Document 1004v0nb8rnG8KDLaBxnqVakZ

. Rirk . 4 m & s ~Vd. 4, ,Vo. 1. 1981 Letter to the Editor Assessment of Leukemia Mortality Associated with Occupational Exposure to Benzene1q2 Peter F. Infante,3 Mary C. White,4 and Kenneth C. mu5 We appreciate the opportunity to respond to the interesting comments concerning the benzene risk assessment. (White et ai. Risk An&sis. 2:195-203. 1982) Dr.Van Raalte of Shell and Drs. Grasso and Imine of British Petroleum have raised issue with the exposure levels and relative risks observed in the various studies of benzene exposed workers. Dr. Gaffey of Monsanto has raised concern about the concept of assuming constant relative risk. and Dr. Chandler of Washington. D.C. is concerned about the justification for selcctmg the linear model for determining a dose response relationship for benzene and Ieukemia. we have divided our response mto four separate areas. namely: ( I ) estimated benzene exposure levels experienced by the study cohorts selected for the risk assessment: (2) results of studies not selected for the risk assessment that were mentioned by the commentors: (3) the concept of relative nsk used in the nsk assessment; and (4) model selecuon. i I j Estimated ben:ene e-rposurefor rhe sta+ cohorrs In regard to the Infante erol."' and Rinsk! et ul.':' stud! of Pliofilm workers. the authors concluded that. based upon a review of available data. benzene exposures to the cohort were generallv Nitfun the limts recommended at the time (Le.. the 8-hour time-weiehted-average (TW.41 exposures ranged between 10 ppm and 100 pprn]. In their letter. Van Raalte er d.questioned this assessment. Thus. we shall summarize benzene exposure for the National Institute for Occupational Safety and Heaith (NIOSH)cohort as published in 1981 by Rinsky et 01.'~' For one Pliofrlm manufacturing facility. information on bcnzcne concentrationsbetween 1946 and 1976 was available from a series of reports by the Industrial Commission of Ohio. the Ohio Department of Health. the company. the University of North Carolina and a NIOSH survey. According to NIOSH investigators. most of the data in these reports appears to have been derived from area samples measured with detector tubes over a sampling period ranging from 2-3 minutes. The company surveys of 1973-75 and the NIOSH walk through survey conducted in 1976. measured concentrations in personal breathing zone air. While the short-tcnn area samples measured over the years indicated that some benzene levels were above 100 ppm. most were below 100 ppm. These latter data points seem to have been ignored by Van Raalte et ai. In addition. the area samples are often measured in an effort to determine point source emissions and may exaggerate actual worker exposure. Furthermore. exposure determined from short-term area samples corresponds to relatively lower personal 8hour average concentrations because workers are mobile and take breaks away from manufacturing The Rinsky er ai.'" report cites documents indicating that workers were required to wear respirators when exposed (even momentarily, LO levels considered above those recommended at the time. These levels ranged from a maximum allowable concentration of 100 ppm in 1941 to an &hour TWA of 10 ppm from 1969 to the present."' One of the several documents does indicate that respirators were not 9 11:-:-43?: u f ) ? ~ I Y I W3~' i ?1 iuY4 L ~ c t irur RI,K inUr,t. 10 used. Ovcrall, the evidence indicates that respirators generally WQt uscd when atposures were ova the ncommcndcd limit.Van Raalte ef af. mentioned only the one report which indicated this was not the case. Van Raaite etaf. further mentioned that the White e r d . assumption that exposures weft 50% higher (i.a,150 ppm, between 1937-40,as compared to 1941, when the dowabie concentration was 100 ppm) was "based on poor data." The 100 ppm level was set in 1941. We thought it was a reasonable assumption based on the data available for the 1940's. Better data are not available. We believe that the actual exposure levels to individual workm comprising the NIOSH study('.2' will never be known precisely. However, much effort has been made to characterize their exposure. resulting in reasonable estimates that are better in quality and in detail than for most historical prospective occupational mortality studies. Only studies of asbestos and arsenic exposed workers have provided better exposure data. Limited information on exposure levels was available from the second location. One report from the Ohio Department of Health dated 1948 indicated t h m wcrc "a few conditions wherein an employee might be mbjtcred to an extremely high conccntration of benzol." but that employees were '*well aware of the toxicity of bcnzol and have been instructed to. and do, wear respirators when they are rtqurrcd to enter" these anas. Environmental data from this location believed to have bten derived around 1957. when the recommended 8-hour W A was 25 pprn. indicate atmospheric leveb ranging between zero and 100 ppm bascd on short-term area samples. Since respirators were used at this facility and management was aware of the toxicity of benzene. the assessment that personal exposure to benzene at this facility as well as at the earlier mentioned plant was generally within the recommended limits at the time of exposure seems reasonable. Upon review of the Rinsky er ai. report. it sums that our assumption in fact overestimates average exposure to the cohort. Dr. Imine argues that a "group of 404 workers were exposed to up to 20 ppm levels of benzene. and yet by the 1975 follow-up date no cases of leukemia had been observed." hence the extrapolations of the risk assessment do not apply. He attributes this state- ment to Hanis and Thorpe on behalf of the American Petroleum Institute. This statement is inconsistent with information presented and discussed 6 years ago.'3' Prior to selection of the cohort. a thorough evaluation of industrial hygene data avaiiable plus the NIOSH walk through survey dctcrmincd arras in which benztne exposure occumd. As stated in lW('):"Data previously requested of indusuy, but never supplied, were introduccd at the U.S. Department of Labor hearings on a proposed bcnzene standard, which alleged that wrposure to benzene existed in anarea not scicctcd for study. No details (Le., sampling location. duration, analpcal proctdurc. or definition of department) were given to permit a valid interpretation of the only data points (0.11.8 and 20 ppm) in that report." Thus, it is difficult to understand how 3 data points of unknown origm can be extrapolated to the statement that 404 workers exposed to benzene levels up to 20 ppm have not developed leukemia. When commenting on the study by Ott era1.(" used in the risk assessment,Van Raalte er af. Cite the highest singie exposure peak (937ppm) measured in the study, and state that they are not told how frequent such exposure oaxus. It is dear from the Ott eraf."' report that the 937 ppm level was the highest and was a rare OCCU~~C~CVCa.n Raalte ef af. cite this single data point to support their opinion that high concentrations of benzcne are crucial in determining leukemia and thus mathematical modelling based on a cumulative dose concept misses "important and rcievant information." The average benzene exposure to workm in this production area was estimated to have bccn 35 ppm. None of the rhm leukemia cases (versus 0.8 expected. RR = 3-75)worked in the area of the plant where the highest utposurc existed. To the contrary, two of the individualswho died from leukemia worked in areas characterized as low bcnzene exposure 2-9 ppm W A . while the third worked in an area characterizcd as very low (Le.. TWA exposures less than 2 ppm). Further assessment of atmosphericbenzene levels in the three production areas from which workers were selected for study by Ott eraf. indicated that exposures to the cohort had always been fairly low. reflecting the use of cioscd continuous systems. In production area 1. estimated TWA benzene exposures based mostly on breathing zone samples ranged between 0.1 and 6.2 ppm for the years 1944 to 1973. In production area 2. TWAs ranged between 0.3 and 14.7 ppm between 1953 and 1972 while in production area 3. the averages ranged between 4 and 35.5 ppm for the years 1952-1974. On the basis of data presented in the Ott erai. study. the average exposure for the entire cohort was about 5 ppm. Thus. in the absence of any mathematical modelling for this study, one could be left with the impression that fairly low average exposure to bentcne results in the development of leukemia. (2) R e d t s of studies not selected for risk assessment Dr.Imine c8mments that one can produce nega- tive studies demonstrating that the projected risk values from White etuf. "do not hold." He sites for his example the study of petroleum refintry workers by Thorpe. The International Agency for Research on Cancer (IARC)`5'reviewed the Thorpe study and in addition to its criticisms of low power and Limited methodology stated "the study suffers from problems of ascertainment. specificity and validity of diagnosis. and the `health worker effect' in the calculation of SMRs" (Standardized Mortality Ratios). More detailed discussions of the limitationsof the methodology used by Thorpe have k n published.""' Dr. Inine also mentions that the extrapolations of the risk assessment do not apply to a small group of individuals who first worked in a benzene-exposed department between 1950-1959. The data do not support his contention. For individuals who were first exposed between 1950-1959 there were two deaths from myelogenous leukemia when the cxpcctcd deaths based on total leukemia was 0.46. For one of these deaths due to acute myelogenous leukemia (AML), however. the hematologist who diagnosed numerous leukemia victims who worked at the plant noted that the discharge diagnosis was changed from AML to aplastic anemia so the widow would receive S1O.OOO in industrial compensation. If the death was from leukemia the widow would have been entitled to no compensation:*' Although this death was not included as attributable to leukemia in the NIOSH study (as appropriate). there is justification for including it in a maximal risk estimate which would result in 2 leukemia deaths versus 0.46 expected. for a relative risk of 4.0 based on total leukemia mortality. or a hgher relative risk baxd on myelogenous leukemia mortality only. The numbers upon which the observation is based are small. but nevertheless they are contradictov to Dr. Irvine's interpretation of the results for these workers. Van Raalte r r d . state that we did not include the Aksoy and Vigliani studies in our risk assessment and that these studies indicate benzene exposure has to be substantial for leukemia to occur. In our paper we previously stated they were not included because we felt the benzene exposure information from the studies was not as detailed as that tn the Infante-RinSky and Ott dol. studies md that the a t i m a t e of risk w c r ~unctrtain.HOW-, wc n c ~ g nize that Akmy, VigIiani and their co-workers have made tremendous contributions in their pioneering work addressing the clinical manifestations of an entire spccvum of blood abnormalities associated with benzene expocwe. Workers described in the Vigliani and Aksoy reports were ex@ to high levels of benzene resulting in the occurrence of over 200 cases of leukemia. However. Vigliani and Aksoy did not have data upon which to &s the carcine genicity of benzene among workers exposed to low leveb. (3) The concept of relative risk used in the assessment Dr. Gaffey has illustrated the confusion that can arise from tbe use of the term rdative risk. Rates of the discase may be urpresscd in units of h e , such as the number of deaths per loo0 pcrsons per year. or may be wrpresscd as proportions which art unitless. such as cumulative mortality. It may be useful to distinguish a ratio of timedependent rata from a ratio of proportions. since both types of ratios are referred to as relative risks. When mortality rates are for a very small interval of time. the ratio of two rates can be viewed as an instantaneous rate ratio. An instantancous rate ratio need not be age dependent. and at least thamtically could be constant over time and across ages. On the other hand. cumulative mortatity expressed as a proportion is upper-bounded by one. A ratio of two proportions cannot exceed one if the denominator proportion is quai to one. In our dcuiations. we treated the SMR as an estimate of the ratio of two cumulative mortalities. As such. the SMR can be thought of as a composite measure of instantaneous rate ratios for all age groups. When one compares SMRs betwan study populations with different age compositions. one is implicitly making the assumption that the instantaneous rate ratio is constant across age groups. The appropriateness of the comparison is dependent on the validity of this assumption. The assumption of a constant instantaneous rate ratio also is made when one uses the SMR from one study to calculate I risk for another population with a different age composition. In our calculations. we expressed the probability of death from leukemia as the cumulative risk over an adult lifetime (i.e.. a proportion). The SMR was treated as an estimate of the ratio of the cumulative .. 12 mortality from leukemia following occupational cxposurc to benzeneOWT the cumuIatiwmortality from leukemia in the abscncc of occupational bcnzene exposure. In our m o d 4 the ratio of cumulative mortalities was dose dependent. Thc use of Abbott's correction assured us that the estimate of the total cumulative mortality from leukemia would not exceed one. Thus. the measure of relative risk that we relied on was the ratio of cumulative mortalities, which could be estimated from the SMRs. The assumption of a constant instantaneous rate ratio was not integral to our calculations. However, this assumption is important to consider when assessing the validity of the risk estimates. (4) Model selection and cumulative dose relationship for benzene and leukemia Dr.Chandler has questioned the appropriateness of using the one-hit, or linear model, and suggests that molecular mechanisms which determine the dose response relationship for benzene toxicity should be evaluated in estimating the risk of benzene. We believe that an assessment of benzene toxicity could be improved by knowledge of biologcal factors such as the dose of the ultimate metabolitqs) in bone m o w (and other organs), the toxicity of these compounds. and the. rates and sites of their transformation as well as the cumulative nature and pharmacokinetic distribution and action along with modifying effects of intrinsic and extrinsic factors. However. after decades of study. the basic mechanism by which benzene affects bone marrow precursor cells is still unclear. Furthermore. even if these factors could be identified. it would be difficult to determine the weight that should be given to each factor in the development of a mathematical model. Nevertheless. selection of this model seemed a p propriate since linear dose response relationships exist for several end points measuring the toxic effects of benzene on the bone marrow. (These are discussed below.) For a number of carcinogens. the linear nonthreshold model has been shown to describe the dose response relationship (Le.. radiation and leukemia. arsenic and lung cancer. and asbestos and lung cancer). This model also has been used for carcinogenic agents by the World Health Organization's Arsenic Working Group. for ionizing radiation on an international basis. by a Task Group on Air Pollution and Cancer io Stockholm in 1977, as well as by rtgulatory bodies in Ihe U.S. such as the carcinogai Assessment Group (CAG)of the EPA.") With spcclfc regard to benzcne and leukemia, OSHA requested sevefal m i e ~ tao ~comment on -the model used by White e f d . for the risk assess- ment.('O' .- In response, Dr. Norman Brcslow, Dr. Philip Cole, Dr. Charles Brown, and Dr. David Gaylor alI thought the linear model was acceptable and that other more complex models were not justified. How- ever. Dr. William Rowe felt other models should be evaluated in addition to the linear model. while Dr. Karrh felt the h e a r model was unverifiable. In our opinion. the issue of model selection can best be summarized by Dr. Charles Brown who stated("': ThC'corrCct'm0dCl is uaknom4 and will rcmrinsountilwe know the wch.llls'ric rdatiomilip bctmcn bal7tne cxp o s w c and leukaniz howm. I do not belien chat the data warrant more sophistication thau the dmpk linear model (towhich the onchit modd is a nry dose approxi- matiw at low rrspoosc rata); in ddition. since tbe raage o ~ e rwhich the d a atrapolatiw i~ Paforwd r p p u u ~10. be rclativcly small (one order of magnitude?). the dose response model should have a small effect on the risk asusmen1 mdu With regard to benzcne and leukemia Dr. Norman Breslow recently stated that dose additivity and low dose linearity should be adopted as biologi- cally reasonable and scientificaly prudent assumptions in the absence of spcclfic evidence to the con- UZUy."O' Van Raalte etd. have the opinion that a cumu- lative dose relationship would not sam to accord with what is known about the toxicology of benzene and. based upon their interpretationof epidemiologic data further allege that both marrow hypoplasia and leukemia only occur in persons exposed to b e w n e concentrations above 100 ppm. In the Ott et ul."' report. there is no cvidcncc that the leukemia cases experienced high peak ben- zene exposures. Likewise. studies of several refineries'11-12' and chemical manufacturing facilities'13' where relative leukemia risks are signifi- cantly elevated and average exposures have been low. do not provide data to indicate that the leukemia deaths resulted only from substantial peak exposures to benzene. The only epidemiolo@cstudy to address maximum peak exposure and cumulativeexposure to ben- zene in relation to cancer was the recently reported industry-wide mortality study of chemical workers exposed to benzene sponsored by the Chemical 1 .. . ... Asscssaent of LeuLcmir Morality 13 Manufacturing Association (CMA).'I3' When data were analyzed by cumulative urposurc. statistically significant dose response relationships were dcmonstrated for leukemia as well as for all lymphohematopoietic cancers. Analyses by maximum peak benzene exposure (Le.. below 25 ppm. or between 25-100 ppm. or above 100 ppm) did not demonstrate a significant trend. However. the latter analysis demonstrated a relative risk greater than 3 when maximum peaks were below 25 ppm. The findings suggest that cumulative exposure may be a valid measure of dose in determining dose response relationships for benzene and leukemia and that maximal peak exposures below 25 ppm may also induce iymphohematopoietic cancer. In order to further evaluate the issue of the level of benzene rquircd to cause toxicity to the bone marrow. one can look to experimental studies where dose and dose level can be controlled. In this regard. the NTP bioassay on benzene demonstrates dose response relationships for multiple cancers as well as for leukopenia in both mice and rats.'"' The low dose in both speciesis equivalent to an 8-hour atmospheric exposure of 20 ppm. Additional experimental studies have now demonstrated that inhalation of 10 ppm benzene causa significant bone marrow depression ( 6 hrsiday. 178 days)('5' and disturbances of immune system function (6 hrs/day. 6 days).'16' Since most leukemias and related disorders in man seem to involve stem cell abnormalities and immune system deficiencies these findings may be highly relevant to benzene and leukemia. Several recent experimental studies have demon- strated adverse effects on chromosomes and bone marrow in relation to low level benzene exposure. Collectively. these studies demonstrate chromosomal damage in bone marrow cells as a result of only one 6-hour exposure to 10 ppm'"' or one &hour exposure to 28 ppm."" or to an equivalent of two 8-hour exposures to 6 ppm."*' Thus. numerous recent findings Seem to contradici [he opinions of Van Raalie er ai. and Irvine that only peak benzene exposures above 100 ppm c3n cause leukemia or bone marrow toxicity. Funhermore. both eppidemiologjc observations and experimental studies suppon il cumulative dose concept for benzene and leukemia. and indicate that even shortterm exposures to 10 ppm are toxic to the bone marrow. REFERENCES 1. P. F. Infante. R A. Itinsky. J. K.Wagoaer. and R J. Young, Leukemia in &nzcnt Workers," LMrn 2 16-18 (19Tl). 2. R. A. Rinsky. R J. Young. and A. B. Smith. "Lculremia in &nzenc Workers." Am. J . I d . .Wed. 2 217-245 (1981). 3. P. F. Infanw. R A Rinsky. J. K. Wagoner. and R J. Young. "&mcw and LcuLunia." Letter IO thc Editor. Lonm 2. 868-869 (1977). 4. M. G. Ott. 1. C. Towaxnd W. A Fishbeck. and R. A. Laagner. "Monality Among Individuals Occupauonallv Ex- posed to &mcne." Arch. Env. H d r h 33.3-10 (1978): 5. IARC Monographs on the Evaluation of'the Carclnogenic Risk of Chemicals to Humans. Some I&irial Chemicals and DvesrU//s.IARC. WHO. Lyon, Vol. 29 (1982). 6. S. M. Brown. "Leukemia and Potcntial Bcmenc Exposure." Letter 10 the Editor. J . OCCLQh. id. 17. 5-6 (1975). 7. P. E Infante and M. C. White. *Benzene: Jipidcmiologic Observations of Leukemia by Cdl Typc and Advmc Health UfmS krsocirced with Low-LCwl Expo~urc."Ew. H d t h Pt~pm5.2 75-82 (19831. 8. Y. J. Sakol. "Clinical Obmation of DiGugIicho LcuLcmir at Oae Division of One Ohio Rant When Workas Wac Expod to Bcmcl3c." Tesriwray I Dcpt. Labor. Docket H-059A TR 285-339 (1977). 9. World Halth o%urpation. Envrmmmral H d t h C n r d 18. Arscaic Gerrtva(19811. 10. Formal Cornmarts Submitted IO OSHA oil &ncnc Risk Awument Docket H-059B. Ex. No. 137, (1983). 11. T. L. Thomrr R J. Wuplaler. R M~c-EnsoS.. Itavk and J. E Raumeni. 'Monrlity PItterns Among Workm in k Texas Oil R ~ f i a c n a "J . O C ~H.d.U.135-141 (19811. 12. R. joy^. "Leukcmaat Shell Wood Rimand Deer Park Manufacturing Complexcs." Later and rcponr of studies to J. D. .MiIlar. Director. NOSH. dated {July 28.1983). OSHA Dockct H-059B. Ex. NO. 142-13. 13. 0. Wong. "An Indusuy-Wick Monalin. Study of Chemical WorLm - a t i d v Exposed KO Banmc" Repon sub- mrted IO C k n i c d Manufacturm Asociatioa (De.19831. 14. NTP Teduncal Rcpon w the Tox~cobgyand Camnogenesir Studies of Bauenc. NTP 339. U.S. DHHS. NIH Pub. No. x4.345 (inpres,. 15. K. A. Baarston. C. A. Snvder. and R. E Albm. "Rcpeatcd Exposure of C57BL .Mice IO 10 ppm Bemcne .Markedly De- p d Erythropoleric Coloav Formanon." Toxtcdop k r e r s 20337-342 (19841. 16. ?A. G. R ~ s e nC. . A Snyder. m d R. E .Uben. "Dcpmsioasin B md T Lymphocyte Miogen-lnduccd Blastogenesis in .Mice Expod 10 Low Concenuauoas of Benzene," to..ncdogv Lrrm 20:(1984). 17. A. D. Kligmnan. G.L. Emon. and J. L. Wilmcr. aDeveiop m a t of Rodent Peripheral Blood Lynphocytc Culture Svs- tern to Detect Cytogmeuc Damage in Vivo," Paper presented at lnremationd Symposium on Sister Chromatid Exchange. Brookhaven ?fad. Lab.. t k . 191131. le. R. R. Tim. J. L Ivctt. SI. J. Saw?. D L. Cosra. R. T.Drew. m d C. P. Cronk~tc."Cytoynctic Mamfestauons of &nzmc Induced Damage in Murine Eone Marrow." Papcr presented at International Conference on Benzene. Collegium R m a z z ~ m . New York. tNov. 19831 19. M. M. Gad-El-Kanm. B. L. Haper. md M. S. Legator. "Modilicatiom 1n b e SIulvcloclastogmc Effect of &nzcne tn .Mice uilfi Toluene. Phenobubltol. ?-lc!elbylcholanrtucne. .U- oclor 1254 and SKF-525A." Hururrun Ra.(in pmsi.