Document B8knY2emqjMqn7oVB7zmNw8ro

AMERICAN 012766 Journal of Epidemiology Reviews and Commentary BENZENE AND LEUKEMIA A REVlEW OF THE I.ITERt1TURE AND A RISK ASSESSMENT' H A R W N D AUSTIN. ELIZABETH D W E L L . AND PHILIP COLE T h e major purpose of this paper is to vicinity of the present occtipational stand- review the evidence. primarily epidemio- ard of 10 ppm. logic. prtnining to the relnt.ian Irctwccn k n t e n e and leukemia. Uenzene ia widely BACK(; HOUNu considered to be a leukemogen for human In 1981, a Working Croup of t h t Inter- beings, but i t s potency is uncertain. Since national Agency for Research on Cancer low-lcvd cxpmrtirw t o Iwnzrnc chrlrnctcrizc cnnc.liirlccl I lint I.licvc wits si1flic-icnl. cvi- mutiy twi-trprirrc~nesl eLtitrg~i,l is important Jence thnt benzene is carcinogenic to hu- to dcterminc to what. extent., if at. all, law mnng ( 1 1. Spwilicnlly. t.licy mndirdcd t.hiit. cxp)niirc*n iric*rcrt.w Icokcmiii risk. Mow- crposure lo hnzcne muy damage the he- over. as w r write. the US Ocmpationd mntnpiet.ic sv9t.t-m nnrl Ilid.wirlcncc from S~I[I.yI m i t i I Iidf 11 A t l i i t i i i i w f rnl itw i w 111 t . l ' i t l ~ . i c i i c , l ~ , # i t . -dcitlic*ri i * w t r i l i l i r t J i c * m i e I W I I W ~ I ~ 1ernpr.ing to reduce workpluce exposures relotion hetweeri lwnzenc nnd m i t e mye- from n i t i*it:I\t Iicwr I iiitc*.wci&c*ci ~ ~ n g tIt.rvtic. Ii*iik(wi1t. I II t.h*!itiiitv rqwri.. ii.WIIW lcvcl ot 10 ptirts per millirm Ippm) to 1 noted that t h e evidence linking hnimne ppm. This paper also includes an evaiua- with cancer other than acute myelocytic tion and critique of various risk assess- leukemia was inadequate. Benzene is in- ments of hentene and leukemia. These risk cluded on the list of carcinogens published assessment; were done for policymaking by the Secretary of the US Department of arid were developed to predict the leuke- Heaith and Human Services 121 and is des- mogenic effect of benzene exposures in the ignated as a suspect carcinoEen by the American Confcrcnce of Governmental In- dustrial Hygienists 13). The present Occupational Safety and Health Administration regulation for ex- posure to benzene was established in 1971 and revised in 1974 14). The rcguintion limits expoaure to un eigIit.-l:i~ur time- weighted average of 10 ppm with a ceiling concentration of 2.5 ppm over a 10-minute period and a peak concentration not lo exceed 50 ppm for more than 10 minutes. '419 This regulation was established because of chromosomd studies is pertinent, since h l n a n e was known to Ln! H h n e marrow these have also conuibuted to the eval~rr- depressant and because of its acuce toxicity; tion of the leukemogenicity of benzene. bn'zene was nor generally recognized as a carcinogen at that time. Indeed. in 1974. A n i d stdks the National Institute for Occupt'iond T h e carcinogenicity of benzene has been Safety and Health t w t d that. although assessed experimentally in rats and mice case reports, chromosomal srudies. and sev- using several exposure methods, such as eral epidemiologic studies suggested il link skin application. inhalation, intragastric, between benzene and leukemia. the evi- and subcutaneous administration (11. The dence was not suilicient to conclude that it International Agency for Research on Can- was a carcinogen (5). However, in 1976, the cer reviewed these and other studies of National Institute for Pccupational Safety animals in 1982 and concluded that 'there and Health recommended that benzene be is limited evidence that benzene is carcin- regulated as a carcinogen because of new ogenic in experimental animals" (1). Addi- epidemiologic evidence (6). In that year, tional studies now have been reported, and the Occupational Safety and Health Ad- they strengthen the hypothesis that ben- ministration issued an emergency tempo- zene is a n animal carcinogen which acts on rary standard for occupational exposure to several organ systems(9-12). However, this benzene which stipulated an eight-hour work has limitations which afect its rele- time-weighted average of 1 pprn with a vance to the evaluation of the relation be- ceiling level of 5 ppm for any 15-minute tween benzene and leukemia in human period. This became the Occupational beings. None of tne experimental work Safety and Hedth Administration stan- pruvides an adequate animd mudel for drrrd in 1978 (7). However, a number of benzene-induced myeiocytic leukemia. Al- organizations representing industrial con- though the studies have demonstrated the cerns obtained an injunction against the induction of nonmalignant hematologic new standard. In 1980, the Supreme Court disorders by benzene, the carcinogenic ef- struck the 1 ppm standard, stating in es- fects are confined largely to nonhcmato- sence that the Occupational Safety and logic tissues. In rafs and mice, for example, Henlth Administration had not demon- benzene appears to be strongly related to strdted that the 10 ppm standard was un- the Occurrence of carcinoma of the zymbal sate ( 8 ) .Thus.a controversy remains about gland a structure with no human analog. the reasonable exposure limit in occupa- An additional limitation of the animalstud- tional settings and especially about the leu- ies is the lack of data on the potential kemogenic effect of low-level exposures. carcinogenicity of benzene exposures below SCIENTIFIC EVIDENCE 100 ppm. The potential adverse health effects of Chromosome studies 1,cnzrnc have I)CC:I evuluuted in cxpcrimcn - Despite several studies. benzene has not tal studies of animals, in investigations of been found to be a genotoxic (DNA- chromosomes, and in epidemiologic studies renctive) agent (13). However, it causes of workers exposed to benzene. Here, we chromosome breaks and other chromo- emphasize the epidemiologic studies he- somal changes in animals, and it has been cause they have provided both the qunlitative and the quantitative estimates of the leukemogenic effect of benzene on which risk assessnients and regulations largely suggested that an excess of chroinosomal aberrations occurs among human beings exposed to benzene ( 1.11). SLudies oCworkers with benzene-induced blood disorders have been bilsed. Nevertheless. '1 brief sum- have coiwistcnLly shown an increased previiiary of the reaults ol' anirniii atudiea and alence ul' chromosomal aberrations in so- mmic reils. Hnwever. stiidies of warken 2 0 to J(Nl ppm fnr wnrkers in the mtngra- w h c i Ii;ii{ tow (,"T, IifwtI Iwir~.c*iic-P X ~ M V W I I ~ * - wire- pliiiii-v. Vicliriiii c*si i i i i t i t t - 1 1 f lid IIN* in ttr what hati no.nven signs of'chronic ben- citience rate of acute leukemia among work- zene poisonine have heen inconsistent with ers heaviiy exposed to benzene was 20 times respect to cytogenetic findings. In addition. manv nf rhcsc srirdics are qiiite small. ;ind r i i i w t I1;IV.C. iiwi tvHItihIcii*iil civIic*it*rit-ic~?c.i l l . clucling inidecptalc assessment of average higher than that of the general population. hilt this estimntp is nnt ndcquntelv docrr- 111~*11cI v l . ' h e epidemiologic st\idieso f 1)enzcncand or cumulative benzene exposure. faiiure to leukemia are summarized in table 1. They adjust for age differences between groups. are reviewed briefly below. and no dose-response evaluations. More- In a hospital-based case-control study in over, there are no data that establish a France, Girard and Rev01 (14) evaluated relation between chromosomal aberrations the association between benzene and leu- and the subsequent development of leuke- kemia and other hematologic diseases. The mia nr nt her disease. cases were hematology service patients, of Epidemiologic studies whom 140 had acute leucosis (aim referrect to as acute leukemia in the study report), Elcvcn cpidemioiogic atudies hnvc evnl- 61 had chronic lymphocytic leukemia, and imted t.he assmiatinn hetween henzene and S6 had myelpid leukemia. T h e n .were also l i ~ i ~ k ~ ~f iI i.Ir i9~ i1'I.'l~c!.uc* d.iulic?twe*n* 1,rc.- 12.1 c w i . r i d n Iic*cpiialiwtl for nonlic*nintc,- ceded by many case reports of leukemia logic conditions. Exposure to benzene and pnt icn1.u wii 11 intiirst r i d Irnzcnc! exposrim tdticnc wnR rletcrmind h y qticstkmingsiih- nnd hy dcscriptions of rreries of leukemia jech HIWcUhe~micals which they hnd iiacd cases associated with benzene exposure. during the 10-year period preceding the two of which have been updated by Aksoy hospitalization. Relative risks for benzene (29)in Turkey and by V i ~ a n(i30)in Italy. or toluene were 3.3 (95,per cent confidence Akaay (291 studied 51 leukemia cases. interval (CI)= 1.2-8.9) for acute leucosis/ identified in 1967-19&.3. among men occu- leukemia. 4.1 f I A - i Z ) for chronic lympho- pationally exposed to benzene in Istanbul. cytic leukemia. and L.8 (O.S-6.6) for mye- Turkey, and he estimated that from 1967 loid leukemia. This study sugge~taa ltrong to 1975 the crude incidence rateof leukemia p i t i v e relation between moat forms of was 13 per 100,OOO penon-years among leukemiaand b e n a n e or toluene. However, shoe workers, compared with 8 u per it does not distinguish between benzene 100,OOO person-years among men in the and toluene, and it provides no estimate of -general population. However, the vaiidity dose response. of t.hese estimates and the meaninEfulness Iahimaru et ai. ( 1.51 cnnducted a stadv nf of the comparison of the leukemia rates of 303 leukemia cases and 303 controls in nhoe workers with those of the general pnp- Nagasaki and Hiroshima. Japan. The study d a t i o n ;ire duuiitfui k a m e ascertainment evnluated cxxttputionnl expoeurcs to bcn- o f leukemia cases was incomplete and was zene and to medical a-ravs: potentiai ex- not cnnducted in the same manner for the posure was determined on the Imis of sub- two groups. jects' occupations. The relative risk of leu- Vigiiani r301 in Italy reported 11 leuke- kemia was 2.5 (95 per cent CI = 1.34.0) mia cases that occurred hetween 1942 and for persons with occupations involving po- 1974 in Milan and 13 cases that occurred tential benzene or medical x-ray exposure between 19.59 and 1971 in Pavia among cnmpared with those with no such expo- workers in shoe 3nd rotogravure factories. sures. Renzcne ex;msiires were thought to have The main limitation of this study is its ranced from 26 to 6(w) ppm for workers use occupnrion as a surrogate for benzene handlinq %luesin 3hoe factories and from exposure. Each occupation classified as in- C -- z2C. zc. . t-. 0 t -i 5 0 $Y ? 2 3i -. vulving p ~ ~ ~ . i i t ibuiilrciie c x p a u r e uruy i~mliiiiurrrryliiidiiigy of u n.tri)s~x!cLivclid- have included workers with no exposure to low-up study of 1.06 workers employed in ' benzene, as well tis workers with exposure the manufacture of rubber hydrochloride to chemicals other than benzene. In sddi- (trade name, Pliofilm) at three plants in tion. the relative risks were not reported two Ohio locaiions. Rinsky et al. (18) pre- separately for benzene and medical x-ray sented additional results from this study in exposures, and no attempt was made to 1981.XI1 workers with at least one day of quantify leukemia risk dccording to inten- exposure m benzene during 1940 through 3 ~ t yor durtlrion 0 1 benzene exposure. 1959 were included, although most of the Thorpe t 161 investigated leukemia inci- results pertained to the 748 men who were '. dence and mortality among 38.000 active workers and annuitants at eight European first exposed between 1940and 1949.These men were foilowed through mid-1975.The affiliates of a large oil company. The obser- investigators used historical benzene ex- - !I vation period was from 1962 to 1971. Em- p u r e monitoring data to describe expo- :!; pioyees were ciassified either as having s u n conditiona at the two'hcations.~Alb\ been exposed to benzene (potential expo- though-these data were sparse, aspdally sure for at least five y e L to refinery for location 2. the inveatigatiii beliwici stremns or petrolcum products containing that exposure tended to be btioo; the reC- at least 1 per cent benzene) or not (no dr ommended limits. 'z only occasional benzene exposure). The as- The mortality experience of the benzene- certainment both of benzene exposure and exposed workers was compared with that 1 of leukemia occurrence was conducted sep- of US white men, unexposed rubber hydro- arately by each of the affiliates. The ex- chloride workers at the study plants (n = pected number of leukemia deaths was de- 398).and white d e fibrous glass workers rived from the general population morrality in Ohio in = 1,447). The standardized mor- rate8 of the countries in which the affbtes talicy mtio for leukemia amongthe748men - were locaced. For workers potentially ex- who were first exposed between 1940 and posed LO benzene, the sundardized mortai- 1949, compared with that of rhe genera1 icy ratio was 121 (95 per cent CI = 37-2051, population, was 560 (95 per cent C1= 225- whereas the scandilrdized mortality ratio 1.153). based on seven observed deaths. was 60 for the unexposed. Using the fibrous glass workers as the re- This study has a number of limitations. ferent. the standardized mortality ratio was Ascertainment of leukemia cases and doc- 473. There were no deaths from leukemia umentation of exposure were inadequate. among the unexposed rubber hydtochlori+ In addition. the report indicates that there workers: the expected number was not re- was considerable uncertainty about the age ported but was probably small. structure of the overall cohort. For these Five of the seven leukemia de& oc- reasons, the validity ol the reported stand- nund among men with five or more years ardized mortality ratios is questionable. of benzene exposure (standardized mortal- The standardized mortality ratio of 121 for ity ratio = 2,100). All of the Ieukemias were leukemia among potentially exposed work- of the myelocyfic (acute, 4; chronic, 1) or vrs in i ~ ~ i r c ~ i ~ i r i t k i Hi lo~wl tc~w. er, the study is monocytic (acute. 1; unspecified, 1) cell iielLlicr pcruuusivcly pusdive iiur ilcgulivc L y p . Aiiioiig I I E I ~ wiio Iiud rtrnt. IWWI'IX Im:ause of the lnck of an analysis of leuke- posed to benzene between 1950 and 1959. Ic.~IniLi tiicirtdit y l ~ yi i i c l t r c - i i c u l Iic*ricd. 1111. 1 l W W~ O I I l .~I L W I~I 1 . n I~U ~ C ~ C I ~ C ~ ~ I C I ~ U C unexplained deticit of deaths from leuke- kemra compared wicb 0.5 expecred. mia among the unexposed and the lack of Rinsky et 31. (19) have recently updated documentation of the methods used to this investigauon. The updated study in- iaicttlif'y I c ~ i k i * i i i ic~li.iuL lis : i i i c l cIc*it*rttrittt- c ~ l ~ l t lI w, I I G 111C'11 WllCI It;wi Iwcn CX~lCI?CcdIo benzene exposure. Lwiizeiie lur a~ Ira1 uiie day durllig IWU- In 1977. lnlante et ai. (171 reported the 1965 and extends follow-up through 1981. . .. .- REN7ZNE AND LEUKEMIA T h e q d t i t e also iraea n jnh-exposure linkage expected. Another deceased worker hnd prrwdirrc m e h r n n t r the crimiilativc hcn- myelomnncxcytic Icirkcmin, hut. this wns no( zene exposure CIL'I he mcn in the study. The certified as the underlying cause of death leukemia skndardized mortality ratio for and was not included in the mortality the overall cohort compared with the gen- analysis. The four leuke'niias were of the eral population was 337, based on nine ob- myelocytic cell type. Ott et al. estimated an .sewed and 2.7 expected leukemia deaths. incidence rate ratio of 4.4 (95 per cent T h e findings of the updated study are dis- CI = 1.2-11) for myelocytic leukemia for cussed further in the next section. benzene-exposed workers relative to the T h e study of rubber hydrochloride work- general population. ers is the strongest evidence available that The average cumulative exposure of co- benzene is associated with myelocytic leu- hort rnemhers can be estimated from data kemia. Even so, the estimate of the stand- presented in the updatGd report. Among a ardized mortality ratio for benzene-exposed total of 242 expected deaths from all causes, workers is imprecise. Moreover, the study the cumulative exposun category wan 0- contains little relevant data on the expo- 499 (midpoint, 250) ppm-months for 131of sure levels d a t e d with ercesu leukemia the expected deaths, 500-999 (midpoint, mortality. Five of the seven leukemia 750) ppm-months for 35, and t.OOO+ (my deat.ha occurred at location 2, for which -months for Tic,. Thru, the aver- ' "rmeasurements on benzene levels were par- ag*2e5c0u'mu ative exposure of cohoitmembers tidariy sparse. Furthermore, location 2 ia estimated .as554 ppm-montSi-orabout was the site of eeveral manufacturing op- 46 ppm-years. The cumulative benzene ex- rhmerations. including tire manufacturing, in pntires were lawcr Lhnn thin_for_ _- .- of addition tu ruL&r hydmhloiidc p d u g - five of the leukemia cases. .x tion, I t is likely that four of the five men The mnjot limitations of (,his stitdy am with lcukeniin at Itmlion 2 bod started i b small size and thus i t s imprecise esti- r;brtine there before 1940. and m e hmi rnntc nF Ihc d f w i or I w n 7 ~ n or n tlw t w - r t i t . ndiirtwl in I!)-I.I. I ~ . I I X I V I V IIW I w w r ~ircc*clIIS rviiw 14.Iviikciiiiti. 'hr~J#crvu~iotthiJlt 1.h . a advent in tire manufacturing therefore. mnnv r9r1 lrc- 9tridv tnc*iiiiwrs I n m ICW:icMiii '4 u' / 't mtly have lxen exposed to benzene in .<,4 departments other than rubber hydrochlo- ',!L Y ride. and their actual exposure levels may i. have been underestimated. Ott et ai. (20) conducted a retrospective follow-up study of 594 Dow Chemical Com- cumulative ben7ene exposure of three of tlw rtvc Iwkc*rnirr c t w x p r t h h l y wrtn Iwclow _. .I the average cumulative exposure of the en- *\2w, ,tire cohort has several pmibte implica;t\.C tions: there may be a threshold effect of :'' . c . I' benzene among susceptibles; t h e leukemias hW. .* ,may not have been c a d by benzene: or the benzene exposure estimates may have 1Q' :' pany employees occupationally exposed to been incorrect. benzene in the prnduction of alkyl benzene. Linos et ai. (22)evaluated the relatilm chlorobenzene, and ethyi cellulose. Men hetween leukemia and benzene in a case- empiayed in these operations from 19.78 to control study of 138 leukemia cases and 276 1970 were identified and followed-up controls. Any mention of a history of ex- through 1973.The study was updated to posure to benzene listed in the medical inciudc nn adtlil ionnl 362 Pxpneed ernpioy- recnrds w:ig t h e solc rritcrion nf l)wizc~ni* e m , and follcrw-up WiIs extcnded through exposure. The relative risk was 3.3 (95 per t h e end of 1952 (21). Cumulative exposure cent CI = 0.6-281. based on four exposed to hentene was estimated for each cohort cases and three exposed controls. Three of member. The expected numbers of leuke- the exposed cases had chronic lymphocytic mia deaths were derived from leukemia leukemia. This studv is largely uninforma- mnrcnlity rates of ail US white men. tive because of the limited information on Four leukemia deaths occurred among henzene exposures and because of its amall benzene-exposed workers, whereas 2.1 were size. Rushton and Aldenon (23)conducted a e n at seven planw. The study gmup in- c w - c u t i i w l study af leukemia within a cluded men who had been exposed to ben- large cohort of workers a t eight oil retiner- zene for at least six months between 1946 ies in the United Kingdom. An earlier ret- and 1977. JobsNere classified as involving rospective follow-up study of these workers intermittent or continuous benzene expo- ( 31I hlrd reported J standardized mortality sure. For workers with continuous erpo- ratio for leukemia of 9-4.The case-control sure. cumulative exposure to benzene =~uCiywcluded 36 leukeixiiii (leutiis w h i c h (pprii-months) wus cnlculutcd by using occurred among men employed between length of employment and the estimated 1950 and 1975 and 216 controls who had eight- hour time-weighted average for each worked at the reheries during the same job. Using the distribution of workers ac- period. hleasuremenu of workplace ben- cording to a cumulative ppm-months ben- zene levels were not available. However, zene measure presented by Wong, the av- study subjects were classified as having erage exposure of continuously expoaed been exposed to low, medium, or high leveis worken is eskmated to be about 366wm- of benzene based upontheir work histories. months (about 30 ppm-yean). "&-corn+ T h e relative risk for medium or high ex- sponds to a time-weighted average of about posurecompared with low exposure was 2.0 P3 ppm, since the average duration of expo- (95 per cent CI = 1.040). The relative risk sure was 10 years. The m 6 i t i E t y j h s o f was not higher among men with high es- e@ worken were compared with dose posure compared w i t h medium exposure, of US men and with those of an'inbmai and, furthermore, relative risk did not ap- comparison group c o m p w ' 6 f 5 ~ G 6 n pear to be related blength of sewice at the who had worked at the -refineries. The relative rirkr for the various .chemical plants but who we.re.-n-ircai&od_ .. typc;-Ol1euIcemia were not p r 7% raaulb of thir -i&?y'ts ~ar#ppnortLiavt&-- ~b~nz~enteo. For the o v e d g-roup o-f erposcd workers. ofa leukemogenic effect ofbenzene-or ottmr them was a siight excess ofle-ukemiadeaths aolventa used in conjunction with benzene. compared with US men (seven deaths, However, the informativeness of the study s t a n d a r d i d m o d t y ratio = 117; 95 per is l i i t e d by the lack of data on erpcwun cent CI = 47-242). Four of thcl.b'-Ie_GE--.i_i.-- leveis and by the absence of any analyses were of the lymphocytic cell type, two were by duration of exposure or by induction myelocytic (bothchmnic), and one waa un- period. soccified Among unexposed workers, these .Decoufli et al. 124) evaluated the mor- were no leukemias, whereas 3.4 were ex- tality expenence of 259 men employed be- pemd (two-tailed p value = 0.07). The tween 1947 and 1960 a t a chemical manu- standardized mortality ratio for all causes facturing plant. Large amounts of benzene was similar for the exposed workers (stand- had been used at the plum, but no meas- ardized monuiity ratio = 87) and unex- urements of benzene levels were available. posed workers ( s u n d a r d i d mortality ru- Employees had also worked with other ti0 = 751. chemicals. These men were followed from Six of the leukemia deaths among ex- 1960through 1977.There were three deaths posed worken occurred among men with from leukemia compared with 0.44 ex- continuous exposure (standardized monal- pected (standardized mortality ratio = 682; icy ratio = 135; 95 per cent CI = 50-295). 95 per cent CI = 141-1.992). The cell types However. for these workers, there w a s no of the three leukemias were chronic lym- consistent trend in the standardized mor- phocytic, acute monocytic, and acute mye- tality ratio for leukemia over categories of lorniinr,cyt ic. either duraiion ol exposure or cumulative Wung ( 2 5 ) coricluctcd a rcirc,hpcctlvr i d - exposure, diliough tlic Iaikr t e s ~lur trend. luw-up study ol 4,602 male chenucal work- using the nonexposed internal comparison BENZPNE A N D LE1'KEMIA 427 gmup. yielded a one-tailed p value of 0.01. sion about the risk of leukemia associated The standardized momlity ratio increased with the low levels of benzene described. w i t h y c t m ninw Fiml.c-xpmiim h m 'IIvnliir Scvcrril cnw-cnnt rol si iiciieg nf leiikrmiri of 0 for under 10 yenrs to 182 (four leukemia and solvenl exposures have lwei1cmcluctetl deaths) for a t least 20 ycan. For men with within the rubber industry (27, 28, 32.33). cumulative benzene exposure of at least 60 Three of these studies were besed on the ppm-years, the leukemin standardized mor- same series of cases from one rubber corn- , tulily ratio wua 271;Imscd on tlrrce dcnths. p m v (27. 28. 321. A foiirth stiidv incliirlrtl .$'4 .i. This study contains some evidence of a this series plus additional cases from t.lirre positive relation hetween benzene and leu- other companies (33). Each study reported ) kemia. However. the overall standardized a positive association hetween lymphocytic m o r t d i t v rat in of 117 for leiikemia is not leukemia and solvents. This association ,' stalislicnlly sigt~ificilnLoiid the slight, c'x- wcin prt!sctil. i n cwily t r i i v o f t lrc- fc,lir ni1hi-r .3' cess corild he due to confounding by expo- companies. and it was not. fmnd for other I sure to other chemicals. Moreover. the leukerniu ccil types. I .qliidv is tnn smiiII In prnvide nn ndcqrinte ''i ?4. 4 nsscwsmc.nt (JC t I c ~ ? r r * - r w p ~I ihi*~c ~* ~.~ i i w1c11~1- . . exposed workers had a deficit of leukemia :d, deaths, no confidence can be placed in stnt.isticd tests nf t.rend which include the T w o of the studies evaluated the reiation !U~l.Wc*cbriIyiriI)hiwv1ic- Iciikc*niiii litre1 igx~ui- sure to specific solvents, including knxenc (27, 29). In the study by Arp et al. (271, exposure to lynzene and to other solvent! \ * - \ ~ l l l 1 ~ x ~ " ' " c ' cElrctli(,. '1'111- r,ic-1 11l111 lllllll' t i r I Ill. wna cic*lr.riiiinrvi Iiy liiikiiig a d ~ i w h w' cirk ' A leukemia deaths among the benzenc-cx- hislotics with hintoriciil inforiiiiih~iiOII !+? lM1MVd w t i r h w wcin IICIII(* tnvcitrcytic Iciikr- spwific solvcnt-q wwfl in vnrioria w x k nrens pl mia further detracts fmni I# cuuwil iiitcq)rt*- crircl prcn:csaw. Sii1~jt~c:tw.u4.n- c*ltwsific~t1l1s i ';. tation of the study re~ults. having 'primary" benzene exposure if they .';'t Tsai et al. (26) r e p o d the mortality had worked in areas where benzene was experience of 454 men who had been em- used and if their jobs entailed direct hand- *t ployed at an oil refinery betwccn 1952 and ling of benzene or benzene-containing so1978 and who had worked in benzene-re- lutions. Subjects were considered to have 5 l a d production units. Review of industrial ysecondary* exposure if they had worked *: hygiene dah for 19.7. -3-.1. 982 indicated that in areas where benzene was used but did the median exposure lwei in these units not have jobs involving direct contact. was about 0.5 ppm. Comparison groupscon- Quantitative levels of benzene exposure sisted of the general US male population ; and a 10 per cent sample of nonbenzene- i , =exposed workers at the same refinery ( n= 1,' 823). There were no leukemia deaths .' :., among the worken exposed to benzene. were not estimated. The relative risk of lymphocytic leukemia was 4.5 for workers with primary benzene exposure and 1.5 for &,workers with secondnry exposure. Esti- --mated reintive risks for exposure to sol- ' -,.,, ' ' T h e expected numbers, based on general .\ 3 : :.$ \\ population rates. were 0.42 for all exposed wnrkcn ; i d 0.29 for workers emploved for at least one year in a benzene-related unit. vents other than benzene were nearly identical to those for benzene. 4.5 for primary exposure and 1.6 for secondary exposure. Checkowav et 31. (2s) studied 11 of tiic here were five deaths from leukemia and lvmphocytic leukemia cases included in the cancer among unexposed work- study by Arp et al. and 1,350 controls. data were presented on the Benzene exposure was determined on the 1 expected numbers of deaths. Although this basis of employment in work areas where - no excess of leukemia, its small benzene was used: no distinction was made size and the relatively small proportion (36 between primary and secondary benxene per cent) of the cohort with an adequate exposures. The relative risk of lymphocytic follow-up period preclude 3ny firm cpnclu- leukcrnia was 2.S for workergwith benzene / Er .+L cxpciure. E1ev;itrtl relative risks were q u t e l y in the rubber industry, an obser- found tor wilrkers expised to il number of vation which a h applies to the pcnchem- tither .wIwtilh. iacludiiig .wecone. carbon ic;rl industry irnd to other industries in disulfide, carbon tetrachloride, ethylace- which there is widespread exposure LO tare, and hexane. manxwlvents. Also, because the major per- These studies do not provide persuasive tinent studies have been small. little con- evidence of a causal relation between ben- fidence can be placed in available quanti- zene and lymphocytic leukemia among rub- mtive estimates of the magnitude oC m y ber workers. Vie contidetiue intervals oC association: Measures of benzene exposure the relative risk estimates are wide, reklect- levels for relevant time periods are ex- ing the imprecision of these estimates. and tremely sparse in all ol the studies. There- they include the null value of 1.0. Moreover. fore, considerable doubt remains about the it is evident that these rubber workers were leukemogenic effect of benzene at specific exposed to solvents other than benzene and exposure levels and especially at low leveis. some of these are associated with a relative risk of leukernin that is at l e u t as large as R ~ S KASSESSMENTS that for benzene. Because exposure c h i - .The expression yyiyk assessment" de- Gcation was not mutually exclusive with scribes the final step in a long process. The regard to solvents, it is possible that the pmcess usually begins with an effort, often observed association between &&ne and b a d solely on experimental findings- in . lymphocytic leukemia is due. not to ben- animals, to describe a dose-response rela-kc zene, but LO other solvents whose use is tion between a carcinogen and a cancer. /' curreluted with the use of benzene. "his plus! of the process uuuully involves In addition to the studies described extrapolation from observed, usually high, above, investigations of large groups of rub- &se levels to hypothetical, usually low. ber workers (32, 34-42). newspaper web dose levels. The process then continues by pressmen (43). and refinery workers (16, the making of a generalization to man on 31,4445) evaiuated leukemia mortality or the basis of the results in animals. This incidence, and some have found excesses generahation involves a number of as- (32,34-37,41,43,46,4831,53,54). Work- rumptiom to quate both th.e -m_anner of em in thue industries & - e a to GI- dosing and the &er.of responding of vents other 1:han benzene d t o other animals and man. These assumptions are chemicals as well. Severai authom have usqdly n?t subject to evaluation. 74:rnoted that because of this, leukernii ex- In other instance:, the $mess o cess~!!illlrong nddwr mi ~ f i i l ~ wr yorkers rrrrsessment dim not employ findhip h l l shodd not be uLLriLutd solely benzene. crniuurls but r a h r UWY cpidciiiiolugic-Iiiiti - In t lit* itggrcg;iie, IIre qiiclemiedcigic evi- ings on human beings. in such instances, a h * i i a ~ i *~ a i ~ ; c * * ~ ~, th 11iaI. In*twiv-ii in.ii/c.iic. ; i i i t l Ihc tliI'iircilhrc 01' I Ire- r i i i i i i i r i l 11) i i i r i i i ge-tl leukecniir. Howcver, thu evidence is noL eralization cue irmlevnnt. Noaecl~elrs~L ,~ I C conclusive because it comes primarily from process may still l e d to results of dubious the single relatively small study by Rinsky validity. There are two common reasom tbr ut id. ( I!)). '1'11~ 411 Iivr cliicicixricilci~ic: cicitri lhis: 1 1 L h uviiilirlrle datu ;Ire s(mrsc iAlltf 21 provide, ut best, weak evidence of n causlrl the pwpie studied expecrieaccd quiic liiyil association between benzene and leukemia. exposures so that some extrapolation is No study rules out the possibility that ob- necessary to predict effects at lower doses. served associatlons between benzene and Efforts to describe a risk assessnient t'ur leukemia are atiribut~l)lea, t least in part, h n z e n e and leukemia are in the second to conlbunding by sohents other than ben- category. There are little meaningful am- zene. As uthcrs h a w pwiierl OUL I2H).this mal data. Data &orhuman bein,1's u e [ e n possibility has n u t been explored ade- and relate to uncertain, but probably rather h i d i . exposrircl levels. Thus. efforts tn de- be true of henzene. That is. there may be a scribe the exteni qmnll increascs in leu- level of Iienzcne exposure below which no kemia risk 3s a consequence of low-level leukemogenic effect occurs. Thwn lifetime exposure to benzene must lead to imprecise of exposure to benzene at 1pprn may entail results. no increased leukemia risk,aithaugh a non- T h e benzene-leukemia risk assessments threshold linear model using a cumulative have used 'linear" and 'exponential" exposure measure might entail an apprecia- models to relate benzene exposures to leu- ble excess risk. A related problem is that of kemia. However. the use of such mathe- peak exposures. No adequate provision is matical models obscures the fact that the made in the risk assessments for the pos- data are too sparse to provide an adequate sibly high, perhaps very high, benzene Iev- description of the shape of the dose- eis that may be uniquely harmful. These response curve. Essentially, the results of intermittentfy high exposures are of special most of the benzene-leukemia risk assess- concern with respect to the Pliofilm cohort, ments can be weil approximated by assum- a group upon which most risk asnesarnents ing that the leukemia "effect" is directly are largely based. There is evidence that proportional to the dose. usually measured these workers were occasionally exposed to as lifetime cumulative exposure. For ex- benzene levels hi& enough to require hos- ample. if 15 yeam of exposure to 10 ppm pitiilizal.ion for aplasLic anemia (56). Thus, (150 ppm-years) of benzene is associated i t is plausible that the excess leukemia rnor- with a twofold excess of leukemia deaths, tality observed among the Pliofilm cohort then it is assumed that 30 years of exposure was due to transient high expzsures. to 10 ppm (300 ppm-years) of benzene We do not .supgcst t.hrit cnrcinngm risk would cause a fourfold excess.-This linear assessments are never useful. However, in model is generally considered 'consiha- . the present situation of a risk assessment ti-" in that itplpbably o v e r i k t i i a h &ea for benzene and leukemia. there are special l e u k e m o ~ n i ceffects at low h.It must problems. There arc few human studies beborne in mind, however, that the validity relevant to an evaluation of the relation of a linear model in the benzene-leukemia between henzene and leukemia. and thnse risk assessments is not established. that exist are small. Thus, estimates of the The ccrncrpt.of a "threshold" is menninp- lerikcmogrnic cffc9c.r of lwnzcnc arc imprrw- ful for Lhe I~cntsnc-leukcmiurisk ossma- cisc. An cvuii more scrious liinitiitioii in ~Iiiit ments and relates directly to the issue of the levels o f henzene to which the workers whrii. constit t i t w u nlcltuingiul memurc of in tliese C f J ~ l f t r ( Sw w c exposed are c.wnticrl- benzene exposure. Moat of the benzene- ly unknown. These particular limitations lerikcmia risk assessments are based on must IH* cnnsiclcred in llie context of thc cumulative benzene exposures: these are more general limitations of carcinogen risk obtained hy multiplying years of exposure assessment.. A reamnahle argument couid by the average ambient benzene levels. be made that the data pertaining to benzene Thus. exposure to 1 ppm benzene for 10 and leukemia are inadequate for the pur- years is considered as the equivalent to poses of a meaningful risk assessment. exposure to 10 pprn henzene for one year. Nonetheless. such risk assessments have This mi-thoci o F Idif nitring n siirnmnry e x - Iwrn :inti protxililv will cvnt inrie f c i lit* dilnr. posure measure is used frequently in car- Five henzene-leukemia risk assessments cinnRen risk assessments. the motivation are reviewed hclow. A n attempt is made to being its simpiicity. Nonetheless. it mav refine and to extend the findings of some not he particularly meaningful. The human of them. LVe also present our own risk hody metabolizes and excretes with no ap- assessrnenr. However. we emphasize that parent harm low levels of toxic substances by dning so we do not endorse the validity thnr are hnrmfiil at hieher levels. This mav of ;I hcnzcnr-lerikcmi;i risk nssessmcnt re- ~ 1 1 1 . .Wc w i 4 irtilv t r r i*iJtisitlcrthc veririus SIISLU~IIL.~ I)y nrcinlws o C Lhc Pliofilm a)- assumptions ul'these risk usessments and hon r x p d in ppm-years (415-1,W LO evulunte the extent to which their results ppm-years according to White et aI.1; and agtee or differ. d, = the cumulative benzene exposure (ex- The White et ai. risk assessmept pressed in ppm-years) for which an estimate of the excess leukemia mortality is One risk assessment has been made by desired. White et ai. (57) of the Office ot'Carcinogen Thus,for example, if 1,000 men were ldentitication and Classitication, US Oc- exposed.in the workplace for 30 years to 10 cupiltloti;ll Safety and Health .Administra- ppm of benzene (do0 ppm-years per ninn). tion. These investigators used the so-called this modified version of the White et al. 01ie-hit, non-threshold model. Although risk assessment predicts that between 28 this is an exponential mathematical model, and 101 of the exposed men will die from it is essentially linear in the range of ben- benzene-induced leukemia. This is as com- zene exposures considered. For clarity, the pared with a baseline of seven deaths results are approximated below by us with among unexposed men. Throughout the re- a simple linear model. The original model mainder of this paper we will use an expo- and our approximation produce almost sure level of 300 ppm-years to compare the identical results. results of the various risk assessments. This risk assessment is based upon the The standardized mortality ratio of 2,100 results of two retrospective foilow-up stud- was obtained from the observation of five ies: the study of rubber hydrochloride work- leukemia deaths compared with 0.23 ex- ers originally nported by Infante et ai. (17) pected among men who had been employed and later updated by Rinsky et al. (18) for five or mom years. White et al. excluded (hereafter termed thb Pliofilm study) and workers with fewer than five years of em- the tirst report of the study of Dow chemi- ployment because the elevated leukemia cal workers by Ott et al. (20). White et ai. risk was observed largely among the long- considered other epidemiologic studies of term employees. This is a prejudicialjusti- benzene and leukemia inadequate for the fication for excluding the short-term em- purlmsc id LI risk nssc-nsnient. ploye~*A~.more defensible basis for their Using the assumptions of White ec ai. exciusiori is ihat over half of rheuc!men lid regarding the findings of the Pliotilm study, been employed for less than one year, and the excess risk of death from leukemia (Pd) therefore had little benzene exposure. resulting I'rum a specified benzene exposure Nonetheless. White et al. do apply their is approximated by the following equation: ri.ik assessment method to exposure durations as short as one year. It is inappro- * priate to derive a modei excluding the ex- perience of those with short exposure du- rations and, yet, to apply it to this very same situation. The ambient benzene levels experienced by the Pliofilm workers were considered to IN*t * t l r i d 1 1 ) I 114. t I r c w m r r t w ~ C ~ ~ ~ I I I I I I I I I ~ I I I ~ ~ ~ I ~ s t a n h r h . This assu~riptiuriis cutiltuvcr- sirrl. Iiiiirritt* et. 111. ( 5 8 ) mnintain that this ushuttrpirori uvcccsiiiuutcs ~Jicuctuul IJCII- zene exposures (and, hence, that their risk i i ~ w ~ ~ i r ~i iii i~t lit iwtn l iriirrl vn t # x t * t ~ ,It~&* r i i i i l rnirtt;ilI~v).Others (59. CiO) Iwlieve that I 11t. I I ~ ' I I I I I ~IIC*IIY.I*IIC. c - x l i c m c t r c * ~w(*ri* (.oIlq(i'l era+ higher than the recommended atan- based nn only three leukemia cases and so dorde. A point of apeement is that the is very imprecise. actual henzene exposures sustained by members of thr.Pliofilm cohort are un- The International Agency for Research on known. as Infante et al. (58) recognize. This Cancer risk assessm-nt uncertainty a b u t the benzene leveis is the The International Agency for Research major limitation of a risk assessment based on Cancer (1) also h3s estimated that ben- on the Pliofilm cohort. White et al. ob- zene levels similar to those experienced by tained a lower bound for the average ben- men in the Pliofilm cohort are likely to zene exposure (415 ppm-years) by multiply- cause an excess of 140-1iO leukemia deaths ing the average recommended standards for per 1.000 men exposed for a working life- t937 through 1954 (83 ppm) by five years, time. The estimate of 140 e x a s s leukemia the minimum employment duration. The deaths was obtained by considering that upper bound for the average benzene ex- those men who had been employed for at posure (1,500 ppm-year) waa obtained by least five years had experienced a 20-old multipiying the average recommended excess in leukemia mortality and that the standard for 1937 through 1975 (50 ppm) risk that a man will die from leukemia in by the maximum length of employment, 30 the absence of benzene exposure is a b u t years. These bounds cover a broad range of seven per 1,OOO. The same assumptions henzene exposures and neither is likely to were made by White et ai. (57). The Inter- accurately de.scribe the benzene exposure national Agency for Research on Cancer ofa typical worker. It is unclear why White considered this estimate a lower b c m d for e t ai. did not estimate individual cumula- the risk nasmiated wit.h a working lifetime tive henzcne exposrtres by multiplying each (45 years) nt similar levels. However, it can year ofemployment of each subject by the be inferred from table 11 of Rinsky et al. standard prevailing in that year and then (18) that the average duration of employ- summing these exposure yean. This f81`1ure ment among men who had worked for at. is a major shortcoming of their risk assess- least five years was about 11yean. Presum- ment. However. this problem was corrected ably. .IS, as opposed to 11, years of exposure in a more recent risk assessment based on a t such benzene levels wauld cause a b u t a these data (see beiow). h t r f o l d higher cxccss. The fikwrc of 170 White et 01. also did a risk assessment was obtained b y considering that the over- ..based upon the results of the first report of all excess in thc Pliofilm cohort. wns 4Gn L I lit- I h w 11tiv. Ih i s risk i i s w ~ w w t t t pv m t t 4i.v.. I lit* rritv of Iwtkmiin nniong yicldcci rcsul1.s sirnihr to those obtained the cohort was 5.6 times that of the general 3 from the Pfiofilm data. As mentioned ear- population ratel and that the average dulier, the men in the Dow cohort probably ration of empiovment of leuhernia caws $ere exposed to considerably lower levels was 83 years. The authors then assumed of benzene (cumulative exposure of about that chis overall excess would increase `v.45 ppm-years) than were the men in the "smoothly" (presumably, linearly) from I P i i o f i h cohort. Although there were two zero just after initial exposure to about leukemia deaths observed. compared with 2.500 per cent after 45 years fi.e., 460 x one expected in the Dow study. White et J5/8.51. The resulting ape-specific excesses al. did not use the resulting standardized were then applied to the age-specific leu- morulity ratio of 200: rather. a relative kemia rates. and the figure of 170 excess incidence rate of 3 . 3 is used. This relative cases nf leukemia per 1.000 men was oh- incidence rate is based upon the occurrence tained. of three myelocytic leukemia cases among In our opinion. their second risk assess- the cohort compared with 0.8 expected. k. ment is wrong. The problem is evident in i.9 emphnsizrd that !his risk assessment is I heir statement th;it "the overnil relative .. mk oC 5.6 in the Rinsky study derivm from However, their methodology can be applid leukaemia cases who had an average of 8.5 m an occupational setting. years of exposure." A relative risk does not The Group's risk assessment was b a e d derive t'rom leukemia cases: it is obtained upon the results of the Pliotilm study (17. i'mm the comparative mortality experience la), the Dow ytudy (201,and Aksoy's study of exposed and unexpused persons. The (29)rFor the Pliotilm study, they used 311 Juthurs should have used the average em- overall standardized mortality ratio UC 7%. ployment duration ut' the entire cohort (3.2 based upon nine observed leukemia cases years, our estimate baaed upon information compared with 125 erpected:,The nine ob- in Rinsky et al. (19)) rather than that ut' served cases include two that were 'known the leukemia cases. Thus, the estimate of to exist" but were not mentioned on the ,, the excess aiter 45 years ot' exposure would death certificate. Since the expected n u m - 3 be about ti1500 per cent tie., 460 x 45/33 ber of deaths is based only upon informa- rather than 2.400 per cent. If this correction tion on death certificates, the inclusion of is made. the result would be about three these two extra cases yields a spuriously iiines as high. high standardized mortality ratio. This The authors judged that the Pliotilm co- .group assumed that the PlioTilm cohort hort members had been exposed to between members had been exposed to between 10and 100ppm of benzene. They state that about 24 and 40 ppm of benzene. However, 'assuming exposure was a t the upper end they erroneously assumed that these men of the range, then it is reasonable to pos- had been employed for 25 to 35 years, tulate thac a working lifetime exposure to whereas their actual period of employment 100 ppm ofbenzene would be likely to result was considerably shorter. If their risk as- in 140-170 case3 of leukemia per 1,ooO e r - sessment is appiied to a workplace exposure posedworkers." However. it is unlikely that of 300 ppm-years, an estimate of about 15 these men were exposed constantly to as excess leukemia deaths per 1,OOO men is much as 100 ppm benzene, and it is cer- obtained However, because of the errors tainly true that they were not exposed for discussed above, no confidence can be as long as 45 yean. Thus,this statement is placed in this estimate. not jurified by the results of the Pliotilm The Carcinogen Assessment Croup also study. According to the White et al. risk based a risk assessment on the resulri 01 rmm~nnicnt(Ai'), 45 yeurn of exposure to the Dow study. In doing so, they this time LtMl piiiii IK'IIXCIICwould priducc I w w c m cormc~lycstiirrilLrl LIIU cliirii~iiw iii trtiiliiiiv. about 420and 1 . 5 0 excess leukemia deaths ment of the cwliort. Appiyiiig their risk per 1.IWW) men. Because of the piobiems assessment to 300 pprn-years of benzene discussed atwve, little confidence can be yields ~AIIexcess ul'47 uiyeicxylic Ieukeiii~;~ placed in the risk assessment done by the deaths per 1O, OO men so exposed. This re- International Agency for Research on Can- sult is similar to that obtained by White et cer, and their methodology is not applied at. ISTI.The Group also attempted ti, usc to an exposure situation of 300 ppm-years. the results of the study by hksuy ("3)LIA a The Carcinogen .Assessment Croup risk a s s y s s r n ~ vti risk assessment. However, it is difficult to interpret the result because oC the many linritiltions of hksoy's study a i d we there- The Carcinogen .4ssessmenr Croup of fore do not consider ihia risk asresamenc the US Environmental Protection Agency meaningful. (61) has also done a benzene-leukemia risk assessment. The Group attempted LO esti- The Rinsky et ai. risk assessment mate Lhe number 01' excess leukemia deaths This risk assessment is based upon the in the general population attributable to update of the Pliotilm cohort ( 19)and ia JII exposure to benzene a i an air pollutnnr. improvement over that done by White e l ai. 157) instifar as a cumulative measure! of both a measure OC the relative rate of dis- bnt.enc! axpnsrire W ~ Lohrained for each en* among thc cxpowd rnrnparedwith the ciihcitc nrcnher. iinexposcd nnd t hcrrforc shoctltl lw siiniliir. Rinsky cc 31. (1'31 report stmdardized I t is apparent that the Riwky et al. model niortality ratios hccording 10 four levels or overestimates the observed standardized cumulnt ive Iwnzene exposure (lessthan 40 mortality ratios in the upper two exposure ppm-years, 40-200 ppm-years, 200-400 categories, especially in the highest. As an ppm-years. and yrmt.er than 400 ppm- alternative risk assessment b a . d on these years). T h e corrcsiionding number of leu- data, we fitted a Poisson regression model kemia deaths and the standardized mortal- using the four reported standardized mor- ity ratios (which have been divided by 100 tality ratios and the midpoint of their four so that thev can he compared with the odds cumulative exposure categories (62). Our ratios presented below) are 2 and 1.1; 2 and predicted standardized m o a i i t y ratios are 3.2; 2 3nd 11.9: and 3 and 66,respectively. 1.2, 2.8, 13.6, and 6 5 respectively. Our Thus, there is a strong positive relation model provides n considerably better fit. LO hetwecn citrnitlntive h r n t ~ n ce! x p o ~ u r cnnd I Iic ciiiwrvrtl s t i i t i c l ~ ~ r c l i t ~iicttiwtniit v rtit iiis leukemia mortality in this study. than does the model of' Iiirisky et. al. Our These investigators. however, did not.use predicted ntimher of CXCCSR leukemia the results of the cohort study to do the deaths for 1.000 men exposed to 10 ppm risk nsscssment. hrit, instcnd. did n ncstod Iwnzcnr fnr nrl vcwrs lwt:int1inI: t i t n(:r* :!O I-IIWI. I - i I t ~ t r t i I~tliiclv I [ )~ t i i i ~ ~ l ~ ~ ~1~n.rl ~ y~aiwi3i tl irt iiw~c l l i~ipoii I Iic I'tkwm rcjiwsnic~i case) and used conditional logistic rems- model is 82. sitin 4 1 ) ctwivr nit cxiH'nirc-rc.?l(wiiin. n*lirtion between cumulative benzene exposure Z ' h Crump and tlUcn rlsk atwssmenl and leukemia mortaiity. Based upon their Crump and Allen (63)also have done a risk assessment. they would predict that benzene-leukemia risk assessment. Their 300 ppm-yean of benzene exposure would risk assessment is based upon the remits result in an odds ratio (aieiative mortaiity of the Pliofilm cohort, the Daw cohort, and rate)or about 44. which we calculate would Wong's study of chemicai worken. yield about 250 excess lifetime leukemia These investigators used two types of deaths among 1,OOO men exposed to 10ppm linear models for their risk assessment. The for 30 years beginning at age 20 years. first, the absolute risk model,assumes that In our opinion. their risk assessment the number of excess leukemia deaths at- should have derived from the results of the tributabie to a specific amount of knzene cohort study rather than from those of the exposure is constant at every age (anaddi- case-cont roI studv. Presumably, an esti- tive excess). T h e second, a miative risk mate of cumulative benzene exposure was model, assumes that the increase in leukc- available for all cohort members. Thus. the mia mortality a t any age for a given level need to restrict the analysis to a subset of of benzene exposure is directly proportional the subjects is unclear. More importantly, to the baseline age-specific leukemia mor- if t hcir mr)ticl is q)pliccf to the d a b 01,- tality rate (a rnultiplicaiive excess). They tained in the cohort study. the estimated also considered three rnensureS of benzene ndds ratios for the four categories of cu- dose: 1) lifetime cumulative exposure. 2) a mulative benzene exposure presented ahove weighted cumulative exposure which con- (usingmidpc,ints of exposure categories and siders recent benzene exposures as most 480 ppm-years for the highest. open-ended relevant (exposures sustained between 2.5 category1 are 1.3. 4.5. 43.8 and 423,respec- and seven years before follow-up) and pro- tively. AI though t he interpretation of an vides progressively less weight to exposures odds ratio is not strictly comparable to that sustained in the more distant past, and 3) of a standnrdized mortality ratio. they are a window exposure method which essen- t i d y - o m all benzene e.poclurcs uu- among the unexpoacd and therefore piafe h i n d more than 15 years in the past. Their lirtle confidence in it. relawe risk m d d used in conjunction with &me UT the discrepancy between the risk lif'etrme cumulative exposure yielded the assessment results on the Pliofdm cohort hlghest risks. Since the other risk assess- reponed by Rinsky et al. and Cmmp and ments discussed above used a d a t i v e risk Allen apparently results from disagreement model with lifetime cumulative exposure as between them as to the levels of benzene to a measure oi benzene dose, we report here which these men were exposed. Cnunp and only these results from Crump and Allen's Allen's estimates of the benzene exposures repon. However. it ISemphasized that their are generally higher (and therefore their risk assessment result based upon the ab- risk assessment results are lower) than solute risk model used in conjunction with those of Rinsky et al., especially for expo- the window exposure method yielded re- sures sustained between 1940 and 1948. sults about one-eighth as high as those However, C a m p and Allen calculated that reported here. This observation demon- if their exposure estimates ani decreased to strates that risk assessment resuits wiil levels more comparable with those of Rin- vary widely depending upon the choice of sky, their risk assessment results would the model and the measurement of benzene i n c r e w by about only 25 per txnt. . dose. Crump and Allen estimated that .u)years ow risk assessment of exposure beginning at age 20 years to 10 We have used the method suggested by pprn benzene would cause 88 excess leuke- Entertine (64) for our risk assessment. An mia deaths per 1,OOO men so exposed. By estimate of l i f e b e excess leukemia deaths simpie linear interpolation, we estimate attributable fo benzene is obtained by di- that their risk assessment method applied viding the difference between the observed Ifor 30years to 10ppm (beginning at age 20 years)would yieid about 68 excess leukemia daaths. These investigators also did a risk as- and expected number of leukemia deaths by the total number of expecteddeaths (ail causes) in the cohort. The underlying as- sumption of this method is that the pro- sessment based solely on the Pliofdm co- portional excess leukemia mortality ob- hort using n relative risk midel with a served during the follow-up period will con- liirLimr cumulutrvc cluae. 'l'hcy did Lhrir tinue until d l cohort members have died. own anaiysrs of the Pliofilm cohort and If a cohort is followed until most of its were able to include additional years of members are deceased, the method will be follow-up. They report eight observed leu- accurdte. On the other hand, if a cohort htp kcmiu &crtha vemw ulwur 3.0 expwud, been fi~liuwedfor only u smuli frnction of yielding a standardized mortality ratio of its life expectancy, this risk assessment 268. Their risk assessment result based method must extrapoiate the prOpoKlOnal upon the Pliofilm cotrort is 63 excess leu- excess to future deaths and hence the result kemia deaths per 1,OOO men exposed to 400 is likeiy to be less valid and less precise. ppm-years of benzene, or about 48 for 300 In the latest report of the Pliofilm study ppm-years. They also did a risk assessment (19). 6.34 excess leukemia deaths were ob- based solely on Wong's study. From these served among the cohort (observed = 9, data. they calculate that exposure to 10 expected = 2.66). The total number of ex- ppm benzene beginning a t age 20 years and pected deaths was 331.6. Therefore, the continuing for 40 years would cause 121 number of excess leukemia deaths per 1,000 excess leukemia deaths per 1.Ooo exposed exposed men is estimated as men (or about 94 for 300 ppm-years). However, they point out that this excess IS due 1.000 X (6.31/331.6)or 19. largely to a deficit of leukemia mortality To apply the method to an exposure situa- !h tian of :IOO ppm-yeon. the benzene expo- result in sures experienced by the Pliofilm workers must be considered and. furthermore. it is (1.9/268.6) - .* nsmimrtl t t l r i i t ! x i - m q l w i k i v n i t r incwtnlitv is directly proportional to cumulative hen- zene dose. It can he inferred from table 2 X (:lo()pprn-l;cnw/.lli ppm-vcom) - -= 46 per 1,000 excess leukemia deaths. of their report that the average cumulative The retrospective follow-up study of re- benzene exposure ;imong mvrniwrs o f the finery warkers hy Tsiii et. d. ( 2 6 ) included Pliofilm cohort was about 69 ppm-years. estimates of benzene exposures. 'Tllcsc T h e excess number of leukemia deaths re- workers were t y p d l v exposed to less than sulting from 300 ppm-years is then esti- 1 ppm of benzene (median. 0.;;ppm). The mated 3s: average duration of employment was about (G.34/33 1.G) seven ycnm. resiilling in an nvcrape curnulative benzene exposure of ahout. 3.5 ppm- x (300 ppm-yeordB9 ppm-years) years. If these refinery work& were sub- = R3 pcr 1.OcM. jected to the same benzene-induced leukemin mnrtnlilv as wr-rr the Plinfilm wntken. We SO iipplicd I.:nt.eriinc'%iircthod t e ) the Pliofilm cohort for follow-up through 1975 (18) so that the result could be comp n r d wit 11 t tint (if W h i h ct. ni. ( 5 7 ) . The t.livir civc*rtill * t ~ i i i i h i r t l i ~ . ~iviilcirtalitv rrit i i t would be a b u t 112. Therefore, the observation that there was no leukemia death ahsewed compnred with 0.42 emected does ol,ncrvud twriilic*r 111' Icitkutniti t l v t r t Irn 11s c d ir111 i v i i i t rtiilic.1 It i i n n-?cidlrc#ifI h i n idwivc- risk rthat time was seven versus an expectation assessments. Tsai's study is rimoly of 1.ST. ' 1 7 ~I.OIJIIniiiiilwr of cxlmtccl deaths was 161.3. We assumed, as did afomrniitltonncdontt.hrcibuhtccnLiurnLchcekxnp7oasnucr-elscutkrwc,rnlaiwri>%',. White et al.. that these men had been ex- risk assessment. posed to between 50 and 83 ppm benzene The study by Wong (25)of chemic3 (say, 66 ppml. The average duration of warken also contains estimates of benzene employment for the entire Pliofilm cohort exposures. There were 3,536 men kontin- was about 3.2 years. The result is a n excess uously' exposed to benzene in this study. of about 51 leukemia deaths per 1.OOO men Their average cumulative benzene exposure exposed to 300 ppm-years benzene. This was about 30 ppm-years. There were six result agrees reasonably well with that ob- observed leukemia deaths compared with tained by White et al. for long-term workers 4.4 expected (standardized mortality ratio only. I t should be mentioned that the cu- = 1351 and the total number of expected mulative benzene exposure level assumed deaths was 613. Therefore, the number 9f hy White et al. in their risk assessment and excess leukemia deaths resulting from ex- the level considered by us in this risk as- posure to 300 ppm-venrs is estimated as: sessment are appreciably higher than the level of 69 ppm-years presented in the lat- (1.57/6131 est report of the Pliofilm cohort (19). EnterIine'.s risk assessment method also can be appiied to the latest report of the Dnw cohort in which there were four leu- x (300 ppm-yead30 ppm-years) = 26 per t.000. The result does not a m well with that of kemia deaths observed v e m s 2.1 expected Cmmp and Allen (63).The discrepancy is f 211. These men had an average cumulative in part related to the observation that there benzene exposure of ahout 46 ppm-years is no consistent dose-response relation be- and the total number of expected deaths tween leukemia mortality and the cumula- was 268.6. Thus. it is estimated that 300 tive benzene dose in Wong's study and in ppm-years of benzene exposure would part to the deficit of leukemia among the '-, uriexpisc~lin lhis . . i t ~ d y .For 1hcse RiISCJllS. The result r t d d hy Rinskv et ill. is we Iwiieve that \.Vtrng'* study is not as considerably higher than the olhers. 11 is iaforrniltive KOil risk assessment as are the diifcuit to understand why and, unfonu- Pliotilm and Dow studies. nately,yq cannot directly attempt to rep- S u m m a n i j i ti& assessments licate their cue-control results. However. we did do two risk assessments based on Five risk assessments for benzene- the results of the cohort study from which induced leukemia have been reviewed. , .their case-cnntrol study derives. Our result T d d e 2 presents ;i summary ol'the results of four of the risk dssessrnents based on using Poisson regression is $2 excess leukemia deaths per 1.000 men exposed LO 3 0 ~ \ \ the l'liotilm and Dow cohorts, along with ppm-years. while our result based upon En- ~ our own estimnces. Some of these investign- terline's method is 83. Since the Rinsky et tors did not directly calculate the estimates al. result does not agree with the others. displayed in table 'I. but we extended including two independent risk assess- their methodology LO this particular expo- ments (ours and that of Crump and Alleni sure situiticin. b a e d upon the same cohort. we hc'l'1eve Our own risk assessment results, those their result is too high. of Crump and Allen (631,and those of An interpretation of these risk assess- White et al. ( S f ) agree fairly well.' The ment results is that the P l i o f h study in- Carcinogen Asesrment Group estimate is dicates an appreciable excess leukemia rink --too low for the Pliofilm study because of an attributable to ben- zene and.-th.a.t-,the other ~ error regarding the duration of employment studies providesome support, or a tleast do of this cohort. as discussed. Their risk as- not conuadict this.. . H.-o--w. ever;'it must be sessment result based upon the Dow study, borne in mind that the o a r +dies were however, agrees well with the other results. .- either very or swee~rii-meth- Given the dil'ferent methodologies and as- dologicat problems. so.that;'dcspitc the suniptions applied by the various risk as- observation that they seem to support the sessment authors. the results, with the ex- results of a risk assessment basedupon the ception oC that of Rinsky ct ai. (19). are Pliofilm study, little confidence can be reasonably c o h t e n t . placed in them. o rTlic- rwiiIIH I l i t - t . l i i c l ~ . , i i i t i I c i ~ i c u. t iiclit-u which we coirxidered in our benzene- lcrikcrnin risk nnse9smcrit iin. ciisliinvcd in I d i l :~I. 't'lic ovcnill whnci~irdirxdmortality ratio from these four studies is 198. A weighted average (weighted according to the expected number of total deaths) of the cumulative benzene exposures sustained by the men in these cohorts is about 42 ppm- years. Thus, an overall estimate of the number of exccss leukemia deaths associ- ated with 300 ppm-years of occupational exposure ta benzene is obtained BE): -(19 9.6)/1.273 x (300pprn-years/42 ppm-years) = 53 per 1,OOO. Using the lower and upper 95 per cent confidence limits of the overall standardized mortality ratio. a lower and upper bound for the number of excess leukemia deaths for 300 ppm-years of benzene exposure are 10 and 11.7 per I.(ww), nspectively. T h e estimates for 30 years of exposure heginning a t age 20 years to the proposed new standard of 1 ppm benzene are one-tenth of these results. slririrIrtrti IS rtn?rtu-trilcvI wit I? 4111 inorriiii;itc* *.excess of leukemia. 1 lic qiitlmnir~logit-litvrnt tiri' r i i i I ~ * i i x t ~ t i v and leukemia supports the inference that benzene causes acute myelocytic leukemia. However, the available data are too sparse, or suffer other limitations, to substantiate the idea that this causal association applies at low levels (i.e., 1-10 ppm) of knzene. Nonetheless, under the assumption that causation does apply at such low levels. a number of authors, inciuding ourselves. have performed risk assessments using similar data but different methodologies. The assessments that we consiaer acceptable suggest that, among 1,OOO men exposed to benzene at 10 ppm for a working lifetime of 30 years. there would occur about 50 excess deaths due to leukemia in addition to the baseline expectation of seven deaths. However, this estimate is speculative and whether or not enough confidence can be placed in it t.o justify a lower occupntionnl benzene standard remains a decision li)r policy makers. SUMMARY Benzene is widely recognized as a leukemogen. and t h e Occupational Safety and Health Administration is currently attempting to limit exposure to it more strictly. T h e proposed new regulation is a limit of a n eight- hour time-weighted average of 1 ppm in piace of the current limit of 10 ppm. The fundamental rationale for the change is a perception that the current c I-- - .___.____________ ..... ...- -.- .. - w of petrniarim reiner?, rarkm. hrksley.CA: Ta. tlmhnr/Cl?oper:alwxwes. 1974. 1s. Thenauk C. G d e t L A morulitv study of nil mfinery vnrkers. J Ocntp M d 1979$!1:367-70. 46. Rushtan L. .4l&rson M R An epidcmiohc survey at' oil dircnhucion centen in Gnu Britain. Surrrv. L'K: Institute of Cancer Research. un- bled. 47. Hnnis YM. Holmes TIM, Shailenberqcr LC,et 31. Epidernioloric s ~ i d yof refinery and chemical plnnr mrkcrs. .I Ocmp i W d 19A2;24:2fX?-12. 88. Divine b.1.Borron V. Kaplan SD.Texncn monel- .SS- Hanir HM. Shnilcnbrrpcr LC. Donaleski D L et d .4 retmsprccive m r u l i t v study of wnrkers In - ,three m p r US refineries and chemical plants. Pan 1. Comparisonswith US papuhtion. J Ocmp Md I%SZi:28.3-YL .- 56. US Department of t b o r in conjunction with the ., --* 5.". * War Pmduaion E n a d Conference on Hcslth .,.s-! Hazards in the Rubber Industry. Akron. OH. May F* . 29. 1942. . "c 5 ,57. White MC. Infante PF.Chu KC. A quantitative , ' ntirnnte nf lmrkrmia mnrtalitv nsmcinicd wth .F' ncctqmtinnrl crposiirc to I m t c n r . Risk Annlysis itv study. I. Mortality among refinery. perrochem- 1982k195-204. icd. and menrctr ulnrken. Houston. "Xi Texaco 58. Infante PF. White .UC. Chu KC. Awurnenr of Inc, undated. 49. Schattcnfeld D.Wamhauer ME &her AC. et ai. leukemia mortality auaci~tldwith occupational e x m u r e to benzene. (Letter.!, Rink Analpin A pmqwctivr study nCmarhidig and momlity in 1904:4:9-13. pcrrnlclim industry employees in the United 59. Van bite HCI. CIrasao P. Imine 0. Tasklinr n States--. preliminmy report In:Pcto R Schnei- malt problem. thtttr.) Ria& Analwir 1984; h a n .W. etb. Qumtificrtion of occuprtion~i &I-2 canrer. Cold Spring Hubar. NY: Cold Spring 60. Tabenhaw IR, LammSH. &nam nnd leukemia. Hnrhar t ( P b 0 ~ r y 1.9111. ttauc?.)Lam lrn;2!!67-R. 51). Wen CP. Tsmi SP. McCtr1l.n WA. et d. Can* 61. A l k n RE Camng.n Anwsment Gmup's find rerm rnomlity smdy of oil rrfiney mrkem. 1. repon on popuiation rin& to untnmt benzene M o m i i t v of hauly md a~luicdworkem. Am J mpomrn Researrh Tnan& Park. NC: US En- Epidemiol 1Sl:11&52642 vimnmenul w o n -cy. OfEn d Air Aurtin SG. Schn.ttcr AR A ahnn mortalitv Quality Planning md Standnrrb. 19'79; Publien- study of prrachcmial workers. 3 Occup M d tion no. EPA-450/5-80-001L I91L3:25:304-13, 62. B m h NE, tubin JH.Mare&P. et ~ 1 M. ultiplic- Wong 0. M o g n RW. Bailey W. et al. An epi- a t i n mod& and mhrt analpin. J Am Statirt dmniologd study d petmhum d i n e r y worken. Aruoc 1983:7&1-12 Br 1 Ind M d 198R4M-17. 63. Cmmu KS. Allen RC. Qiantitntiw estimates nC MmWn IIW. Wtrng 0. A n rpitlrcninl~wicnnnlyria ria& nC Imiknnin Imm trrcqrntitmnl cqrmtrr 10 of rhc m o d t y d Mdril Oil Corpo- mtiinn cmpkyem at the Ehumont. Texm. refin- benzene. Oceup.tiotm1.SdetV a d tlcnlth Atlmin- intmrim. clnckcr H A W R. rxktlit 192. May 1W. cry. Onklnnd. CA: Favimnmental Hmlth A . d - M. Enterline I'R. I i f r t i m c mkr 4 kitinq m m r a m . Inc.. 1984. c a d by d i n e cnrrmated frnm animal and 54. McCmw DS. .hyner RE. Cnle P. Excess leukemia human sttwiim. P m d Inr the Mntnr Vehicle in n rrtincn pqrilnruin. .I k q t Mwl tWL5; MnniiInrtitrrn Amurn-mcicin ed r l i c * 1 hcitcvl Slri8c.u ? f : Z ! !2. Lltitd Ntivcmlwr ;U). I1Jrc( ~ i i n ~ n i l h h * l ) .