Document ByzKYjOZx1eXxZqNNk1QDeKRk

320. R. H. Burgess, ed., "Manufacture and Processing of PVC," McMillan Pub. Co., New York. 1981. 320A. r. w. Coutant, "Laboratory Evaluation of Commercially Available Passive Organ c Personal Monitors," EPA-6Q0/4-82-Q31, April 1982, PB82-234261, L. A. Wallace and H. R. Ott, J.A.P.C.A.. 32 601 <1982), 3. B.Perkins. N. H, Price, L. Eggenberger, and J. A. Burkart, "Evaluation of Passive Organ c Vapor Monitors," December 1981, PB 83-221028. 3206. Z. 0. Bell, Jr.p 3. C. Lafleur, R. P. Lynch, and 0. A. Work, Chem. Eng. Prog,. 71 (9) 45 (1975). 321. F. R. Mayo, C. Walling, F. M. Lewis, and M. F. HuUe, 70 1523 (1948). Ahk Chem. Soc.. 322. S. J. Tunkel, CEP, 79 (9) 50 0983). 323. D. W. Klondike, Prof. Safety. 28 (4) 17 (1983). AP00020745 82-9.2 Risk Assessment for Vinyl Chloride in Perspective John T. Barr Air Products and Chemicals^ Inc. ALLENTOWN; PENNSYLVANIA For Presentation at the 75th Annual Meeting of the Air Pollution Control Association New Orleans, Louisiana June 20-25,1982 AP00020746 AP00020747 *2-9.2 Introduction The COTbtnatfon of cl rctastames which found the carcinogenic hazard of vinyl chloride (VC) being discovered at about the sane time as the science of risk analysis uas undergoing rapid development. and the great commercial interest and long history of ase of the substance has resulted In body of literatere end pharmacological data greater than one can expect to have for Most substances. It is therefore instruc ts* to review the many risk assessments which have been prepared for VC against the available biological Information to determine if we can evaluate the extrapolation methods used, and to discuss the current regulations for VC in light of this eoaperlson. Hazards of Vinyl Chloride It is necessary to decide first which of the hazards presented by VC should be the basis for the risk estimation. The substance presents the acute hazards of frostbite from exposure to the liquid, of anes thesia at concentrations over 6,000 ppm and suffocation at higher concentrations (von Oettfnger, 1955). It also forms explosive mixtures In air above 3.7$ volume percent, and sd the efforts to control the physical safety of operations generally preclude exposure to acutely toxic concentrations. These control efforts were reinforced Id the aid-1960's where It ms discovered (Sucta, 1963) that workers who hod been exposed to very high levels of VC developed "vinyl chloride disease.*' the prfMry manifestation of which was acreosteolysls (AOL), a degenerative disease of the bone tufts in the hands, and more rarely af the feet and luabar reglon. Although crippling to some degree, this disease is not fatal, and Is at least partially reversible if exposure is oliainated (Greniger, Walker and Ward, 1960). Almost ten years later ft ms found that some of the workers having eleilar exposure also were developing angiosarcoma of the liver (A5L), a rapidly fatal disease. Oddly enough, there is only one possible case af a worker developing both AOL and ASl (Stafford, 1991) Wong cas*5 * *L 90-plus cases of ASL now known worldwide, although both are diseases of the vascular system. Several large epidemiology studies were conducted on workers exposed to VC (Baxter and Fox, 1976; Chiazze, 1960; Puck. Carter and Coombu, 1975; Equitable Environmental Health, 1978; Fox and Collier, 1977; FrenUel-Beyae, Schnitz, and Theiss, 1978; Theriault red Allord, 1981), and ASl was the only fatal disease found consistently to be in excess in these 2 persons. Animal studies have shown on excess of tumors at other sites, but the lowest exposures at which these occur are considerably higher than that for ASL- For example, Hal ton1 (1979) reported the following data: Site Concentration Far Significant Elevation Forestomoch papillomas: Heurobl astomas: Zyabal gland carcinomas; Nephroblastomas: Liver angiosarcoma male: female: Hammary adenocarcinoma: 30,900 ppm 10,000 ppm 10,000 ppm 250 ppm 200 ppm, SO mg/kg 50 ppm, 16.7 mg/kg S ppm The low concentration for onset of summary tumors was af concern when a preliminary study of fabrication employees reported an excess of breast tumors (Chiazze, et al., 1979) but a fo11r-qJ case-controlled study (Chiazze, 1980) found no association between the cases and VC exposure. The largest study of VC-PVC workers in thm United States reported slight excesses of brain and lung tumors (Equitable Environ mental Health, 1978), but this was not seem in the other studies referenced above. The excess af brain tumors was small, and not dose er exposure-related. The overall excess of lung tumors resulted from an excass in one plant only, and reexamination of those cases also showed no association with VC exposure (Waxweiler, 1978). Vinyl chloride has been found to be active In several in vitro outa genetic tests with bacteria and yeasts (Hopkins, 1979) end It appears to cause chromosome abnormalities in exposed workers, but these changes are reversible when exposure Is reduced (Kansteene, 1978) and several studies of neighborhoods around PVC plants have failed ta show a supportable association with birth defects (Edmonds, 1975, 1976). It is not a teratogen in rodents (Johns, 1977). Therefore it appears reasonable to assure that If there Is any signif icant chronic risk other than ASL, it Is considerably smaller than that for ASL, and that an adequate risk assessment can be based on only the liver tumors. 82-- MOTE TO EDITORS Under the now federal copyright law, putMIcbtlon right* to this paper *>* retained by the authors). 3 AP00020748 teyfcw of Bisk Assessments 1. Schnfidermen, 1975 One of the first attempts to utilize inlwl dot* to estimate risks at very low exposures was that of Schneldeman, Mantel and Brown (1975). They used preliminary Kaltonl results to compare the estimates obtained from three possible mathematical models. Thefi99K assurance level of a "safe" dose at a Kfetlae risk of 10 was estimated from several extrapolation models as follows: Log Problt (slope = I) logit (slope 3.45) Logit (slope 2.3, one-hit) 73 ppb 119 ppb 2.1 ppb The authors discussed the recognized difficulties of extending these rat data to humans and of providing animal experiments that could answer satisfactorily the question of hueaii risk at very low doses. 2. kuzmadc and McGaughy. 1975 The EM was the first group to atteept a human risk assessment for vinyl chloride (Kuzpack and McGaughy, 1975). This pioneering effort attempted to use both animal and human data, and to show comparative results from both the linear and log-problt mgdels. It concluded that there was an Individual risk of 71 x 10 per pp* of lifetlne exposure to VC by the linear extrapolation method, and that the log-problt results were one-tenth to one-hundredth of that. This effort Is subject to several serious criticisms. The exposure data used for human experience was that froa a group with less than average exposure, while the ASl rate was chosen from only those plants which did report cases, and Ignored the remainder of the population. Thus, their incidence rate of 7.5X compares to an actual figure of about 0. IX. They used as their primary method a linear extrapolation of rat data, which often has been seen to overestimate the actual rates by at least two orders of magnitude, and they assumed the total cancer rate to be twice that found for A5L. This same estimate was used by the EPA (1979) to estimate the concentration of VC In drinking water which would produce various levels of risk. These estimates are, of course, subject to the same criticisms. Nlsbet (1970) challenged the estimate of Kuzmack and McGaughy (1975) when It was used by Wllsan In testimony before the OSHA hearing on Its generic cancer policy. Nisbet stated that his calculations showed the risk to be 10*30 times greater, by the same calculation method. Wilton (1978) suggested several flaws 82-S.2 4 In the Nlsbet procedure, Including the fact that he chose for his extrapolation one point at 25 ppm from Kaltonl experiment BT-15, and that this point is not In good agreement with the whole body of data. Further, he chose to use total cancer Incidence in the rats, including those at zywfcal glands, which have no counterpart In humans, loth Wilson and Kucmack and McGaughy had used a factor of two times ASL to account for possible cancer at other sites. Wilson did acknowledge mathematical error which made his results half the proper mariier. Albert (1978) applied this same general procedure to other poten tially carcinogenic air pollutants In the United States and calculated the expected annual cancer deaths as follows: Arsenic Benzene Cadmium Coke ovens VC (after regulation) 15.6 77.B 26.2 149.5 1.0 3. Gehring, 1979 Gehring, et el., (1979) applied an experimentally derived biotransformation correction (Gehring, et al., 1978) to rat data and estimated the incidence In humens at two different exposures by means of four different extrapolation models. Their estimates at 500 and 200 ppm TWA bracket the observed experience for huaws when derived from the probit and the unconstrained linear models. The linear-through-zero and one-hit models consistently over estimated the incidence. Although not considered by the authors, the linear and probit models match rather closely the total U.S. experience of occupational ASt at an assuaed 1,000 ppm exposure. The linear model predicts no incidence below 99 pgm.in humans. The problt model predicts a human risk of 1.5 x 10 at 1 ppm. Thus, a mechanism for adjusting for the difference In metabolism between animals and humans appears to be useful. A limitation of the Gehring procedure is that It uses partial Kaltonl data, and tests the results egelnst the CMA epidemiology study. That study was mot the "end of the experiment"; It stopped at the end of 1973, and several deaths have occurred since then. Neither did it cover the entire population, but only the employees of these plants which met certain criteria for data retention and length of operation. The Stafford (1981) data does cover the entire population and extends the history for seven years. The size of the population is not known, hut a reasonable estimate, based on normal worker turnover rates and the number of plants not Included In the CMA study, Is certainly not less than 25,000. This would give a gross Incidence of about Q.1X. Of these, the number actually exposed to substantial exposures would be about 25-30 per plant at any one time. Nultipl(cation by 25 plants, and a factor of three for the turnover during this period, would give about 2,000 highly exposed persons, for an effective Incl- 82 5 AP00020749 t fence of Just over 1%. Portonel experience would Indicate that, for the period prior to 1964. when all or the first exposures of the fatal 2( cases had occurred, the average exposures of this highly exposed group certainly was In excess of 1,000 ppm for the working day. Maltonl (1979) found a IX incidence at about 1*10 pps in rats. Calculation of the dose equivalent to IX incidence in rats gives 0 ppo by the linear method and 7.5 ppa froai the log-prebit equation far the ccabined Hal tori Inhalation experiaents. This crude and subjective estimate would then say that man is about 100 tiaes as resistant as tha rat to VC inhala tion. a figure generally In agreement with other estiaates (NCAB, 1979). 4. Food Safety Council 1979, 1990 The Food Safety Council has rCo--ended (FSC, 1978) the use of the ga--a auUi*hlt modal because of its flexibility In handling dose response data of varying eurvtlfnearity at low doses. It has calculated (FSC, 1908) the aaxtaua likely aid lower 97.SX Halt doses for substances at various risk levels and with different models. Fer VC, at 10 * risk, these resalts are as follows (based ea early Maltoflf data): One-hit Arnltage-Doll Welbull Multi-hit 2.9 x 10 } ppa 2.0 x 10 ppa 2.1 x 10 ?QPpa 3.9 x 10 ,,rppo Fer this substance, the goodness of fit of the Welbull aodei (0.S6) was superior to that of the awlti-hlt (0.32). Neither of the other two Models gave acceptable fits. This was In part because of the concave shape ef the curve, which included all of the high doses in the dose response data. 5. Dow, 1979 A Dow Heath Team performed a relative risk estlaation for several compounds danger, et el., 1979) Which considered probable expo sure, the consequence of exposure, the physical state of the substance during processing, and the current exposure standards. , This resulted in a value ef 480 far VC (n a "closed system but with employees la the vicinity." The same procedure assigned hazard rating values to some other substances as fellows: benzene, 10; phosgene, <10; hydrogen sulfide, 5; arsine, 9,700; and bis-chloromethyl ether, 59,700. In a batch operation with occasional manual handling, the hazard rating for VC Increased to 9,700 by this method. 6. Hehir, 1990 Hehlr. et a!., (1980) conducted series of .tests for the Consumer Product Safety Comatsslon, a part of which consisted of exposing rats and mice to a saries of short, high exposures, rather than 02-9.2 6 the usual extended tow dosage, They Included one-hour exposures to rats and alee at 50, 500, 5,000, and 50,000 ppa, 10 and 40 hour exposures at 500 ppa, and 49 and 100 one-hour exposures at 50 ppa. After lifetime observation they found no effects on rats, or their offspring, nor on aico exposed to less than 500 ppa. Those exposed to over 500 ppa developed pulmonary adenomas, but they also had suffered from pneumonitis. They considered the published date en animal exposures and concluded that thtre was a lifetime dose below which no oncogenic response Is seen. This was estimated to bo 5,000 ppm-hrs fer alee and greater than 50,000 ppa for rats, regardless of whether the dose was admiaistered ever a short or long period. This concept of equality of effectiveness for all modes of exposure does not have general acceptance and would not appear to be correct, based on our present understanding af carcinogenesis. Oose-rate effects are, of course, well known. However, the degree -to which this can be extended to all types of effects is not known. These authors also used the Crwap-Ceess aodei (Cruap, Cues* and Deal, 1977) te evaluate their data an aouse pulmonary cancer, and estimated' that exposure to 5,004 ppa VC doubles the probability of cancer, while $0,000 ppo Increased the risk nine-fold. In view af the fact that pneumonitis was present In all animals exposed above 500 ppa. It is questionable if this was direct oncogenic response, or the result of an nongenetic event because of severe lung damage. Haltoni (1979) also reports an Increase In lung tuaors in mice, but not in rats or hamsters. Thus, the significance ef this finding to risk in humans is questionable, 7. Anderson, 1968 Anderson, at a!., (I960) extended the work ef Cehring, et el., (1978 and 1979) te incorporate the amount of metabolic products from VC which actually was bound to the 0NA ef exposed rats, (Gehrlng and llau, 1977) rather than the total amount metabolized. They assigned various values to the parameters In a Michael isHentan equation depicting the kinetics of the metabolic process, and coapared the results froai extrapolation to low doses by log-problt and multi-hit models. They faund that the two extrap olation models responded quite differently to these variations at very low doses, and that it was not possible to select one model as the more appropriate from the high-dose data. Use of the values of Cehring for the primary parameters7gave estimates of the dose equivalent to lifetime risks of 10 of less than I ppm for the probit model and less than 2 ppm for the multistage model, a correspondence which the authors pointed out was better than the precision of interspecies comparisons. 2-9. 7 AP00020750 8. EM, I9M) The final version of the water quality criteria document for VC (EM, 1980) used a different approach far risk estimation. The slope of Uw incidence of all tumors at the lowest doses of Haltonl experiment BT-1 was adjusted for the fraction of exposure, the equivalent feeding level te give the sane blood concentration of VC as by Inhalation (see Vlthey and Collins, 1978), and the ratio of the surface area of humansvs. rats, te produce an estimate that a tlfetloe risk ef 10 9 would be caused by drinking 2 1/day of water containing 20 g/1. There Is sons confusion In the mathematics given in the report, and the assumptions on which the adjustments are eade are far fro* having general acceptance, although generally following MS recommendations. It appears that this procedure overstates the risk by several orders of magnitude. 9. MS, 1988 The Rational Academy af Science (1977) calculated the upper SSX confidence Unit for risk from drinking water containing vinyl chloride from the probabilistic multistage model and early Haltonl rat data. They report (HA 1980) a lifetime risk of 10 * as being equivalent to 3.8810 mg/kg/day. For a 70 kg person consuming 2 1/day, this would calculate to an acceptable level of 1 g/1. The difference between the EPA and NAS mmbers comes from the different curve-fitting methods for the animal data. 10. Gaylor and Kodell (1980) applied linear "Interpolation" to the same early Haltonl data used by the Food Safety Council (1978) to arrive at a predicted maximum risk of 18 . The upper 97.SK confidence liuit of the animal dot# was taken as one point on the Interpolatlve lino, and aero incidence ot zero exposure as the other. Jbis produced a lower 97.5% confidence limit dosage of 7.1 x 10 m ppm for a lifetime risk of 19 6 in rats. Their_appli cation of the Arnitaoe-Ootl multistage model geve 5.2 j,10M ppm as the dosage at 10 lifetime risk compared to 2 x 10 t by the Food Safety Council. The difference Is due to alternative assump tions on the value of the exponential dose term. 11.. Crump and Guess (1980) reviewed some of the earlier risk estimates for vinyl chloride In drinking water, and recalculated the risks, using the one-hit and multistage models. They arrived at an tqtper 95% confidence limit of lifetime risk for drinking water containing 1 g/1 of VC of 4 x 10 , based on early Haltonl Inhalation data. Using the assumption that a 0.2% incidence of A5L In workers had resulted from a lifetime emosure of 70 g/kg, they obtained a maxiiwm likelihood risk of 10 9 from 0.34 g/1 by both the multistage and linear models, with 95% lower confidence Halts of the same risk ot 0.24 g/1. These two models reduce to a linear form when used at very low doses and with the assumption of no threshold value. 82-9.2 a These authors cite EM data on the occurrence of VC In public water supplies which by their methods yield o lifetime risk of 3.7 x 10 , or 12 deaths per year from this cause In the United States. None of those has been observed, despite the accumula tion of 15 years' data on ASL deaths (Popper, 197ft). 02-9 12. Scott (1981) ascribed the decreased Incidence of tumors In rats at the higher doses to a coll killing process, and adopted the Velbull model te account for this. Application of the model to some eerly Haltonl data produced a curve which fit the data froo 50-10,000 ppm. He did not attempt te extrapolate to doses beyond the experimental range. 13. Carlborg, 1981, else applied the Velbull model to 31 bioassay reports on a variety of animat carcinogens. Ha concluded that the one-hit model was not appropriate and that carcinogens could be divided Into categories according to the shape of the curve, e.g.. (oncavs or convex. He found that the early Haltonl data on VC fell into the former category. Application of Ms parameter estimates to those {lata, assuming no spontaneous incidence of ASL, gives 2.$ x 10 ppm for a lifetime risk of 10 8 for rats. Later calculations Including all of the ptdillshed Haltonl data did not change the results significantly (personal coamunlcatlon). He found the Velbull shape parameter to bo approximately 0.S, which Is assumed to be the number of stages for tumor Initiation. IhU Is consistent with the finding by Gehrlng (1977) of satur able metabolic path which produces the proximate carcinogen. It also suggests that the number of "stages" is the number of finiterate steps before the rate-limiting step, there eiy be other stages following, but they ore not rate controlling. Actually, there appears to be at least two saturable mechanism involved In the pharmacokinetics of VC, the metabolism to the ultlmte carcin ogen and the detoxification by suHhydryl grwpi. 14. One further evaluation of human risk can be made from the experi ence of persons residing near VC-PVC plants. The EPA estimated (Kutmack end NcGaughy, 1975) that five alllfon persons lived within five miles of these plants, and were exposed to an annual average concentration of 17 ppb. The present distrlbetloo of plants was generally well-established by 1959, thus we have 22 years of history, or about 110 million person-years. About five or six of these plants, with 1-2 million neighbors, go back another 20 years, but these data are not firm enough for inclusion. The fact that no case of ASL has been confirmed as arising from these ambient oxpesures places the upper bound of risk at less than 2.7 x 10 per ppn-yr. It is believed that the exposure data were overestimated by EPA. and thus this result may be too low, but It Is In the sane general range as that arrived at by Gehrlng (1979) and Anderson (1980) after staking corrections for pharmacokinetics. 9 Extending this crude calculation, these five Million persons are now supposed by EM to be exposed to 0.2 ppb (probably a high figure), which would predict no more than 0,0003 deaths per year, or one per 3,700 years in that whole population due to VC exposure. But H also must be recognfted that with approximately 20 cases per year of ASt in the general population, there can be expected from a purely statistical basis that there should be one case very two years or so among this group or S pillion plant neigh bors. The results of these estimates discussed above ere compared In Table 1, after conversion to a uniform 10 lifetime risk. Estimates 5, (Oow 1979) and 12 (Scott, 1901) wot* not in form to permit this comparison. See OSHA, (1980), for references to a few other estimates that were not considered here. It can be seen that the results fall Into two najor categories, those which project that the risk of 10 occurs at exposures or greater than I ppm, and those which find that risk in the ppb range. The estimates uhlch yield the higher allowable exposures are based on human data (Not; 3, 7 and 14) or use a log-probit extrapolation model (Ho. 2, second estimate), or predict e threshold (No. 6), The remainder generally are based on thd linear, noivthreshold model, and make no biological correction. The result is a difference of 3 or 4 orders of magnitude. The estimates which yield the higher allowable exposures are (n better agreement with human experience than art those of the other group. Additional Oata All of the extrapolations reported boro have used for the original Naltonl data from bis experiment W*l. He has now reported (Maltonl, 1979) three other cooporable Inhalation experiments on the sane strain of rats, and one an another strain. In addition to two Ingestion studies. The results of these experiments are shown In figure 1. an a log*prob1t scale. It can be seen that they all follow a similar pattern, but that there are large variations in slope between the various data groups. Table III contains the log-probit equations calculated from some of the individual experiments, and various groups of experiments. Excellent fits are obtained for a single experiment, as would be expected from the small number of data points, but adequate fits are ebtained for the group as e whole. Inclusion of the historic control data on ASL (Q.09% spontaneous incidence) did not effect the fit sutatantiatly, except for the very low dose data. Inclusion of the 0,0 (origin) as a data point did give significantly poorer fits. The combined experiments Indicate that a lifetime rfsfc of 10 * for rats is obtained from a dose In the 1-2 ppb range. Similar variation Is seen with the other mathematical expressions, such as linear or exponential equations. 82-9.2 10 O o N o 82- kegulatory Status The current regulatory status of vinyl chloride is summarized In Table II. The first regulatory action on VC was taken in 1973 when the Bureau of fax. Alcohol and Firearms prohibited the use of rigid MC as liquor containers. This was based on It being present as an adulterant, and not on any consideration of risk. The Consumer Product Safety Commission (CPSC), the Food and Drug Administration. (F0A), and the EM all acted te bon the use of VC as an aerosol propellant thus establishing a zero risk position. The FDA proposed (FM, 1975) to withdraw the prior sanction status of rigid PVC as a food package component because ef the concern for residual VC that might migrate. The fOA has taken no further action on this proposal, and now Is considering a "constituent" policy which would pemlt a lifetime exposure at some acceptable risk level. This risk has been proposed recently to be 10 lifetime for the gluttonous consumer. As was discussed above, the EPA required a best available technology approach which reduces the aeerage exposure to those within S miles of a plant to about 0.2 ppb, by EPA estimates. OSHA established a rule In 1974 which set I ppm for 8 hours as the maximum permissible exposure, and also set 0.5 ppm as an action level below which most features of the regulation did not apply. These were chosen as feasible levels, ami not necessarily "safe1* doses (OSHA, 1974; EPA. 1976). The EPA has established an exposure te the general population only 0. IX of that allowed In the workplace. The CPSC has required zero exposure, and the FM has considered that approach. Depending on which method of estimation the FDA may choose. Its allowable exposure could b either greater or less than those currently set by CPA and OSHA. it has been estimated that the maximum amount of VC Ingested by the average European, who uses much more plastic packaging than we, is less than Ze. fl/day, (CEF1C, 1976) which would be in the order of a 10 or 10 lifetime risk by even the most conservative models. There have been various estimates made of the cost-effectiveness of the Federal regulation for vinyl chloride. Graham and Vaupol (1981) estimated that the OSHA rule cost $7.5 million per life saved, and $490 thousand per life-year saved over the option of leaving the exposure Unit at 50 ppm. Luken and Hiller (1901) state that the iaputed value of a life from the OSHA standard is $4 million. Nerrell (1982) uses an annual cost of $20 million and an annual benefit of 0.1 life saved to derive a cost/beneflt ef $200 million per life for the OSHA rule. The EPA has reported (EPA, 1979) that the cost of compli ance with its VC standard was $296 million through July 7. 1901, and will be an additional $470 million during the next five years, all in 1977 dollars. If the EPA estimate of up to 20 dealths per year were correct, this would be a cost f $4.7 million per life. However, as discussed here, there Is no evidence that any lives have been saved by this rule. There are many difficulties In obtaining accurate estimates of this type, and serious probleos In determining the proper value to be assigned te a life, nevertheless, the doubtful nature of the claims 11 AP00020752 for any signtftcent benefit frmu these rules suggests that at best, these regulations are excessively costly to society. Therefore, we ust attempt to Iegrove bath our data base and our methods for inter* preting and applying the data. Discussion What can be learned froa this exorcise other than the already recognized fact that various extrapolation models can yield very different results? In this case, at least, there are several points which are worth considering. 1. Vinyl chloride Is no exception tb the rule that hum data always wist be incorporated whenever possible. The epldealc of occupatlonatty Induced ASL which was feared In 1974 has not Materialized, probably due to the steps that were taken In the early 1%0's to reduce exposure because of the discovery of AOl. Mo Instances of ASL fro* exposure to VC in the general population have been substantiated, the ovarpredlction of occiqiationat cases was due to the underestimation of worker exposure and overreliance on raw animal data without proper pharmacokinetic adjustment. We are not now able to extrapolate reliably between stellar species and certainly not From rodents to humans, without much additional data. 2. The regulations far vinyl chloride were not based primarily on scientific data, bet on socioeconomic and political decisions. This Is no surprise (Crandall and Lave, 1961), but is a fact which should be acknowledged openly, along with the understanding that this position will continue to penalize good science. 3. Mathematical extrapolation model* are not adequate in themselves for predictions of risks much beyond the experimental range, ns matter how good the fit Is to the data In the observed range. The variability of relatively small experimental groups adds to the error range. Thus, bloassayi intended for quantitative risk assessment applications should be at as low doses as possible, and as large as possible, and should be interpreted very cau tiously. 4. The current state of the art is iuch that quantitative risk assessments stay be useful for deterialniog relative risks from similarly acting carcinogens, but are not suitable fer acrossthe-board application to all mechanisms of carcinogenesis. This is not to say that we should dbondon efforts at developing mare effective risk assessment methods. We must, however, recognize the problems inherent in blind application of mathematical models without proper assessment of the available biochemical data, or an understanding of how applicable the experimental data are to humans. 92-9. 12 92-9We have available to us at least as much data regarding vinyl chloride . as we have for any other substance, and we still have difficulty In deriving a suitable expression for risk from a purely mathematical or statistical basis. Only when human relevance Is considered can we arrive at a prediction that approximates actual experience. The regulators are faced with a tremendously difficult task when they are presented with a few pieces of animal data which suggest the need for concern and potential regulation. We must develop a suitable program to obtain and use as much relevant data as possible to assure that rational regulations are possible. The vinyl chloride experience can help us understand the kind of data which are needed. 3785-A1 2/19/82 Ibh 13 AP00020753 References Albert, R. E., letter to I. S. Nivnh, EPA, "CiwptrfiM of vinyl chloride carcinogenic risks with risk fra* other pollutants'*, Washington. OC, 16 June 1976. Anderson, M. V., et al., Tom. AppI. Phare. 55, 754 (1960). Baxter, P. J., and A. J. Fbx, lancet, 1976 2(5. Carlborg, F. V., Fd^ Cosmet. To*. 19 255 (19B1). CEF1C CoMsittee for the Toxicity of Vinyl Chloride, "Vinyl Chloride Toxicity and the use of PVC for Packaging Foodstuffs," Brussels, Feb. 1976. CMazte, l., J. Occup. Wed. 22 (TO) 677 (I960). Chlazze, L.. Jr., V. E. Nichols, and 0. Wong, J. Occup. Med.. 19 623 (1977). Crandall, B. V., and 1. lave, "The Scientific Basis of Health and Safety Regulations," Brookings Institution, Washington, 1981. Crump, K. S., end H. A. Guest, "Drinking Water and Cancer", 2681-128167, December. 1980. Cruep, K. S., H. A. Guess, and K. L. Oeal, Biometrics. 33 4370451 (1977). Duck, B. W,, J. T. Carter, and E. J. Coonbes, lancet 1975 II. 1197. Edaonds, L., "Birth Defects and Vihyf Chloride", Proc. Conference on Women and the Workplace. Washington, OC, 1976, also Teratology. 17~T37 (RrtB). Edmonds, l. 0., H. Falk and J. E. Rissim, ttw lancet 197S 1098. Environmental Protection Agency, Standard for Vinyl Chloride, 41 Fed, tea. 46,560 (1976). Environmental Protection Agency "Vinyl Chloride, Ambient Water Quality Criteria", PB-292446, Washington, DC, 1979. Environmental Protection Agency "Ambient Water Quality Criteria for Vinyl Chloride," EPA 440/5-00-078, October, 1980. Envimomenta! Protection Agency "The Cost of Clean Air and Clean Water", Annual Report to the Congress, Deceeber, 1979b. Senate Document No. 96-38. U.S. Government Printing Office. Washington, DC. 82-9.2 14 Equitable Environmental Health, Inc., "Epidemiological Study of Vinyl . Chloride Workers, Final Report". Prepared for Manufacturing Chemists Assoc., Washington, DC, January, 1978. Food and Orug Adeinistration, Notice of proposed rulemaking, 40 Fed. Reo-. 40.529 (1975). ------- Food Safety Council Final Report "Proposed System for Food Safety Assessment", Washington, DC, June, 1960. Food Safety Council, Scientific Committee "Proposed System for Food Safety Assessment." Food and Cosmet. Ton. 16 Suppl. 2 Oeced>er 1978. Fox, A. J., and P. F. Collier, Br\ X lnd. Hed. 14 1 (1977). Frentzel-Beyme, R.. T. Schmitt, and A. ft. Thtess. Art). Socfalmed. Pravent.. 13 218 (1978), -------------------------- Gaylor, 0. -V., and R. L. Kodell. R. 1.. J. Environ. Pathol. Tox. 4 305 (I960). -------------------------------------------- Gehring, P. J., and G. E. Blau. J. Environ. Pathol. Toxicol. 1 163 (1977). -------------------------------------------------- Gehring, P. J., P. G. Watanabe. and C. M. Park. Tox. and AdoI. Phare. 49 IS (1979). ---------------- ------------------- Gehring, P. J., et al., Toil AppI. Pharmacol. 44 5B1 (1978). Graham. J. D., and J. W. Vaupel. Risk Analysis 1 89 (1981). Graniger, R. G., A. E. Walker, and A. N. Ward "Vinyl Chloride Monomer-induced Disease; Clinical, Radiological and lamenotogical Aspects," Chapter 11 in "Induced Disease; Drug, Irradiation, Occupation". 1. Preger, ad., Grune and Stratton, London. 1980. Hansteene, l.-L., et al., Mut^ Res^ 78 211 (1978). Kahir, R. M., et al., "Toxicology, Carcinogenicity and Reproductive Effects of Single and Nulttpla Exposures to Vinyl Chloride in Rats and Mice". Pre-publication draft, Feb. 6, I960, U.S. Consuaer Product Safety Commission. Washington, DC. Hopkins, J., Fd^ Cosmet. Toxicol.. 17 542 (1979). John, J. A., et al., Toil and AppI. Pharmacol. 39 497 (1977). Kuiaack, A. M., and R. E. NcCaughy. "Quantitative Risk-Assessment for Community Exposure to Vinyl Chloride", U.S. EPA, Washington, OC, December 5, 1975. Longer, R. R., $. *. Nonfood, 6. E. Socha, and H. R. Hoyle, Aai. tnd. Hyg. Assn. (12) 1039 (1979). -------------- 82 15 AP00020754 token. 1,8.. and S. 8. Hiller. J. Air hi. Control Assoc. 31 1254 (1981). ' a-BtU.--------------------- ----------- Maltoni, C., G. Lefemine, A. Clllbcrti, G. Cotti, end ft, Cerrettt, "Vinyl Chloride Carcinogenicity Btoaisayt, (8 T Project) as an Experinental Hade I for Sisk Identification and Assessment in ... Environmental and Occupational Carcinogenesis" Presented at "le Club de Cancerogenese Chimique", Institute Curie, Paris, Nov. 10, 1979. National Academy of Science, "Drinking Mater and Health", National Acadeny Press, 1977. National Acadeny of Science, "Drinking Water and Health", Vol. 3, p. 38, National Acadeny Press, 1980. National Cancer Advisory Board, "The Relation of lloassay Dose to the Assessment of the Risk of Carcinogens far Humans under Conditions of Low Exposure", Subcommittee on Environmental Carcinogenesis, Draft of 1979, citing data from Heselson and Russell, 1979. Nisbet, I.C.T., Post-hearing statement to OSHA docket 090, September IS. 1978. Occupational Safety and Health Administration, Standard for Exposure to Vinyl Chloride, 39 Fed* ReJL 35,890 (1974). Pepper. H. et al,. Am. J. Pathol. 92 349 (1978). Schnelderman, H. A., N. Mantel, and C. C. Brown, Ann. NY Acad. Set.. WS 237 (1975). Scott. B. R.. Bull. Mathematical Biology, in press, 1981. Stafford,. J., private complication (1981). Suciu, J., J. Drejman, mid M. Valaskat, Med. Intern, 15 967 (1963). Theriault, G., and P. Allard, J. Occup. Med,. 23 (10) 671 (1981). von Oettingen, W. F., Public Health Service Publication No. 414, II.S. department of Health. Education and Welfare, Washington, 0C, 1955. rfarweiler, R., et al., "An Epidemiological Investigation of an Excess Lung Cancer Risk In a Synthetic Chemicals Plant", Presented at the * nineteenth International Congress Tor Occupational Health, Dubrovnik, fug*., Sept., 1978. rfilton, R. "Response to cooments of I. C. Nisbet," Post-hearing record 35HA Docket 090, 1978. Vithey, J. Rand B. T. Collins, Tox^ Envir. Health, 2 311 (1976). 82-9. 16 FIGURE 1 GRAPHICAL REPRESENTATION OF TABLE HI LOG-PROBIT PLOT LEGCNO iMHALario* BT-1 BT-2 are BT-tS BTII 8T-27 17 table i SUMMARY OF QUANTITATIVE RISK ASSESSMENTS FOR VC 92-9.2 sSTIMATE NO. AUTHOft RASE SPECIES EXPOSURE POR 10* LIFETIME RISK COMMENTS 1 SCHMCIOERMAM, 1070 RAT 73 ppb 119 ppb In* LOG-PROMT LOGIT SLOPE 3.45 LOGIT SLOPE 2J. 1-H4T 2 KUSMACK ft McGAUGHY. RAT. HUMAN 14 ppb U71 140-1400 ppb LINEAR THROUGH ZERO Loc-pROerr 3 GEHRMG. 1070 RAt. HUMAN >1 PP* MOTHANSFORMAL DATA AND LINEAR OR LOG-PROBIT 4 POOD SAFETY COUNCIL. 1900 RAf 2 X 10^ppb WEI0ULL HEH1R. 1MO RAT. MOUSE THRESHOLDS SEEN IN 00TH SPECIES 7 ANDERSON, I960 RAT. HUMAN >1 ppm ONA BINDING ERA. 1900 RAT 4 pG/DAY FOOD OR WATER NAS. 1MO RAT ' 3X105MGIKGrt>AY WATER 10 6AYL0R ft KOOELL, 1900 RAT 0.7 ppb 03 ppb UPPER 97.5% CONFIDENCE LIMIT OF LINEAR MODEL ARMITAGE DOLL MODEL It CRUMP ft GUESS. 1900 HUMAN RAT 0.7 GrtJAY aSpG/DAY APPLYING WORKER DATA TO WATER.UPPER 95% CONFIDENCE LIMITS 13 CAALBORG, 1001 HAT 2.5 X 10"^ ppb W1EBULL 14 THIS PAPER HUMAN >1 PPM NEGATIVE EPIDEMIOLOGY IB TABLE II REGULATORY STATUS OF VINYL CHLORIDE AP00020755 Z< V IT UJ Ul H <E o O IS Q < Ui IM Ul m ? Q. Ul a o < izS 0c Ul Q 2* 02 <J AC > o o _i oz UJ X> o ui UJ _i I- UJ 111 .1 03 co 3< Ul < iu > < fe Ul m f2 6 o CJ a. a Ul .J a o z m o t s ui DC >O Ul r> -j g a fUtiJi _i v> < a. < o. Y ec Ul O < o. S Z3 CZO o o o O o u. aa z 2z S A3 a a 5 cs X o< >- s UJ i-1 t< 5 r- > or ui o a Q tco x -1 Xa: Ul Ul Kz z <b2 o (C e2 Z z z U|U1 n Q.1- O < < to g a. ao u CO CD =>o < -- < >- zo Ul O u. 1< o to CL < o < aa u u_ < <X a. Ul too TABLE III EQUATIONS FOR CURVES FITTED TO VARIOUS SINGLE AND COMBINED MALTONI EXPERIMENTS EXPERIMENT BT1 PLUS CONTROLS ST-2 PLUS CONTROLS 6T-16 PLUS CONTROLS ALL INHALATION STUOIES (4) PLUS CONTROLS ALL INGESTION STUOIES (2) PLUS CONTROLS ALL STUOIES (61 PLUS CONTROLS ALL STUDIES; LOW DOSES ONLY PLUSCONTROLS LINEAR y-ix * b b r 0.26 0.27 3.06 1.52 6.5 S.26 0.26 0.27 1.75 1.74 0.26 0.29 3.35 2.47 -5.59 1.13 1.06 0.30 3.07 2.B0 0.15 0.20 3.60 3.41 0,97 0.97 1.0 0.68 1.0 0.99 0.91 0.91 0.99 0.99 0.73 0.72 1.03 0.97 0.79 LOC PROSIT P-alNDOSEb CONCENTRATION AT 10* RISK, (LOO-PROBIT),ppm a br 0.39 2.76 0.99 0.03 1.60 0.86 1.0 23 0.69 3.46 0.34 0.27 2.93 0.92 0.002 0.30 3.22 0.88 0.002 0.27 3.0S 0.82 0.001 0.91 2.53 0.75 0.17 2.71 0.49 0.42 0.0002 V l-1 2 AP00020756