Probability of causation for occupational cancer after exposure to ionizing radiation

Statistical modeling of PC for radiation cancer

In the National Cancer Institute-Center of Disease Control (NCI-CDC) model, PC is basically a calculation of excess relative risk (ERR) as a function of radiation dose for each exposure using the following formula [9]:

[TeX:] \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {\displaystyle \begin{array}{l} PC= risk\; due\; to\kern0.17em radiation\kern0.17em exposure/\left( baseline\kern0.17em risk+ risk\; due\; to\kern0.17em radiation\kern0.17em exposure\right)\times 100\%\\ {}= ERR/\left(1+ ERR\right)\times 100\\ {} Where\; ERR\left( Excess\kern0.17em relative\kern0.17em risk\right)= excess\kern0.17em risk/ baseline\kern0.17em risk\end{array}} $$\end{document}PC=riskdueto radiation exposure/baseline risk+riskdueto radiation exposure×100%=ERR/1+ERR×100WhereERRExcess relative risk=excess risk/baseline risk

ERR is estimated after adjusting for cancer type, sex, age at exposure, attained age, and radiation dose. The sensitivity to radiation by cancer type depends on the dose coefficient in PC model which corresponds to ERR per unit dose(Sv) for exposure age(e) 30 or older and attained age(a) 50 or older for most cancer type. When we calculate the ERR/Sv at e = 30 and a = 50, the ERR/Sv in male is higher in the order of leukemia, lung cancer, thyroid cancer and cancer in kidney and other urinary organs. In female it is leukemia, cancer in kidney and other urinary organs, bladder cancer and breast cancer. Applying different age at exposure and attained age may change the order of sensitivity. The younger the age at exposure, the higher ERR. It is same for the attained age. The dose-response relationship is different according to the cancer type and the characteristics of exposure. Solid cancer and leukemia from high-linear energy transfer (LET) exposure or chronic low-LET exposure assume a linear-dose response relationship. On the other hand, leukemia from acute low-LET exposure assume the linear-quadratic dose response. In case of multiple exposures, each ERR is calculated separately and added together.

International trend of PC for radiation cancer

Since the first PC model for radiation cancer was proposed by the NIH, there have been several models such Biologic Effects of Ionizing Radiation V [10] and United Nations Scientific Committee on the Effects of Atomic Radiation [11]. Most are based on the mortality of Japanese atomic bomb survivors without adjusting for many uncertainties. Furthermore, other limitations remain such as the extrapolation from one population to another and difference in PC according to the choice of model (additive or multiplicative model). The revised PC model proposed by the NCI-CDC in 2013 [9], based on Japanese atomic bomb survivors focused on the evaluation of the distribution of uncertainty. It adapted the random mixed model considering the basic uncertainties and was rated the most reasonable and objective model. Using the NCI-CDC model, the NCI-Interactive Radio Epidemiological Program (IREP) was developed after adjusting for the population of the United States [9] and the National Institute of Occupational Health (NIOSH) developed the NIOSH-IREP, a revision of NCI-IREP [12]. A risk model for radiation cancer based on a life span study of Japanese atomic bomb survivors is updated regularly by the Radiation Effect Research Foundation [13].

Development and recent revision of PC for radiation cancer in Korea

The first PC program for radiation cancer in Korea was KORRAS in 1999. It used point estimation, which was revised in RHRI-PEPC by evaluation of the uncertainty using the NCI-IREP model and Korean baseline cancer incidence in 2003 [7]. In the RHRI-PEPC model, the ERR per unit radiation exposure (mSv) was estimated mainly based on the Japanese atomic bomb survivors study, for the 30 cancer types in the NCI-CDC report (Table 1) [9]. To adjust the uncertainties, the systematic and random errors of radiation exposure dose measurement were statistically revised. Uncertainty due to population transition was adjusted using the random linear combination model. In this transition, the Japanese baseline cancer incidence was from Hiroshima and Nagasaki [12] and the Korean baseline incidence was from 1993 to 1998 [14]. DDREF for chronic low-LET radiation allowed discrete distribution. For acute low-LET radiation, the starting dose at which DDREF applied was randomly decided by log-uniform distribution. In the case of lung cancer, the interaction between radiation and cigarette smoking was added as a mixed additive and multiplicative model. The relative risk of lung cancer by smoking level was obtained from an NCI-CDC report [9]. Other interaction factors such as the effect of race on skin cancer and the first full term delivery on breast cancer were not considered. The minimum latent period of cancer was assumed to be 1 year for leukemia, 2 years for thyroid cancer, and 4 years for other solid cancers. The cancer risk was phased by an S-shaped function to avoid rapid increase from 0 immediately after exposure to full value after a transition period. Therefore, minimum and maximum values were attained at 1 and 5 years for leukemia, 2 and 8 years for thyroid cancers, 4 and 11 years for most solid cancers, respectively.

Table 1

Cancer types for Radiation Health Research Institute-Program for Estimating the Probability of Causation

In 2015, OSHRI and RHRI jointly revised the RHRI-PEPC to Occupational Safety and Health (OSH)-PC for the assessment of uncertainties and baseline cancer incidence in Korea. Among the NCI-CDC [9], a radiation risk assessment tool for lifetime cancer risk projection (RadRAT) program [15], and Japan atomic bomb survivor [16] models, the NCI-CDC model included a large number of cancer types and an uncertainty evaluation method for each cancer. Finally, the OSH-PC program included risk models for 29 cancer types (Table 1) from RHRI-PEPC except digestive system cancers. Among the solid cancers in Table 1, bone cancer, connective tissue cancer, eye cancer, endocrine glands cancer except thyroid cancer, and other, ill-defined cancers do not have individual cancer risk models. For these cancers, the residual solid cancer model from NCI-CDC was adopted. Male breast cancer used the female breast cancer model. For malignant melanoma, the non-melanoma skin cancer model was used. While NIOSH-IREP and RHRI-PEPC used the age-standardized baseline cancer incidence rate to assess population transfer uncertainty, OSH-PC adopted the age-specific cancer incidence rate for a 5-year interval and therefore gave us more accurate PC values for each case. It is well known that lung cancer risk depends on the interaction of radiation and smoking. If the worker is a smoker, the contribution of radiation to his lung cancer is lower than in non-smoker. In the new PC model we reflected the smoking-related adjustment factors in Korea for each lung cancer type (squamous cell, adenocarcinoma, small cell, other types) and the factors were derived using lung cancer relative risks by smoking category in Korea [17]. Differences between RHRI-PEPC and OSH-PC were summarized in Table 2.

Table 2

Summary of Differences between RHRI-PEPC and OSH-PC

PC of cancer cases after radiation exposure in Korea

In the 10 years since 2005, 16 claims of workers’ compensation for cancers after radiation exposure used PC (Table 3). Half of these were lymphohematopoietic cancers (7 leukemias and 1 lymphoma). The remainder included three thyroid cancers, and one case each of breast cancer, auditory canal cancer, rectal cancer, multiple cancer (gastric and pancreatic), and unknown origin cancer. Leukemia cases consisted of three lymphatic leukemia and four myeloid leukemia (three myelocytic, one myelomonocytic, and one myeloblastic) cases. Seven cases involved health care industry workers such as a radiologist, radiological technologists, and a dental nurse. Others worked in nuclear power plants (3 cases) as operators or in radioactive waste treatment, non-destructive tests (3 cases), the semiconductor industry as maintenance workers (2 cases), and sales of medical devices such as the installation of X-ray instrument (Table 3).

Table 3

Cancer cases after radiation exposure in Korea (2005–2014)

Assessment conducted using RHRI-PEPC when the claim was filed (PC1), and by OSH-PC after revision of the PC program (PC2) (Table 3). Because the RHRI-PEPC and OSH-PC dealt with all types of leukemias except chronic lymphocytic leukemia as leukemia, the PC of the 7 cases were assessed as leukemia (Table 3). The cumulative radiation exposure dose for the leukemia cases was 1.7–204 mSv, and the upper 99th confidence limits of PC1 and PC2 were 2.3–65% and 1–58.9%, respectively (Table 3). Workers’ compensation was offered in two leukemia cases with more than 50% PC in the 99th confidence level. A radiologist with 9.1% of PC at the 99th confidence level was accepted for workers’ compensation because his cumulative radiation exposure dose was not believed to correctly reflect the actual exposure. Considerations included inappropriate protective glove during the procedure, higher cumulative dose of co-workers, and testimony of co-workers that they used to work without a film badge. Work-relatedness of a leukemia case was not decided because of lack of objective exposure level data (Table 3). A non-Hodgkin’s lymphoma case with 15.7% PC at the 99th confidence level was accepted for compensation because the exposure dose was believed to be underestimated.

The PCs of three thyroid cancer cases at the 99th confidence interval were 1.2%–33.3% with 3.8–245 mSv of cumulative dose. Workers’ compensation was not given for any of these cases, while a case of carcinoma of unknown primary site was accepted with high PC (67.3%) and cumulative dose (1870 mSv). The PC of a squamous cell carcinoma in the external auditory canal was analyzed under ‘the rest of unclassified cancers or unclear cancer’ category, and was shown to be 16.7% with 99th confidence intervals and not accepted for compensation. The PC of rectal and breast cancer was lower than 1.9% and 12%, respectively, with 72.2 mSv and 16.4 mSv dose, respectively. The PC of multiple cancers assessed as a multiple effect (gastric cancer + pancreatic cancer) was 10% at the 99th confidence level (Table 3).

Looking at the three cases (number 2, 9, 11) although case 2 was exposed to more radiation dose, case 2 (1870 mSv) and case 9 (204 mSv) have similar PC values. This is presumably due to the fact that leukemia has a higher radiation risk than ‘the rest of unclassified cancers or unclear cancer’ and that the age at first exposure of case 9 (24 yr) was younger than case 2(30 yr). Leukemia in case 11 (51 mSv) has a lower exposure than case 9, but the PC value is quite high. The reason is presumably because he was exposed to radiation in a relatively short period (10y1m).

PC2 assessed by OSH-PC was higher than PC1 assessed by RHRI-PEPC in three cases (number 5, 6, 12). PC2 in other cases were lower than PC1 except three cases with the same PC (number 3, 14, 16). The highest variation was seen for a carcinoma of unknown primary site (number 2). In case of the 95th and 99th PC, the highest variation was observed for acute myelocytic leukemia (number 5, Table 3). In thyroid cancer, PC1 and PC2 showed the same results because ERR was transferred independently based on baseline cancer incidence.

The exposure dose of some cases, especially number 3 and 13 based on the estimation based on the statements by the patient and their co-workers, the PC of them might be distorted from actual level.