Document YrE9D6dD633QZenZj9p7qzboN

American Petroleum Institute .-4?1220 L Street, Northwest Washington, D.C. 20005 memo Date: From: To: Re: August 24, 1993 ~ Mary Paxt"/ff~ Benzene Task Force (BTF) Minutes of August 11 Meeting and ACGIH Submission Enclosed .J.re draft minutes of our BTF meeting on August 11. The final version of our submission to ACGIH is enclosed. You will also find the remarks from Sheryl Bartlett of Dan Krewski's office, w1ich unexpectedly ~rrived in time to include with our submissio 1. Pat Beatt; is preparing the overheads for his presentation at ACGIH on .)eptember 13. They will be distributed at the beginning of next w~ek in time for our conference call on Thursday, September 2, at 2:30 EDT. Pat will be holding a- "dress (or shorts?) ::-ehearsal" of his presentation at 1:00 on Sunday, September 12, at tne Holiday Inn Crowne Plaza in Rockville, MD; if you ar~ able to attend, your input would be appreciated. The next Eull meeting of the BTF will be on Thursday, September 30, in Sa:1 Francisco. Details and an agenda will follow. Alsc inclJded is a newly published article by Cody, Strawderman, and Ripen, which we helped support. If you have comments on the draft by Tony Cox that was distributed at the BTF meeting, please convey them to me to pass alor:.g to him. a-tt: minutes. 805 acgih.813 Bartlett letter Cody et al. BP-00020066 BENZENE TASK FORCE (BTF) Wednesday, August 11. 1993 American Petroleum Institute, Room 939 1220 L Street, NW Washington, DC 20005 Wednesday, August 11, 1993 Meeting Minutes - Subject to Approval Attendees Pat Beatty Peter Craig Barbara Divine Anita Ducca Carol Fairbrother David Mongillo Ramona Panson Mary Paxton Gerry Raabe Dave Steup Dale Strother Li Yang Affiliation Chevron Mobil Texaco API Exxon API SlJW API Mobil Shell BP America Jl.RCO Int:::-oduct ion Pat Beatty convened this meeting of the Benzene Task Force (BTF) at 9:10. The minutes for the last meeting (6/22/93) were accepted without change. The primary purpose of this meeting was to EinaLze the s-ubmission to be sent to ACGIH on Friday, so oth~r rna~ ters were briefly reviewed first. Regulato::v Affairs U.S. EPA: A journalist had informed us that EPA had reactivated_ its reconsideration of the benzene p0tency factor. (Responses to inquirie:; received the day after the meeting indicated that EPA is devehping a position statement on the need for updating the benzene )otency factor. This should be available by the end of the year.] California EPA: Pat Beatty reported that he will be requesting funding in 1994 from WSPA to fund a workshop and a rev1ew paper of devel::1prnents since the Cal. EPA developect their benzene potency factor. D E A F I' 1 August 24, 1993 BP-00020067 ProJect Updates ITRI: If a contract has not been settled within two weeks, the $150K requested in ITRI's revised proposal will be sent in the form of a gift. Kalf: Kalf has been told that we will support a graduate student in reseaching the hydroquinone receptor. Bob Drew is looking for money to cover the full three years of support from this year's bu :lget. Cox: His draft and summary comments of a journal article summarizing his feasibility study of biologically-based risk assessment (BBRA) modeling of leukemia using data for the wellstuc.ied chemotherapeutic agent cyclophosphamide (CP) were distributed for comment.. It was formatted for a special issue of Risk Analysis to which he was asked to contribute. Richard Faul, staffer of the American Automobile Manufacturers Association (AAMA) , sent their integrated- summary of how the biological data support a threshold for benzene's leukemogenic effect. Mary Paxton will arrange for a few BTF members to meet with P..PJ~Ji'. representatives to determine how AAJIJ.?:; might contribute s.upport to the BBRA modeling effort. Genotoxicity: Irene Jones of Lawrence Livermore National Laboratory reviewed the two articles from Legator's lab (Au et al. 1991 and Ward et al. 1992) . Mary Paxton discussed possibilities for research in this area with her at the Gordon ResE=arch Conference on genetic toxicology. The Gordon Conference was largEly concerned with defining the background mutation rate _and deve~oping a profile of spontaneous effects. The attendees focused their risk assessment concerns primarily on somatic effects cancer) and discussed why quantitation of heritable dama.ge wuuld probably entail much more effort than the actual dangers rlerited. Mary Paxton also noted that she met Ken Turteltanb of Lawrence Livermore Labs, wh-o is applying ultrasen! i ve detection (e.g., one molecule of 14C in 10- 1" mole) by tandem accelerator mass spectrometry to DNA adducts. His work to date has included some work with benzene exposures. Pap_er on biological basis for model selection: In planning to have thi.o paper written, note should be taken of an extensive review o: dose-response modeling by Zeise et al. (1987 Environ. Health P~rspect. 73:259-3-08) supported by API, Cl".L.ZI., and Monsanto. Benzene exposure in gasoline studies: The summary of the Nottingh.:J.m workshop is now available. The study of the Imperi-al Oil cohort will be presented at a- meeting in Nice in September. API's pat"ticipation in the May 1994 IARC meeting will be limited, but Lesl~y Rushton will be making a presentation. David Mongillo will inform the Industrial Hygiene Oversight Group about this issue at its meeting at the end of August. Jack Hir:ton will then develop a scope-of-work statement that reflects D F: A F T 2 August 24, 1993 BP-00020068 the objective of reconstructing benzene exposure from gasoline ln a generic fashion. Finatncial Impact Study: Anita Ducca suggested doing this study in two pa~ts, rather than just rev1.s1.ng our previous RFP for the full study. Phase I will identify potential federal and state modifications in benzene regulation and what changes would be t-ric;rgered ~by events (such as a reduction in the TLV to 0.1 ppm) and prioritize the likely magnitude of their impact on the industry. In the next 4-6 weeks, Anita will prepare a scope-ofwork and approach several contractors for cost estimates; the anticipated cost is in the range of $10-20K for about a month's proJect. After this effort to more precisely define the problem, we. ~;houlc be better able to determine exactly what should be included in our RFP for the study itself (Phase II). ACGIH Strategy AftE~r a treak from 10:25-10:50, the group reconvened to discuss plans for the meeting with ACGIH on September 13 in Rockville, MD. Jennifer Galvin learned from Dr. Sharma that, if ACGIH does not approve the 0.1-pprn TLV and does not wish to fall back to the current 10-ppm TLV, the entire process will have to be repeated for any .:lternative figure they select, i.e., there would be a formal pl.blished Notice of Proposed Change followed by a yearlon9 corrnHent period before another value for the TLV could go into effE:ct. Subxniss-ion for September 13: A letter with comments for attachrner't to our sutlmission has been rece1ved from Vern ChinchilJi, and one is expected next week from- Dan Krewski and Sheryl Bctrtlett, which we will send along after the submission. Exxon Biomedical has made available a critique of the Brown and Spi:rtas cmalysis, which will also be attached to our submission. The group went over the prepared draft and noted several points tha~: needed to be added or expanded. After breaking for lunch at noon, thf, BTF -members composed the necessary insertions. Plalo.s for September l-3: Pat Beatty will give a dry run of his presentation for -all who can attend at 1:00 on the afternoon before a+ the meeting hotel. API's delegation to the meeting will be ~at Beatty, Jack Hinton, Vern Chincfiilli, Bob Drew, and -Mary Pax:~on. Kenny -Crump and Gerry Raabe have also been invited independ1:ntly by ACGIH. Future-meetinas A c::mfer~nce call will be held 2:3 0 EDT on Thursday, September 2, to go over the Overheads Pat Beatty will use in his presentation at the September 13 ACGIH meeting. DRAF T 3 August 24, 1993 BP-00020069 All interested BTF members may attend a practice session for the presentation to be held the afternoon of Sunday, September 12, at the Holiday Crown Plaza in Rockville, Maryland. The full ETF will convene Thursday, September 30, in San Franciscc. The meeting was adjourned at_ 2:00. Resr:~ectfully submit~ ~~~~~ I Mary Burr Paxton, Ph.D. D R A F 'I' 4 August 24, 1993 BP-00020070 Health and Welfare Sante et B1en-etre social Canada Canada Health ProtectiOn Branct' D1rect1on generale de Ia protectron de Ia sante Room B43 Environmental Hea11h Centre Tunney's Pasture Ottawa, Ontario KIAOL2 July 26, 1993 Dr. Mary Paxton American Petroleum lastitute 1220 L Street, Northwest Washington, D.C. 2(XX)5 Dear Mary, As requested, I have read over the material you sent us on benzene risk estimation by C. Brown~ R. Spirtas as well as the reply by K. Crump. My comments follow. Spii"tas and Brown1 were asked to review Paxton et a1.2 Specifically, they addressed three issues: (I) the risk-group stratifiCation, (2) the dose-response function and (3) combining data from Akron and St. Mary's locations. K. Cruqf commented on the differences between analyses by Paxton et a1.2, Crump4 and Brown and Spirtas1 As pointed out by Crump3, a higher log-likelihood does not necessarily mean a better fit. It only means that in total, tbe curve is closer 10 the observed points. However, this does-not address potential Jack-of-fit, particularly systematic departures of the estimated model from the observed data, and there could be other models with even higher log-likelihoods. One method for assessing lack of fit is 10 add more parameters to the model. lf-:here is a significant increase in the log-Ukelihood, then the original model is not fitting the data as well as the new model. It would also be helpful if Crump, had provided standard-errors for the parameter estimates of tbe -expanded models.- 1 Brown C. and Spirtas R. (1993) A re-analysis of the leukemogenic risk associated with occupational benzene exposure in th pliofilm cohort. Unpublished. 2 Paxton M., Chinchilli V., Brett S, Rodrick& J. (1993) Leukemia risk associated with--benzene -exposure in -the pliofilm cohort. I. Monality updata and exposure diSlribution. and II. Risk estimates, to appear in Risk Analysis. 3 Crump K. (1993) Reconciliation of results of-three Slalist.ical analyses of the risk of benzene exposure based on data from the pliofilm cohort. Unpublished. Crump K. {1993) Risk of benzene-induced leukemia derived from pliofilm cohort: effect of additional follow-up and new exposure estimates. j. Tox. Environ. Health. submitted. -Canadall I BP-00020071 M!~re Wa%ker of the Biostatistics Section graphed the five curves fit by K. Crump' for cumuJati"Ye doses up to SO ppm-years~to view the differences-for low doses and up to 3<XXl ppm-years to obtain a more global view. In the fust figyre, three of the models are quite similar. The exponential curve is somewha! natter than these three though starts rising quickly at around 1000 ppm-years. The power curve raises quickly from the beginning am is much higher than the other four until 2500 ppm- years when the exponential curve overtakes it. It is puzzling that tbe power curve is so different from the others especially since the cumulative exposure for cases is DOt likely to exceed 3<XXl ppm-years. A gra(il of data would go a long way to clarify this issue. Wha1 about a plot of the relative risk in SOO ppm-year dose intervals? If you can provide this data, we_ would be~happy to do some graphical oomparisoos. My FAX number is (613} 952-9798. The method of stratifying risk-group did not lead to appreciable differences in parameter estimates for the exponential model and some small differences for the power function model (Paxton5, 1993). For all three sets of exposure estimates and for both models, the Spirtas strata yielded higher maximum log-likelihoods. It would be useful to know oow the differenres in parameter estimares translate into differences in estimated relative rists at specifiC cumulative exposures. The on1y comment we have about the Akron and Combined analyses is tbat we believe that acute myelogenous leukaemia is the response of interest aod would prefer to focus on that response. I hope these comments are helpful to you. Please give me a call witb any questions tha~ arise. A{~- -Sheryl Bartlett, PhD Head, Biostatistics Section Enkonmental Health Directorate (613) 954=7787 cc. Dr. D. Krewski ~, Paxton, M. (1993) Exponential vs. Power function as dose-response in proportional hazards model. Unpublished table. BP-00020072 Benzene: relative risk vs. dose ~l co ~~ Expo~ential: r=exp(J3x) ............. Power: r=(1+x)~ ----- Linear: r=1 +~x ---. Linear-quadratic: r=1 +~1 x+~r2 - - , - Expand~d power: r=(1+ax) ......._.....-----...... ..... <0 ~ rn ;:; Q) > 0::: (tJ Q) X: v C'i ........................- ........- .........-..... .....- .. .......' .I 0 0 ttl '"0 I 0 0 0 N 0 0 -..] C,N 10 20 30 40 50 Dose, x (ppm-years) g J.. Benzene: relative risk vs. dose -- 'r I 1 ..:. Power: r=(1 +x)li ----- Linear: r=1+J3x ---;"' Linear-quadratic: r=1 +Ji1 x+~:zX2 - - - Expanded power: r=(1 +ox) 8 -"- 'r ~ 1) > :I p:0 c 0 lO -.........-............-...... ..-.-'... ..' //_.---- . --~==-===~-- ' ~~,/ ----~--- --'--1 0 0 500 1000 1500 2000 Dose, x (ppm~yeIars) ttl '"0 I 0 0 0 N 0 0 -..] +:- 2500 ------. 3000 OM Pogram Surveys He~ mann et ai H_ernatologic Effects of Benzene Job-Specific Trends during the First Year of Employment among a Cohort of Benzene-Exposed Rubber Workers Ronald P. Cody, EdO William W. Strawderman, PhD Howard M. Kipen, MC,-MPH Benzene is a bone marrow depressant that, after significant exposure, may eventually cause peripheral blood cytopenias due to bone marrow depres-sion in occupationally exposed hu- mans.'"" The data that demonstrate Hemarologic surveilfance data from 1940 to 1975 were analyzed for a benzene-exposed cohort of 459 rubber workers. The present analyses-are restricted w 161 workers with "preemployment" counts done before exposure this in humans are mostly from individual members of industrial cohorts of the 1920s and 1930s or from uncontrolled exposures without even and rely on their subsequent counts from the first 12 months ofemployment. basic protective measures. Exposures I While blood cell counts declined approximately 1000 cells/mm3 over the first were regularly in excess onoo ppm. 4months ofexposure. Using repeated-measures analysis ofvariance, workers Except for individual cases, surveilexposed above the median benzene exposure at the plant had significantly lance of modern cohorts followed for lower av~~rage wh1ie and red blood cell counts at each month during the first the clinical effects of benzene -expo- year ofwork when compared with workers exposed below the median. These sure in refineries and chemical plants decreased counts suggest that clinically detectable bone ma"ow depression have not yielded significant non- accompanied-the onset of work in this plant during the 1940s and support malignant hematologic findings, per- exposure assessments that favor higher benzene levels in the 1940s when haps because of relatively lower compared with subsequent decades. The-general utility ofrepeated-measures benzene exposures.'-' We have retro- analytic techniques for medical surveillance data is also demonstrated by spectively analyzed hematologic mon- this analysis. ' itoring data from 1940 to 1975 on an Ohio cohort of-benzene-exposed rub- .ber workers known as the Pliofilm Cohort. They are acknowledged to have approximately a fivefold excess of leukemia mortality and to be the best cohort upon which to base risk assessments for the malignant effects of benzene exposure.916 These anal- yses of hematologic effects are not only unique in their depiction of the widespread bone marrow depressant effects of benzene, but provide insight into discrepancies in quantitative es- timates of the quantity of benzene exposure that the cohort experienced in its early years. From-tile UMDNJ-Robert Wood Johnson Medical School Department of Environmental and Community Medicine, Piscataway, New Jersey (Or Cody, Dr Kipen); Rutgel\, The State University of New J.ney, Department or Statistics, Piscataway, New Jersey (Dr Strawderman); and lbe Environmental and Occupational Health Sciences Institute, Piscataway, New Jcrxy (Dr Kipeu). Addrt!$ correspondence to: Environmental and Community Medicine, EOHSl, 681 Frelinghu~o Road, PO Box I 179 (Dr .Kipen), Piscataway, NJ 0885S.ll79 0096-1736/93/3508.{)776$03.00/0 CopyriaiJt @ by American College of Occupational and Environmental Medicine Our previous.publications on the St Mary's Pliof!lm cobort 10 11 showed a significant increase in white blood ccU (WBC) and red blood~ cell -(R:B) counts of the entire group ofsut"Veyed workers in the 1940 to 1948 time period, -corresponding to apparently declining benzene exposures in the BP-00020075 ., ,\. ' ! i"t "! I:~ Si! ~ ~: -.-~,:mpb)~ment :pre: . .,,._.,,ur~l blOC"-! .:nunts" 1 ' a.nal~stl .f e;~ch worker's md1>tdual experienn beginmng wah a preemp!o1 men! count ~eemed to he ;J Llsefu! approach to further explore and define any exposure-blood count relationship. The main analyses reported here describe blood count depressions in the months following new-onset ben- zene exposure in previously unex- posed workers. Blood counts for groups of individuals- are followed prospectively for 12 months after each worker's onset ofexposure. in contrast to our previous cross-sectional ap- proaches, which did not account for a worker's prior exposure history. These analyses an: based on repeated-mea- sure analys:is of vanance for counts starting from an unexposed baseline in previously unexposed workers. To minimize interindiv1dual variability, this technique establishes unique baselines for each sJbject and then serially tra,:ks subsequent measure- ments. This has not been used previ- ously in sc:rial analysis of medical monitoring data for correlation with tox.icant exposure and may provide a valuable approach tc future analysis of longitudinally collected medical monitoring data. Methods For our analyses, described in greater detail previously. 1w- we used hematologic: surveillance records of 459 rubber hydrochloride workers from the PI:iofilm production deparTments of an Ohio rubber products manufacturer. For each person, we obtained a cover sheet that showed demographic data and-a chronologie work history of his rubber hydrochloride jobs with beginning and ending dates, as well as records of his blood counts. This data set consisted of handwritten medical department records of 17,800 blood counts perfbrmed on the 459 men. The date of sampling (from 1940 to 1975) and handwritlen results were legible for 17,279 of the counts. Sampling was performed at approximately :nd:-Hltai !~:k~. L'iUJl~ Ll:u!J ~~ Jnkr~~ 1i1n>ugh uu: prrs.>n\ 1\o.rk hi~l(fl k J wwflc pb at a spec:fi.: ttme and through that to either of two d!f1ertnt e.xposure estimate schemes: one developed by the National Institute for Occupational Safety and Health P and the other for the Occupational Safety and Health Administration by Crump and Allen.'" Each estimate scheme assigned a certain number of parts per million of benzene to each job for each different year the plant was in operation. ln the present studies, only WBC and RBC counts were completely analyzed because they provided the most reliable data for the time period in question before automated analyzers. Platelet counts were not pe-rformed on the original sample. All statistical analyses were performed using SAS software.' 9 Further examination of the hardcopy hematelogic monitoring records resulted in a total of 161 workers on whom actual preemployment blood tests were performed and recorded on the material provided to us. This is 120 (preemployment) counts more than we used in our original analysis of 17,2'79 counts. The additional blood counts had been recorded on hiring and personnel forms rather than on ''medical data recording sheets" from which we initially transcribed the recorded counts-of an individual. We had not routinely abstracted these forms with predominantly personnel information during the initial process of data entry used in our previous studies. and would not have considered these preemployment counts in our previous analyses because they were not linked to a particular job and exposure. The lack of preemployment counts on the remaining 298 St Mary's workers is an unfortunate gap in the available data. which is probably due mostly to the incomplete nature of the combined employment and medical records which we obtained. Secular Preemployment Trend To address the issue that secular trends in the blood counts, independ- :~. 1 .;l Tdgn .1 . .. .~un ,.:~, ,: ~,li.~c.Lr ~li.li ;10nilJn ~: th~ pre. .ousl:- rcp,:t<:'d aggregate lil crea~s 1n WBC and RBC .:ounts fJJr the cohort. preemplo~ ment WBC and RBC counts were regressed against time for the entire 1940 to 1970 time period and specifical!v for the !940 to 1950 time period. where we have previously reported substantial effects. Blood Count Changes in the First Year of Work Blood count depressions relative to an unexposed baseline for that worker were sought within the first year of starting work with exposure to benzene. We reasoned that effects ofbenzeneinduced marrow depression on workers' WBC counts would be expected within weeks to months of the onset of exposure and that effects on RBC counts might take longer to manifest because of the circulating RBC's longer lifespan. Confounding by job transfers due to observed low counts was anticipated to potentially diminish any detectable effect of benzene based solely on an analysis from time ofonset ofexposure. This is dealt with in the subsequent set ofanalyses. All workers who actually had a~ baseline count or had a blood test "'ithin 2 months of the start of work were eligible for this analysis, and initially no stratification was done by level of subsequent exposure. The 2month window is used because only 161 workers have a true preemployment count recorded in our data set. We then computed the number of months (rounded to the nearest month) from onset of exposure to the time of each blood sample during his first year of work. For example, a worker whose first blood test was 2 months from the start ofworlc would not contribute _to the mean count at the 0 and 1 month mean counts. When- the results of more than oneblood sample were recorded in a !month period, the mean of an~mple results within that month was used. The mean WBC and RBC count across all eligible workers by month was then comp1.1ted. We restricted this analysis to workers who had five or more blood tests in their first year to BP-00020076 :~.. h!S l iiSl p_ :...[- 1hi~ m;.nth-h:\-:-~ ;;l~.\(trnoarall\1~ , ~J >~!' 1.\;ts appilc''~ ,_. ,clu<>I\t' 1\ to the ()46 thro,tgh 194G :n'l~ mt~nal. Thts 1nter.a! was chaser t.J ma~tmize the rtumber of eligible subJects (new hires\ and mmimize the number of years over which data were examined, Large numbers of worker. were hired in this period. a~parently because of the combined effects of the end of World War 11 and the rapid expansion of production with a tripling of plant capacity,'~' and we were thus able to identify a relative!: large number of workers with baseline counts. Simi- larly, we chose the 4-year interval be- cause a more narrow interval would not provide-sufficient new .workers to permit detection of significant trends. An overly broad interval might result in considc:rable bias as well as hetero- geneity of the anticipated effect be- cause late hires would be exposed to lower benzene e~<.posure than would earlier hires if, in fact. exposure was declining in this period. as we have previously suggested. Regression analysis was used to test for-trencb in WBC and RBC counts in the first year of work. The entry criteria offive counts or more in the first year and the ..time window of -enrollment" were ,aried in a sensitiv- ity analysis to determine their effects on the n:sulting hlood count versus time curves. Exposum-Biood Count Relationships in the First Year of Work The primary goal of this extended analysis l)f the Pliofilm hematologic data was to explore more thoroughly any relationship between exposure and blood count distinct from therelationship between blood counts and time. The aim ""as to account for effects of secular confounders and job transfers due to effects of pre~ ou~ exposure. Because of uncertainty about inhaled and dermal benzene ex- '~ :l lttn I.; : j;_ ' .1 . .. Jart . 'cl- tim,, f L '.iS , ,. ~-m.p .; -.~-~ rt>(nrJ-; -fl1 , ~ ::;" .,]mu"t 75~ r the ~~>)ri\c:r< ' "ei' hir~l betv.een !9411 and .,,1, ~ h.1nged .pecific JOb classific.w.'n' lar.d thus -potentially benzene e\posurei at least once in the first ;ear. Thts is a probable explanation ofwh.y analyses by specific job classification here and tn earlier analyses failed to demonstrate consistent patterns in blood counts (unpublished data). Because of the frequent changes in specific job classification, a new approach was devised. For each of the time periods from 0 months to \2 months in our previous analyses. the median benzene exposure as estimated by Crump was computed.' 011 Based on job assign- ment and year, the median exposme, as estimated by Crump. was computed for all the workers who were sampled at each of the 0 to 12 month periods. The median expo- sures ranged from a low of 30 ppm (month 0) to a high of 54 ppm (months 4 and 5). The sampled workers were then dichotomized into those whose estimated job/exposure was below the median and those who were above the median for each month postemployment. A two-way analysis of variance (time by exposure category) with time as a repeated measure factor was performed. Blood counts for each level of exposure were also compared at each month using t tests. To demonstrate the robustness of the approach. the analysis was repeated with changes in the two parameters discussed above: the time window-~nd the minimum number of-blood tests in the first ~ear requined for inclusion. A narrow window of 1948 to 1949 and a wide window of 1946 to r955 were used. We also varied the minimum number of blood tests required on each worker in the first year of work from 3 up to 11. Identical firstyear analyses were also performed using exposure stratifications based on those estimatedby Rinsky et al. 1516 This same analysis was performed on workers hired between !955 and 1960 as a comparison to the earlier, probably higher exposure, time pe- Recsuits Secular Preemployment Trend -\ r-:gress1on of the 16 I preemplo!ment \l, BC and RBC counts against time over the period !940 to 1966 showed an almost random scatter. The,: values for WBC and RBC ver- sus time are .0168 (P = .10) and .0412 !P"" .01 ). respectively. When the pre1950 time period is examined, where the strongest correlations between time and all blood counts was previously demonstrated, 10 11 the number of analyzable counts decreases to 93. The r1 values are .044 (P = .04) and ,046 (P = .04) for preemployment WBC and RBC versus time. respectively. Determination of aDepressant Response to Benzene Exposure in the First Year of Work Figure I shows aggregate plots of mean WBC count versus the number of months in the plant during each analyzable worker's first year in the plant. Only subjects hired between 1946 and 1949 with a blood test performed either before the stan of work or within the first 2 months of employment are included. Thus, for month 0 (true preemployment} the number of workers is 34. for month I it is 57. and for month 2 it is 61. Subsequent months, by definition, all have less than or equal to 61 workers, ranging from 50 to 60 in months 3 to 6, 40 to 50 in months 6 to 10, 34 in month 11, and 25 in month 12. Ex amination of the plot for WBC count shows declining counts for the-first 6 months, followed by a slight rise back toward the baseline. No job title re striction or classification is employed in the analyses. Regression analysis on the. first 4 months of data shows that the decrease in WBC count was statisticallv significant (r1 = .79, P = .04). Incluci. ing the first 6 months of data results in an r1 value of .51 (P = .07). The lower r 2 value for the 6-month period is likely a result of the higher value at BP-00020077 JOM Volume 35, Number 8, August 1993 WBC 9,500 9,000 8,500 '-----'---' ._ _L___L_ _ t _ _ j _ _ _ . L_ I __j__.L___j_-.L___l 0 2 3 4 5 6 7 8 9 10 11 12 Months on the Job F"~g- t White blood cell (WBC) count versus months on the job. For workers beginning work from 1946 through 1949 the mean of all WBC values for a given number of months after beginning worlc is shown. At month 0 (troe preemployment). 34 workers were sampled. After I month on !he job. 57 workers were sampled (23 forthe first time): at 3 months n = 61. Error bars I SE. 5 months. The data for months 6 through II show a statistically significant increa.se (r 2 .65, P = .05 ). Similar analyses for RBC count failed to show any consistent or statistically significant pattern. Relating Ex.posure to Blood Count Changes Although it is interesting to show a decline in WBC count for the first 4 to 6 months of exposure, we sou_ght stronger evidence of a direct expo;. sure-blood count relationship, taking account of differences in estimated exposure bt:tween workers. Figure 2, A and B, clearly shows the effects of exposure wlhen it is dichotomized-as above or be1ow the median for people sampled in that month. The median estimated exposures ranged from 40 to 54 ppm-f:xcept for month 0. In the case of-WBC, the high exposure means are consistently below the means of the low exposure group, except for months 0 and 12. A twoway analysi!; of variance with repeated measures showed that the mean WBC count was highly significantly different as a function of the exposure level (F"' !2.31, P = .0005). The plot of RBC count versus months on the JOb shows an even more impressive effect. Tile high exposure group is consistently below the low exposure group for the entire time period. Analysis of variance for RBC count showed a sig- nificant exposure effect (F = 61.03, P < .000 I). 1 tests conducted-for each of the months 0 through 12 showed a significant difference for WBC count at months I and I0 and for RBC count at months I and 4 through 12. The RBC count difference at time zero is not statistical!y significant (P > .05). Because the-baseline (month 0) RBC count in the high exposure group was lower than in the low exposure group, we reanalyzed the RBC count by month relationship with all values in the high exposure group adjusted by the baseline difference. The effect ofthis adjustment was to reduce the overall differences between the high and low exposure groups. The effect of .exposure in the two-way analysis of variance was still signifi- cant however (P = .03).-Examination ofthe adjusted values suggests that the RBC count differences occur in the later months. Dividing the 12 month period in.half, 0 to 6 months versus 7 to 12 months, we see no exposure differences in the 0 to 6 month period (P = .88) compared with a highly significant difference in the 7 to 12 month period (P= .004). A-similar baseline adjustment was not made in the WBC count by month analysis because the high exposure group had a higher WBC count at baseline and an adjustment would have increased an already highly significant difference. We present the more conservative nonadjusted analysis. By narrowing the time window to include only 1948 and 1949 hires, the exposure effects could no longer be seen. However, this narrow window reduced the number of workers to between 10 and I5 for most months. Choosing a larger window, 1946 to 1955, showed effects similar to those ofthe 1946 to 1949 window.-Varying the requirement for a minimum number of blood tests in the first year of work showed essentially the same exposure effects for both WBC and RBC counts as the five-test minimum discussed above. One further analysis was performed using workers hired between 1955 and 1960. Exposure effects seen in the earlier time windows were not apparent on the WBC and RBC counts versus months on-the job plots. The analyses were rerun using the 1946 to 1949 cohort, substituting exposure estimates made by Rinsky et al'6 in place {)f the Crump exposure estimates. When this was done, no separation in WBC or RBC count by median exposure category could be detected. Discussion These data strongly support the use of hematologic surveillance data to improve understanding of the benzene exposures of the Pliofilm cohort, and should eventually provide- improved risk assessments for- benzene leukemogenesis based upoit this co- hort. Yet the use ofblood counts such as these are not a complete or ideal tool for the retrospective inference of benzene exposure. Theoretical problems include the insensitivity of the marrow at lower doses and potential BP-00020078 .. ~~ :ocT {'I r"'iOuSh :,f'I(..,; \\jlt._ C<:un;' ac: tmer>el) ,-,yrelakd v.ith esttma1iJ ~ 1.soc - benzene con.:enlrdcions for the enttre cohLm from 194(1 to 1948. w' The present repeated-measure<:- analysis of variance demonstrated consistent ph vsiological response to benzene exposure within individuals over a defined exposure period. where secular effects are mmt mal. We sought to analyze workers who were not ex- 8,500- posed to high concentrations of ben- 8,000 _ _ _ j _ __l_ 0 12 _ l __ 3 _ J _ _ . . . L __ 45 _ L _ _ j l _ _ _ J __ 678 Months on the Job _ _ L_ 9 _ _ L __ 10 _,__ 11 12 zene before they s1arted work in the Pliofilm plant and came under hematologic surveillance. This may not have always been the case because it ~Low Exposure o High Exposure- is believed that some of the workers starting in the St Mary's Pliofilm op- workera hired between 1946 and 1949 with at leaet 6 blood teats in the linl year eration that we studied may have worked previously in benzene-exposed jobs. To the extent that was ~ true, it would have biased the results of the present analyses toward the null, making it more difficult to show decreased blood counts in the first several months of work. because our 5.1 ~- 4.9<)).11 (. 0 0 'J o o 00 ! baselines would not truly-reflect unex- 1I posed values and would possibly reflect already pathologically depressed I bone marrow. This was likely less of :: l _ _ - - '~_,__.____j__.l_--L_...L_____l._....JCL--r'-; a problem with 1he large number of truly new workers hired in the postwar expansion than in the pre-i946 workers w.ho were excluded from these 0 2 3 4 5 6 7 B 9 10 11 12 analyses. Months on the Job To verify the observed decrease in WBCs in the first 4 to 6 months of -e- Low Exposure ') High Exposure employment, we ran the analyses with different constraints. We varied the- Workers hired between 1946 tnd 1949 wi1h at lUll 6 blood ttall in the lirat year F"~t Z. Pkns of mean WBC (A) and RBC (B) for workm beginning work from 1946 to 1949 afte~ a specified number of months at work divided according to whether an individual worker was cstimat1:d to be above or below the median e~posure (a~~ro~imately -40 ppm) for -that month at the: time of sampling. The difTerence between high and low e~posure groups calculated by two-way analysis of variance with repeated measures was highly signiftcant (P < .0005) for WBC and RBC (P< .0001). number of blood tests we required each worker to have to be included in the analysis from 3 up to 11. The same basic "V" shape emerged, regardlessofthe setting of this parameter. When more counts were required for a worker's inclusion. the scatter became greater because of the smaller number of workers included. For example. when we required only three counts resistance to an- effect in selected procedural changes, and the admix- per worker, there were 81 person- workers. Recovery of function over ture of surveillance and clinically in- means that were averaged. When we time as <:xposure decreases is also dicated counts without labeling differ- required 9-t~ts in the year per worker. problematic for interpretation. Prac- ences. Yet on the pesitive side. blood the 2-month mean was based on 31 tical difficulties include the clinical count depression is a relatively spe- workers. The other parameter of in- nature of the datil collection, the cific effect of benzene exposure in an terest was the time window of 1946 to sketchiness of protocol details, uncer- exposed cohort, especially as this co- 1949. We varied this from a low of 1 tainty at<Jut laboratory personnel hort is said to have not had exi>OSure year to a high of I0 years. In each changes in those performing counts, to other hematotoxins. 14 We have pre- case, we could identify a decrease in BP-00020079 the fir.;t 4 to 6 mon: h ~'! n,)! loqe< ' 'J There alfe a fe\\ por,:nttal npLl~1d UH .1ut~ -:ur~. e). P~~-.11~:'.: '...... ~l~ ld r ' 1r.e~)( 1 :_1; a~.',C 1 art' r:l .! >':11gtr~-::.1. tions for the apparent rncreaS< Jn l1e m 1he i 2(Lda) iile<.pJn of th: FH( !T: : h~ t :n:m ber;hJr's ac:tu:;i exp.-:it.trt. WBCs ov;!r mooths fi to ! I. It could creating a much mu-r-e L'ompk:' ~ .,,. m '"':lassHica!lon possibly be due to a physiologic ad- cokine1ic interaction v.1th txnzenc ex- \!though the correlations between aptation to continued benzene expo- posure. JOb change. and sampling preemplovment \VBC and RBC counts sure, although there is little biologic frames compared with that of the and time are both statistically signifi- basis to suppon such an adaptation. mostly more short-lived WBC. In ad- cant lor 1940 to 1950 (P < .05 ), they More likely it could be attributed to dition. there is some indication that are quite small (r 2 "' .044 and .046. some bias in the data, such as sam- blood transfusions may have been respectively) and do not represent a pling bias due to removal from expo- used to treat some "anemic" workers. meaningful upward time trend in the sure of workers w1th more severely This would probably not impact WBC preemployment counts for 1940 to affected counts. For example, there counts, except possibly immediately 1950. nor for the entire 1940 to 1975 was a policy that workers with low after a transfusion. but would dearly period. This can be seen when their counts were to be transferred to other wreak havoc with a sequential re- magnitude is compared with the r2 areas in the plant where there would peated-measures analysis of worker.;' values. of all WBC and RBC counts be less exposure to benzene or they blood cell counts. vmus time (;2 for WBC count versus were given medical leave.20 It was somewhat of a surprise to see time= .88; r 2 for RBC count ver.;us The potential importance of a sam- the dramatic separation in the RBC time=- .53 for the 1940 to 1948 time pling bias such as this is underscored counts by exposure category because period). 10,11 by tbe 75%--rate of job change in the the original plot of RBC counts by lim year of hire. Any bias due to job changes for the purpose of reducing months on the job failed to show any consistent trends. A possible explana Conclusions benzene e~posure would again bias tion for this is that the RBC count Using different approaches to ret away from the significant blood count remains fairly stable for some months respective exposure assessment, we depression:> we repon and toward a at a specific exposure level, eventuaJ.!y have been able to demonstrate repeat- rising trend in worker's counts as time declines, a.~d then remains persist- edly a relationship between various passed. Another possible reason for an ently low in higher exposed workers surrogates of benzene exposure and increase in the later months is that once they have been moved to a lower WBC counts in the workers from the regardless of the date of hire (from exposure environment. This is again St Mary's Plioftlm cohon. These sur- 1946 to 1949), there might, on aver- consistent with the kinetics of RBC rogates include duration of work in age, be a lower exposure in the last 6 production and senescence, in that red the plant, calendar year, and job .des- months Cl>mpared t-o the first 6 cell life span is 120 days, whereas the ignatiun. RBC count effects appear to months ifthe trend toward lower ben- life span of granulocytes is only a few be more complex-when time factors zene exposure in the entire 1946 to_ days. are introduced, but were clearly pres. 1949 time period is real. Although There are two likely explanations ent for analyses stratified by j~re such changes did occur from time to for the absence of demonstrable ef- lated exposure. We have also clearly time, one would not expect them pre- fects in theJ 95 5to 1960 hiring period: demonstrated -the small size of any dictably to be of great magnitude on I) there were very few hires between secular trend in preemployment an ongoing; basis for each new hire, 1955 and !960 {approximately 10 counts. and this is not a likely explanation for workers per monthly mean) and 2) These results again suggest that the the effect we have demonstrated. because of the reduced range of ex- Crumpet a!. exposure estimates were The approach that makes a median posures in these later years due to more realistic than were the Rinsky cut of exposure for aU workers who engineering control, the separation estimates. They also begin to docu- contributed to the monthly mean re- between high and low exposure was ment that newly hired workers as a sulted in a successful separation of the not great enough to show a difference. whole in this facility experienced clin- WBC and RBC ver.;us months on the The inability of Rinsky estimates of ically detectable bone marrow depres- -job curves. We see this as evidence exposure to discriminate between sion likely due-to benzene exposure. that those workers in the higher ex- high and low exposed individuals even These data further suggest that sensi- posure areaiS had consistently lower in the 1946 to 1949 period is consist tive surveillance systems need tostrat- oounts thafl did their lower eKposed ent with our previously .published ify workers based upon exposure to colleagues. The differences at months results, which demcmstrated strong . facilitate detection of effects, which I and 10 were-sigmficant for WBC WBC count_-versus Crump relation- may only be significant in those with count (P < .05), but more'important ships-and weak or nonexistent rela- higher exposures. Many of the limi- is the consistency of the difference tionships between Rinsky exposures tations of this data set are due to its over the e~tended initial exposure and WBC count. 11 We believe this is incomplete nature and poor 'docu- period from I month to 11 months. lik-ely due to -the narrow range of mentation of the actual data collec- No consistent trends were observed exposure pennined in the Rinsky tion protocol. We hope that currently BP-00020080 1:' ~ 1, r~ , .... . ~-. r.. 5.. "'If." -~~; L niLi!H~~ ::Jt\;fC'I.lOI~~<"' : 1>1 -'''lGt Sl': . '' '':n :ratt: upon .::" ~ ,.g ol1ndl, ;,:. ua xorkers fmm J,:>b t<: .1.\b. It i~ r~L "g11Zed thot the Srrtdli nurnber of data points m the early :-ears of the cohort "'1:1 alway:; limit analysis of the pres- ent data S<!1 for temporal changes of benzene effects on blood cell counts. Acknowledgments We thank Dr Daniel Wanenberg, Dr George Rhoads, Derrick Petersen, Dr Sandra Mohr, and Dr Bernard Goldstein for advice and assistan~. We a!'f' grateful to Patricia Hutty for e~pen assistance with manuscript preparation. These analyses were funded under a contract from the American Petroleum lnstJtute and Dr K.ipen a!~ received support from NIEHS Center Grant ESOS022. References I. Goldstein BD. Hematotoxicity in humans. J Toxico/ Enviror: Health. 1977;2 (suppl):69-105. 2. Goldwater U. Disturbances in the blood foUowing exposure to benzol. J Lab Clir~ Med. 1941;26:957--973. 3. Greenburg L, Mayers MR. Goldwater L. Smith AR. Benzene (benzol) poisoning in the rotogravure printing industry in New Y1Jrk City. I Jnd Hyg Toxicol. :1 l "'1(""-:,''~' : ;c:~, !~~ , Br .' fr:L: ~1J' ':(;-! :~ 2Y0 ' .dm~ .U. C--Jr~~er P FLtdlan\.k BR =:~1~Lerda~ P~. \.11r RS. Braun J ; q.;d, Jfthe hemaroiogK etTecrs of chrlDi: k,.,.. e,eJ exposure tu hcnzene J Ouup !>ft>d. .'191. 33:619-6::6. 6. Tsai SP. Wen CP. We1ss \'S, Wong 0. '.1'cClellan WA. G1bscn RL. Retrospec- tJve mortality and medical surveillance studies of worker> in benzene areas of refineries. J Otnp .\fed 1983:25:685- 692. 7. Townsend JC, Ott MG. Fishbeck WA. Health e~am findings among individuals occupationally exposed to benzene J De- cup Med.. 1978;20:543-548. 8. Fishbeck WA, Townsend JC, Swank MG. Effect of chronic occupational ex- posure to measured concentrations of benzene. J Occup Med. 1978;20:539- 542. 9. Austin H, De!zeU E, Cole P. Benzene and leukemia: a review of the literature and a risk assessment. Am J Epidemio/. 1988;127:419-439. 10. Kipen HM, Cody RP, Goldstei11 BD. Hematologi effects of benzene: a thirty- five year longitudinal study of rubber workers. J Toxicof ltui Health. 1988; 4:411-430. II. Kipen HM. Cody RP, Goldstein BD. Use oflongitudinalanalysis of I!Cripbcral blood counts to validate historical recon- structions of benzene exposure. Environ :~liS ' 1 ~ ~-: . ' ' ;1 K rOC"n HM. C.och RP, <:roldstein BE) . 'hnatologK effects of ber.zene. a t.hi.rtY-'' li.t ,ear longnuctinal study of rubber , worker> flener). J Toxicol lllfi)iealtlu i,. i~89:5:Jl56-1158. , ,, an 14. Infante PF. Rinsky RA., Wago?~r J~r Young RJ. Leukemta in ~nzeoe work ers Lar~cet. i978;ii:76.:.78 RA:1~. Rinsky Young Rl, Smith AB. Leukemia [n benzene ~orkers. Am J Ind Med 1981;2:217-24~. ' 16. Rinsky RA, Smith AB,Jiomung R, Filloon TG, Young.RJ,-OkJI~(lli. Landri- gan RJ. Benzene and leu~~111ia: an epi- demiolOgic risk assessment: N Eng/ J Med. 1987;316:1.044-1050:~- 17. Rinsky RA, Smith AB, Honlung R, Fil- Ion TG, Young I,U, Okun RI, Landri- gan PJ. Benzene .and leukena; an epi- demiologic risk aS-menL fSDHEW. PHS, CDC. NIOS,H 1987. 5 IB. Crurnp-K, AUen' B. Quantiljltive esti- mates of risk ofleukemia.fr01n occupa- tional exposure to benzene.1 Occupational Safety and Health Admilistralior~ /984;Docket H.()5PB. - 19. Cody RP, Smith JK. Applied ..Statistics and the SAS Proiramming ~ third ed. Amsterdam: Elscvie~ North- Holland; 199 L ! ' 20. Wilson R.H. Benzene poisorun} in industry. J Lab Clir~Med. 1942;2l:IS17- IS21. Oh, To Be Rich wl never dreamed when I manied a Rockefeller that I would winfup :spending half .-r my time asKin~ people for money." (Sharon Rockefeller on her fimd-raising_duties as - president of a-Washington PBS station.) : From Overheard. Newsweek. 1993:' In:l, p If . - ~ - "" r .. . \ a [ s tl B BP-00020081