3, 6.9) 2.7 (2.1, 3.2)

Abbreviation: IMRT = intensity-modulated radiotherapy.

The other approach to comparing the risks in this study was examining the ratio of the risk between treatment modalities. This approach is more limited than the absolute risk estimate because it does not offer any information about the severity of risk of fatal secondary malignancy, only the relative likelihood compared with other treatment modalities. For optimal results, the ratio of the risk information must be combined with epidemiologic data that offer insight into the absolute risk. However, by considering the ratio of the risk between treatment modalities, it may be possible to resolve significant differences in the risk between two treatment approaches, which could not be done by evaluating the absolute risk. With the ratio of the risk, not only is the confidence interval much smaller than that on the absolute risk, but many of the unaddressed sources of uncertainty may also become less relevant. For example, the age at exposure and survival time after treatment become less relevant because they will likely be similar between different radiation treatment options. Patient age is doubly complicating when estimating risk because it not only plays a role in overall sensitivity to

Fig. 2. Ratio of the risk of fatal secondary malignancy as a function of effective-dose equivalent of secondary fatal malignancies. The solid line is the ratio of the risk assuming a linear dose response, the dashed lines are the 90% confidence intervals, and the points along the line are the risk estimate data from Table 4. Risks have been normalized to conventional therapy (0.23 Sv).

second cancers, but specific organs may also become especially less sensitive to cancer induction with age; for example, the thyroid and breast are principally sensitive in younger patients (1, 6, 28). As a result, conditions must be met for the confidence intervals presented in Fig. 2 to hold true. First, the patient age must be consistent between the different treatment modalities; and second, the dose distribution must be uniformly different between the treatment modalities at each sensitive organ (i.e., if the effective dose equivalent is double from one treatment modality to the next, the dose equivalent is double at all sensitive organs). An additional source of uncertainty that is minimized by considering the ratio of the risk is the dose-rate effect. Dose rates between different radiation treatment approaches are much more similar than between radiation therapy and an acute exposure such as the atomic bomb. Furthermore, because a higher out-of-field dose typically results from a higher dose rate, and higher dose rates are more biologically damaging, any increase in dose rate would simply increase the risk of the high-dose treatment relative to the low-dose treatment. That is, any dose-rate effect would only accentuate the differences between treatment approach options, provided that there is comparable treatment fractionation. The net result of all of the above caveats is that similar treatment approaches can be compared, and statistically different risks can be calculated if the dose is different by 50%. The uncertainty in the ratio of the risk may be larger if the treatment approaches are not similar; for example, a pediatric treatment versus a prostate treatment, photon therapy versus proton therapy, or even prostate therapy versus headand-neck therapy. However, this work offers reasonable guidelines when similar treatment modalities are examined, such as a variety of potential prostate treatments. There are still other sources of uncertainty that would ideally be taken into account when comparing the ratio of risks. Uncertainty in the dose plays a substantial role in comparing high-energy treatments with low-energy treatments as the uncertainty in the neutron relative biologic effectiveness is not accounted for. Furthermore, as stated originally, there is no account of second cancers in the higher dose regions. Because of these complications, great care must be used to draw a conclusion about the significance of a difference in

1270

I. J. Radiation Oncology d Biology d Physics

the ratio of the risk. However, as seen in Fig. 2, for a specific set of conditions (in this case, second cancers in the low-dose region induced by prostate cancer radiotherapy), significant differences may exist between the risks of different treatment options. The implication of this study, namely, that it is possible to detect significant differences in the risk of secondary cancer after radiotherapy, emphasizes the importance of using risk models as part of the treatment planning process. Although the potential for induced malignancies is a secondary concern that is unlikely to justify the sacrifice of tumor control, this risk is a reasonable consideration when different treatment options are otherwise comparable. If multiple comparable prostate or head-and-neck cancer treatments are available or are being considered for a patient, the risk of induced cancer as calculated by existing risk models may be a meaningful criteria with which to select the optimal treatment approach. CONCLUSIONS Risk models for estimating the likelihood of developing secondary malignancies after radiation therapy can be used for evaluating radiotherapy treatments. This work provides

Volume 68, Number 4, 2007

estimates of the uncertainty in such calculated risks of induced cancer. Very large uncertainties were associated with the absolute risk model that precluded drawing statistical conclusions about the absolute risk estimates of the different treatment modalities examined here. The ratio of the risk between different treatment modalities was found to provide a more statistically significant comparison than the absolute risk. The risk of fatal secondary malignancy may be statistically different when there is a difference of at least 50% in the effective dose equivalent. Such a difference may be encountered between reasonable clinical treatment options such as those examined in this study, including the risk between the 6-MV and the 18-MV IMRT treatment. When evaluating any radiation treatment in the proper framework (i.e., ratio of the risk) the calculated risk of induced malignancy generated from risk models may provide meaningful insight into the potential for late effects. Because this risk of late effects may be different between different treatment approaches, risk models should be used in the comparison of different radiation treatment approaches. As longterm follow-up on IMRT patients accrues, the actual risk of secondary malignancies can be measured and compared with calculated estimates.

REFERENCES 1. Beir V. Committee on the Biological Effects of Ionizing Radiations. Health effects of exposure to low levels of ionizing radiation. Washington, DC; 1990. 2. Preston D, Shimizu Y, Pierce D, et al. Studies of mortality of atomic bomb survivors. Report 13: Solid cancer and noncancer disease mortality: 1950–1997. Radiat Res 2003;160:381–407. 3. Boice JD, Day NE, Anderson A, et al. Second cancers following radiation treatment for cervical cancer: An international collaboration among cancer registries. J Natl Cancer Inst 1985;74: 955–975. 4. Brenner DJ, Curtis RE, Hall EJ, et al. Second malignancies in prostate carcinoma patients after radiotherapy compared with surgery. Cancer 2000;88:398–406. 5. Do¨rr W, Herrmann T. Second primary tumors after radiotherapy for malignancies. Treatment-related parameters. Strahlenther Onkol 2002;178:357–362. 6. Ron E. Ionizing radiation and cancer risk: Evidence from epidemiology. Radiat Res 1998;150(Suppl.):S30–S41. 7. Ron E. Cancer risks from medical radiation. Health Physics 2003;85:47–59. 8. National Council on Radiation Protection and Measurements. Limitation of exposure to ionizing radiation. NCRP Report 116. Bethesda, MD: National Council on Radiation Protection and Measurements; 1993. 9. ICRP. ICRP Publication 60: 1990 Recommendations of the International Commission on Radiological Protection; 1990. 10. U. S. Environmental Protection Agency. Estimating radiogenic cancer risks. EPA 402-R-93–076. Washington DC; 1994. 11. Kry SF, Salehpour M, Followill D, Stovall M, Kuban DA, White RA. Rosen II. Out-of-field photon and neutron dose equivalents from step-and-shoot intensity-modulated radiation therapy. Int J Radiat Oncol Biol Phys 2005;62:1204–1216. 12. Howell RM, Hertel NE, Wang Z, Hutchinson J, Fullerton GD. Calculation of effective dose from measurements of secondary neutron spectra and scattered photon dose from dynamic MLC

13. 14.

15.

16.

17.

18.

19.

20.

21.

22.

IMRT for 6 MV, 15 MV, and 18 MV beam energies. Med Phys 2006;33:360–368. Mutic S, Low DA. Whole-body dose from tomotherapy delivery. Int J Radiat Oncol Biol Phys 1998;42:229–232. Followill D, Geis P, Boyer A. Estimates of whole-body dose equivalent produced by beam intensity modulated conformal therapy. Int J Radiat Oncol Biol Phys 1997;38:667–672. Kry SF, Salehpour M, Followill D, Stovall M, Kuban DA, White RA. Rosen II. The calculated risk of fatal secondary malignancies from intensity-modulated radiation therapy. Int J Radiat Oncol Biol Phys 2005;62:1195–1203. Hall EJ, Wuu C. Radiation-induced second cancers: The impact of 3D-CRT and IMRT. Int J Radiat Oncol Biol Phys 2003;56: 83–88. Hall EJ. Intensity-modulated radiation therapy, protons, and the risk of second cancers. Int J Radiat Oncol Biol Phys 2006;65: 1–7. Verellen D, Vanhavere F. Risk assessment of radiation-induced malignancies based on whole-body dose equivalent estimates for IMRT in the head and neck region. Radiother Oncol 1999;53:199–203. Meeks SL, Paulino AC, Pennington EC, et al. In vivo determination of extra-target doses received from serial tomotherapy. Radiother Oncol 2002;63:217–222. Chibani O, Ma CC. Photonuclear dose calculations for highenergy photon beams from Siemens and Varian linacs. Med Phys 2003;30:1990–2000. Little MP, Weiss HA, Boice JD Jr., Darby SC, Day NE, Muirhead CR. Risks of leukemia in Japanese atomic bomb survivors, in women treated for cervical cancer, and in patients treatment for ankylosing spondylitis. Radiat Res 1999;152: 280–292. Little MP, Boice JD Jr. Comparison of breast cancer incidence in the Massachusetts tuberculosis fluoroscopy cohort and in the Japanese atomic bomb survivors. Radiat Res 1999;151: 218–224.

Uncertainty of calculated risk estimates d S. F. KRY et al.

23. Little MP. Comparison of the risks of cancer incidence and mortality following radiation therapy for benign and malignant disease with the cancer risks observed in the Japanese A-bomb survivors. Int J Radiat Oncol Biol Phys 2002;78:145–163. 24. Schneider U, Zwahlen D, Ross D, et al. Estimation of radiation induced cancer from 3D-dose distributions: Concept of organ equivalent dose. Int J Radiat Oncol Biol Phys 2005;61: 1510–1515. 25. U. S. Environmental Protection Agency. Estimating radiogenic cancer risks addendum: Uncertainty analysis. EPA 402-R-99003. Washington, DC; 1999.

1271

26. National Council on Radiation Protection and Measurements. Uncertainties in fatal cancer risk estimates used in radiation protection. NCRP Report 126. Bethesda, MD: National Council on Radiation Protection and Measurements; 1997. 27. Brenner DJ, Sachs RK. Estimating radiation-induced cancer risks at very low doses: Rationale for using a linear no-threshold approach. Radiat Environ Biophys 2006;44:253–256. 28. Land C. Low-dose radiation—a cause of breast cancer? Cancer 1980;46:868–873. 29. Pierce DA, Preston DL. Radiation-related cancer risks at low doses among atomic bomb survivors. Radiat Res 2000;154: 178–186.