Document KJKOK5wQpxoL88pL1L4K4d6zN

MANUFACTURING CHEMISTS ASSOCIATION 182b CONNECTICUT AVENUE N.W WASHINGTON, n C 20009 (202)483-6126 \\ November 19, 1974 To: Technical Task Group on Vinyl Chloride Research Subject: Draft of EPA Briefing Report Environmental Aspects of Vinyl/Polyvinyl Chloride Gentlemen: Today I secured a copy of the subject draft (about 235 pages) dated October 15, 1974. Much of this copy is not of quality suitable for reproduction, but the title page. preface, table of contents, the summary and conclusions and the chapter on Biological and Statistical Consideration in the Assessment of Risk have been reproduced, and are distri buted herewith. The file copy of the full text will be available for review at MCA offices. Sincerely, KDJ/mb Enclosures cc: Mr. A. W. Barnes D. P. Duffield, M.D. Dr. Tiziano Garlanda Kenneth D. Johnson, Ph.D. Technical Project Manager Vinyl Chloride Research BOR 013006 BRIEFING REPORT DRAFT r,Tr DO NOT QUOTE OH CITE External Review Draft ENVIRONMENTAL ASPECTS OF VINYL/POLYVINYL CHLORIDE NOTICE TMs ft b (It r K#r dfjf* T * V T '' \ t ^ r V nr' r r * : ' i * i * > t'> rn f>rMt A|*n v r>r'ify. It i* ) r|tn; fCr ll'j'f J *r\- f'rr- ~f J- f ()fl Jfg tPcHlirai A>'rur,d.i\ and policy i^rliralions. U.S. environmental PROTECTION AGENCY NATIONAL ENVIRONMENTAL RESEARCH CENTER RESEARCH TRIANGLE PARK, NORTH CAROLINA 27711 October 15, 1974 BOR 013007 PREFACE This report was prepared by a Task Force convened under the direction of Dr. John F. Finklea, .Director, National Environmental Research Center (I1ERC) in June, 1974. In a preliminary ar essiiont of the environmental problems associated with vinyl chloride and nolyvinvl chloride, an EPA Task Force under the direction of the Of nee cf Toxic Substances deter mined that emissions of vinyl chloride monomer was primarily an air pollution problem. Accordingly, the Office of Air and Solid Waste was given the responsibility for an in-depth evaluation of the problem. This report was proposed as a part of this evaluation. The objective was to review and evaluate the current knowledge of vinyl chloride and polyvinyl chloride emissions into the environment as related to possible deleterous effects upon human health and welfare. The units of parts per million (ppm), in lieu of metric units, have been used in tnis report to be consistent with other agencies currently involved in the national assessment of the vinyl chloride problem. The following members served directly on or contributed to the '1ERC Task Force. James R. Smith, Chairman Kenneth Bridbord Paul E. Brubaker J. Bufalini David Coffin R. Boksleitner Jo Cooper nERC/RT_P Antiiony V. fr 1 ..rci J.t. Davis D. Denny Joan French J.H.B. Gamer OAQPS John Crenshaw Gordon Ortman Bruce Turner Choudari Kommineni B. Lonnc.iian F.P. Scaiingel 1 i Mike Jones EH_S R. Drew ORDTHQ P A r G a11 r h v BOR 013008 PREFACE 1 . SUMMARY. hMj t'f: : MS IONS----------1.1 SUMMARY........................ .......... .......... .. 1..: CONCLUSION----------------------------------- i;i: i" uk't: r*j- 3 CULM: '.L ,V .. HIYSICAL PROPERTIES----------------3.1 I'iiy leal PROPERTIES-------- -----------------------------3 . ? u it Ml CAL 'ilOP;"'TIi :------------------------------------ iT.ASMRlM: N i 'LCPIQUES--------------------------------1 LNV! I'SSrif : ' 1;. A IP.--............. -................... ........ M.s r<r.n-r>' ns: , ------------------------------------------------- 5 ENV1PO 'M ' "!. A^RAISAL --........................ --1.1 SOURCEj- - ------------------------------------------1.2 OVi.RV I r L Ai ,' s0 C L L. .>----- - - - - - - - - - - - - - - - - - - - - - - - - - - '3 CONCENTf-A TION I.--------------------------------------------.. -i ESTIMA-' or AIR QUALITY fOr-U.NTRAriONS3.5 TRAN'-1 'PL': '-N, TRANSPORT. AND Rp-'OV.p-- 6. ENVIRONMINiA: /n05URC AND RECEPTOR PIf''- n THRl > HUMAN EC111 PS 1 i V ! CON i " `l APPLV ; .jr>iK rjj, r-A - ROM VINYL C IlORIDL A! r I. \ i>' tI 11 . r. '> ^ - -- 1j i ' / BOR 013009 DRAFT ro f;0T QUOTE 00 CITE 1. SUMMARY AND CONCLUSIONS 1.1 SUMMARY This report represents a review and evaluation of the available current scientific data relative to the health and welfare implications of environ mental pollution resulting from the production and use of vinyl chloride and polyvinyl chloride. New information about this compound has become available and important new data may be forthcoming in the near future. Vinyl chloride monomer (VCM) was first synthesized in 1837. The vinyl chloride monomer is a synthetic chemical derived from petrochemical feedstock and chlorine. Its principal use is in the production of a wide variety of useful plastic materials such as floor tile, phonograph records, pipes and electrical insulation, although it has also been used in other ways, for example, as an aerosol propellant. The pro duction of vinyl chloride began in the United States in the 1930's, the first important use was in the manufacture of synthetic rubber. Production levels increased rapidly after World War 11 --the beginning of the industrial chemical era which has produced over 20,000 new chemical products, yinvl chloride production in the U.S. was less than 45 million kilograms (kg) in 1943 but exceeded 2.9 billion kg in 1973. The annual growth rate in this industry Is expected to exceed 10 percent per year through the lWs. in the United States vinyl chloride monomer is produced at 15 plants and polyvinyl chloride (PVC) is produced at 37 plants. - / BOR 013010 t f-' * -r ! Approx imutely 1500 workers are project in the* production of vinyl chi on dm, and approximately 5000 are engaged in the production of poly vinyl chloride. The demand for products and components manufactured from polyvinyl chloride is extensive due so its widespread use, thus the impact of control actions will be felt beyond the VCM/PVC industry, .housands of companies, large and small, and hundreds of thousands of workers are engaged in the manufacture of and/or use of plastic products made from PVC, Only a very limited amount of VC!-1, emission data from: industrial st';rces is available. VCM loss estimates of approximately 6 percent >ave been reported, based p-imariy. m j^teriei balance studies. Losses to the outdoor atmospheres from industrial sources may occur at a large number of points in the manufacturing processes and will vary depending upon the manufacturing facility. Currently, emissions of vinyl cnlonde from VC: 1 and PVC plants are estimated to exceed 90 million kg annually. It is estimated that 90 percent of all vinyl chloride atmospheric emissions are believed to emanate from nnlwinvl chloride Diants. Monomer slants emit less than 10 percent of the totEmissions of VCM from fabricating plants and from fabricated products may also occur, but at present there are no oata to ouantifv what those emissions may be. The concentration of residual "on oner in PVC powder that is fabricated into final products is an important lete1** `nant './fu in hnt.h these casts, Fugitive emissions contribute a significant fraction to total VCM emissions, parti - cularly in PVC plants, and these emissions are an important limiting facton in determining the degree of emissions control that can be achieved. BOR 013011 A' \ *i Technology may currently be available to reuuce vinyl chloride emissions from VCM plants by as much as 90 percent and from PVC nlants by as mmb as 75 percent. Control of emissions from PVC plants is a more difficult problem which may require complex process changes. Means of control!inn omissions from PVC plants and .from fabrication processes are yet to be determined. Polyvinyl chloride plastics usually are not readily biodegradable. Incineration (without scrubbinq) of polyvinyl chloride plastics results in fht emission of nydrogen chloride gas, but not VCM [xnerimental studies indm m that vegetational injury symptoms for ethylene and vinyl chloride are identical, however, no known information showing vegetational damaue around VCM manufacturing or processing plants exists. Vinyl chloride is a chlorinated olefinic hydrocarbon monomer whirl) is a gas at ambient temperatures and atmospheric oressure. It is normrllv shipped and stored as a liquid under pressure. Jt is flammable, explori , and only slightly soluble in water. VCM is about two times heavier tlm.n air. Anal;-is of vinyl chloride usually reveals trace amounts of oroanif impurities, such as acetylene, 1,3-butadiene, methyl chloride, vin'Midine, and vinyl acetate. Polyvinyl chloride contains residual entrapped VC' in the larts per million range. The entrapped concentration is dependent, upon the production process and can ranoe from 9.1 to several (5-8) thousand ppm, wnich can be liberated during fabrication, particu1srly wnon heated. The production of VCM/PVC involves the use of a wide variety of chemicals other than VCM which also may contribute to adverse health effects under occupational circumstances. BOR 013012 . - . ,, ' ' , .. , ' 1 : v."' ' - iir v .`ii "w ii.i: on 'j;im-t- levels of visyi rl,lie-id in ariDient ..ic. Tilt: wori p 1 < f f' peak p'.poMjrp' m the past, i;-i y at times have ex- eeded thousands nt ppm, however, the average coin.enti titiot. would have hern less. Limited ai.mosnheric vinyl chloride concentration measurements n ivr been m.ule in the vicinity [0 to 8 kilometers (km)] of VCM/PVC pro duction -jOurr , Tn over 90 percent of the cases the peak concentrations have been be! ov ' j>nn;hcwever, one peak value (qrab sample) of 33 ppm lias been observed. /\ few twenty-four-hour average values of 1 to 3 opm, it distances of j.8 to 8 km *rom the source, have also been measured, ism.n' 1 y in Hit- downwind plume, although over 90 percent of 24-hour measurements were in:1,-.-, 1 ppin. because of the sampling and analytical procedures used, the accuracy of the >o inea .urements may lie no better than + 100,.'. Available atmospheric VCM data have been obtained us inn a variety of Sdii.piinc; and analytical techniques with varying degrees of sensitivity and d^curu.y. Consequentl y, the data are not directly comparable. Standard sampling and analytical procedures have not been established and practiced. Continuous monitoring methods suitable for field use are not availalle, Measurement methodoloav is mil iuat. , hut rias not yet neen applied to the problem of atmospheric nut ,ource samnling tor VCM. The nature of VCM/PVC manufacturina faciht' s,particularly the older plants, is sucli that conventional source monitoring techniques may not be aoolicahle d ,{ t,Ci the -iiscontim r.us nature of "missions. Using an atmosnheric BOR 013013 X ' ' IV In* I : .* ' ` dispersion model, estimates of VCM concentration in a downwind plume indicate that hourly integrated concentrations of 1 to 30 npm might he expected depending upon atmospheric wind and stability conditions. The measured values and the values predicted bv the model are the same order of magnitude. Only limited laboratory studies have been made regarding photo chemical reactions of VCM. Vinyl chloride does undergo atmospheric reactions in the presence of nitrogen oxides and solar radiation; although the reaction rate is slower than with other hydrocarbons known to be in tne atmosphere. Reactions products of VCM photooxidation include CO, formaldehyde, formic acid, formyl chloride and hydrogen chloride. In audition, VC may indirectly contribute to the buildup of ozone. The extent to which VCM contributes to these other components in lhotorherd cal smog is not known. The estimated half-life of VCM in the atmosphere is about 6 hours. The principle route of human exposure to vinyl chloride is tliou- n' to be through air initiation, although exposure could occur from inn^Mr-n of food and waier. and from skin contact. There is no evidence to in !ic.:te that vinvl chloride exists in normal drinking water, or in foods, ex' opt possibly in special cases involving leaching of VCM from wrapping and storage materials. Use of vinyl chloride as a propellant in aerosol products has also recently been discontinued so that this source of exposure, though significant in past years, is no1 anticipated to repr-^s^nt a problem in the future. Acute animal toxicity to VCM was first reported in 1938. Toxic mani festations in experimental animals and man included eye irritation, increased motor activity leading to tremor and loss of muscular BOR 013014 i - -J i t j I (oordi ii it i on , find finally iwi'iums, an:! cardie' i rrogulari tie.,. Exposure cnnren1 r., l i His in ihp'-`> studies ranged up <o 400,000 ppm for periods n_\tending from 40 ennui':; ;o daily exposure of several hours. Short-term acute human experiments (intermittent 5 minute exposures separated by 6 hours over a period of 3 days) with concentrations ranging up to 20,000 ppm produced acute toxic effects at levels above 8,000 ppm. Chronic toxicity effects due to VCM in experimental animals include cancer, damage to the liver, spleen, kidney, lungs, brain and nerve bundles. Some of the pathological lesions observed in these animal experiments were similar to those later observed in humans encaged in the production and handling of vinyl chloride. Our present knowledge of undesirable health effects associated with vinyl chloride exposure in man comes primarily from recent occupational observations, complemented by additional animal data. Between 1949 and 1966 an increased incidence of excessive liver damane and acrnnsf0olwsis, a degenerative disease affecting hones and finqertips were reported amoriq vinyl cnloride woiur's in Europe. Studies in Germany revealed evidence of liver philology in an abnormally high percentage of PVC production workers with a history of employment ranging from 1.5 to 21 years, but exposure levels responsible tor tins damage are not known. Since early occupational health studies often reported acute toxic effects, similar to chose found in the human experiments previously mentioned (dizziness, headaches, nausea, etc.), it can be assumed that peal exposure levels of several tiiousand ppm were experienced at times. Available air monitoring data in PVC plants during the period 1950- l'l-yj indicates that the highest time weighted average exposures in thes facilities were in the range 120-385 ppm. Studies in Europe BOR 013015 I *U | o I L- <' v ^ ' i * w and the United States since 1966 tend to confirm the earlier findings in Europe. These studies include observations of liver damage among workers not directly Involved In the actual production of PVC. The frequency and severity of liver pathology among PVC workers has been related to the length of exposure; i.e. being most common in workers with an exposure history in excess of 10 years. In one study, the degree of damage did not appear to decrease with increasing intervals of time between the last exposure and taking of biopsies. To date 15 cases of liver angiosarcoma have been reported among workers with a history of exposure to vinyl chloride in the United States and 10 such cases have been reported from Europe. Most, but not all, of these reported cases have been among workers involved directly in PVC production. Cases of liver angiosarcoma have been reported in 1 U.S. and 3 European workers exposed to VCM, but not directly involved in PVC production. These cases suggest that exposure to vinyl chloride at lower levels than usually encountered in PVC production plants may be capable of causing liver angiosarcoma. Two community cases of liver angiosarcoma have also been reported in persons whose residences were in the vicinity of industrial VCM emission sources, which raises the question as to whether or not ambient air levels of VCM may, under certain circumstances, contribute to this disease. Additional studies are, however, necessary to confirm this possibility. Cased upon the present reporting methods, angiosarcoma is a rare form of liver cancer in the for all practical purposes, general population, and/is invariably fatal. The latent period for liver BOR 013016 . v >w , ' ' L . ! > i *'i o i!d lio- ! * *e r* i ind t ed dl d! *..u t i'll year's, based upon medical : up-! li end I exposure i fV.O". The levels and durations of ; n. ure to i n du. r liver angiosarcoma in the occupational or the fo.vi.il popiild i ion living in the vicinity of emissions sources are not ,i. isely known. Compared to ihe general population, the relative risk of liver ang 11.sarcoma aim.' workers exposed in the past to high levels of vinyl ill.,tide in estimated nt approximately 3,000. Such a relative risk ..pi. ..nits .. .iril.MK] si d t. i s t ir a 1 ly significant difference (p '0.01) ni i m- f,i-.|iiemy of liver angiosarcoma among those exposed to high levels I "in I cliltn ide numpaied to those in the general population not exposed i 1 . , n; e..posed to much lower levels. ml- tin' fu< us of attention has been on liver angiosarcoma, it should ,i t-d that ,i number of industrial studies indicate that the risk of ii_ i .Mi . j o th I* Cdiicers her, ides liver angiosarcoma, particularly lung and i .i . iai,i-r, i disc 11_ 1 a fed to exposure to vinyl chloride. The multiple i ci i isl. a .ci. intod wim vinyl chloride also is supported by the available 1 * t III L I ! It i I chi dm,. 1.11a ic etfecLs (if VCM have been studied in a variety of animal .. as. \ 111 j i' j ', a i c o 111 a of the liver lias been observed in rats, hamsters, and studies. , .-xp'.'.ed to vinyl ihmride. Angiosarconia was not observed in all animal / , ,, of iats and mice, liver angiosarcoma has been produced by .,i...iiv, a:, i a s 50 ppin. The frequency of liver angiosarcoma in experi- td,, I a, 1111,11 s d, i p c a i' i to be dose-dependent above 50 ppm, but the shape of : ,.,sc respoiibrt rurvo below eO ppm is not known. Duration of exposure , in t 11 hwwn to dfreci i-ia- tumor response in animals. Other damage, / BOR 013017 DRAFT DO NUT OUOTl on Giro including tumors of the lung, Spleen,and kidneys has been observed in animals exposed to vinyl chloride, VCM/PVC workers are exposed to a variety of chemicals which may be carcinogens and/or liver toxins in addition to VCM. Such a complex exposure pattern makes it difficult to draw final conclusions regarding the specific role played by vinyl chloride in the development of liver cancer. However, the results of animal experiments demonstrating liver angiosarcoma from exposure to VCM in 3 species, coupled with occupational data implies that vinyl chloride is a causal factor in the development of liver angiosarcoma. Although actual VCM exposure levels responsible for liver angiosarcoma and/or other cancers in man are not precisely known, limited measurements around VCM/PVC production facilities indicate that contiguous populations are being exposed to low levels of vinyl chloride, which may impose a health risk. The presence or importance of chemical co-factors besides vinyl chloride in the etiology of liver angiosarcoma is not well defined, though other chemicals thorotrast and besides VC11; i,e./arsenicals have been associated with liver angiosarcoma in man. Health implications relating to PVC dust particles containing well residual VCM have not been/studied. Data in animals and man for the lower end, less than 50 opr, 0f t.hp VCM dose response curve are not available. Attempts to extrapolate animal dose response curves to define a presumed "no-effect" level are fraught BOR 013018 W11,1 \r> N'Ji : viK whld> intact upon the state-nf-the art in can^r research, and tnc availability of resources re.ini rod to conduct long time chronic stulu-s. iiibi iaily it is difficult to extrapolate data in experimental animals Mi t.utly to man who may be more or less sensitive than animals to given ch structures. These problems reflect important gaps in our knowledge concerning environmentally related cancers. The mechanism for producing liver angiosarcoma by the inhalation of VC has been postulated but has not been confirmed. It is also not known whether the mechanism can be activated by intermittent peak exposures or whether frequent.or essentially continuous,exposure to low concentrations is sufficient to cause cancers to develop. i.t cnncidsrofis Precise data that indicate the degree to which the general population is exposed to vinyl chloride, and its consequent effects, are not available. However, available data supports the following tentative conclusions: 1. Vinyl chloride in the atmosphere in the vicinity of uiipsion soui'. os is a potential health hazard. p. Occupational cases of angiosarcoma have been -bserved among production workers predominantly with long-term exposure (qreater than 20 years) to VCM at unknown, but suspected high,concentrations. However, cases of liver angiosarcoma have been reported among workers expose to VCM but not directly involved in PVC production, raising the question of effects at lower levels of exposure. 3. Observations among workers and in experimental animals indicate ,vwL there is a multiple cancer risk from exposure to vinyl chloride. BOR 013019 rrv IV i i ti r | f, r ^ ' Mil'* " r-'' * r * Uc< ll J I m.'U I L v.'. .., and liver angiosarcoma and other .r< - 4. The mechanism and dose-response relationships between vinyl chloride/ at community exposure levels are not known. 5. Persons living in the immediate vicinity of VCM/PVC plants have been exposed an unknown number of times to 24-hour average concentrations of VCM of at least 1 ppm with occasional peak exposures of 30 ppm, however, over 90/ of the observations have been less than 1 ppm. 6. Available monitoring data indicate that exposure to VCM around VCM/PVC plants is a local problem confined to within about an 8 km radio'.. Data in the vicinity of PVC product fabricating plants and other snur''cs are not available. 7. Interim methodology is available for monitoring VCM in the atmosphere, but a standard monitoring system has not- been developed. 8. Air inhalation is the primary route of human exposure to VCM. 9. VCM is a primary pollutant, but is atmospherically active and hence a precusor for other pollutants. Little is known about transforma tion, transport and removal processes. The half-life in sunlight is about 6 hours. 10. VCM in drinking water or food presently does not appear to be a problem. 11. There are no known natural sources of VCM. 12. The emissions data identifying specific point sources in I'VC and VCM plants is based on calculationsand estimates and is not sufficiently accurate to serve as the basis for the formulation of a quantitative emissions control strateqy. 13. The practicality and effectiveness of carbon sorption systems, survei1lence-maintonance programs, and deep stripping of the polymer BOR 013020 r ' r ' ' ' , y r- l v u'j f l'.' * > ' i ,, .. ji i L roi'Min In be (]ei'Kinst.rit(v| .is t.niiime-c ii 1 ly visible control techniques. f^1. Theie cite no Known studies ol ilanidqp to venetcition in areas surround i ny VC'1/PVC plants. BOR 013021 INTRODUCTION Historically, national and international commerce hoc estahlished markets for new products rarely with due consideration being given to the environmental consequences of the manufacture, use, and disposal of tfie new products. Consequently, air, water, soil and biota have been contaminated with a wide variety of natural and snythetic chemical compounds tint may threaten public health and welfare. The vontrib ..t orv role of a number of chemicals in the production of cancer and other chronic degenerative disease is well known. In the absence of appro priate pro-market testing, assessment of environmental health hazard: for many chemical compounds orten depends upon retrospective analy-.es after these products have attained broad multi-media distribution. Establishing prudent standards of environmental gualitv depends n > hravailability of o oread integrated data base that is sufficiently quantitative to nor-.it ciopropriate risl- and benefit assessments to he ir,a do . Until r-' ci.tlv, the principal environmental concerns assoc i 11 r-d with the pi -Stic irdir * co has been waste-water effluents from 1 ri. I m t- m 1 facilities and the solid waste problems associated with accumulati as md disposal of various plastic materials. BOR 013022 DRAFT DO NOT QUOTE OR CITE m 19/1 the release of phthalate plasticizers from flexible plastic material not only aroused public health concern but served to elicidate a mechanism of transporting potentially hazardous material in a wide-spread fashion. More recent attention has focused upon serious occupational health hazards associated with exposure to vinyl chloride. The basis for this concern is evidence of vinyl cnloride carcinogenicity in experimental animals and man. The predominate commercial importance of vinyl chloride lies in the manufacture of polyvinyl chloride resins which are subsequently manufactured into a large number of useful plastic products. Vinyl chloride may be disseminated on a broad scale as an unreacted monomer entrapped in finished products such as polyvinyl polymers ana co-pulymers similar to that of the phthalate plasticizers. Durinq the past thirty years, vinyl chloride production has increased from less than 45 million kg in U43 to more than 2.4 billion kg in 1973. estimated loss from industrial facilities (bot.i monomer and polymer production) have been i 1 act'd at over go million kg in 1 973. The primary purpose of tnis report is to provide an interpretive and where possible, quantitative summary of available biomedical effects of vinyl chloride. In this regard, attention is given to caps in the existing data uase . BOR 013023 Emphasis has been placed upon recent health effects developments, efforts have teen made to review and place into perspective the older literature as well. BOR 013024 V. Applying the results of toxicological investigations cut. J" i-'-d in the laboratory to the population of interest generally involves two :.'.nds of extrapolations. The first, which is more difficult to deal with uui more important with respect to human exposure, involves predicting, the inhabit effects on one species from the results of experiments pertot j'l. 1 .m another. The second kind of extrapolation is that of extending dose-rcapons., curves beyond the limited range of observation to determine the dose coi;expending to an extremely low incidence of adverse effects on the organism tested. Tills report concerns itself primarily with the task, of loping objective technical information needed to make decisions on the com sc of action to be followed to insure the safe use of chemicals. It J:..:;, rot, except for a general statement of principles, go into the socio~political .spects of decision-making. Terms such as "toxicological insignificance," "safe," ' t-. tolerance," "no effect level," and negligible risk" have been in rather common u..All of these contain in one way or another value judgments or technical -- pli cations which have no place in an objective asaeasment of risk. 'lh."e is no substance which, under certain circumstances, cannot be dangeruU; -"..d unsafe. There is no battery of tests, however elaborate, which can prove ley. ..d chal lenge the complete safety of a chemical. For the toxicologist tu isp^ly the terms "toxicologically insignificant" or "negligible risk" to a set of obser vations mokes a premature Judgment in the wrong arena by the vre i. person as to insignificance or acceptability. An attempt has been made to eliminate such terms from this report. BOR 013025 ;jrwAr i r.D MOT QUOTE OR CITE Anolhet common ti'iai of wide usage !.; r he "no effect level." This i:. statistically meaningless and Lheiefore of limited value since it merely i .. ,iiis that; no elfeet was observed .in studies using a group of animals of articular size, Such an observation is completely compatible with the precenc<. of an adverse effect, which in further studies with larger sample sizes or with dii'ierent types of observation might lead to a positive outcome. We prefer the usage of the term "no observed effect," which should always carry with it a qualifying statement as to size of the group in which no adverse t i , : was observed. 1 *><t > To most instances it will be imperative to develop a dose-response r j - t iunship , and because man;- toxicological techniques are relatively insensl'ive, hicli doses (whi.ch produce high incidence of effects) are frequently . i; mI. These can be and have been called "unrealistic" or "inappropriate." So.ii _.;posures may be well above, sometimes many orders of magnitude above, lil-ely levels of exposures to human or wildlife systems. Nevertheless, they :r.' i Lcn an essential part of practicable laboratory studies which necessarily use M mated numbers of animals. The underlying challenge to the toxicologist .. to use these points on the dose-response curve as a means of quantifying tespouses, and to devise, with suitable margins of safety, appropriate means for extrapolating to realistic, actual exposure conditions. The biological as, clf. of this extrapolation will be discussed first, and the statistical considerations will be developed Inter. B. RIOI.UdlCAI. C.0N>1 DURATIONS Tor clarity in the following, it will be assumed that we are concerned with extrapolation from the laboratory situation to human populations. i\ U'.f I h- W BOR 013026 r r o oo hot draft QUOTE or CITE Except in the ease of lifesaving drugs, only very low risks will normally la accepted in chemical usage. The acceptable risk will, however, vary with the benefit anticipated. In the case of risk of death from a chemical of trivial utility, the acceptable risk would be essentially zero. Put in explicit terms, this might mean 1 death in 100 million persons. As noted in the section on statistics below, extrapolation to such risk levels from experiments on small numbers of animals Is extremely uncertain. Again, and as noted repeatedly in this report, the gravity of the effect is a major determinant in an overall assessment of risk. At one extreme lies a fatal outcome, and at the other, a temporary functional alteration producing no disability or discomfort and lying fully within the range of physiological compensation. The susceptibility of human populations varies widely since genetic background, age, prior or co existent disease are all important determinants, and part of the toxicologist's task is to identify susceptible groups in the population as the basis for estab lishing limits of exposure. Another and vital factor constantly facing the toxicologist is the often striking biological differences between the effects of chemicals on labora tory species and on man. it has been repeatedly shown that no one species (including non-human primates) has responses parallel to the human over a wide range of the effects of chemicals. The choice of species must then be based on a determination of the biological similarity in the responses to the chemical under study. In extrapolating from animals to man the transfer is often made on a dose,per unit weight (milligram per kilogram) basis. This practice overlooks the well demonstrated (Freireich e_t_ al_., 1966) observation that dose per unit surface area (mg/m^) is generally a better transfer parameter. BOR 013027 m .... U|WI " f'3' QUOTE OR CITE ' There has been much loose talk about "thresholds." The term threshold menns "the entrance or beginning point of something." This implies (and is normally so used) a discontinuity in the slope of the dose-response curve. True discontinuities in biological phenomena are rare. However, they do occur. One example i6 the threshold for glucose excretion by the kidney. Most biological dose-response relationships appear to be smooth functions and in absence of con crete evidence dose-response curves should probably be assumed to be smooth. Many dose-response curves have an "S" shape with a much lower slope at the low end of the curve than in the mid-range. This could be regarded as a "quasi" thres hold. The steepness of the dose-response curve is an important consideration tor predictive purposes. A steep dose-response curve implies a sharp cutoff (again, a "quasi" threshold) with decreasing dosage. Some dose-responae curves appear to be linear, especially when atten tion is limited, to relatively low incidence rates. One example of this is cigarette smoking and lung cancer (Doll, 1967); there are many experimental situation where this appears to be the case. Despite the above comments, there are some biological reasons for anticipating chat with some chemical agents there may be something approximating a true threshold. The biological basis for this is twofold: (1) the possibility of a relatively greater effectiveness of repair mechanisms at low dose levels; and (2) the possible presence of competing biochemical processes which could convert the chemical to harmless products at low dose levels. It is difficult to generalize on these mechanisms since they can be .* expected to depend on the chemical and the species. Unfortunately, investiga tion of these questions has rarely been undertaken. BOR 013028 n r DRAFT C3 MOT QUOTE OR CITE 1 In tlm pant, these complex considerations have been dealt voith pi..;;matically by the use of arbitrary "safely factors." C. STATISTICAL CONES1DEKATIONS The need for a proper consideration of statistics in the design of toxicological experiments and in the interpretation of 11.* results cannot be overemphasized. First, before meaningful results can be obtained, attention must be given to identifying and reckoning with possible sources of error. Second, statistical techniques are available which can give meaningful esti mates of the level of exposure to chemicals corresponding to the level of risk which the decision-maker considers acceptable. 1. Experimental Error and Sampling Krror The outcome of an experiment is normally dependent upon innumerable factors, only some of which are known and even fewer of which are controllable. In dose-response experiments with animals, for example, some identiiiablu factors influencing the outcome include (1) the composition of the particular batch of test preparation, which typically represents a significant source of variation in independent repetitions of the experiment, (2) animal variability, (_)) technic Ian reliability, and (4) the precision of laboratory techniques such as dilution techniques or dose preparation. Since such factors influence the dnsc--response re J a t 1 onsh i p , they represent sources of experimental trior, and hence independent replications which randomly sample the levels el these factors are necessary in order I1 i nn r.: 'U i , . . BOR 0X3029 P>'"i * f~T i. L w | I :r Q'JOii cr.; c/rE ! stimuli: their contributions in the measure of experimental error. Tf n>...'0: sources of variations arc not considered in such replication, then sta tistical precision as reflected in the width of confidence intervals, for ^.p.'.ple. nay he grossly misleading. When several major sources of experimental error can be identified, a conceptually simple experimental design would consist of independent repli- i . _cs :,t each target dose level randomly sampled with respect to all sources . , .'ri.~r.ion. If batches from the chemical manufacturer represent a source ..f v.'-'"'-ation, for example, then this design might assign each animal at each . vel to n different batch of chemical from the manufacturer. If dilution errors ure non-negligible, then dilutions to target dose should be Independent, . , :oiy among dose levels but also among animals within dose levels. Wien . cell sources of errors exist, this conceptually simple, completely ivr.tHT.ized experimental design clearly becomes impracticable, and blocking "ccces a more feasible means of conducting the experiment. Thus, each batch of the chemical from the manufacturer might be administered to a group of u Imais at every test dose to produce, in effect, a separate dose-response : urw for each batch. For any one batch, the proportion of animals responding at u given test dose is subject to sampling error due to such factors ns animal "'ii!crences and possible errors in dilution which would result In each animal : coivirg a slightly different test dose. At any given dose level, the pro- portico responding also varies among batches; thus, the average proportion responding at a dose level is subject to both sources of error, namely, the rc.. plirc. error within batches and the variability among batches, which together cojprice experimental error. A valid statistical analysis should utilize the ^^opriate experimental error and not merely its sampling error component. h - tO , r\ r\ ri . BOR 013030 draft noI-'l * A11'r 7 narirfr CITE Since standard statistical methods o biousuay often are addressed only to Lite analysis of sampling error, there is need for caution in applying these methods. 2. Estimating bovT F.ffeet bevels Estimation of low effect levels poses a difficult problem. Direct experimental estimation of the level affecting one percent of Lite population may require several hundred animals to obtain adequate statistical precision. For many reasons, particularly in human populations, much lower risks than one percent are desired. A true no-effect level cannot be observed experimentally. Any observed level has meaning only for a particular sample size. The observation of no-effect for a group of animals may arise from one of two reasons: (1) the dosage level may indeed be below the theoretical no-effoct level; or (2) the number of animals tested may have been inadequate to give a high enough probability oi detecting a biologically important change. For example, a test on 20 animals may show no deleterious effect, but a test on 100 animals, tested under the same conditions, may stow one or more animals exhibiting deleterious effects. Similarly, for a graded response, a small sample may fail to provide enough statistical precision to detect a change from baseline, whereas a larger sample may. Thus, an observed "no effect level" lias no absolute meaning since it depends on sample size and poorly estimates the theoretical "no-effect level"; a better term would be the "no observed effect level." However, data from experiments in which no effects are observed are useful in placing limits on the probable incidence of effects. For example, if no 'animals out of 100 animals displayed a deleterious effect, it can be stated with 997. confidence that fewer than 4.5*. of animals tesLed under these conditions f r, II BOR 013031 would exhibit tie 1 c l e r i ou-s ellects. If LIi s level of risk is too high, more inim.ils can be irsled. For example, observance of zero ,inim.ils directed out 1;'titi tested results in an upper 9h'/ confidence limit of only 0.467,. To ii.icli an extieiuiy low acceptable risk level in ibis manner generally requires ut'ob i b i t i ve l y large number of animals. The past practice of selecting some arbitrary fraction ni "no effect level" as a limit for exposure leaves one with no estimate oi risk. However, i a fairly conservative estimate of the risk can be made by employing the s.i-bjL (one-parttclo) theory (Food and Drug Administration Advisory Committee . a Protocols for Safety Evaluation, 1971). This theory states that for low lies,o-.es, if an experimental dosage is divided by a factor f, then its upper . iMiCence level ol the risk is also divided by the factor i. Such an approach 'I ; i1 o lien result in near-zero dosages for extremly small acceptable risks. ; . e> aple, if zero deleterious responses were observed in 450 animals ; a dose d, it can be staLed with 997. confidence that the true response rate is less than 1% (one out of 100). The predicted dose for risk of -4 i 'i.e out oi 1,000,000 would then be 10 d . An alternate means of estimating low risk exposure levels involves extrapolation from parametric dose-response curves. Many different empirical mathematical models may be fitted to a set of experimental data (Finney, 1964 ). The problc and logistic curves have been commonly used in biology, for example, and noth curves may fit equally well in the region of experimental observations 12"i to 987, response range) but give widely different estimates for extapolated resjnn.es. For example, the problt curve will predict a dosage level approximately 140 times higher than the logistic curve for extrapolation to a dosage expected to elicit one response in 1,000,000 animals. In some instances II IX BOR 013032 DRAFT DO MOT QUOTE OH CITE .t I (Doll, 1967) linear dose-response curves have been reported; often, however, these cover only a relatively limited response range. There is no assurance that the dose-response curve observed in the experimental range of dosages will apply at extremely low response levels. Mantel and Bryan (1961) suggest the use of a presumably conservative slope of one probit for each factor of 10 in dosage level for extrapolating to low levels of carcinogenic risk. A more recent approach to the extrapolation of laboratory findings to the establishment of standards or limits for human populations (Albert and Altshuler, 1973) has taken into account age at the time of the appearance of the adverse effects as well as the frequency of Its occurrence. In the ca.->e of cancer from external sources, for example, It has been shown, both experimentally and in humans, that with lower doses cancer appears later, that is, at increasing ages. Under this concept, and assuming the availability of reliable data, it should be possible to establish limits which would place the earliest occurrence of malignancy at an advanced age, e.g., no more than 10% incremental likelihood of cancer at age 95. D. SUMMARY In the past, toxicologists have not only made the laboratory assessments of toxicity, but in many instances they have made the final judgment as to the social course to be taken on the basis of a particular.set of findings. Instead, the technical experts should be charged with securing an objective Independent determination of the extent, nature, and frequency of adverse effects. They should be asked Lo explain the relative gravity of these effects for t he targcL o' ifnt' Jf 3 i BOR 013033 rO *`r Ei * >. i i.i! L feystem, be it humans or wildlife, Similarly, other qualified technical experts should be requested to make an objective assessment of the benefits of use and of alternative materials or processes. However, the final judgment as to a tradc-.-off between an adverse health effect and a desired benefit is a social decision and should be made with the participation of those who are affected. This is not to say that technical experts using their technical expertise will not participate, but it does state that they should not be the sole judges of determining the balance between the benefit and the risk.. ^ The dose-response curve is a valuable tool for assessing the safety of a chemical compound. Estimates of low effect levels are part of the information leading to the ultimate designation of safe and acceptable levels. The statis tical problems of extrapolation from experimental dose levels to very low levels ^nd the estimation of appropriate errors are particularly troublesome but can be handled if care is taken in the design and analysis of the experiments and the interpretation of results. Without supporting experimental evidence, however, statistical analysis will never be capable of making the critical extrapolation from laboratory animals to man. A^ s> BOR 013034 ---------4 r r DRAFT un:"AT"BI <:m:" "Q .".'T QUOTE OR CITE Albert, R.E., ami B. Altshuler, 1973. Cons i derat ions Relating in the Fomin lal ion of Limits i or Unavoidable Population Exposures In Environ mental Carcinogens, pp. 233-253. In C.L. Sanders, R.li. Busch, U.E, Ballou, and D.D. Mahlum, Eds. Had ionuc 1 i dc Care i noddies i s . l'rm . l?Mi Ann. Hanford Biology Svmp. AEC Symp. Ser. #29 CONI -720305. Nai ional Technical Information Service, Springfield, Va. Doll, H. 1967. Prevention of Cancer: I'o infers from l.p i do m i n 1 ogv . Nuffield Provincial Hospitals Trust, London, 144 p. Finney, D.J, 1964. Statistical Method in Biological Assay. 2nd Ed. llafner Pub. Co., New York, 668 p. Food and Drug Administration Advisory Committee on Piotocols for Saletv Evaluation. 1971. Panel on Carcinogenesis report on cancer testing in the safety evaluation of food additives and pesticides. Toxicol. Appl. Pharmacol. 20:419-438. Fre ire fell, E.J., E.A. Cohan, D.P. Hall, L.ll. Schmidt, and II.E. Skipper. 19b6. Quant 1 tative comparison of toxicity of anticancer agents in mouse, rat, hamster, dog, monkey, and man. Cancer Chcmocherap, Hep. 50:219-244. Mantel, N., and NCR, Bryan. 1961. "Safety" testing of care ieouenic agents, 1. Nat, cancer Inst. 27:455-470. e _ JQ - <5 BOR 013035 M I I f'l Nf I S * <r 1. S< fine iderman , M.A. Mou\<' tn Man-Extrapolat ion of I abor.i t nr y Results to Human 11 i sen \e . Presented to: I In' Working (troop on Tlu' 1 ox u i I y of Vinyl (hioride - Polyvinyl Chloride. Tho Now Yori Academy of Se ientes, Now York, New York., May, 1974. R. Woil, C. s. Statistics vs. Safety factors and Scientific .lodgement, m [valuation of Safety for Man. fox. Appl. Pharm. ?1 ; 4S4-4C3, 1 97?. 3. Weil, fiuidelinos tor Experiments to Predict the Doqree of Safety of a Material for Man. Tox. Appl. Phans. ? I : 194-199, 1973. 9, food and Drug Administration Advisory Committee on Protocols for Safety evaluation; Panel on Carcinogenesis Report On Cancer Testinrj in the Safety [valuation of food Additives and Pesticides. Tox. and Appl. Pharm. 20:419-438, 1971. b. Symposium on the [valuation of the Safety of food Additives and Chemical Residues. Tox. Appl. Pharm. T6:^9C>-5?0, 1970, ft. The Efforts on Population exposure to Low LEvels of Ionizing Radiation (Pier Report), In- the Report of the Advisory Committee of the Bioloqica! Effects of 1 on i z ing Rad lation. National Academy of sriern.es, National Rese n c h Council, Washington, 1). C. IJ. S. Government Print mu Off up, Publication No. f) - > i 9 - 7 9 7 . 1 97?. 7. Interim Report on [x11apo1 ation of Risk of Cancer from Animal Data Committee to coordinate toxicology and Related Programs. Department of health, [duration, and Welfare. Personnel Communication. May, 1974. 8. from Principles for Evaluating Chemicals in the Environment. A report of the Committee for the Working Conference on Principles of Protocols for Evaluatmq Chemicals in the Environment.: Environmental Studies Board, National Academy of Sciences - National Academy of CM fc> BOR 013036