Document qa9Yn8O40n68rOakmDM6EdV65

Increasing Trends of Multiple Myeloma Mortality in England and Wales; 1950-79: Are the Changes Real? 1 - 2 ~ 3 I PLAINTIFF'S '*Ramon Velez, M.D., M.Sc.,4B5e6Valerie Beral, M.R.C.P., and Jack Cuzick, Ph.D. ABSTRACT-National statistics for England and Wales for the period ol 1950-79 were examined to study the secular trends in multiple myeloma mortality and to evaluate the extent to which these changes may be due to biases associated with improvements in medical care. Age-adjusted mortality has increased more than fivefold during this 30year period. Contrary to the impression obtained from clinical series, 1 slightly greater increase in age-adjusted mortality has occurred in males. Examination of age- and sex-specific mortality trends showed lhal the greatest apparent increase in myeloma mortality occurred in individuals over 70 years of age. The data are consistent with either a cohort effect, indicating a true increase in mortality. or an age-dependent diagnostic effect in which a greater percentage of deaths among lhe elderly are now being certified as myeloma. Analyses of mortality data by social class and region point toward improved case ascertainment being responsible for at least part of the apparent increase in reported mortality. The increase in reported mortality in the younger age groups may reflect a true increase in the incidence of this disease. This increase, however, would account for only a proportion of the total Increase. even when extrapolated to the older age groups.-JNCI r 1982; 691387-392. The apparent increase in incidence and mortality from iirultiple myeloma has become the subject of increasing interest. In the United States the reported increase in incidciice of myeloma has been proportionally greater than that fur any other site of cancer (1-3). However, a recent study of the population of Olmsted County, Minnesota, which has been under careful medical surveillance by the Mayo Clinic for several decades, has failed to show any increase in the iricidence of the disease ( 4 ) . The purpose of this study was IO determine whether a similar trend in myeloma mortality lias occurred in England and Wales and to assess the extent tu wliich these changes may be due to change in diagnostic factors. Our analyses are based largely on routinely collected, published, and unpublished vital statistics for England and Wales. Comparisons have also been made with data from a cancer registry in Texas, where secular trends are tabulated for 3 ethnic groups (5). MATERIALS AND METHODS Mortality data were obtained from the Registrar General's Statistical Review for England and Wales (6) for the period of 1950-73 and from the OPCS Mortality Statistics (7)for the period of 1974-79. Unpublished mortality data by region for England and Wales were also obtained from OI'CS. Mul~iplemyeloma has been separately classified under the IC11 rubric 203 since the sixth revision of the ICD in 1940. There have been no changes in this classification in subsequent revisions. The denominators for rate calculations for England and Walcs are based on census year populations and intercensal estimates published by the Registrar use of the World Health tion standard. The data social class were method as that used by Holman et al. (11). RESULTS Review of vital statistics.-In 1979 there were 1,602 deaths attributed to multiple myeloma (ICD 203) in England and Wales (7). This statistic represents approximately 1% of all cancer deaths and 18% of deaths from cancers of the lymphatic and hematopoietic system. Although these numbers by themselves are not particularly striking, the proportional increase in mortality from myeloma since 1950 is noteworthy. Since the case fatality rate for multiple myeloma is high, changes in mortality should reflect changes in incidence rates over time. ABBREVIATIONS USED: df=degree(s) of freedom; ICD=International Classification of Diseases; OPCS=Ofice of Population Censuses and Surveys (England); SMRastandardizcd mortality ratios(s). 'Received C-tober 27, 1981; accepted February 26, 1982. * Supported i part by the Milbank Memorial Fund and the Imperial Cancer Research Fund. 3This work was initiated by Dr. Velez while in the Department of Medical Statistics and Epidemiology, London School of Hygiene and Tropical Medicine, and in the Department of Community Medicine, St. Thomas' Hospital Medical School, London, England. ' Department of Medicine, Duke University Medical Center, Durham, N.C. 27710. Address rcprnf rcqucs~sfo Dr. Velez, FlSR & D and Clinical Epidemiology Unit, Durham Veterans Administration Medical Center, 508 Fulton St., Durham, N.C. 27705. Recipient of a Milbank Memorial Fund fellowship in clinical epide- 'miology. Department of Medical Statistics and Epidemiology, London School of Hygiene and Tropical Medicine, Keppel SI., London W C l E 7HT. *Cancer Epidemiology and Clinical Trials Unit, University of Oxford, 9 Keble Rd., Oxford, OX1 3QG, England. We thank Sir Richard Doll (Oxford University), Professor Geoffrey Rose and Mr. Peter Sniith (London School of Ilygiene and 'Tropical Medicine) for their advice and encouragement. We also thank Miss Naonii Klein for editorial assistance, Miss Carol Hermon and M n . Irene Stratton for computing assistance, and M n . Celia A. Murphy for preparation of the manuscript. JNCI, VOL. 63, NO. 2, AUGUST 1982 388 'Vekz, Beral, and Cuzick TEXT-FIGURE I.-Death rates for multiple myeloma in England and Wales, 1950-79. Age-adjusted to the European standard population. 0 9 0 0 (wr a W I- a4 I I- 4 8 0 5 2 0 2 3 2 1 0.9 0.8 0.7 0.6 C-L-4 Males Females 0.5 I 0.2 I-. . I~ 1 - , - , - , - , . , - , . , . , . , . , . , . 1950 52 52 56 58 60 62 62 66 68 70 72 72 76 78 YEAR The age-adjusted mortality trends for both sexes from 1950 to 1979 are shown in text-figure 1. The proportionate increase during this 30-year period was 515% for males and 398% for females. The sex ratio of age-adjusted mortality rates shows an excess of males which, remarkably, has increased in recent times. In 1950 the ma1e:female ratio was 1.15, whereas it was 1.26 in 1960, 1.31 in 1970, and 1.42 in 1979. Text-figure 2 illustrates the changes in age- and sexspecific multiple myeloma mortality rates for the 5-year periods 1953-57, 1963-67, and 1973-77. Although some increase has occurred in all age groups, most of the increase in myeloma mortality has -been concentrated in the older age groups. Consequently, the shape of the age-mortality curve has changed over time: In 1953-57 peak mortality was at ages 65-69 in males and 70-74 in females, but in 1973-77 the mortality rates continued to increase with advancing age for both sexes. Social class.-Text-figure 3 summarizes S M R by social class for occupied males and females for the periods of 194953, 1959-63, and 1970-72. During 1949-53 a high SMR was found among social class I males and females, while other groups were similar and no clear trend was seen acros social classes I1 through V. The same pattern was apparent during 1959-63, but the excess SMR was smaller in social class I. By the last 5-year period excess SMR had almost disappeared, and the rates among different social classes became similar. It is useful to compare these observations with the change for leukemia which occurred at an earlier time. Table 1 compares mortalities for all types of leukemias and for all cancers for occupied males by social class for the periods of 1930-32, 1949-53, 1959-63, and 1970-72. The general pattern for leukemias is similar to that for myeloma, although a clear trend is not seen for high SMR among the higher social classes (I and 11) and low SMR in the lower social classes (IV and V). These differences were most marked for the period 1930-32 and again have diminished over time. In contrast, the pattern for mortality from all cancer by social class shows a gradient in the opposite direction, with excess mortality among classes IV and V and lesser mortality among the other social classes. For all cancers, the trend with social class has become more marked in recent timesa reflection, for the most part, of the greater reduction in 3 decades in England and Wales. MALES FEMALES 1973-77 z 20. 0 4c 10: za3 5: 84 3: 0- 0 2. fK aW 1: cW K4 0.5- r 0.1 ~ c 4 0.3- % 0.2- ..0 . l J r . . . . . . , . 35- 40- 4s- so- 5s- 60- 6s- 70- 75- 80- as+ 1973-77 * 35- LO- 15- so-5s- 60- 6s- 70- 75- eo- as. AGE GROUP AGE GROUP JNCI, VOL. 69, NO. 2, AUGUST 1982 Trends in Multiple Myeloma Mortality 389 200 SUR 150 19L9-53 l lMales U Married females 100 50 n SMR 100 I ' II 111 I V 1959-63 V 50 0 150 1 SUR 100 50 I II Ill IV 1970-72 V 0 I II IllN lllM IV V SOCIAL CLASS TEXT-FIGURE 3.--Standardized mortality ratios for multiple myeloma by social class in England and Wales. smoking which has led to less lung cancer among social classes I and 11. Geographic variations.-Text-figure 4 shows the SMR for multiple myeloma for the standard regions of England and Wales (1976-78). There was some variation in SMR by region with the highest mortality in the Southwest (SMR-113) and Southeast (SMR-105) and the lowest mortality in the Northwest (SMR-88) and East Midlands (SMR=91). Cohort analysis.-Text-figure 5 portrays mortality by birth cohort. This is plotted on a log-log scale so that a power law age-incidence curve appears as a straight line. This simple type of age-incidence curve is suggested by general TABLEI.-Standardized mortality ratios by social class: occupied males in England and Wales during four 5-year periods" Period, yr Social class' I I1 IIIN IIIM IV V Leukemias 1930-32 1949-53 1959-63 1970-72 153 125 96 94 107 98 104 93 106 100 103 97 113 100 107 101 104 ALL cancers 1930-32 1949-53 1959-63 1970-72 83 92 99 102 94 86 104 95 73 80 104 102 75 80 92 113 116 Source: Occupational Mortality Tables (8,9). N = nonmanual occupations; M = manual occupations. 85 89 108 95 115 113 139 131 principles (12) and provides a good fit for cancer at a number of sites (13).T h e fit to the present data is also quite good. The shifting of the curves to the left with more recent. cohorts is what one wouid expect if successive cohorts were experiencing a greater mortality from this disease. The increases are largest between the earlier cohorts but continue until the 1905 birth cohort when the curves appear to have almost stabilized. A more mathematical analysis ( I f ) can be based on a model that includes terms to account for both cohort effects and calendar year effects (that could arise from an age-independent change in diagnostic or certification practices). Such a model attributes the bulk of the increase to a cohort effect. A model that corrects for age and birth cohort fits the data extremely well (x2for error -42.07 on 30 df), and no further improvement is obtained if calender year terms are included (x2for error -38.65 on 27 df). A model that includes only age and calendar year terms does not fit so well h2for error =561.8 on 40 df), and a much better fit is obtained when the cohort terms are included h2for error =38.G5 on 27 df). However, the calendar year variables do not accommodate an age-dependent improvement in diagnosis and certification (which could masquerade as a cohort effect), and this likely possibility must be considered in the assessment of the validity of this model. This model assumes that the logarithm of the expected mortality rates in each 5-year calendar period and 5-year age group is a linear combination of indicator func- 95 -104 SMR for England and Wales = 100 TEXT-FIGURE 4.--Standardized mortality ratios for multiple myeloma by region io England and Wales. Males and females aged 2 2 5 yr, 1976-78. JNCI, VOL. 69, NO. 2, AUGUST 198' -- 3,.Velez; Beral, and Cuzick tions for 5-year age groups, 5-year birth cohorts, and 5-year calendar periods, and that the variations from the expected rates are independent Poisson random variables. A more AGE AT DEATH 1 log. scale I TEXT-FIGURE 5.-Mortality from multiple myeloma in England and Wales in different cohorts born in 5-yr periods between 1880 and 1915. 7 61 5- z 0 L- <I- J a13 3- 2 *e 0 0, 00 *I0 0' s0 2 - 0 a aW 0 o b 0 / 2 4 I- K 0.9- W Vzw 0.80.7- 0' 00 ,/' e. 0 0 M0 I A o'2] A .O . l J , . , , , . , , , , , , , , . , 1950 52 54 56 58 6 0 62 6 L 66 YEAR T -EXT FIGURE 6.-Age-adjusted multiple myeloma incidence rates by ethnic *--.,group in Texas, 1950-1966. 1960 U.S.population used as standard. m,whites; non-whites; &----A, Spanish surnamed. JNCI, VOL. 69, NO. 2, AUGUST 1982 detailed discussion of this model based on data from a number of countries will appear elsewhere. Texas cancer reg.istly.-Shown in text-figure 6 are the ageadjusted multiple myeloma incidence rates for white, Spanish, and nonwhite populations living in a defined geographic area in Texas. The slopes of the rise of the regression lines of logarithm of incidence versus calendar year are steeper for Spanish and nonwhite populations compared to the white population. DISCUSSION The clinical presentation of multiple myeloma varies widely, with initial symptoms ranging from vague migratory bone pain, back pain, pathologic fractures, pneumonia, neurologic symptoms, and/or anemia to acute renal failure (14, 15). Diagnosis requires the aid of various tests, which have changed in their type and availability in the last 50 years. The first diagnostic test was the examination of urine for Bence Jones protein (16). Unfortunately, this test is positive in less than 50% of proven cases of this disease (17, 18).Thus other ancillary tests, such as bone marrow exam- ination, skeletal X-rays, serum, and urine electrophoresis must be employed when this diagnosis is suspected. Beforc these procedures became widely available, the diagnosis of myeloma was often confirmed only at autopsy (19). The gradual introduction of these diagnostic tests has spanned nearly 4 decades. Although bone marow exarnina- tions were first performed in 1929 ( 2 4 , it was not until 1938 that the value of this procedure was recognized in the diagnosis of myeloma (21).It remained an infrequently used procedure, however, until the late 1940's, when a greater understanding of the cell types in normal marrow made the identification of abnormalities easier (22). Needle biopsies, that could be taken easily under local anesthetic, as opposed ' to the more diflicult and painful trephine cuts used previ- ously, became available in the 1950's and early 1960's and also led to the wider use of bone marrow investigation (23, 24). Similarly, serum electrophoresis was first described as being of diagnostic value in myeloma in 1938(25).However, the method was laborious and remained a research tool until the 1950's (26-28), when, with the development of paper and cellulose acetate electrophoresis, this procedure became more widely available. We inquired at 10 London , -teaching hospitals to establish the time when serum electro~ phoresis became available. The responses gave the average year when this occurred as 1956, with a range from 1950to 1962. fA diagnosis of multiple myeloma is highly dependent an 4the use of medical facilities and laboratory tests. The effects of these factors must be taken into consideration when one ;* interprets mortality statistics (29).The sharpest increase in reported mortality rates occurred between 1950 and 19rfl (text-fig. 1). This increase coincides with the period of the most rapid advances in the laboratory diagnosis of this `1 disease. 1The age-adjusted male-to-female mortality ratio reveals a slight excess of males in the 1950's that has incrcascd in the next 2 decades. This finding is in striking contrast to the 1iimpression gained from clinical series in which a largcr male P" (1' thl te: ha ha lai th. N: m: di. ch thl no hii an of in Stl ([e act ' P" CI I9 un Trends in Multiple Myeloma Mortality 391 preponderance was seen in the early part of this century (19) and is less apparent now. In 1961 Martin (30)reported die male-to-female ratio of cases of myeloma seen in London teaching hospitals as 3:2. More recently, Waldenstrom (31) has pointed out that in his recent experience the sex ratio has.shifted toward a preponderance of females. The early large male excess is possibly partly an artifact due to a bias hat used to exist (particularly before the introduction of the National Health Service) in favor of hospital referral of iiiales rather than females for the investigation of obscure diseases (32),and its disappearance may be partly due to changes in referral patterns and in the sex distribution of die aged population. A greater percentage of females are now alive in the age groups in which myeloma incidence is highest, and the early case series were based on numbers and not rates. For example, in 1979 64% of the population of England and Wales above age 70 were female, whereas ill 1950 only 59%were female. This bias can be removed by rtudying age-adjusted rates instead of cases for the two sexes (text-fig. l), where the size of the normal population is accounted for. Kyle et al. (33) examined the death certification of 31 patients in whom myeloma had been diagnosed at the Mayo Cliiiic in Olmsted County, Minnesota, during the period of 1945-64. He found that myeloma had been recorded as the underlying cause of death in 23 cases (74%). By contrast, in wdyina the recorded cause of deaths in a sample of incident from the Third National Cancer SurLey, 1970-71, et al. (34)found that 96% of the 699 incident cases of ma were certified as dying of this disease. This suggests n increased awareness by clinicians of multiple myea, associated with changes in diagnostic habit and death contribute to the apparent increase in acMahon (35) has previously commented on the mortality from myeloma in social class I in A trend in the same direction has also en for the leukemias. The proportionate increase in t y from leukemias earlier this century has been exonly by greater increases in lung cancer and coronary rtality (36).Although radiation exposure ted as a causative factor in both leukemia and myeloma (38),the increase in leukemia has been buted primarily, although not exclusively (37),to imagnosis and better access to medical care which may well be true also for myeloma. rther support for the hypothesis that underascertainis related to social class and access to medical care is examination of the data from a cancer ( 5 ) , where cancer incidence in white, sh, and non-white populations has been analyzed. the Spanish and the non-white group represent socio- antaged populations. Over time both of benefited from greater access to medical ease in incidence of myeloma has been more pronounced in these 2 ethnic groups (text-fig. 6). Studies of leukemia incidence have indicated that there k a consistent selective underascertainment among the el- derly (39-41). Meadors (40) has suggested that this is the mult of attributing leukemia deaths in the elderly to dis- eases of old age and infections. If correct diagnosis is a factor in myeloma mortality rates, as laboratory investigations increase, the greatest increases in registered mortality would be among the elderly where underdiagnosis is most prevalent. It is apparent from text-figure 2 that while myeloma mortality rates have increased in all age groups, the change is far greater in the older age groups. Regional mortality data for England and Wales also support the view that access to medical care may be important in the diagnosis of myeloma. Mortality from all causes and all cancers is highest in regions in the North and lowest in the Southeast (42), but the reverse is true for myeloma (text-fig. 4). This observation is consonant with the disproportionately greater availability of medical facilities in the South (43). In conclusinn, I: aijpears that technical improvements in the diagnosis of myeloma, as well as greater access to medical care especially among the elderly, have contributed to the increased recognition, and hence to the reported mortality, from this disease. Similar arguments were put forward to explain the reported increase in lung cancer earlier this century (44, 4 3 , yet the increase in lung cancer was real. In the case of lung cancer, the increases were first noted in males, suggesting a specific etiologic agent much of which was later identified as cigarette smoking (46).For myeloma, the increase occurred in both sexes and has been only slightly greater in males, suggesting that any major new putative factor must have been introduced across broad sections of the population. The extremely good fit of the data to a model with only cohort effects also suggested that the increase may be at least partly real. Although much of the apparent increase in incidence of myeloma may be due to changes in diagnostic practice and referral, myeloma is certainly not as rare as it once was thought to be. Our knowledge of the etiology of this disease is meager and requires further investigation. REFERENCES ( I ) BLATTNERWA. Epidemiology of multiple myeloma and related plasma cell disorders:An analytic review. In: Potter M, ed. Progress in myeloma. New York: Elsevier/North- lolla and, 1980:1-65. (2) DEVESSAS, SILVERMAN DT. Cancer incidence and mortality trends in the United States: 1935-74. J Natl Cancer Inst 1978; 60:545-571. (3) FRAUMEJNFIJR.Environmental and genetic determinants of cancer. J Environ Pathol Toxicol 1977; 1:19-30. ( 4 ) LINOSA, KYLE RA, O'FALLONWM, KURLANLLTY. Incidence and secular trend of multiple myeloma in Olmsted County, Minnesota: 1965-77. JNCI 1981; 66~17-20. (5) MACDONALELJY, HEINZEEB. Epidemiology of cancer in Texas: Incidence analyzed by type, ethnic group, and geographic location. New York: Raven Press, 1978. (6) Registrar General. Statistical review of England and Wales 19501973. London: Her Majesty's Stat Off, 1952-1975. [Separate annual publications]. (7) Oflice of Population Censuses and Surveys. Mortality statistics: Cause 1974-79 (series DH2 No. I). London: Her Majesty's Stat Off, 19771981. [Separate annual publications]. (8) Registrar General. Decennial supplement. England and Wales 193I , Occupational mortality tables. London: Her Majesty's Stat Off, 1938. (9) Registrar General. Decennial supplement. England and Wales I95 I and 1961. Occupational mortality tables. London: Her Majesty's Stat Off, 1957 and 1971. JNCI, VOL. 63, NO. 2, AUGUST 1982 392 We!ez, Beral, and Cuzick (10) Ofice of Population Censuses and Surveys. Decennial supplement. England and Wales 1970-72 (series DS No. I). Microfiche. Iandon: Her Majesty's Stat ON, 1978. ( 1 1 ) HOLMACND, JAMES IR, GATEY PH. ARMSTRONBCK. An analysis of trends in mortality from malignant melanoma of the skin in Australia. Int J Cancer 1980; 26:703-709. (12) ARMITACPE, DOLL R. Stochastic models for carcinogenesis. In: Neyman J, ed. Proceedings of the Fourth Berkeley symposium on mathematical statistics and probability. Berkeley: Univ Calif Press, 1961:19-38. (13) DOLL R. T h e age distribution of cancer: Implications for models of carcinogenesis. J R Statist Soc 1971; A, 134:133-166. (14) SCHWARSTOZ, CATALDMO. Some clinical caprica of multiple myeloma: 15 case reports from a study of 75 cases. Ann Intern Med 1953; 39: 1267-80. (15) DOLL DC, WEIS RB. Unusual presentation of multiple myeloma. Postgrad Med 1977; 61:116-121. (16) JONES HB. On a new substance occurring in the urine of a patient with mollities ossium. Philos Trans R Soc London (Biolj 1948; 138:55-62. (17) ADAMWS S, ALLINC EL, LAWRENCE JS. Multiple myeloma: Its clinical and laboratory diagnosis with emphasis on electrophoretic abnormalities. Am J Med 1949; 6:141-161. (18) HOBBSJR. Immunoglobulins in clinical chemistry. Adv Clin Chem I97 1; l4:2 19-3 17. (19) GESCHICKTECRF, COPELANDMM. Multiple myeloma. Arch Surg 1928; 16:807-863. (20) SNAPPER I, TURNER LB, Moscovrrz HL. Multiple myeloma. New York: Grune, 1953. (21) ROSENTHAL N, VOCEL P. Value of the sternal puncture in the diagnosis of multiple myeloma. J Mt Sinai Hosp 1938; 4:1001-1019. (22) VAUCHAN SL, BROCKMYRF.E Normal bone marrow as obtained by sternal puncture. Blood 1947; 1:54-59. (23) BRODY JI, FINCH SC. Bone marrow needle biopsy. Am J Med Sci 1959; 238:140-145. (24) ELLIS LD, JENSEN WN, WESTERMAMNP. Needle biopsy of bone and marrow. Arch Intern Med 1964; I14:213-221. (25) LONGSWORTLGH, SHEDLOVSTK, YMACINNEDS A. Electrophoretic patterns of normal and pathological human blood serum and plasma. J Exp Med 1938; 70:399-413. (26) WALDENSTROM J. Abnormal proteins in multiple myeloma. In: Advances in internal medicine. Vol 5. Chicago: Year Book Med Publ, 1952398-440. (27) RUNDLES RW, COOPER GR, WILLETTRW. Multiple myeloma. IV. Abnormal serum components and Bence Jones protein. J. Clin Invest 1951; 30:1125-1142. (28) REINERM, STERN KG. Application of electrophoretic studics on ihe distribution in the serum of multiple myeloma patients. Acta Hemato1 1953; 9: 19-29. (29) NEWTON PA. Multiple myeloma: Uncommon or uncommonly diaR nosed. Postgrad Med 1977, 61:107-1 IO. (30) MARTIN NH. T h e incidence of myelomatosis. Lancet 1961; 1237239. (31) WALDENSTRJO.MDiagnosis and treatment of multiple myeloma. New -York: Grune, 1970:3. (32) . Malignancy of the hematopoietic system and ageing. In: Engel A, Larnon T, e&. Thule international symposium No. 2. Cancer and ageing, 1967. Stockholm: Nordiska Bokhandelna Forlag, I 968: 155-166. (33) KYLE RA, NonREcn IT,KURLANLDT. Multiple myeloma in Olmctrd County, Minnesota, 1945-1969. Blood 1969; 33:739-745. (34) PERCY C, STANEK E, GLOECKLELR. Accuracy of cancer death cenifi- c a t a and its effects on cancer mortality statistics. Am J Public Health 1981; 71:242-250. (35) MACMAHOBN. Epidemiology of Hodgkin's disease. Cancer Res 19%; 26: 1189-1200. (36) O f i c e of Health Economics. Leukemia: Towards control. London: Office of Health Economics, 1980:1-47. (37) COURT BROWNWM, DOLL R. Leukaemia in childhood and yoirnd adult life: Trends in mortality in relation to etiology. Br Med J 1961; 2:981-988. (38) CUZICKJ. Radiation-induced myelomatosis. N Engl J Med I9RI; 304:204-210. (39)MCPHEDRANP, I ~ E A L T HC W JR, GARCIA JS. Racial variation in leukemia: IncidenFe among the elderly. J Natl Cancer lnst 1970; 45:25-28. (40) MEADORGSF. Epidemiology of leukemia. Public Health Rep 1956; 7I :103- 108. (41) HEWITTD. Geographical pathology of leukemia in England arid Wales. Acta Un Int Cancr 1960; 16:1643-1647. (42) O f i c e of Population Censuses and Surveys. Area mortality decennial supplement. England and Wales 1969-1973 (series DS No. I). London: Her Majesty's Stat ON, 1981. (43) WEST PA. T h e nation's health and the NfIS. King's fund project paper No. RC14. London: King's Fund Centre, 1980:32-34. (44) GILLIAAN. Trends of mortality attributed to carcinoma of the lung: Possible eNccts of faulty certification of deaths to other respiratory disease$. Cancer 1949; 3:3 I 1-320. (45) HEADY JA, KENVAWAEYL. The increase in deaths attributed to cancer of the lung. Br J Cancer 1949; 3:311-320. (46) DOLL R, H ILL AB. A study of the aetiology of carcinoma of the lung. Br Mcd J 1952; 2:1271-1286. JNCI, VOL. 69, NO. 2, AUGUST 1982