Document RpzbjQ6wrvpbZx9k7ojkq7RmV

! iI I I I I ! ~ , ~ , i r , , ,a~ndl nEn~t~~mdninentalMedicine 1996;53:357-360 357 d CORRESPONDENCE Riqk of myelngenous leukaemia a n d ,,,,,ltiple myeloma in workers exposed ,o benzene li.irRr--~e recent report by Wong rrccents valuable new analyses concerning bcn,ene and lymphohaematopoietic cancers Jm,,ng workers in the Pliofilm cohort.' In Gr,te of a series of analyses of this r,,pulati~n,' no previous analyses have fr,,vided specific risk estimates for acute ff,,eloid leukaemia (AML) although the cchon has been widely viewed as providing evidence most pertinent to that outcome. lhe effort to examine patterns associated ,,.lth leukaemia subtypes is clearly ,,,rt~lwhile, for the reasons presented by n`ong. ' rlowever, we would like to challenge two aqpects of this report: (a) the claim that ,`Specificity is one of the major criteria for analysis" @age 383); and (b) the Jts&on that "by lumping all cell types into a $ingle category, the misconception that benzene can increase -&e risk of other cell t ! y ~o~f leukaemia may be created." (page ?P2). Specificity was proposed by Hill some ye3rc ago as one of several considerations in cvalristing causality,' but even then with mnng caveats: "We must not, however, ,.,veremphasise the importance of the characreristic" (page 297). Subsequent experience and evolution of epidemiological methods has led to virtual abandonment of this as a r~cefulcriterion for causaliw except insofar 3s it suggests a pattern of bias in self reported exposure data or incomplete follow up for disease. Given the established multiplicity of consequences of ionising radiation, tobacco smoke, asbestos, oral conuaceprives, physical activity, and fruit and vegetable consumption, for example, it is .icriia!ly rather difficult to identify any bio- Iwicallv active exposure that is specific in its consequences. V'ith inferences from the Pliofilm cohort pertaining to leukaemias other than AML, \X`ilnc correctly asserts that the numbers of CW-S of individual cell types are so small as !(I preclude nieaningful analysis, but the niiniber of total non-AML cases (admittedI \ . a heterogeneous group) is sufficient to 3nnlyFe. To examine whether the association IT-n\-een benzene exposure and total Istuknemia observed previously in this <-ohon`is driven by AML cases, we integrat- ?LthIe results &om the two reports (table)." These data indicate that the association is :rrmger for AML than for total leukaemia, brlt the differences in association for AML, nnn-AhlI, and total leukaemia are modest. :!lco. movement of a single case from AML I o another cell type would considerably diminish this pattern and movement of two m e s would obliterate it, which serves as a reminder of just how imprecise these stanh d i s e d mortality ratios are. Wong empha- sises the distinction in dose-response patterns found for AML u total leukaemia, but the numbers of cases available from this study preclude making such subtle distinc- tions. Given these results, the claim that they point specifically towards AML as the only type of leukaemia associated with benzene exposure in this cohort is unwarranted. DAVID A SAVrrZ KURTIS W ANDREWS The Unwenity of N o d Carolina ai Chopd H a , CE 7400. McGowon-Gnmbng HaR; C8bpdHi& NC 27599-7400. USA I Wong 0. Risk of acute myeloid leukaemia md multiple myeloma in workers exposed to benzene. Occup E n v i m Med 1995;52: 380-4. 2 Infante PF, Rinsky RA, Wagoner JK,Young RJ. Leukaemia in benzene worken. Lnncn 1977;11:76-8. 3 Rinsky RA,Young RJ, Smith AB. Leukemia in benzene workers. Am J Ind Med 1981;Z: 217-45. 4 Rinsky RA, Smith AB, Homung R, Fdloon TG, Young RJ, Okun AH, Landrigan PJ. Benzene and leukemia, m epidemiologicrisk assessment. N End 3 Med 1987;316: 1044-50. 5 Paxton MB, Churchill VM, Brett SM, Rodricks JV.Leukemia risk associated with benzcne exposure in the PIiohlm cohort. I. Monality update and exposure distribution. Risk Anal 1994;14:147-54. 6 Paxton MB, Churchill VM, Bren S.U, Rodricks JV.Leukemia risk associated uith benzene exposure in the PlioHm cohon. II. Risk estimates. RUC Anal 1994j14155-7. 7 Hill AB. The environment and disease: association or causation? hceedings of the Royal Socury of Medicine 1965;58:295-300. 8 Rochman KJ. hfodern epidemwlogy. Boston: Little, Brown: 1986. Author's repl'Savitz and Andrews raised two points about my recent paper "Risk of acute myeloid leukaemia and multiple myeloma in workers exposed to benzene."' Firstly, they questioned whether specificity of disease should be included as a criterion for causation analysis. Secondly, they argued that the data from the Pliofilm cohort indicated that exposure to benzene could result in an increased risk of leukaemia cell types as WPJ! 2s acute myeloid leukaemia (AN). With regard to the tint point, S a d z and Andrews might have misunderstood what "specificity" means. Specificity of disease refers to being "distinct," and does not imply "non-multiplicity" or "exclusiveness." In my paper, I did not claim that because benzene can cause A\& therefore benzene cannot cause other types of leukaemia or other diseases. Certainly n e know that, given sufficient exposure, benzene can cause both AML and aplastic anaemia. However, both AML and aplastic anaemia are specific (or distinct) diseases. Therefore, specificity does not contradict multiplicity, as long as the diseases involved R, rdtsfor ,41\.fL,non-AAfL, and toral kukaonia in Miofilin cohon studv ` P I C . tofdra_t_h_______ I L (31 leukaemia -\ \ \ I I1 \\lL Obtenvd death 14 6 8 Expected dcathr 3 89 119 2 70 ~~ SMR (95% CI) 3 60 (1.97-6 04) 5 03 (1 84-10 97) 2 96 (I 25-5 84) are specific and meaningful diagnostic entities. To support their first point, Savia and Andrews cited a statement in Hill's 1965 paper on causation that we must not over emphasise the importance of specificity.' What Savitz and Andrews have omitted from their citation is the example given by Hill on this issue. Hill provided an example that occupational exposure to nickel can cause lung as well as nasal cancer (page 297). Again, both lung and nasal cancers are specific diagnostic entities recognised by the medical professions. Thus, Hill`s paper was hardly an endorsement for the practice of combining heterogeneous disease categories for statistical analysis. Most importantly, specificity of disease is not a statistical issue, but should be based on the biology of the disease. It makes little sense to lump different diseases into a single category for causation analysis. Before any statistical analysis, one must review and consider the underlying biological mechanisms of the disease process. As stated in my paper, it has become increasingly evident that the diseases collectively known as leukaemia are several distinct m a l i g n a n c i e ~ . ~B~ased on their letter, Savia and Andrews themselves seem to recognise the heterogeneity of various leukaemia cell types. Given this recognition, it is not clear why Savitz and Andrews would endorse combining heterogeneous leukaemia cell types for analysis. With regard to the second point, Savia and Andrews provided a risk estimate for non-Ah&. Such a statistical analysis is meaningless, as non-AML is not a recognised diagnostic entity. Although Savitz and Andrews themselves admitted that such a category was heterogeneous, they nevertheless justified such a statistical analysis on the ground that "the number of non-AML cases in aggregate (admittedly, a heterogeneous group) is sufficient to analyse." Mere sufficiency of cases does not validate an analysis. Any statistical analysis which totally disregards our current understanding of the underlying biological mechanisms is meaningless. Finally Savia and Andrews concluded: "the claim that these results (results which I provided in my paper) point specifically toward AML as the only type of leukaemia associated with benzene exposure in this cohon (the Pliofilm cohort) is unwarranted." I did not claim in my paper that the data from the Pliofilm cohort showed that benzene did not cause other types of leukaemia besides AMI.,. I simply stated the following: "For cell types other than AML., the Pliofilm study does not provide sufficient cases for any meaningful analysis. The specific cell type with the second largest number of cases in the Pliofilm study was chronic myeloid leukaemia (ChZL), consisting of only two deaths. One of the two deaths from CML was employed at the plant for one month in 1948, and died two years later in 1950 at the age of 29. His cumulative exposure was 0.10 ppm-years. Clearly this case could not have been associated with exposure at the plant." Therefore, the PlioNm study offers little useful information on the relation hctween benzene exposure and leukaemia cell types other than AML. The evidence for the lack of an association between benzene and other leukaemia cell types comes from recent laboraton. invecrientinnc' and nrh:r 358 epidemiological studies,' I' which were cited in my previous paper. Laboratory investigations have shown that benzene in vivo and hydroquinone (a benzene metabolite) in vitro alter the recruitment or stimulation of myeloid progenitor cells, thereby increasing the number of cells at risk of developing leukaemia. Furthermore, this effect is selective for myeloid cells on1y.O Similarly, epidemiological cohort and case- control studies based on analyses of specific leukaemia cell types did not detect any increased risk of other leukaemia cell types in people exposed to benzene.'-" None of the evidence from these laboratory or epidemiological investigations was con- sidered by Savitz and Andrews. In my paper! I provided an account of why different leukaemia cell types were combined for analysis in some epidemiological studies in the past and why such an analysis would not be appropriate or meaningful given recent developments in laboratory and epidemiological research on the subject. In particular, I showed that ignoring the heterogeneity of leukaemia cell types would underestimate the risk of AML at high concentrations of benzene exposure. Although these high exposures are uncommon in developed countries, in some parts of Asia and eastern Europe they have not been totally eliminated. Advocating an analysis which ignores the underlying biological mechanisms of the disease on one hand and underestimates the risk of occupational exposure on the other is a d i s s c h e to the workers. OTTO WONG ADdied Health Scrmur. 181 Sec%dAoenuc, Suiu 628 Sun Mareo, C d g m i a 94401 USA 1 Wong 0. Risk of acute myeloid leukaemia and multiple myeloma in workers exposed to be&ene. Occup b w i m n Med 1995;52: 3804. 2 Hill AB. The environment and disease: associ- ation or causation? Rocpedings 4 the Royal Society o/Medicinr 1965;58:295-300. 3 Unet MS. The leukoemiar: epidemiologic arpccr~. Oxford: Oxford University Press, 1985. 4 Heath CW. The leukaemias. In: Schottenfeld D, Frayneni JF,eds. Cancer cpidpmiorogY and preuention. Philadelphia: WB Saunders, 1982. 5 Wmtrobe MM, Lee GR, Boggs DR, Bithell TC, Forester J, Athens JW,Lcukens CIinuol hrmatdagy, 8th ed. Philadelphia: Lea and Febiger, 1981. 6 Ford AM, Molgaard HV,Greaves MF, GouId HJ. Immunoglobulin gene organization and expression in hemopoietic stem cell leukaemia. EMBO 1983;2:997-1001. htaher VE, Gill L, Townes PL, Wallace E, Savas L, Woda BA, Ansell JE.Simultaneous chronic lymphocytic leukemia and chronic myelogenous leukemia. Cancer 1993;71: 1993-7. Irons RD, Stillman WS, Colagiovanni DB, Henry VAS. Synergistic action of the benzene metabolite, hydroquinone, on myelopoietic stimulation activity of granulocyte/macrophage colony stimulatingfactor in vitro. Roc NadAcadSci 1992;893691-5. Wong 0 , Hams F, Smith TJ.Health effects of gasoline exposure. II. Mortality patterns of disuibution workers in the United States. Envimn Health Pmpect 1993;101(suppl 6 ) : 63-76. Brandt ,.I Nilsson PG, Mitelman F. Occupational exposure to petroleum products in men with acute non-lymphocytic leukaemia. BjM I978;1:553. McKinney PA, Roberts BE, O'Brien C, Bud CC, Richards IDG, Alexander ,F, et al. Chronic myeloid leukaemia in Yorkshire: a case-control study. A m Haematd 1990;83: 35-8. Ijnet his, Stewart EF, Van Natta ML, McCaffrey ID, Szklo M. Comparison of methods for determining occupational cxposure in a case-control interview study of chronic Igmphoq.tic leukaemia. J Occrip Med IOR;:?~.IW .+I 13 Mnlone KE, Koep;ell TD,Ddhg Weiss NS, Mom8 PD, Taylor p,a al. Chronic lymphocytic leukaemia in relation to chemical exposure. A m 3 Epidomor 1989;130: 1152-8. 14 Wong 0, Raabe GK. CeU-me-specific leukemia analyses in a combined cohort of more than 208 000 petroleum workers in the United States and the United Kingdom, 1937-1989. R e p i Toxicd Phannacd 1995; 21:307-21. Comment on EQM Testmate OP cholinesterase klt Editor-We read with interest the report by London et al, on the EQM Testmate OP cholinesterase kit.! We have previously reported good reproducibility of repeated measurements of haemoglobin adjusted erythrocyte cholinesterase with this kit under field conditions.? T h e Pearson correlation coefficient for duplicate measurements on the 23 samples in that study was 0.93 (in a reanalysis of data from our previous study). The correlation between duplicate erythrocyte cholinesterase measurements conducted recently on 20 blood samples, with a production model of the field kit, was 0.98 (Amaya A, unpublished data). In epidemiological studies, the kit has been found to be sensitive to subtle differences in cholinesterase activity based on exposure.' We do not offer an explanation for the poor performance of the kit, in the study of London et al, under conditions apparently similar to those under which our evaluation showed good repeatability. However, we recently have found that the temperature adjustment of cholinesterase activity by the field kit can be a source of s i g d c a n t error. Lq m e experiment, we measured haemoglobin adjusted erythrocyte cholinesterase activity according to the manufacturer's specifications, at different ambient temperatures, on blood @om the same unexposed person. A sample of blood was drawn in heparin the night before testing and stored at 4C overnight in six separate aliquots. Each aliquot was defrosted 30 minutes before testing. Figures 1 and 2 suggest that the temperature adjustment by the kit is not accurate, as measured activity on the same sample varied considerably with temperature both for plasma and for erythrocyte cholinesterase. In another experiment, we examined the rate of thermal equilibration of the kit by moving a Testmate kit from a cool room, where the kit's internal thermometer and an independent external thermometer read 8"C, to a room where the external thermometer read 40C. T h e internal thermometer of the kit (the basis for the colorimeter's temperature adjustment) registered only 30C after 120 minutes at ambi- Figure 2 Temperatureadjutted plarma cholinesterase activities as afunclion of temperature, as nteasured by the E Q M Testmare OP kit. ent temperature of 40C. Moving the kit from an air conditioned vehicle to a non-air conditioned field site has been, in our experience, a common scenario. Such rapid changes in temperature immediately before measuring cholinestersse would result in considerable error. Although we have found repeated cholinesterase measurement to be highly reliable, in contrast to the experience of London et al, we think that variable temperatures under field conditions result in significant error in the accuracy of cholinesterase measurement by the kit. This problem might be solved if the manufacturers were to provide accurate temperature adjustment factors, and if they were to measure directly the temperature of the reagent solution, rather than the temperature of the colorimeter. ARACELJ AhlAhYA Aimnamow thvunily of The Stau ofh4cttm. Tduca, Mrxiw MATTHEW KEIFER Uniwnity of ICbrhmpn, Scanh, Wwhmgwn ROB MCCONNEIJ. Pan Amenian Cenrnfor Human E a l o ~ aynd Iiealth, Pan American Health Orgonizanbn, Metepc, hfe.ncc 1 London L, Thompson ML,Sacks S, Fuller R, BaChmann OM,Myen JE. Repeatahilin.and validity of a field kit for estimation of cholinesterase in whole blood. Occup Encvmn Med 1995;52:57-64. 2 McConnell R, Ccdillo I, Keifer ht, Palomo MR. Monitoring organophosphate inwcticide exposed workers for cholincaerase depression: new technology for office or field use.J Ocmp Mcd 1992:34-7. 3 Ciesielski S, Loomis P, hlims S, Aur A. Pesticide exposure, cholinesterase dcprce- sion, and symptoms among Nonh Carolina migrant farmworkers.lournal of the Amcncan Public Health Assoclanon I994;84' 446-51. NOTICES 2nd Advanced C o u r s e in Occupationnl - and Environmental Epidernlology: Principles of Exposure Assessment. 17-22June, 1996. The Netherlands. 1105 L The Department of Epidemiology and Public Health, and Department of Air 5 k Quality, University of Wageningen, The Netherlands, in collaboration with The 09.1 20 34.2 24 20.3 15.7f'CI 0 3 6 9 12 151hl Netherland Institute for Health Sciences (NIHES) have organised this course. T h e Figure 1 Temperature-adjwted m t h n u y t e cholinesterase activities as a finctia of tmrperarure, as measured by the EQM Testmate OP F.r course is divided in plenary morning sesions and parallel afternoon sessions. The parallel sessions include two modules: module 1 ' environmental epidetninbm: module