Document reReqwby0yj0OM7qZgKmgN5Jr

I" 1 1 Acute Myeloid Leukemia in Adults: A Case-Control Study in Yorkshire R. A. Cartwright,' C. Darwin,' P. A. McKinney,' 6. Roberts,2I. D. G. Ri~hardsa,n~d C. C. Bird4 I I Leukaemia Research Fund, Centre for Clinical Epidemiology, Leeds; Depanments of ?Haematology and 'Community Medicine. k e d s General lnfirmaq, Leeds; and 4Depanment of Pathology, University of Edinburgh, U.K. This paper reports the results of a casecontrol analysis of 161 cases of acute myeloid leukemla and 310 matchedhospitalcontrols. The patients were interviewed between 1982 and 1986. The study shows a weak association for cases with previous malignant disease. Furnace workers show excess risks. Urticariaand vertigoare Inexcess, as well as some aspectsof family medical histories, including multiplesclerosisand cases of leukemlallymphoma in blood relations. S._ INTRODUCTION c EVERAL problems obscure the clear understanding of the epidemiology of the leukemias; these include difficulties of classification, particularly by cell type. Geographically based surveys attempting complete case ascertainment remain rare in this area of study. This paper examines the diagnostic subgroup of AML from a large case-control survey of leukemias and lymphomas in Yorkshire (1). METHODS New cases of A M L in people aged 15 years or older were found as a result of visits made to hematology clinics in hospitals throughout Yorkshire. excluding South Humberside. Residents of Yorkshire who were diagnosed as having AML between October 1979 and March 1986 were eligible for the study. Dianostic details of the majority of AhlL cases were ascertained from the Regional Centre where subclassification of AML was achieved by morphological (FAB) (2). conventional cytochemical (peroxidase, specific and nonspecific esterases) and immunophenotypic criteria. Immunological studies undertaken included TdT and membrane CDl, CD7. and CD19 for exclusion of lymphoblastic leukemias. Myeloid leukemias were routinely examined with a panel of monoclonal reagents including HLADr,CD13, CD14, and CD33 as previously described (3. 4). In this way fewer than 5% of the cases could not be satisfactorilycategorized, and these were excluded from the study. The small proportion of AML cases diagnosed at peripheral hospitals used morphological and cytochemical techniques and similarly excluded undifferentiated leukemias. The interviewing of AML patients only commenced in 1982. For each patient interviewed, two control patients were sought from hospitals in the health dismct of residence of the patient. Such controls were of the same sex and within three years of the age of the patient at diagnosis, and were currently in hospital for reasons other than malignant disease. A wide range of conditions was used, but accidents and "cold" surgery cases form the bulk of the controls. Hospital notes were perused for drug use and past medical history for both cases and controls. The general practitioners of the cases and Received March 3. 1988. Accepted May 9, 1988. Abbreviations: RR. relative risk: CI, confidence inierval. Correspondence to: Dr. R. A . Cartwright, Leukaemia Research Fund Centre for Clinical Epidemiology. 17 Springfield Mount, Lee& LS2 9NG (053243517). U.K. 0887-6924188/0210~687$2.a)/0 LEUKEMIA Copyright e 1988 by Williams & Wilkins LEUKEMIA. Vol2, No 10 (October),1988: pp 687-690 controls were contacted and their notes were perused for further past medical and drug ingestion information. A combination.of reponed and medically verified information forms the base for analysis. Most interviews took place on wards or in outpatient clinics. Identically structured questionnaires were used for cases and controls covering many aspects of the past medical, social, and employment history, along with details of illness in the immediate family. All the data were coded, input into the k e d s University mainframe Amdahl computer (with an anonymous format), and, following verification procedures, were analyzed using the SPSSX statisticalpackage (5). Further statisticalanalysesused the methods of Rothman and Boice (6) to obtain estimated relative risks. Cornfield's exact confidence intervals were computed for the risk ratios presented in this paper, by using the algorithm of Thomas (7). The results given are those from a pooled analysis. .._ _ . 1 RESULTS One hundred sixty-one patients and 310 controls were interviewed; 12 controls were not found in the time available for the study. During this time (October 1979 to March 1986), 529 histopathologically confirmed cases of A M L occurred in the study area: 69% were not interviewed. There was an even geo-graphic spread of non-interviewed patients, and 92% of these had died prior to an interviewer's approach. The time lapse between study start date and the beginning of AML case interviews (October 1979 to March 1982) accounts for this. However, in 1985 and 1986 special efforts were made to incorporate all cases, and here the successful interview rate was more than 6 5 8 . For the small proportion of non-interviewed live patients who were approached (8%), various reasons explained their exclusion-refusal by patient or consultant, language problems, too ill to interview. The distribution of interviewed patients by AML subtype and age at diagnosis is given in Table 1 along with the sex ratio. Resuls are given for data pooled on age and sex, and, although FAB types were defined, the numbers were too small to stratify on this variable. Risk assessment for medical vari- Table 1. Distribution of Interviewed AML Patients by Age and FAB Subtype. Group * case* <60 Years Years or Over Total (column %) AML subgroup M1 9 13 22 (14) M1/M2 13 12 25 (16) M2 14 7 21 (13) M3 8 8 16 (10) M4 20 9 29 (18) M4lM5 M5 M6 UnclassifiedC Total 0 7 1 11 83 `See text for diagnostcCntena. -1 1(<I) 8 15 (9) 1 2 (1) 19 30 (19) 78 bSexratio Male:Fernale = 1.1:l. category excludes undifferenbated leukemias. PLPUNTFPS EXMEm EE: 81 81 PL LP 92 69 ZP ez 01 PZ ZL LC LZ 01 ZL PE 81 PL 6 Le LE OE LL 26 ZP E6 OV LP 12 9 Ez5 61 82 EZ L LZ 8 29 sic cp LZ SL P 92 SL 9 Le 19 L9 LL 01 SL 9 EC 61 op 81 EL LL 5 6L 6 91 01 ZL L ELL 65 LL PZ P1 SZ 9 w 8Z L 82 EL LE ' 8 69 LL 21 2 SI 9 LZ L S L L LL L LE L8Z 09L 01 6 9P 81 L L 92 6 <s ZL E 11 S 01 S c8 \. EL L 99 ZE 01 68 5P 11 6 S2 SE 21 6 SL 8 M: PZ 7.9 et 11 c 26 6C -. Table 4. AML in Adults: Case-Control Studv-Risks Ratios Greater Than or Eoual to 2.0 or Statisticallv Sianificant - Face- ti0 0' Cases No. 01 Controls RR 954. CI Two-tailed 0 Past medical nisrory Previous malignancy Vertigo Allergy Urticaria Basal cell caronma Family history Leukemi%lympnoma Multiple sclerosis Skin cancer Infectious mononucleosis Social charaaensucs - ' .--* Smokers Occupation Electricalworkers Furnace and forge workers 11 9 40 6 6 11 7 5 5 91 13 8 4 5.6 1.6-24.5 4 4.5 1.2-20.4 109 0.6 0.4-0.95 2 6.0 1.1-60.8 6 2.0 0.5-7.5 0.004 0.02 0.02 0.04 0.38 4 5.6 1.6-24.5 0.004 5 2.8 0.7-11.2 0.14 4 2.5 0.5-12.5 0.32 1 9.9 1.1469.8 0.04...- . . 208 0.6 0.4-0.96 0.04 11 2.4 0.95-6.0 4 4.0 1.0-18.4 0.06 0.04 . .- I three female genital tract cancers. three breast cancers, one prostate, one larynx. and two mixed parotid cancers. The four controls were malignancy of bladder. bowel. cervix. and breast. It is noteworthy that the risks for past therapeutic radiotherapy are not significant (cases = 9, controls = 10, risk ratio = 1.8, 95% CI 0.7-4.4. p = 0.2). This includes some patients who had radiotherapy for nonmalignant conditions. The cases showing a past history of vertigo were all con- f m e d by perusing National Health Service records and occurred 3-25 years before diagnosis of AML. with a mean of 11 years. Four cases and one control were known to have been prescribed Stemetil: no drug treatments are recorded in the rest. The medical record of one control was incomplete. The results for skin conditions gave a single significant excess associated with past urticaria. These cases were all confirmed by medical records and were first recorded, on average. 73 years before diagnosis of AML. No common treatment pattern emerged. nor was there any FAB type of AML in excess. Table 4 also gives details of aspects of family history. The strongest association is \vith blood relatives with either lymphoma or leukemia. These are of all types. including four cases of Hodgkin's disease and four of ALL: only two cases might have been AML and neither could be properly confirmed. The multiple sclerosis family association is striking but not confined to blood relatives; of the seven cases, five were in first degree blood relations, in one case the father-in-law and sister-in-law were both affected. and in the final case a close friend of many years was affected. Skin cancer in the family achieved a pooled risk of 2.5, but this is not statistically significant. In addition, Table 4 gives some details of occupations and the cigarette smoking risks. In the latter case there is an overall significantly negative risk which is mainly accounted for by women (RR = 0.53.95'%CI 0.29-0.91. two-tailedp = 0.02). The occupational excesses in this table are all for small groups of people. The risk for furnacemen and for forge and foundry workers is 2.5 but this is not significantly in excess. The electrical workers do represent a more substantial statistical excess, with a risk ratio of 2.5 ( p = 0.03). No one clear occupational group is recorded among the electricians, although all were involved with wiring and installations, often for many years. On average the workers started their employment as electricians 31 years before diagnosis of AML. There is no preponderence of one FAB type. The excess in the chemical industry is confined to women. Of the five women, two were munitions workers, one a laboratory assistant for a dyestuff manufacturer, and one made fireworks at home. CASE-CONTROL STUDY OF AML IN ADULTS The crude risk for solvent exposure of any type (RR = I .24, 95% CI 0.79-1.95, two-tailed p = 0.35), the age-stratified male and female risk (male: RR = 1.45, 95%CI 0.84-2.57, two-tailedp = 0.18, female: RR = 0.76,95%CI 0.23-2.51, two-tailed p = 0.65) and the risks by FAB types all fail to show any statistical excess. The same is true of occupational contact with ionizing irradiation, although the crude risk is higher at 2.9; this is nonsignificant (two-tailed p = 0.2). \ DISCUSSION This study is part of the larger case-control adult survey of lymphoma/leukemias.in Yorkshire. This AML analysis differs somewhat in that the proportion of cases interviewed is the smallest of the groups analyzed so far. This is partly due to the later start in interviewing and partly due to the greater prevalence of rapid deaths. However, the patients interviewed were evenly distributed among the health authority areas in Yorkshire, although the study will inevitably be deficient in older cases and those subtypes resulting in rapid deaths. The analysis itself produced large numbers of low risks less than 7.0 which were not statistically significant. The relatively few positive results are clearly shown in the accompanying tables. The two tradiational risk factors for AML: solvent exposure and ionizing irradiation are not represented directly here. Ionizing irradiation cannot account for the excess risk among people with radiotherapy for past malignancies (ca = 5, co = 2 ) , as the situation is confused by past chemotherapy (ca = 3, co = 0). The past chemotherapy was with alkylating agents and this excess has been suggested in.case reports and surveys by several authors (8, 9). Radiotherapy for cervix cancer has not produced any excess risks of leukemia (lo), although the irradiation of ankylosing spondyIitis is linked with a leukemia excess (11). Occupational contacts with irradiation have produced little to link them with excess AML (12, 13). This is the fmt.observation of a previous history of vertigo increasing the relative risk of AML, especially in men. Further studies are necessary to indicate whether this is a true relationship. One biological explanation of this finding may be an iatrogenic element. Although no significant relative risk was found for anti-nausea drugs, it may be that our exclusion of treatment spanning less than 3 months has concealed a positive risk factor. Another expianation is a possible viral link. Although vertigo is a symptom rather than a disease, and is likely to have several causes, recent indications suggest a viral etiology for Menier's disease. - 689 - .n -n 'IZ 'OZ '61 '81 'L1 '91 'SI .-- ...