© Springer-Verlag 1994

Int Arch Occup Environ Health (1994) 66:71-75

Bernard L Cohen

Dose-response relationship for radiation carcinogenesis in the low-dose region

Received: 2 May 1994 / Accepted: 31 May 1994

Abstract Evidence that low-level radiation substantially enhances the effectiveness of repair mechanisms is summarized This finding destroys the theoretical basis (there is no other basis) for use of a linear-no threshold dose-response relationship to estimate the cancer risk of exposure to low-level radiation Such a methodology will exaggerate the risk This conclusion is further supported by epidemiological evidence and by studies of the effects of radon exposure in the home, which are reviewed. Key words Low-level radiation · Cancer risk · Doseresponse relationship · Radon Epidemiologic studies

The number of initiating events is then proportional to the number of particles of radiation, which is proportional to the dose. A potential problem with this reasoning is that it gives no consideration to repair mechanisms which render harmless the vast majority of initiating events The L-NT theory tacitly assumes that the efficiency of repair is not affected by radiation. A great deal of evidence has recently been developed indicating that LLR substantially enhances the effectiveness of repair mechanisms, probably by increasing production of repair enzymes and/or stimulating the immune system We summarize some of this evidence here.

Introduction Effects of LLR on biological defense mechanisms Essentially all experimental information on radiation carcinogenesis is derived from the high-dose region, where effects are large enough to be easily observable However, essentially all practical applications are in the low-dose region Since these low-dose applications, such as limiting releases of radioactivity into the environment, protecting the public from accidents in reactors and in transport, management of radioactive waste, regulating uses of radiation, and clearing up contamination, are costing many billions of dollars each year, it is most important that every scientific effort be exerted to understand the doseresponse relationship in the low-dose region. With only minor exceptions, since the 1950S official groups such as ICRP, NCRP, UNSCEAR, and U S National Academy of Sciences committees, have consistently used a linear-no threshold (L-NT) dose-response relationship in estimating the cancer risk of low-level radiation (LLR) The basis for this is the widely accepted concept that a cancer is initiated by a single genetic mutation which can be induced by a single particle of radiation.

B L Cohen University of Pittsburgh, Pittsburgh, PA 15260, USA

Many studies have shown that preexposure to LLR substantially reduces the number of chromosome aberrations induced by a later exposure to a high radiation dose Clear demonstrations of this have been reported for human lymphocyte cells by Shadley and Wolfe ( 1987) and by Shadley and Dai ( 1992) It has been demonstrated for bone marrow cells in mice by Cai and Liu ( 1990), Liu et al (1990), Jiang et al (1992), Yu et al ( 1992), Gaziev et al (1992), and Farooqi and Kesavan (1993) It has also been reported for mice spermatocytes by Cai and Liu (1990) and Cai et al (1994), for mouse embryo fibroblasts by Wolfe (1993), and for mouse lymphocytes by Wojcik and Tuschi ( 1990) lalthough this was not found in another mouse strain by Wojcik et al (1992)l. Many studies have shown that the frequency of gene mutations induced by exposure to a high radiation dose is substantially reduced by preexposure to LLR This was observed for mutations at the hprt locus in human lymphocyte cells by Sanderson and Morley (1986), Kelsey et al ( 1991), and Rigaud et al (1993), in human HL-60 cells by Zhou et al (1994), and in mouse SR-I cells by Zhou et al ( 1993) Perhaps the most dramatic demonstration of such effects is the observation that low-dose irradiation of Drosophila substantially reduces the number of dominant

72 lethal mutations induced by later exposure to high radiation doses (Fritz-Niggli and Scheppi-Buechi 1991). The straightforward explanation for this impressive body of observations is that LLR stimulates repair processes for DNA damage There is little difficulty in explaining such an action It is well known that many genes are activated by LLR and that these can stimulate or inhibit production of various enzymes and proteins Excesses or deficiencies of some of these proteins (e g , cyclin B) can delay cell mitosis, thus giving more time for repair Also, some of these enzymes and proteins are the principal agents of DNA repair For example, it has been shown that proteins produced in mouse cells by irradiation, when added to cultures, reduce the frequency of chromosome aberrations from exposure to radiation in cells grown in that culture (Liu and Li 1994) A similar effect in human cells, involving the protein XIP 269, was reported by Boothman et al (1989). Another mechanism for explaining the enhancement of repair processes by radiation is radical detoxification. Cells must remove toxic radicals by mobilizing enzymes, and the levels of these enzymes may be increased by LLR This ensures the availability of scavengers to cope with radicals produced by later irradiation. All of the above evidence leads to the conclusion that the efficiency of DNA repair is affected by radiation, clearly destroying the theoretical basis for an L-NT theory The dose-response relationship at low dose and dose rate is then the sum of the cancer initiation effect, which is L-NT, plus the effect of improved repair efficiency. An additional factor, the effects of LLR on the immune system, is also worthy of consideration It has long been recognized that radiation can have important effects on the immune response (Anderson and Warner 1976; Anderson 1994) High doses suppress the immune response (Kennedy et al 1965), but there is now abundant evidence that low doses usually stimulate it As a recent example, Liu et al (1992) have reported that LLR of mouse splenocytes substantially increases a variety of indicators of implaque-forming cell ability, mune system function mixed lymphocyte reaction, reaction to concanavalin A, natural killer cell activity, antibody-dependent cell-mediated cytotoxicity, and interleukin-2 and interferon secretions. The involvement of the immune system in cancer development is not well understood, but there is much evidence to suggest that it has effects As one direct example, white cells cultured in the presence of interleukin-2 acquire the capacity to kill tumor cells (Smith 1990) Other examples involve experiments in which low-level irradiation of animals is found to augment the effects of immunization in reducing tumor growth Immunization is tested by response to injection of tumor cells Improved resistance to the development of these tumor cells was reported for fibrosarcoma (Anderson et al 1986), Lewis cell carcinoma (Li et al 1994), and squamous cell carcinoma (Miyamoto and Sakamoto 1987) in mice There is evidence for such effects in human lymphoma patients; radiation caused an increase in some types of T cells, and tu-

mor remission was found even when the irradiated area did not include the tumor (Takai et al 1992). Suppression of tumor growth by stimulation of the immune system may be a third factor that must be taken into account to understand the dose-response relationship for radiation carcinogenesis in the low-dose region It may be a very long time before all of these effects are understood and sorted out, but it is surely reasonable to conclude that there is no reason to continue believing that L-NT, extrapolated from data at high doses, should be valid in the lowdose, low-dose-rate region In fact, most of the evidence reviewed above strongly suggests that this procedure substantially overestimates the effects of LLR.

Epidemiologic evidence An alternative approach to studying the dose-response relationship is through epidemiology Unfortunately, epidemiology in the low-dose region is plagued by severe statistical problems: effects are small, leading to a requirement for massive amounts of data to achieve statistical significance Moreover, the errors introduced by even small biases in selecting control groups and in adjusting for confounding factors can be very important As a result, any one study can hardly provide conclusive results. However, there have been a large number of studies, and the effects of these problems should tend to average out when all of them are considered The data have recently been reviewed by Luckey (1991); we summarize very briefly here his most relevant points on radiation-induced cancer. With regard to gamma ray radiation, studies of cancer rates in high natural radiation vs low natural radiation areas give a negative correlation high radiation areas have lower cancer rates in the United States, India, and China. All of these were somewhat sophisticated investigations involving substantial efforts to eliminate confounding factors Less elaborate studies give weak evidence for a positive correlation in Japan and a negative correlation in Austria An important problem in these studies is that natural gamma radiation, according to L-NT theory, is responsible for only about 1 % of all cancers, but actual variations in cancer rates among regions of the United States are about 20 % and we have no understanding of what causes them. There have been several studies of radiologists and their technicians in the United States, Japan, and Britain, and except in the early years of this century, when exposures were very high, they have generally experienced lower cancer rates than comparable medical groups with specialties not involving radiation exposure. Taken together, studies of radiation workers in nuclear power plants and in government-operated facilities in the United States, Canada, and Britain have shown no evidence for increased cancer rates, in spite of occasional media reports to the contrary Comparisons between workers in nuclear and fossil-fuel power plants indicate lower cancer incidence in the former Studies of people living

73 There have therefore been more than 50 epidemiology studies of the relationship between radon exposure in homes and lung cancer Most of these have been cursory and tentative; only about a quarter of them have included actual measurements of radon levels in homes We discuss here the three case-control studies that are generally regarded as significant. Blot et al ( 1990) studied radon levels in houses of 308 lung cancer cases and 356 controls, all female, in Shenyang, China Median radon levels were 2 3 pCi/l and 20% of homes were above 4 p Ci/l Results were controlled for smoking and indoor air pollution The cases and controls in each category were: < 2 p Ci/l, 91/95 ; 2-4 pCi/l, 131/148; 4-8 pCi/l, 60/77 ; > 8 pCi/l, 26/36 This indicates, if anything, a slightly lower risk for those most exposed. Schoenberg et al ( 1992) studied radon levels in houses of 480 cases and 442 controls among New Jersey women. Unfortunately for the study, radon levels were generally very low, reducing the statistical power Results were controlled for smoking They report a positive trend, an increase in case/control ratio with increasing radon level, but this is almost completely due to the very few houses with > 4 pCi/1-six cases vs two controls (compared to 86 cases vs 107 controls in the Blot et al study). Pershagen et al (1994) measured radon levels in houses of 1281 cases and 2576 controls, selected from 109 municipalities in Sweden, correcting odds ratios for smoking, job type, and urban vs rural living They found a positive association between radon exposure and lung cancer, corresponding to a relative risk increase of 3 7 % per pCi/l (95 % confidence interval 0 4 %-8%) One possible problem with this study is in their selection of municipalities, since both lung cancer rates and average radon levels were known in advance If they preferentially selected high lung cancer municipalities with high radon levels, and/or low lung cancer municipalities with low radon levels, that would give a bias toward a positive result A similar problem arises in their selection of cases and controls: if their cases were preferentially selected from high radon areas and/or their controls were selected preferentially from low radon areas, there would be a bias toward a positive result. Studies involving radon If we accept the Swedish study (since it has far better statistics than the others), the result is within the range of The need for large numbers of subjects to satisfy statisti- predictions from L-NT based on the miner data, but well cal problems in the low-dose region suggests the use of below its midpoint On the other hand, it still leaves a 2 % natural radiation There are at least two serious problems probability for zero slope. here: ( 1) usual risk estimates indicate that only about 1 % Although this author can hardly claim to be unprejuof all cancers are due to gamma radiation from natural diced, it is my opinion that our study is by far the most sources; and (2) with the exception of a few small, low- convincing From various data sources we determined the population areas, geographic variations in natural gamma average home radon level, r, in 1729 U S counties, inradiation are generally rather small; for example, in the cluding 90 % of the U S population With or without corUnited States, the variation among the 50 states is only by rection for smoking prevalence, we found a strong negaa factor of 2. tive correlation between r and lung cancer mortality, m. These problems are much reduced if one studies ef- This is in sharp contrast to the strong positive correlation fects of radon in homes; according to usual estimates this expected from L-NT, a discrepancy in slope of the m vs r causes about 10% of all lung cancers, and state average regression of 20 standard deviations While ours is an radon levels in the United States vary by a factor of 5. "ecological study," we showed that the "ecological fal-

near nuclear installations in France, Britain, and the United States indicate lower cancer rates among them than among controls, with some statistical significance. Occupationally exposed men with body burdens of plutonium have experienced fewer cancers than expected from the L-NT theory Studies of former dial painters with body burdens of radium report statistically significant deficiencies of bone cancer in the low-dose region when compared with L-NT predictions from the high-dose experience Numerous studies of the tens of thousands of U.S , British, and Canadian military personnel involved in nuclear bomb tests generally show a statistically significant reduction in cancer incidence among those exposed to radiation (contrary to media-publicized reports). Experiments on animals show a clear deficiency of cancers in the low-dose region from the numbers expected by application of L-NT to higher dose data This has been found, at least for low-LET radiation, in several different studies involving both external radiation (e g , Ulrich et al 1976) and radiation from injected radioactivity (e g , Finkel and Biskis 1968). To summarize these studies, there is a large body of data supporting the thesis that L-NT substantially exaggerates the cancer risk from LLR, and there are few studies indicating that it gives the correct or too small a risk. There is even a sizeable amount of data that can be interpreted as indicating that low-dose radiation is protective against cancer Luckey (1991) seems to be completely convinced of the latter interpretation, but BEIR committees, ICRP, NCRP, UNSCEAR, and similar groups have not accepted that interpretation. On the other hand, these groups have generally accepted the position that L-NT substantially exaggerates the cancer risk from low doses This is usually expressed in terms of a "dose and dose-rate effectiveness factor" (DDREF), a factor by which predictions of L-NT are to be reduced in the low-dose/low-dose-rate region This is usually estimated to be in the range 2-10, but taken to be 2 for reasons that this author interprets as conservatism lInquiries to individual members of these committees generally elicit larger numbers (Cohen 1991) l

74 lacy" does not apply in testing an L-NT relationship, and that other problems with ecological studies are not relevant We studied possible effects of uncertainties in radon levels and in smoking prevalence and found that they can do little to explain the discrepancy We stratified the data separately on 54 different socioeconomic variables, and on seven physical characteristics, but got the same result for each stratum: our results apply if we consider only the urban counties or only the rural, only the richest or only the poorest, only those with the best medical care or only those with the worst medical care, only the warmest or only the coldest, only the wettest or only the driest, etc , etc They also apply to all strata in between, as for only counties with average wealth, average temperature, etc , etc The same results are obtained if we consider only counties in a given geographic area, or only counties in a given state Factors known to correlate radon and smoking, e g , urban people smoke more but are exposed to lower radon levels than rural people, were studied in detail and found to be inconsequential Other potential explanations for the discrepancy were considered and found to provide little help We conclude that L-NT fails badly, grossly overestimating the risk of LLR.

Summary Our conclusion is that there is very little evidence supporting the linear theory down into the low-dose region, and there is abundant evidence that doing so substantially overestimates the cancer risk from LLR.

References Anderson RE (1994) Effects of low dose radiation on the immune response In: Calabrese EJ (ed) Biological effects of low level exposures to chemicals and radiation Lewis, Boca Raton, Fl. Anderson RE, Warner NL (1976) Ionizing radiation and the immune response Adv Immunol 24:215-335 Anderson RE et al (1986) Low dose irradiation permits immunization of A/J mice with subimmunogenic numbers of Sal cells Br J Cancer 54:505-509 Blot WJ et al (1990) Indoor radon and lung cancer in China J Natl Cancer Inst 82: 1025-1030 Boothman DA, Bouvard I, Hughes EN (1989) Identification and characterization of X-ray induced proteins in human cells Cancer Res 49:2871-2878 Cai L, Liu SZ (1990) Induction of cytogenetic adaptive response of somatic and germ cells in vivo and in vitro by low dose Xirradiation Int J Radiat Biol 58:187-194 Cai L et al (1994) Induction of an adaptive response to dominant lethality and to chromosome damage of mouse germ cells by low dose radiation Mutat Res ¶l¶: q:l-1 ll Calabrese EJ (ed) (1994) Biological effects of low level exposures to chemicals and radiation Lewis, Boca Raton, Fl Cohen BL (1991) Catalog of risks extended and updated Health Phys 61:317-335 Farooqi Z, Kesavan PC (1993) Low dose radiation-induced adaptive response in bone marrow cells of mice Mutat Res 302: 83-89 Finkel MP, Biskis BO (1968) Prog Exp Tumor Res 10:72 Fritz-Niggli H, Schaeppe-Buechi C (1991) Adaptive response to dominant lethality of mature and immature oocytes of D. melanogaster to low doses of ionizing radiation Int J Radiat Biol 59:175-184

Gaziev AI et al (1992) Radioadaptive response in mice chronically exposed to low dose rate of gamma rays In: Sugahara T et al. (eds) Low dose irradiation and biological defense mechanisms. Elsevier, Amsterdam ISBELLES (1994) International Symposium on Biological Effects of Low Level Exposures to Radiation and Related Agents, Changchun, China 1993 Chin Med J Jiang T, Chen D, Wang C (1992) Adaptive response in mouse bone marrow cells to chronic beta and gamma-rays In: Sugahara T et al (eds) Low dose irradiation and biological defense mechanisms Elsevier, Amsterdam Kelsey KT et al (1991) Human lymphocytes exposed to low doses of X-rays are less susceptible to radiation-induced mutagenesis Mutat Res 263:197-201 Kennedy JC et al (1965) Radiosensitivity of the immune response to sheep red cells in the mouse as measured by the hemolytic plaque methods J Immunol 94:715-722 Li X et al (1994) Effects of low dose radiation on pulminary dissemination of Lewis cell carcinoma in mice In: ISBELLES, International Symposium on Biological Effects of Low Level Exposures to Radiation and Related Agents, Changchun, China 1993 Chin Med J Liu SZ (1992) Multilevel mechanisms of stimulatory effect of low dose radiation on immunity In: Sugahara T et al (eds) Low dose irradiation and biological defense mechanisms Elsevier, Amsterdam Liu SZ, Li XK (1994) Induction of proteins in lymphocytes by in vitro and in vivo radiation with low dose X and gamma-rays. In: ISBELLES, International Symposium on Biological Effects of Low Level Exposures to Radiation and Related Agents, Changchun, China 1993 Chin Med J Liu SZ, Cai L, Sun JB (1990) Effect of low dose radiation in repair of DNA and chromosome damage Acta Biol Hung 41:149-157 Luckey TD (1991) Radiation hormesis CRC Press, Boca Raton, Fl. Miyamoto M, Sakamoto K (1987) Anti-tumor effect of total body irradiation of low doses in WHT/Ht mice Jpn J Cancer Clin 33:1211-1220 Pershagen G et al (1994) Radon and lung cancer in Swedish homes N Engl J Med 330:159-164 Rigaud 0, Papadopoulo D, Monstacchi E (1993) Decreased deletion mutation in radioadapted human lymphoblasts Radiat Res 133:94-101 Sanderson BJS, Morley AA (1986) Exposure of human lymphocytes to ionizing radiation reduces mutagenesis by subsequent ionizing radiation Mutat Res 164: 347-351 Sasaki S, Kasuga T (1986) Life shortening and carcinogenesis in mice irradiated at the perinatal period with gamma rays In: Thompson RC, Mahaffey JA (eds) Life-span radiation effects studies in animals: what can they tell us? CONF-830951 Schoenberg JB et al (1992) A case-control study of radon and lung cancer among New Jersey Woman In: Cross FT (ed) Indoor radon and lung cancer Battelle Press, Richland, Wash. Shadley JD, Dai G (1992) Cytogenic and survival adaptive responses in G-1 phase human lymphocytes Mutat Res 265:273-281 Shadley JD, Wolff S (1987) Very low doses of X-rays can cause human lymphocytes to become less susceptible to ionizing radiation Mutagenesis 2:95-96 Sugahara T et al (eds) (1992) Low dose irradiation and biological defense mechanisms Elsevier, Amsterdam Takai Y et al (1992) Anti-tumour effect of low dose total (or half) body irradiation and changes of the functional subset of peripheral lymphocytes in non-Hodgkin's lymphoma patients after TBI In: Sugahara T et al (eds) Low dose irradiation and biological defense mechanisms Elsevier, Amsterdam Ulrich RL et al (1976) Radiat Res 68:115 Wojcik A, Tuschi H (1990) Indications of an adaptive response in C 57 BL mice pre-exposed in vivo to low doses of ionizing radiation Mutat Res 243: 67-73 Wojcik A, Bonk K, Streffer C (1992) Indications of strain specificity for the induction of adaptive response to radiation in mice In: Sugahara T et al (eds) Low dose irradiation and biological defense mechanisms Elsevier, Amsterdam

75 Wolff S (1993) Communication to the UNSCEAR Secretariat Yu WR et al (1992) Pre-exposure of mice to low dose or low dose rate ionizing radiation reduces chromosome aberrations by subsequent high dose of radiation or mitomycin C In: Sugahara T et al (eds) Low dose irradiation and biological defense mechanisms Elsevier, Amsterdam Zhou PK et al (1993) Cultured mouse SR-1 cells exposed to low dose of gamma rays become less susceptible to the induction of mutagenesis by radiation as well as by bleomycin Mutagenesis 8:109-111

Zhou PK et al (1994) Adaptive response to mutagenesis induced by low dose y-rays in human HL-60 cells and its molecular basis In: ISBELLES, International Symposium on Biological Effects of Low Level Exposures to Radiation and Related Agents. Changchun, China 1993 Chin Med J