Document rB8pezr8KpM9G6dmN6G38nkee

204 THE NEW ENGLAND JO'JR&IAL OF MEDICINE droxylysyl glycoside levels and their relationship IO collagen mcmbolism. J Clin Invest. 1970; 491497-1509. 32. Kivirikko KI. Urinary excretion of hydroxyproline in health and dis- ease. In1 Rev Connect Tissue R a . 1970; 593-163. 33. Askenasi R, Rao VH. Lkvos A. Peptide-bound hydroxylysine and large polypeptides related to collagen ajnthess. Eur J Clin Invest. 1976; 6361-3. 34. Hamazaki H, Hotta K. Purification and characterization of an a-glucosidase spccifii for hydroxylysine-linked disaccharide of collagen. J Biol Chem. 1979, 2549682-7. 35. Sternberg M. Spiro RG. Studies on the catabolism of the hydroxylysine-linked disaccharide units of b w m e n t membrane and collagens: isolation and characterization of a rat kidney a-glucosidase of high specificity. J Biol Chem. 1979; 254: 10329-36. 36. Judisch GF. Waziri M, Krachmer JH. Ocular Ehlers-Danlos syndrome with normal lysyl hydroxylase activity. Arch Ophthalmol. 1976; 94:1489-91. 37. Eyers PH, Holbrook KA, McGillivray 8 , MacLeod PM, Lowry RB. Clinical and ultrastructural heterogeneity of type IV Ehlers-Danlos syndrome. Hum Genet. 1979; 47141-50. 38. Vogel A, Holbrook KA, Steinmann B. Gitzelmann R. Byers PH. Abnormal collagen fibril structure in the gravis form (type I) of Ehlers-Danlos syndrome. Lab Invest. 1979; 40:201-6. 39. Tilsley DA, Beard TC. Epidermolysis bullosa simplex Lancet. 1963; 2905-7. -40. Meyer UA. Strand U.Dorr M. Rces AC, Marvcr acute porphyria demonstration of a genetic defect I gen metabolism. N Engl J Med. 1972; 2861277-82. 41. Mustajoki P. Red cell uroporphynnogen I synthet.oe in .ep tent wrohyria. Ann Clin R a . 1976; 8 Suo01 17133-8. 42. Bauei EA,-Coopcr TW. Tucker DR. Esteiiy NB. of rc~cssivedystrophic epidermolysis bullose: clin posed mechanism of action on collagenase. N Engl 776-81. 43. Bauer EA, Fiehler WK. Esterly NB. Increased glywrsmi cumulation as a genetlc charactmstic in cell cultum of dominant dystrophic epidermolysis bullosa. J Clin Invest. 19 9. 44. Pearson RW. Studies on the pathogenesis of eplde Invest Dermatol. 1962; 39551-75. 45. Stenn KS, Madri JA, Roll FJ. Migrating epidermis producu A 4 - 2 lagen and requires continual collagen synthesis for movement. N- 1979 271229-32. 46. Mur& JCTStingl G, Kleinman HK. Martin GR, K a u SI. E p G d cells adhere preferentially to type IV (basement membrane) collrgmJ Cell Biol. 1979 80197-202. SPECIAL ARTICLE 4 i RADIATION-INDUCED MYELOMATOSIS 7. ' I!I JACK CUZICK, PH.D. I 5 Abstract It is well known that radiation can cause observed, 10.71 expected; P = 0.012). All other groups myeloid leukemia. However, no excess of chronic lym- combined had a highly significant excess (50 ob li phocytic leukemia has been observed. Myelomatosis, served, 22.21 expected; P = 2x10-7). The largest rela like chronic lymphocytic leukemia, is a tumor of B tive risk appeared among persons receiving internal :.!I :? lymphocytes.To determine whether this disease has a doses of a-particles (14 observed, 3.24 expectradiogenic origin, we surveyed all cohorts of persons ed; P = 2xlO-5), but a significant excess (13.- i: exposed to radiation for which data on cancer-related served, 6.33 expected; P = 0.026) was also found In mortality are available. patients receiving only therapeutic or diagnogtlc An excess of myeloma was found in most cohorts. yrays or x-rays. Most cases occurred 15 to 25 However, a striking deficit was found in two groups ir- years after exposure. (N Engl J Med. 1981; 304: radiated intensely for uterine neoplasms (three cases 204-10.) ,' THE evidence that radiation causes myeloid leuke- small numbers of myelomas reported incidentally ih mia is now overwhelming, and current interest cohorts in which other cancers (most often leukemia) has centered on the shape of the dose-response were of major interest. Although myelomatosis and curve.'-' Radiation also produces cancer in many myeloid leukemia are both cancers that present in the other tissues of the body, although the latency period bone marrow, they are very different diseases. Mye- is usually much longer.' Chronic lymphocytic leuke- loid leukemia affects granulocyte and macrophage mia, however, is one exception; an excess of this dis- precursors, in contrast to the B-cell lineage of the neo- ease does not appear to result from exposure to radia- plastic plasma cells in myeloma. Radiation-induce tion.'.*J leukemia usually becomes evident between two and 15 Since chronic lymphocytic leukemia is predomi- years after exposure, whereas the scanty data avail- nantly a disease of B lymphocytes6(the remaining few able suggest that any excess in the incidence of mye- -. cases have T-lymphocyte markers'), it is of interest to loma does not begin to appear for at least 10 years examine the evidence for the radioinduction of other after exposure and may continue for 30 years or more. tumors of B lymphocytes, particularly myelomatosis. Perhaps because similar mechanisms govern the peri- The case for radiation as a cause of myelomatosis is od between exposure and manifestation in spontane- generally thought to be much weaker than that for ra- ous cases of these two diseases, their age distributions dioinduction of myeloid leukemia, and is based on are different. Spontaneous myeloma tends to occur later in life, and unlike myeloid leukemia, myelo- ma has an incidence that grows as approximately the ' From the Cancer Epidemiology and Clinical Trials Unit, University of Oxford, 9 Keble Rd.. Oxford OX1 3QG. England, where reprint requests should be addressed. Supported by the Imperial Cancer Research Fund. fifth or sixth power of age in common with many epithelial tumors.*,9 The purpose of this study was to bring together RU)L-\TION-INDUCED AdYELOMATOSIS - CUZICK 205 SOM to update) information from as cohorts apovd to radiation as possible, in o obtain a m ~ c ccomprehensive understanding ential rbt In some instances, expected numyelomas M not directly available, and it was ry h & n - e x p o s e d cohort in which a d m incidence or cancer-related morMETHODS ing r a b o o a wth myeloma appears N et al '' o(1 thc atomtc-bomb survid include workers in the nuclear inn,or orha d t t o n s RESULTS ned a total of 22 cases of myed to thc atomic bomb who reiroshima on October 10, 1950. f no radiation-induced risk, the adjusted for sex, age, s a-e given in Table 1 Table 1. Cases of Myeloma in Atomic-Bomb Survivors.* VARIABLE RADIATION EXF'OSURE No. of observed cases 7 10 5 No. ofexpected c a m t 10.45 9.95 I .59 -Risk ratio $ I .o I .5 4.1 *P * 0.028 for comparison between groups receiving < I rad and > 100 rads. P 0.55 (not significant) for comparison between groups rcaiving < I rad and 1-99 rads. 2 = 2.75 and P = 0.006 (normal approximallon) lor analysis of trend. Data are from lchimaru et al." thpected numbers of casu were calculated under the assumption that the 22 observed caws were distributed among the population wlrhout regard 10 exposure. $Risk ratios assumed unity risk in the group exposed to < I rad. number of 1.59. These cases occurred 14, 18, 20, 23, and 27 years after exposure. A nonsignificant excess was also seen in the group exposed to 1 to 99 rads (P = 0.55). Full use of these data may be obtained from an analysis of trend; this method showed a highly significant increasing risk with dose level (P = 0.006). The relative amounts of y-ray and fission neutrons differed between the two Japanese cities, and Ichimaru's analysis suggested that the excess of myeloma was associated more closely with 7-irradiation. Thus, strong evidence was obtained for the radiogenesis of myeloma by high levels of certain types of radiation. However, the data are equivocal in the important 10-to-50-rad range, and because of the neutron flux, these data may not be directly applicable to the more common industrial and clinical exposures to y-irradiation and x-rays. Occupational Exposures to Radiation Nuclear Workers Reports of cancer-related mortality among workers in nuclear processing plants at Hanford, Washington, and Windscale, England, have shown an excess of myeloma on the basis of a small number of cases. Eleven cases have been reported among all employees at Hanford. The original method of analysis used by Mancuso et al." is open to many valid criticisms, however, and their results should be discounted. A more appropriate, though indirect, analysis by Andersont5arrived at an expected number of 8.1 myelomas among all employees. More recently, Gilbert and Markst6further analyzed the data on these workers. Among white men, nine myeloma-related deaths had occurred after 1965, whereas 6.3 would have been expected from the appropriately adjusted national rates. Gilbert and Marks also performed an internally standardized analysis by exposure category; in that analysis, three deaths from myeloma were recorded among white men who had been employed for at least two years and who had had an accumulated exposure level of at least 5 rem two years before they were at 206 T H E NEW ENGLAND J O U R N A L OF LLEDICINE Jan. 22, 1981 Table 2. Observed and Expected Cases of Myeloma in Atomic-Bomb Survivors and Cohorts Receiving OCCUpatiOnal EXPO. sures to Radiation. - CUHURT M4JOR TWE OF R4Dl4Tlo\ NU. OF CASES OBSERVED W l l UP C4\ES E\PtCTED PVNLE (TWO-SIDED) YO0 CUNFIDEKE I N T E R V A L FOR RISK R ~ r i u REFERENCE Atomic-bomb survivors exposed to > 100rads ,-rays, fast neutrons 5 1.59 0.03 I .3-11.0 lchimaru et al." Hanford workers ( > 5 rem) Windscale workers Radium-dial painters Uranium millers Uranium miners American radiologists British radiologists All ofcuDational exwsures y-rays. a-panicles y-rays. a-panicles a-panicles a-particles. ways u-particles. y-rays x-rays. 7-rays x-rays, y-rays - 3 4 6 1 0 I1 0 25 1.10 I .00 0.86 0.19 0.038 O.ooO5 0.20 0.36 0.07 - 7.91 0.28 I .@I - 12.18 0.002 0.74-7.0 1.4-9.2 3.0-13.8 0.26-23.7 042.9 0.8-3.0 0-2.9 I .4-2.9 Gilbcn and Marks16 Dolphin!' Stehney AF. Personal communication; Polednak et al." Archer et al.!! Wagoner et al.22 Matanoski GM. Personal communication Smith and Doll" myeloma actually had considerably higher total doses (34.9, 29.4, and 20.0 rem) that had been accumulated many years before myeloma was diagnosed (total duration of employment, 20, 19, and 17 years, respectively), so that a test for trend or a lonser lag period between exposure and risk period would serve only to enhance the level of statistical significance. An early report from Windscalel' also suggested a link between radiation and myeloma. This study omitted retired workers and exworkers, and in view of the long latency period and late age distribution, some of the myelomas may well have been missed. However, four cases were diagnosed in workers while they were still employed. Only one myeloma would have been expected; the observed number is significant at the 5 per cent level (P = 0.038), although it was not reported as such. Furthermore, three of the four workers had had prolonged exposure to radiation (total service, 25, 1 1 , and eight years) and appreciable cumulative doses (59, 3.3, and 4 rem, respectively). ' 4 d i u m - D i a l Painters In a registry of 3600 radium-exposed persons that was compiled at the Argonne National Laboratory, eight deaths from myeloma were found (Stehney AF. Personal communication). These cases were identified only by examination of death certificates, and it is possible that osteoporosis and bone necrosis associated with the direct skeletal effects of radium influenced the diagnosis of myelomatosis; however, all deaths occurred after 1959, so that any diagnosis would probably have been confirmed by serum paraprotein identification or bone-marrow puncture or both. None of these deaths occurred among the subgroup of 634 women in a cohort of radium-dial painters, ascertained solely from employment lists, on which Polednak et aL'* recently reported. However, six myeloma-related deaths occurred among dial painters who had worked in the industry before 1929, and five were among women, who constitute the vast majority of dial painters. The other two persons who died were a chemist and a woman who became a dial painter in 1944. Polednak et a1.I" stated that about 1400 female workers in radium-dial plants had been identified and that the United States Bureau of Labor Statistics had estimated that "not more than 2000 persons" had worked in radium-dial plants from their inception to 1929. If one assumes that all myelomarelated deaths among these workers have been recorded, that the age distribution in the cohort reported represented that in the entire work force, and that deaths from myeloma in the general population constituted 35 per cent of the deaths in the group with lymphomas and myelomas coded according to the International Classification of Diseases (ICD) as 200, 202, or 203, one would expect fewer than 0.86 deaths from myeloma among all workers in radium-dial plants. It may seem surprising that no deaths occurred in the cohort ascertained from employment records. However, if the total population of workers is assumed to be 2000, the probability that no deaths would occur in that cohort by chance alone is slightly greater than 10 per cent; thus, the finding is not highly exceptional. Uranium Miners and Millers The excess of lung cancer in uranium miners and millers has been studied in some detail. 19~20However, little information is available +bout other cancers in this group. In a study of all deaths from cancer among uranium millers, Archer et a1.*' listed one death due to myelomatosis and 1.02 deaths expected among all the hematopoietic and lymphatic diseases coded as ICD 200 to 203 and 205 (sixth revision). In 1965, myeloma accounted for 20 per cent of these deaths in American Vel. -304 NO. 4 RADIATION-INDUCED MYELOMATOSIS - CUZICK 707 men, so that 0.20 expected deaths from myeloma is the estimated figure for this cohort. An earlier report from the same research group also considered uranium miners.22No deaths from myeloma were reported among white uranium miners with five or more years uperience, whereas an estimate of 0.07 expected L._,chs from this cause was obtained in the manner outlined above from the report of 0.6 expected deaths due to leukemia, lymphoma, and myeloma (ICD 200 to 205). Radiologists Lewis5 first reported an excess of myelomatosis in American radiologists. His analysis showed five myeloma-related deaths as compared with 1.01 expected deaths. Matanoski et al.z3reaffirmed this observation in a large study of radiologists and other physicians, including the radiologists studied by Lewis, and presented combined totals for myelomas and lymphomas. Dr. Matanoski has generously supplied further details about the myelomas (personal communication). In all cohorts studied from 1920 to 1969, 11 deaths from myelomatosis occurred among radiologists; this figure can be compared with that of 4.58 expected deaths when the standard taken is the group of American white men, or with 7.91 expected deaths if the comparison is standardized internally within the cohort of all 30,132 physicians, including 6518 radiologists. (This expected number also includes deaths from mycosis fungoides [ICD 205, seventh revision], and thus it is slightly too large. In fact, two such cases were observed in addition to the 11 myelomas.) Ber - w e myeloma is more likely to be ascertained in phy- ins that in nonphysicians, the larger value is presented in Table 2. (The difference between these two expected values is partly due to the fact that the latter value arises from an internal standardization. The expected number of myelomas in radiologists is 6.8 when all other physicians are taken as the reference population. This figure may still be high, as some of the other physicians may have had higher than average exposure to radiation.) Court Brown and Doliz4reported on cancer among 1381 British radiologists. In follow-up to 1977, Smith and Doll2Sfound no myeloma-related deaths, compared with an expected figure of 1.04. This number was compiled with the correct national death rates from 1951 onward and the death rates from 1951 to 1955.for earlier periods; in view of the steady increase in reported deaths from myeloma, the figure of 1.04 is likely to be slightly too high. However, one radiologist is known to have died of myeloma less than a year after the close of the study, and another had myeloma in 1972 when he committed suicide. Both these members entered the study before 1921, when exposure%-. radiation was probably much higher. Diagnostic and Therapeutic irradiation The data accruing from diagnostic and therapeutic exposures are summarized in Table 3. Patients Receiving Thorotrast Over 4000 patients given thorium oxide (Thorotrast) as a contrast medium for angiography have been followed up in Danish, German, and Portuguese cohorts. According to Kaul and N o f f ~a, 2~5~ml injection of thorium oxide produces a continuous dose of 9 rads per year in the form of a-rays to the bone marrow. The main concern of these studies was the large excess of liver cancers. However, seven myelomas were reported in the three cohortsz'-z9(Table 4). FaberZ7reported four of these tumors in the Danish group, as compared with 0.77 expected cases. Expected numbers were not given in the other cohorts. However, since the age distribution and follow-up time in the cohorts were similar, expected numbers can be estimated (Table 4) from the cohort sizes and from differences in national death rates. Data from Spain .. I 1 Table 3. Observed and Expected Myelomas in Cohorts Receiving Diagnostic and Therapeutic Radiation. COHORT Thorotrast angiography Denmark Germany Portugal Spondylitis Bntain Germany Fluoroscop) b l g n gynecologic disorders. Connecticut h0 rddium Radium Metropathla hemorrhagica Uterine cancer, Connecticut k i c a l cancer MAJOR TIPE OF RAUIATIO~ No Ot CASES OBSERVED No (it Cut, ExPcrrEn a-particles a-particles a-particles x-rays wparticles x-rays 4 2 1 3 0 0 0.77 0.93 0.16 1.87 0.25 0.39 x-rays -,-rays x-rays >-rays. x-rays -,-rays. x-rays 3 7 5 I 2 1.10 0.81 2.09 2.92 7.19 PvAILt (Two-SlotD) 0.015 0.48 0.30 -- 0.19 0.44 0.12 0.42 0.03 90% CO'.FIDEVt I'.TEI\AL FOR Risk RATIO REFEREICE 1.8-11.9 0.4-6.8 0.3-30.0 0.44.2 0-12.0 0-7.7 Faber" van Kaick et aLZn da Motta et al.'9 Smith and Doll (ref. 31 and personal comunicacion) Spiess et al.`: BoictJ` 0.74-7.0 0.4-72 0.94-5.0 0.02-1.62 0.0~-0.81 Wagoner" WagonerJn Smith and Doll (ref. 39 and personal communication) Wagoner'" Boicc and Hutchison40 Y I THE NEW ENGLAND J O U R N A L OF M E D I C I N E Table 4. Deaths from Myeloma in Patients Given Thorotrast (Thorium Oxide). Couorr Vo. o~ PtriEws ?Jario\+L No.OF No.OF FOLLOW- M ~ E L O WEAXPECTEDO B S E R ~ E D L P D E t r H D E A T H S DEITHS D A T E Ram REFEREWE x /o-.:,v. Danish German 1026 1964 Portuguese 1052 2.95 1.86 0.59 t 0.77 0.93 0.16 4 -3176 Fabcr'' 2 9115 van Kaick et al.28 1 12/76 daMotta et al.29 *4vcragc of rata from 1968 10 1974. *Since data from Portugal were not Jvadablc. comparable data from Zaragoza. Spain. were used. were used for Portugal, since detailed Portuguese mortality data were not available. Rates represent an average of the rates from 1968 to 1974, standardized to the European population. Data were obtained from World Health Organization magnetic tapes. Similar estimates were obtained when adjustments were based on incidence data from the Danish cancer i! registry, from Hamburg and Saarland combined for Germany, and from Zaragoza, Spain, for the Portu- `r .) guese series.jOThe low death rate used for the Portu- guese series suggests underascertainment of myeloma :! in the general population of this country, and the ex- 11. pected numbers estimated in Tables 3 and 4 might % * prove to be too low in this series if the recorded causes \. .*'i of death in members of this cohort were determined I*, 16 by a more thorough investigation than is normally conducted. *..:I . Irradiation for Spondylitis Court Brown and Doll2studied a British cohort of 14,111 men and women given x-ray therapy for ankylosing spondylitis. Two cases of myelomatosis were found when the group was followed up to 1963. Further follow-up to January 1, 1970, after a single \ course of radiationJ'uncovered a total of three cases of i myeloma, whereas 1.87 would have been expected from the rates in England and Wales (Smith PG, Doll R. Personal communication). These cases occurred ---. --eight, 16, and 17 years after treatment. A smaller cohort of 612 adults treated with radium 224 for tuberculosis, ankylosing spondylitis, and a few other diseases was investigated by Spiess et a1.I2 N o cases of myeloma were manifest, and expected num- bers were not given for myeloma, but data on expected cases are available for other sites. With the expected number of intestinal and rectal cancers and the age-adjusted rates from the Hamburg cancer registry used for standardization,'O an estimate of 0.25 expected myelomas was obtained. Similar results are obtained if the estimate is derived from the expected number of cases of cancer at another common site, such as the stomach or lung. Fluoroscopy Boice" reported on a cohort of 1047 women with tuberculosis who had been monitored weekly by fluoros- _-- copy after treatment with pneumothorax. Individual exposures averaged 1.5 rads, and the mean cumda. tive dose was about 150 rads to the breast tissue. N~ myelomas were seen, whereas 0.39 would have been expected if myeloma had accounted for 24 per cent of the expected 1.6 myelomas and lymphomas that are coded as ICD 200 to 203, as it did in the Connecticut cancer registry.M Irradiation for Gynecologic Disorders Several cohorts of women irradiated for gynecolog, ic disorders have been described. An excess of myc- loid leukemia has been found in most large cohortJ given uterine irradiation for benign conditions, but surplus has been found among women given a higher dose of intense radiation for uterine cancers. SmithB has reviewed these data and discussed the explam- tion for this phenomenon, proposed earlier by Hutch- ison, that potentially leukemogenic cells are killed by intense radiation. Further support for this theory, which implies that a reduction in dose from high to in- termediate levels will not produce a proportional re- duction in incidence, was provided by Mole's analy- sis of other cancer and by the experiments of Major and Mole'' on myeloid leukemia in x-irradiat- ed mice. The available evidence on the incidence of myelo- matosis is consistent with this hypothesis of nonline- arity. Excess cases were found among patients given moderate levels of radiation for benign conditions, but not in patients irradiated for uterine neop1asms:In a study of 1893 women in Connecticut who were irradi- ated for benign gynecologic disorders from 1935 to 1964 and followed until 1967, WagoneP reported 2 myelomas (as compared with 0.88 expected cases) in patients treated with radium (mean marrow dose, 40 to 126 rads of y-rays) and three observed cases (com- pared with 1.10 expected) in patients given x-ray ther- apy (mean marrow dose, 100 to 300 rads). Smith and Dolli9 reported three cases of myeloma among 2068 women given x-irradiation (mean marrow dose, 70 to 190 rads) for metropathia hemorrhagica. Further fol- low-up to November 1 , 1979, (Smith PG. Personal communication) yielded a total of 5 myelomas, where- as 2.09 such tumors would have been expected. These cases occurred 19, 22, 22, 23, and 27 years after radio- therapy. Thus, moderate levels of radiation produced 10 cases of myeloma in these cohorts, although 4.07 would have been expected (P = 0.018). However, women given higher levels of radiation for uterine cancer had a deficit of myelomas. Wagoner" found one myeloma (as compared with 2.92 expected cases) in an irradiated cohort of 7835 women with uterine neoplasias in Connecticut (mean marrow dose, 300 to 1500 rads). In a large international sur- vey of almost 30,000 women irradiated for cervical cancer, Boice and Hutchi~on`r~eported two cases of myeloma, whereas 7.79 would have been expected. Follow-up continued for only 10years, so that the lack of a n excess is not unexpected. However, the deficien- cy of cases is of interest, since there were three cases I i i ! t .I ! 1 I I I 4 I I i i1 i iI I I I 1 1t Vol..304 No. 4 -RADIATION-INDUCED M Y E L O M A T O S I S CUZICK - ~ 209 observed and 10.71 expected (P = 0.012) in cohorts of women intensely irradiated for uterine cancers. DISCUSSION iution is needed in interpreting this analysis, since there are many possible sources of error. Numbers of expected cases were estimated indirectly, although when alternative methods were available I tended to act conservatively and choose the larger values. Any resulting inaccuracy in the expected numbers will be small and will not materially affect the overall conclusions. 4 more serious concern involves possible biases in- troduced by a more complete ascertainment of deaths due to myelomatosis in the exposed cohorts than would be found in the general population. The possibility that ascertainment of deaths from myeloma has been incomplete in the general population is suggested by improvements in diagnostic procedures over the past 40 years owing to bone-marrow puncture and electrophoresis, and that possibility is reinforced by the social-class gradient" (standardized mortality ratio for Social Class I (professional] men aged 13 to 64 years, 143) and by the sharp secular trends of the past 30 years. Thus, populations under greater surveillance for cancer may have misleadingly higher rates. This bias may be an important factor in radiation .workers, dial painters, British radiologists, and possibly Portuguese patients given Thorotrast for whom the low national death rates also suggest underascertainment in the general population. However, p 7 within the cohort of radiation workers at Han- and Windscale, persons with myeloma had higher cumulative exposures, and it seems unlikely that they would have received more thorough diagnoses than did their fellow workers. I compared American radiologists with other physicians, so that ascertainment would be similar in both groups, though possibly more complete among physicians than in the general population. Expected numbers of cases were determined internally within the cohort of atomic-bomb survivors; thus, ascertainment bias should not have occurred in that group. It is difficult to assess the degrees of bias in the other groups, although they are likely to be minimal, since the cases were not in general followed particularly intensively, and since myeloma was not an expected consequence of the exposure to radiation. Furthermore, myelomaR h t e d deaths in these cohorts were published incidentally with minimal further comments. Bias associated with the selection of favorable cohonS was minimized through attempts to include every cohort of radiation-exposed persons for which a list of observed cancers has been published. The data are summarized in Table 3. There is a clear excess of myeloma among persons exposed to radiation (53 cases observed, 32.92 expected; p = 0.0016). This conclusion is not dependent on the results from any one cohort, and it will withstand the r' Tion of any single set of data. The most striking ~ .' aaly is the deficit of myeloma observed in two Table 5. Summary of Data on Observed and Expected Myelomas According to Type of Radiation. GROUP NO.OF NO.OF PVALUE CASES CASES (TWO-SIDED) O ~ S E R V EEDXPECTED 90% CONF~DEHCE ISTERVAL FOR RISK RATIO All uterine cancers All cohorts receiving appreciable airradiation All cohorts receibing diagnostic or therapeutic -,-rays or x-rays. except uterine cancers All cohorts receiving only x-rays All cohorts except uterine cancers All cohorts 3 14 13 II 50 53 10.71 3.24 0.012 2x10-' 6.33 0.026 0.08-0.72 2.6-6.8 I .2-3.3 5.45 22.21 32.92 0.017 ?xIo-0.0016 1.1-3.3 ---1.7-2.8 --. 1.3-2.0 large cohorts of women who received intense irradiation for uterine carcinomas (three cases observed, 10.17 expected; P = 0.012). In view of the highly significant increase in myeloma in all other cohorts combined (50 cases observed, 22.21 expected; P = 2 X lo-'), the decrease in this group cannot be ignored. It remains uncertain whether the decrease is attributable to a protective effect of intense cytotoxic irradiation or to other factors, such as exclusion of patients with myeloma from the cohort, incomplete ascertainment of myeloma, or a lowered susceptibility to this disease. Nevertheless, this deficit of myeloma, the absence of an excess of leukemia in these cohorts, and the decreased leukemogenicity of radiation at high doses in animals all suggest that the carcinogenic po- tential of intense radiation is different from that of less concentrated exposure, and these factors provide a justification for treating these cohorts separately. The data from all other cohorts provide strong and internally consistent evidence for the thesis that radiation can cause myelomatosis. Although accurate estimates of the degree of marrow exposure are unavailable for many cohorts, the variations in the 90 per cent confidence intervals for the risk ratios are, for the most part, .in qualitative agreement with the likely differences in exposure. The highest risk ratios were found among persons receiving internal doses of a-emitters (14 cases observed, 3.24 expected; risk ratio, 2.6 to 6.8). In this group, which consists primarily of dial painters and patients who received Thorotrast, the effect of a small total dose of radiation may be magnified considerably, since the radionuclides involved are in fact deposited in bone tissue, so that high-linear-energy-transfer a-emission effectively irradiates the marrow. This means of exposure may also be an important factor among workers in the nuclear industry, especially in plutonium-reprocessing plants, for whom the risk ratio (seven cases observed, 2.10 expected; risk ratio, 1:6 to 6.3) was higher than one might predict from the recorded exposures to external y-rays. Some evidence (13 cases observed, 6.3 expected; P = 0.026) was also found of induction of myeloma by moderate thera- ! \ ,1 i i I t 31 210 T H E NEW ENGLAND J O U R N A L OF MEDICINE Jan. 22, 1981 peutic and diagnostic levels (70 to 300 rads to the marrow) of y-rays and x-rays, whereas the lesser average exposure likely among radiologists accords with a minimal excess risk that is not statistically significant and could well be negligible. Although these data do not permit an accurate computation of excess risk as a function of time from exposure, a review of the cases for which the time from exposure is available suggests that radiogenic myelomatosis usually does not appear until at least 10 years after exposure. It will be interesting to see whether the deficit of myeloma in patients with uterine cancer persists with longer follow-up. I am indebted to Drs. ,Matanoski and Stehney for permission to use unpublished data and to Professor Sir Richard Doll and Mr. Peter Smith for helpful criticism of earlier drafts of this paper. REFERENCES I. Ishimaru T, Otake M,lchimaru M.Dose-response relationship of neu- trons and ?-rays to leukemia incidence among atomic bomb survivors in Hiroshima and Nagasaki by type of leukemia, 195&1971. Radiat Res. 1979; 77377-94. 2. Court Brown WM, Doll R. Mortality from cancer and other causes after radiotherapy for ankylosing spondylitis. Br Med J. 1965; 21327- *. JL. 3. B.E.I.R. The effect on populations of exposure to low levels of ionizing radiation. Washington, D.C.: Nauonal Academy of Sciences, 1972. 4. UNSCEAR. Sources and effects of ionizing radiation, with annexes. New York United Nations, 1977. 5. Lewis EB. Leukemia, multiple myeloma, and aplastic anemia in American radiologists. Science. 1963; 142:1492-4. 6. Salmon SE, Wligmann M. B-cell neoplasia in man. Lancet. 1974; 2 1230-3. 7. Bnard CW, Costman J, Jaffe ES.Malignant lymphomas as tumours of the immune system. Br J Cancer. 1980 42:l-20. 8. Doll R. The age distribution of cancer: implications for models of car- cinogenesis. J R Stat Soc. [A]. 1971: 134133-66. 9. Idem. An epidemiological perspective of the biology of cancer. Cancer Res. 1978; 38~3573-83. IO. Armitage P. Statistical methods in medical research. New York: John Wiley. 1971. 11. Mantel N. Chi-square tests with one degree of freedom: extensions of the Mantel-Haenszel procedure. J Am Stat Assoc. 1963; 58690700. 12. Peto R, Pike MC. Conservatism of the approximation \'(O-E)'/E in the logrank test for survival data or tumor incidence data. Biometrics. 1973: 29579-84. 13. lchimaru M, lshimaru T. Mikami M, et al. Multiple myeloma among atomic bomb survivors and controls in Hiroshima and Nagasaki, 1950. 1976. Hiroshima, Radiation Effects Research Foundation. 1979. (Radiation Effects Research Foundation technical report no. 9-79). 14. Mancuso TF, Stewart A, Kneale G . Radiation exposures of Hanford 4_ wwkers dying from cancer and other causes. Health Phys. 1977; 33:36985. 15. Anderson TW. Radiation exposures of Hanford workers: a critique of the 5lancuso. Steuari and Kneale report. Health Phys. 1978; 3SiJ3-50. 16. Gilbert ES, Marks S. An analjsis of mortality of workers in a nuclear facilit). Radiat Res. 1979; 79: 122-48. 17. Dolphin G W . h comparison of the observed and expected cancers of the hacmatopoietic and lymphatic systems among workers at Wmdscale. London: Her Majesty's Stationery Oflice. 1976. 18. Polednak AP. Stehney AF. Rowland RE. Mortality among women fim employed before 1930 in the U.S. radium dial-painting industry. Am J Epiderniol. 1978: 107:179-95. 19. Wagoner JK. Archer VE. Lundin FE Jr. Holadaj DA. Lloyd JW. Ra- diation as the cause of lung c a n e r among uranium miners. N En@ j Med. 1965; 273181-8. 20. sevc J. Kunz E, Plafek V. Lung cancer in uranium miners and low- term exposure to radon daughter products. Health Phys. 1976; 3 0 ~ 3 3 - 7. 21. Archer VE, Wagoner JK, Lundin FE Jr. Cancer mortality among m- nium mill workers. J Occup Med. 1973; 15:11-4. 22. Wagoner JK. Archer VE, Carroll BE, Holaday DA, Lawrence PA. Cancer mortality patterns among US. uranium miners and millen, 1950 through 1962. J Natl Cancer Inst. 1964; 32787-801. 23. Matanoski GM, Wltser R, Sartwell PE, Diamond EL, Elliott EA. The current mortality rates of radiologists and other physician specialists: specific causes of death. Am 1 Epidemiol. 1975; IOl:199-210. 24. Court Brown WM. Doll R. Expectation of life and mortality from cancer among British radiologists. Br Med J. 1958; 2181-7. 25. Smith PG. Doll R. Mortality from cancer and all causes among British radiologists. Br J Radiol. (in press). 26. Kaul A, Noffz W.Tissue d o x in Thorotrast patients. Health Phys. 1978; 35:113-21. 27. Faber M. Twenty-eight years of continuous follow-up of patients injected with Thorotrast for cerebral angiography. Environ Res. 1979; 183743. 28. van Kaick G . Lorenz D. Muth H, Kaul A. Malignancies in German Thorotrast patients and estimated tissue dose. Health Phys. 1978; 35127-36. 29. da Motta LC, Horta JdS, Tavarn MH. Prospective epidemiological study of Thorotrast-exposed patients in Portugal. Environ R a . 1979; 18152-72. 30. Waterhouse J. Muir C, Correa P, Powell J, eds. Cancer incidence in five continents. Vol. 3. Lyon: International Agency for Research on Cancar, 1976. 31. Smith PG, Doll R. Age and time dependent changn in the rates of ndiation induced cancers in patients with ankylosing spondylitis follow- -ing a single course of X-ray treatment In: Late Biological E,ffccts of Ionizing Radiauon. Volume 1. Vienna: International Atomic E n m Agency: 1978:231-49, 32. Spiess H, Gerspach A, Mays CW. Soft-tissue effects following IZ`Ra in- jections into humans. Health Phys. 1978; 3561-81. 33. Bow J D Jr. Multiple chest fluoroscopies and the risk of breast cancer. In: Margimn GP. ed. Advances in medical oncology. research and #I- ucation. Vol. 1. Oxford Pergamon Press, 1979 147-56. 34. Smith PG.Leukemia and other cancers following radiation treatment of pelvic disease. Cancer. 1977; 391901-5. 35. Mole RH. Ionizing radiation as a carcinogen: practical questions and academic pursuits. Br J Radiol. 1975; 48:157-69. 36. Idem. Carcinogenesis by Thorotrast and other sources of irradiation, especially other a-emitters. Environ Res. 1979; 18:192-215. 37. Major IR, Mole RH. Myeloid leukaemia in X-ray irradiated CBA mice. Nature. 1978; 272:455-6. 38. Wagoner JK. Leukemia and other malignancies following radiothera- py for gynecological disease. [Thesis]. Boston: Harvard School of Public Health, 1970. 39. Smith PG. Doll R. Late effects of X irradiation in patients treated for metropdthia haemorrhagica. Br J Radiol. 1976; 49224-32. 40. Boice JD, Hutchison CB. Leukemia in women following radiotherapy for cervical cancer: ten-year follow-up of an international stud). 1 Natl Cancer Inst. I980 65:I 15-29. 41. Registrar General's decennial supplement: England and Wales 1961: occupational mortality tables. London: Her ma jest)'^ Stationery Office. 1971.