Mortality among Radiological Technologists in Japan : Updated Analysis of Follow-up Data from 1969 to 1993

A retrospective cohort study was conducted for 12,195 male radiological technologists who received the occupational exposure to low dose radiation over a long term. A total of 1,097 deaths including 435 from cancer were ascertained by Koseki and death certificates from 1969 to 1993. Cancer mortality among the study population was basically compared with that of whole Japanese men. The significant low SMRs were obtained for all cancers, stomach and lung cancer partly due to Healthy Worker Effect, unlike the results of the early reports with some inappropriateness in the methods. Apparent high risks of lymphatic and hematopoietic cancers were observed, although none of site-specific cancers revealed the statistically significant increase. For these cancers, the SMRs among old sub-cohort were somewhat higher than those of young sub-cohort, whereas similar SMRs for solid cancer were obtained between the two subcohorts. The SMR for leukemia reached statistically significant level of 1.75 (95%CI:1.07-2.71) when using whole professional and technical workers as a standard population. The study results might suggest that the chronic exposure to low-dose radiation enhanced the risk of lymphatic and hematopoietic cancers. J Epidemiol, 1999 ; 9 : 61-72


Cancer
Mortality among Radiological Technologists in Japan : Updated Analysis of Follow-up Data from 1969 to 1993 Shinji Yoshinaga 1, Takashi Aoyama 2, Yasuhiko Yoshimoto 1, and Tsutomu Sugahara 3 A retrospective cohort study was conducted for 12,195 male radiological technologists who received the occupational exposure to low dose radiation over a long term.A total of 1,097 deaths including 435 from cancer were ascertained by Koseki and death certificates from 1969 to 1993.Cancer mortality among the study population was basically compared with that of whole Japanese men.The significant low SMRs were obtained for all cancers, stomach and lung cancer partly due to Healthy Worker Effect, unlike the results of the early reports with some inappropriateness in the methods.Apparent high risks of lymphatic and hematopoietic cancers were observed, although none of site-specific cancers revealed the statistically significant increase.For these cancers, the SMRs among old sub-cohort were somewhat higher than those of young sub-cohort, whereas similar SMRs for solid cancer were obtained between the two subcohorts.The SMR for leukemia reached statistically significant level of 1.75 (95%CI:1.07-2.71)when using whole professional and technical workers as a standard population.
The study results might suggest that the chronic exposure to low-dose radiation enhanced the risk of lymphatic and hematopoietic cancers.J Epidemiol, 1999 ; 9 : 61-72 radiation, cancer mortality, radiological technologists, healthy worker effect, standardized mortality ratio Medical radiation workers have been receiving the occupational radiation throughout the long term in medical practices such as X-ray examination and treatment.Especially, those who started working in the early period seem to have received substantial doses of radiation when the conditions of the radiation protection were poor.Such medical radiation workers represent one of the earliest occupational groups exposed to external radiation.Epidemiological studies of these workers have an advantage to provide direct findings on the effects of low-dose protracted exposures of radiation over long periods1).
Radiation dose limits, safety standards of radiation exposure for public or workers, have been established mainly on the basis of extrapolation of epidemiological data of the atomic bomb survivors who got single exposure to the high dose and high dose-rate of radiation 1, 2).Epidemiological studies of medical radiation workers would also help our understanding of the carcinogenicity of ionizing radiation.Epidemiological studies have been carried out in several countries to assess the cancer risk among medical radiation workers.In the previous investigations of pioneering radiologists in UK and USA who are supposed to have been heavily irradiated, the higher mortality of leukemia and skin cancer were observed compared with those of control groups 3-9).Studies in China and Denmark also demonstrated the elevated risks of several sites of cancer among medical radiation workers 10, 11.12) A retrospective cohort study of radiological technologists in Japan has started in order to evaluate the health effects of the low-dose radiation exposure 13.14)The study populations include the early radiological technologists with the potential high exposure to occupational radiation in poor protective conditions.based on the observation from 1969 to 1993.Among 12,133 radiological technologists the standardized mortality ratios (SMRs) for some site specific cancers, such as leukemia exceeded unity though they were not statistically significant, while statistically significant low SMRs were calculated for all cancers, stomach, and lung cancer partly due to the Healthy Worker Effect15).
The main purpose of this paper is again focused onto the comparison of the cancer mortality among radiological technologists in Japan with that of general population.The comparison of cancer mortality between two cohorts by birth year will be also stressed in this paper.This paper will address some methodological issues in the early reports 13,14,15) One of the issues is related to the calculation of expected number of deaths.It will be clarified that there was some misleading from inappropriateness of the calculation in the early reports.
Another issue is related to Healthy Worker Effect.Trying to eliminate the Healthy Worker Effect, SMRs for several causes were calculated using not only whole Japanese but also whole workers or professional and technical workers as a standard population.

Study population
The Ministry of Health and Welfare of Japan introduced a new license system for radiological technologists in 1968.Most of medical X-ray technicians who were working at that time under the previous system applied to the new license.Basic information such as name, date of birth, and permanent address were described in the application for receiving the license.It should be noted that the permanent address here does not always mean the place of birth or the current address but that given in their family register as a domicile.A total of 14,611 technologists had got the licenses by the end of 1975 (Ministry of Health and Welfare of Japan, unpublished data).Using the copies of their application records, we set a fixed cohort of the 12,195 technologists with the eligibility criteria that they were male, born in 1950 or before and with Japanese nationality.
In Japan, Radiation Hazard Prevention Act was enforced in 1958, and the individual monitoring of radiation doses by dosimeters such as a film badge was legally imposed.The introduction of personal monitoring of occupational dose influenced on the decrease of radiation exposure of radiological technologists as well as of other radiation workers .The technologists born in 1933 or before are likely to have started their work in poor protective conditions without monitoring of dose'.
Considering that such technologists had probably got the substantial radiation doses, the study population was classified into two groups by birth year.These two groups were cohortl constituted by the 4,595 technologists born in 1933 or before, and cohort2 by the 7,600 technologists born in 1934 or after.
The birth years of the two cohorts distributed from 1897 to 1950 with a mean value of 1936.6.The proportion of subjects in each category increased with the increase of the birth year as shown in Figure 1. Figure 2 shows the distribution of the calendar year when the study population got their licenses; that is the year at entry into the study.The majority of those in cohort l got their licenses by the end of 1969 just after the new license system was introduced, which implies that most of them were engaging as a radiological technologist at that time or before.Figure 3 shows the distribution of age at entry for the two cohorts by birth year.In cohort I, it had the peak in 40-44 years while it had in 25-29 years in cohort2.The average age at entry was 45.1 years for cohortl, 27.6 for cohort2, and 34.2 for the total cohort.

Ascertainment of death
The ascertainment of the vital status of the study population was carried out using Koseki that is the unique family register in Japan.The inquiry through each municipality public office that controls the family registry of each individual enabled us to ascertain if they were alive, dead or emigrated.Japanese people sometimes but not frequently change their permanent addresses on such occasion as marriage or other ones.If some of them had already transferred the permanent address into another, the relevant municipality office was checked to get information from the latest depository three times at maximum.The cause of death was confirmed by a copy of death certificate from the Legal Affairs Bureau in each relevant region when they were dead.The underlying cause of death was identified and encoded according to the rule of the International Classification of Diseases 9th revision; ICD916)

Calculation of SMR
Person-years at risk of the study population were calculated for each category by calendar year (each from 1969 to 1993) and age at risk (20-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64, 65-69, 70-74, 75-79, 80-84, 85-).Expected number of deaths for each category was calculated by the person-years at risk in each year of 1969 to 1993 multiplied by the corresponding standard mortality rates.The standard rates used here were the age specific mortality rates of whole Japanese men from the vital statistics of Japan 17) .Then the SMRs were calculated as the ratios of the observed number to the expected number of deaths.The 95% confidence intervals of SMRs were calculated assuming the number of deaths followed the Poisson distribution 18).
Basically the SMRs were used to compare the cancer mortality among the radiological technologists with that of general population.However, mortality rates for some causes of death are available by occupation of employees in every 5 years of census ; 1970, 1975, 1980, 1985, and 1990 19).The age specific mortality rates for all cancers, stomach cancer, lung cancer and   leukemia reported by 10 age categories; 20-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64, and 65-of male employees were also used as the standard rates for calculation of SMRs.Mortality rates by occupation in every year from 1969 to 1993 except for the census years were estimated by the following two methods.First, they were estimated by substituting the rates in the nearest census year.Next, these rates were estimated through linear interpolation or extrapolation from those of two adjacent census years.But the latter data were not shown because the two methods had similar estimates.
The information on current employment status of the study population was not available.Therefore, the age-specific rates by occupation were applied for all age categories regardless of the actual employment status of the study population.Furthermore, assuming the radiological technologists would retire around the age of 60 as most of Japanese do, hybrid rates of whole Japanese for age of 65 or over and whole workers or professional and technical workers for age of 20-64 were also used for calculation.

Results of follow-up
Follow-up was carried out from 1969 to 1993 for 12,195 subjects, which yielded 272,043 person-years at risk with the average follow-up period of 22.3 years.Figure 4 shows the distributions of person-years by age at risk for the two cohorts by birth year.In cohortl, there was a peak at 55-59 years and For all malignant tumors of lymphatic and hematopoietic tissue combined, the SMR was at almost statistically significant level (1.36,95%CI:0.99-1.82).

Comparison of SMRs between two cohorts by birth year
Table 3 shows the results of comparison of SMRs between two cohorts by birth year.The SMRs for all cancers were similar between the two cohorts while the SMR for all causes of death among cohort1 was 30% higher than that among cohort2.The SMR for all malignant tumors of lymphatic and hematopoietic tissue combined was significantly high (1.59,95%CI:1.12-2.20)among cohort 1, while that was only 0.86 (95%CI:0.40-1.64)among cohort2.On the other hand, the SMRs for all solid cancer combined were almost equal between the two cohorts.

External comparison with whole workers or professional and technical workers
Table 4 shows the comparison of SMRs by different standard mortality rates applied in the calculation.It should be noted that the SMRs based on whole Japanese in Table 4 were somewhat lower than those in Table 2-3.This is due to the difference in methods of calculation of expected number of deaths; age at risk was classified into 14 categories in Table 2-3, whereas 10 categories in Table 4.The summarized standard rates were used for age of 65 or over in which the background mortality rate is extremely high.The difference of age categorizing might affect the calculation of SMRs because the age structures were not assured to be identical between the standard population and the study population Using whole workers or professional and technical workers as a standard population, the expected number of deaths decreased and SMRs increased for these five disease categories.The SMRs also increased by adapting the hybrid rates of whole Japanese and whole workers or professional and technical workers.The increase in the usage of the hybrid rates was somewhat smaller compared with that based on professional and technical workers.The SMRs for stomach cancer in the total cohort remained to be significantly low regardless of standard mortality rates.For leukemia, the SMRs in the total cohort reached statistically significant levels of 1.75 (95%CI: 1.07-2.7 1) on the basis of professional and technical workers as a standard population.

Discrepancy from early reports
The statistically significant high SMRs were reported for cancers of all sites (1.32,95%CI:1.11-1.56)and large intestine (2.22,95%CI:1.36-3.43) in our early report based on the observation from 1969 to 1982 on 9,179 technologists13).Of the ed numbers.Furthermore, the calculation in the early reports was carried out on the assumption that proportionality between age-specific mortality rate from all causes and that of a specific cause in the middle year was held for other years, although it did not actually hold in this case.Because of lack of adjusting the cohort effects derived from temporal change of age-specific mortality rates and the proportionality of mortality rates of specific causes and all causes, the expected number of deaths was underestimated in the early reports.Although the early reports 13.14) exhibited the significant excess of brain cancer, it was also due to another inappropriateness of the calculation.Since the published data didn't provide the background age-specific mortality rates of benign and malignant types of brain tumor separately, the calculation of the expected number was tentatively based on the malignant type of the disease 15).This led to the underestimation of the expected number of death and the overestimation of the SMRs in the early reports.Although a total of 9 cases of brain tumor were observed in this study, these deaths were excluded from the analysis of the cancer mortality.Efforts are now being made to obtain the separate background rates of benign and malignant types of the disease.

Interpretation of the results
In the total cohort, the SMRs of site-specific cancer exceeded unity for lymphoma, multiple myeloma, leukemia, colon, skin, and bladder, although none of them was statistically significant as shown in Table 2. Considering the findings from other related studies, the possibility that the occupational exposure to radiation increases the risks of lymphoma, multiple myeloma, leukemia, cancer of colon, and skin will be discussed below.
Numerous epidemiological studies have consistently provided an evidence of elevated risk of leukemia after exposure to radiation.Of all subtypes of leukemia, the risks of chronic lymphatic leukemia (CLL) and adult T cell leukemia (ATL) are not associated with radiation exposure 1.20).A total of 20 leukemia cases observed in this study included 1 case of CLL and 3 cases of ATL.Detail information in the vital statistics was not available for calculating expected number of deaths excluding the cases of CLL, ATL or other sub-types of leukemia.Since diagnosis criteria of ATL have been established recently and causes of death were not always verified histologically, the information on ATL might have been underreported on death certificates.
Previous studies to date other than that of American radiologists failed to reveal the increased risk of malignant lymphoma among medical radiation workers.Of malignant lymphoma, Hodgkin's disease is not thought to be induced by radiation exposure 21).For non-Hodgkin's disease, in addition, there is no convincing evidence of association between radiation exposure and the disease 1.22).Of 17 deaths from malignant lymphoma observed in the study, no case of Hodgkin's disease was included.The SMRs of malignant lymphoma for cohort l was 2.6 times higher than that of cohort2 although these values were based on too small number.This finding might be due to the possible high radiation doses in the poor radiation protective conditions.
The mortality of malignant lymphoma is rather higher in southwestern part of Japan, especially in Kyusyu area than any other parts of Japan 23).This is associated with the infection of human T lymphotropic virus-1 (HTLV-1) which causes adult T-cell leukemia/lymphoma; ATL.The proportion of those individuals with the permanent addresses in an ATL endemic area, Kyusyu was 13.1% (1,594 out of 12,195) for this cohort and 12.0% for the general population on the basis of the current addresses by the census in 1985.There was no evidence that the geographical distribution was substantially different between the study population and general population.The geographical factor regarding ATL was unlikely to affect the mortality of leukemia or lymphoma in the study population.
With regards to multiple myeloma, elevated risks have been demonstrated in many studies of population exposed to radiation w.Recent combined analysis of nuclear workers in UK, USA and Canada 25) also showed a significant risk of multiple myeloma.Similar SMR of multiple myeloma was obtained for the two cohorts by birth year.In addition, these elevated risks were somewhat higher than those of leukemia.For multiple myeloma, however, epidemiological data from atomic bomb survivors as well as from other populations imply that the minimal latency is rather longer, the relative risk smaller, and the distribution of age at death older than for leukemia ~~.Further follow-up would provide additional information to determine the risk of multiple myeloma due to radiation exposure.
The SMR for colon cancer was apparently high, although there was no large difference between the two cohorts by birth year.The risk of colon cancer can be increased by intensive irradiation in humans as demonstrated in the study of atomic bomb survivors 21).No elevated risk of colon cancer, however, has been shown in previous studies of radiologists in UK, USA, China or Denmark 27).There was no explanation for the elevated risk of colon cancer in this cohort.Although this might be a chance finding, the possibility of such biases as misclassification of cause of deaths should be examined.
Increased risks of skin cancer among medical radiation workers have been consistently reported in UK 9), USA 8), and China 1D.In this study, the SMR of skin cancer exceeded unity with 2 observed number of deaths but not statistically significantly, both of which were observed in cohort 1.The scattered radiation on their skin surface that they received in the course of medical practices such as photofluorogram or indirect radiogram might have caused these cancers.As for non-lethal cancers such as skin or thyroid cancer, data on incidence would be much informative than those of mortality to evaluate the health risk of radiation.

Occupational radiation exposure of radiological technologists
The lack of individual dosimetry in the early period of poor radiation protection is one of the most serious limitations to preclude the usefulness of epidemiological studies of medical radiation workers 1).A retrospective dosimetry is one of the useful methods to estimate occupational doses in the early periods 28,29,30).One of the retrospective dosimetry demonstrates that the early radiological technologists had been exposed to the radiation at a high rate, even several Sv per year in the poor working situation 31).At the present, however, the published data indicates that the recent level of radiation dose is as low as 0.51 mSv per year in 1993 32).Therefore, radiation doses in poor protective conditions seem to be dominant in the cumulative doses and recent doses could be negligible for early technologists.
Another retrospective dosimetry based on the work history demonstrated that the mean cumulative dose was 470 mSv (for the mean work years of 38) for 3,461 individuals born in 1933 or before whose work history was available among our study population 30).Although the work history was not obtained for those technologists born in 1934 or after, the mean cumulative dose can be estimated by the dosimetry to be 132 mSv (for the mean work years of 22) on the assumption that they had been working in the general hospital with ordinary protection from 21 to 60 years old.This difference of the estimated doses could be sufficient to explain the high risk of lymphatic and hematopoietic cancers in cohortl compared with that of cohort2.Because of much uncertainty involved with the estimate due to systematic and/or non-systematic errors in that dosimetry, further efforts are expected to reduce the estimation errors and to use the estimates for epidemiological studies.
There is no central registry system in Japan for medical radiation workers that controls the individual radiation dose, while there is one for nuclear workers.The system for nuclear workers facilitates the large-scale epidemiological study based on individual dose 33).It is important to introduce a central registration system on the radiation dose of the medical radiation workers for risk assessment and occupational safety.

Healthy Worker Effect in study population
In this study, the SMRs on the basis of the mortality rates of whole Japanese men were statistically significantly low for all causes, which implies that there was a strong Healthy Worker Effect in the study results.The lower SMR for all causes among cohort2 than that of cohortl could be explained by the Healthy Worker Effect.The SMRs for all cancers, cancer of stomach and lung were also significantly below unity, although the Healthy Worker Effect is unlikely to manifest in the study of cancer mortality 34).These findings suggest that the study population consisted of more health-conscious people than general population.The study population may have special medical knowledge related to the prevention of diseases and can get more medical profits than general population.Apart from such accessibility to medical services, the significant low SMRs for stomach and lung cancer might be explained by some lifestyle factors among the study population.
Our findings were not *healthy hire effect' due to selection of healthy people at the first stage of employment, one of possible factors to generate Healthy Worker Effect 35).Generally it is thought to decrease with the increase of working duration S. Our follow-up started for each subject at the time when he got the license several years or even decades after the start of his work.In fact, the analysis excluding the first 2, 5, or 10 years of follow-up did not reveal any substantial difference of SMRs not only for all causes of death but also for all cancers.The SMRs with the lag-time of 2, 5, and 10 years were 0.66, 0.66, and 0.68 for all causes and 0.82, 0.81, and 0.81 for all cancers.

Comparison of SMRs
Regarding the birth year as a surrogate variable of the radiation exposure, the SMRs were compared between two cohorts by birth year.Actually, Kondo et al. 36) reported that there was a good correlation between occupational dose estimates by the retrospective dosimetry and age for approximately 1,000 radiological technologists.However, the comparison of SMRs between several exposure groups has been criticized by some epidemiologists 18,37).In the calculation of SMRs, a standardization of mortality rates is carried out by weights of the distribution of person-years at risk.Even if the SMR of one exposure group is higher than that of another exposure group, this difference couldn't be explained by the effect of the exposure to the risk factor of interest alone.Caution is needed for the interpretation of the results of comparison of SMRs, because the distribution of person-years by age at risk was considerably different for the two cohorts by birth year in this study as shown in Figure 4.The extension of follow-up will provide sufficient person-years at risk for old age categories for cohort2 to facilitate an appropriate internal comparison.

Other methodological issues
The endpoint in the present study was mortality based on Koseki and death certificate, which enabled the good completeness of the follow-up.However, the accuracy of underlying cause of death based on death certificate is a matter of concern.Accuracy of diagnosis is not always assured in death certificate.Kodama et al 38) demonstrated that the agreement between death certificate and autopsy diagnoses is rather low for specific sites of cancer among atomic bomb survivors .Caution should be paid in interpreting the results based on mortality data.The possibility of availability of data on cancer incidence is now under consideration .
The potential confounding factors such as smoking and drinking could not be controlled, although the effect of such factors should not be ignored in studies of low dose exposure .An investigation on life style factors will be carried out for a Another weakness of the present study is that appropriate control groups such as medical workers other than radiological technologists could not taken.Although the Healthy Worker Effect seems to be reduced by applying the mortality rates of occupational groups, some incomparability still remains between the study population and whole Japanese, whole workers or professional and technical workers as a standard population.An internal comparison within the cohort will provide additional important findings.
In conclusion, apparent high risks of lymphatic and hematopoietic cancers were demonstrated, although none of site-specific cancers revealed the statistically significant increase among the total cohort.For these cancers, the SMRs among the old sub-cohort were somewhat higher than those of young sub-cohort, whereas similar SMRs for solid cancers were obtained between the two cohorts by birth year.These findings suggest the carcinogenic effect of occupational radiation exposure throughout the long term.The average age of the survivors in the study population was 56.2 years at the end of follow-up, 31, Dec. 1993.The background mortality of the cancer is still increasing in the study population.The extension of the observation period would improve the statistical power.
Further study is expected to offer the important information for the risk evaluation of the long-term exposure to radiation.

Figure 1 .
Figure 1.Distribution of birth year of the cohort.

Figure 2 .
Figure 2. Distribution of calendar year at entry for two cohorts by birth year.

Figure 3 .
Figure 3. Distribution of age at entry for two cohorts by birth year.

Table 3 .
Comparison of SMRs between two cohorts by birth year.0: Observed number of deaths CI: Confidence interval a) Cohortl were born in 1933 or before and cohort2 were born in 1934 or after b) Including 2 cases of ATL and 1 case of CLL c) Including 1 case of ATL

Figure 4 .
Figure 4. Distribution of person-years by age at risk for two cohorts by birth year.
Aoyama et al 15) recently reported the updated results Received July 31, 1998 ;accepted September 24, 1998.' Division of Human Radiation Environment , National Institute of Radiological Sciences, Chiba, Japan.'Department of Experimental Radiology , Shiga University of Medical Science, Otsu, Japan.

Table 2 .
Observed and expected number of deaths, SMR and 95% CI for all causes and main sites of cancer.

Table 4 .
Comparison of SMR by different standard mortality rates.a) Standard rates except for the census years were estimated by substitution of the nearest census year's rate.b) Hybrid rates of whole workers for age categories of 20-64 and whole Japanese for age categories over 65 were used.c) Hybrid rates of whole professional and technical workers for age categories of 20-64 and whole Japanese for age categories over 65 were used.d) Cohort] were born in 1933 or before and cohort2 were born in 1934 or after.
ogists 14).These 4,052 technologists were all born in 1933 or before.On the contrary, SMRs were generally lower than unity in the present study that covered 12,195 and periods of observation.In the early reports 13,14)only a data set of life-table and age-specific mortality rates of the middle calendar year in the observation period (i.e. that of 1975 for the observation from 1969 to 1982 and 1985 for the observation from 1983 to 1986) was used for the calculation of expect-