Document bOY0ZNLj2deroKK7JYkn8xqNg

O i-i W ^ L* il X Li, tl'iiii ;U VUL ^ **' *k ,h. ^ w j VMtkt. ^ , urcccl to investigate) its co ri^ht status *----jr----- ri'---- Ji""1" '" "' ~ *- '"" I "Safety" Testing of Carcinogenic Agents1 NATHAN MANTEL nnrf W. UAY TlUYAN, llinmctry Branch and Laboratory* of Viral Oncology, National Cancer Institute,* flcthrsda, Maryland <i SUMM YHY Tfio pruldrm of iJctrrmmlng what dose 9 X 10-8 mg per mouse when n statisti IctrU of an agent arc Rnfo, e.g., non-# cal assurance level of 99 percent and u caioinngoiiir, rntinol hr rrs^yrd unless conservative prohit slope of l normal t one fir*J <?< fine* iovt-j of nermis- deviate per log for 0x1 rapola I ion arc I xil'h:.-*LUm fioinn^l rr_Junv jnivaiU rather used* The prinriplrs given arc of gen \i / than on ;b u>hnc_a_fety. Holh eral applicability in other safely-testing In cause* of practical considerations aiul problems, the point of emphasis being | Mat Nt ieal variation, the determination ilia t y mrrjlireelouLu-urvnlioj i yj\ inml he i 4 of Iow*-ri*k dose levels, for example 1/100 made that the risk at some do*e level is I million, cannot he made directly hut cleurlv"Tow, TnTlIreet fansrrvaJi^c^ro- L must he hv eytrtJPJlli.' * >t>n front oh_*rrt rd mlurrs for iho determination of low' !i cfntA conservative up[r;ich for risk levels must he made. Hie arbi doing ro ir given. In addition lo an trary risks and definitions for so doing arbitrary definition of "virtual safety," may change with eircum^taiu'es. The il in nc^r^sary to define nn ni'lntrarilY procedure docs not require `.pecifiral irn high nlaI Nliral nuur:iiiir_ level and a of nn experimental protocol; I lie 4<*\afcT rule foi ex l rapola t ion hv use of an m hi- dose is determined niJ.lheJ.M>i>? of what trniilj *'h;,Uow tdopr. JllnMrutive data ever data are available. iNlinimmu pro hy llryan and Shtmkiti (J. Nat. Cancer tocols mj, Jnwc*cr, he dohablr, since Inst. 3: r03-331, 1913) oik the carcino greater amounts of data w ill ordinarily genic action of mcthylcholanlbrenc permit specifying large ".safe" levels.-- yield a "hafe," 1/100 million, dose of J. Nut. Canter lust. 27: -l-la-tTU, 1961. ALTHOUGH IT is not definitely known whether nil chemical compounds that induce cancer in experimental animals will also cause cancer in man, it has bean fairly well established that, with the possible exception of arsenical compounds, every chemical oi-phvUcal agent. known.to produce cancer in man will Iikcvdse do so in one or more species of lov er animals 01---- ' Of necessity, the potential deleterious effects of chemical compounds. mu3t bo tested in laboratory animals. Tho reaction of a particular species,', of animal does not, constitute proof thatliumans wall react_similarlv. but;' 1 Received loi publication March r?, 1001, 1 National Instituted of Health, Public Health Service, U.8. Department of Health, Education, and Welfare. 455 667873--01-----20 ''T y-.a.... BOR 014494 ,,cntion is drawn to Cue urced to. lavcstigrato its cor *ifTht_status. ____ '""I"' *' Vfc--KBMaHIL. " ~- I - I , I I C. I * 456 ^ MANTEL AND BRYAN t the only recourse tho investigator has in implementing test programs for \ ncute toxicity control of the human environment with respect to injurious chemical or physical agents is to proceed under the assumption that any substance i (see for s' of method'. f( harmful to animals is polentdallyjiamiful-te man. Also in carrying out time, such ns control tests in animals one must proceed, initially, ns if protection of tho with the csli' . animal population were the actual problem. When reliable estimates of the doses tolerated (with specified probability) arc achieved_for -Urn \ i chronically n Tho jmrpo' Wdmal population, tho problem then becomes one of juunmcnl, based on in the dovrh accumulative experience, uyjransferring tiie implications of the results "safety level-- t^man7" Kesulls oftosUng with a variety of animals could suggest how extrapolation could bo nrndo properly to mammalian species of higher I I metric proved known carctn , ; order or larger size, Many chemical compounds may bo harmful at certain concentrations, )1 tors which or . results in nn; i! though beneficial at others, Tho control of "toxic" or "harmful" sub-* t are not reatn |j stances therefore does not imply the necessity of their complcto or ab- to compound i solute elimination, which in some cases would bo either impossible or disease. i economically infeasible, but their reduction to concentrations that can : ho tolerated by essentially all individuals of tho population at risk, * SOME THO j In rapidly acting toxic substances that are either quickly eliminated i i from the body or are readily transformed to less harmful compounds i through metabolic processes, the estimation of tolerated levels is not. too i i In a safel. dilficult, Wi< h a reasonable allowance for an extra margin of safety iutror i cols and the ! duced in the form of an arbitrary "safety factor," tho results obtained in be dotermw i laboratory animals can be successfully projected to humans. For the, h apart from i most part, modern pharmacology is based on just such usage of laboratory Certain is animals. These relate i There are other compounds, however, for which tho results obtained we are wilh ! in laboratory animals cannot, bo so confidently projected to man. These neecssnty o- substances arc not readily excreted or metabolized, and because of their to problems weak solubility in aqueous solution, nmy remain in the body or on body 1) llo'c ` surfaces for very long periods. Other substances may be metabolized to denionstruf ' some extent, but. because of their selective affinity for cells of certain only to c-r-t types thc.y may accumulate selectively within these cells to yield harm some level fully high concentrations if supplied on a continuing basis. Most chemi whether or cal compounds that cause cancer, c.g., polycyclic aromatic hydrocarbons, that a rid; naphthylamines, azo dyes, and some steroids, have properties falling into Other arb; one or the other'of these categories. Still other compounds have an immediate initial effect, which is not t conditions Incident considered harmful, and are rapidly eliminated or metabolized; yet, acceptance during their brief sojourn in the body they produce some critical intra > strnting at cellular damage or change, which does not manifest itself until Inter, st ration, t Urethan, which interferes with nucleic acid metabolism, is a classical moderated example of a compound of this type. Injected parcntcrally, urethan acts ever, only immediately to induce transient anesthesia, but single anesthetic doses the true ri- may cause lung tumors in mico many months later. Fortunately, this 1,000 an in drug was never approved for use ns an anesthetic for man. 2) ll'Viu: Biometric methodology has been highly developed for the study of experimen JOURNAL OF THE NATIONAL CANCER INSTITUTE VOL. 27, N< BOR 014495 . vCiiCiOJH til' ^ u V'J CuC ,,> U 1 ^ A v J U W \ * W w ** s .-urged to Investigate Its.o' /rifTht status. A * ' -- 4J * or or co ut 30 of JO >n ts w 2r r Q 4 1 3 ) A ! ( I I I TESTING OF CARCINOGENS 457 acute toxicity and rapidly developing biological rcaclions of other types (gee 2-4 for review). Somo progress has been made in tho (levoloj)incnt of methods for tho analysis of responses that arc greatly- prolonged in time, such as cancer (see 5 for review); however, the problems associated with tho estimation of tolerated levels of carcinogenic ngonts and other chronically acting compounds have not yet received adequate study. The purpose of this communication is to discuss the major problems in the development of methods for estimating limits of tolerance, or "safety levels," of carcinogenic compounds and to describe certain bio metric procedures, based on available data, that are applicable to presently known carcinogenic compounds and experimental animal systems. Fac tors which one must consider in transferring the inferences derived from results in animals to man arc. also discussed. The suggested procedures arc not restricted to carcinogenic agents, but arc applicable, in principle, to compounds which cause various other types of harmful reactions or disease. SOME ritOBLEMS IN PLANNING AND ANALYSIS OF SAFETY STUDIES In a safety-testing program both the design of the experimental proto cols and the nuUhod of analysis or interpretation of resulting data must be determined. Tho experimental design cannot be properl}- determined apart from the plan for analysis of the data. Certain issues in a safety-testing program must be resolved in advance. These relate to what, we moon by safety and what kind oT feasible results we ore willing to accept as proof of safety. Settling of these issues, if necessary on some arbitrary but conservative basis, may permit answers to problems that would otherwise be insoluble. These problems me: 1) flow safe is safe? Absoluto safct}r can never bo unquestionably demonstrated experimentally. Father, experimental results can be used only to establish limits on the risk involved. With the specification of some level of risk, no matter how small, the possibility of determining whether or not that risk is exceeded opens. We may, for example, assume that a risk of 1/100 million is so low as to constitute "virtual safety." Other arbitrary definitions of "virtual safety" maj^ bo employed as conditions require. Incidentally, an inflexible requirement for absolute safely may lead to acceptance of high levels of hazard. The impossibilit}- of really demon strating absolute safety lends to the acceptance, ns a satisfactory demon stration, that no hazard was observed in an experimental protocol of moderately large size, 100 or even 1,000 animals. Such evidence, how ever, only provides assurance, at the 99 percent probability level, that the true risk is under 4.5 percent in the 100 animals or 0.4G percent in the 1,000 animals. 2) IWiaf constitutes proof of safetyt In principle, one could use an experimental protocol sufficiently large to demonstrate that "virtual von. 27, NO. 2, AUGUST 1961 BOR 014496 tcntion is drawn to the possibility that this text may *o copyrighted. The reader, .-urged to investigate its_.cc_^.rlght-Status.------ - 458 MANTEL AND BTITAN safety" obtained. For this purpose it must be realized that an observed ii scaling [or t outcome of no tumors among 100 million treated mice does not necessarily - According to demonstrate clearly that treatment was either absolutely or even vir | slopes tually safe. This outcome could arise with a probability of 1 percent, ' dn-o-n -pon-i even if the risk involved were ns high as 4.0/100 million. It would in | rithmie ran:: fact require a total of some 4G0 million tumor-free mice to demonstrate : occur and I' at. Die 00 percent assurance level that "virtual safety" obtained. Simi ; slopes ure ot larly, tumor-free results for 10,000 mice, would only indicate that the risk ' rithmie or t< was less than 1/2,200 and it would require tumor-free results in a total of riot given he some 450 mice to establish with high probability' that the risk was under i The till-or-nc 1 percent.. Studies of fensibh size can be used to establish directly only t of antibiotic risks of the order of 1/100 or higher. Data from such studies can ho used | din Dower, o to ascertain the treatment level consonunt with a proscribed risk or to ; slopes on the establish limits on (he risk for n particular treatment level. The deter ( it mny be tl i mination of "safe" levels can he made only by indirect methods extrap particle cur'- i olating from the data obtained in a feasible study. The use of extremely large studies to establish safely may well be self- ^ wotdd appe' low ns one defeating. The almost certain occurrence of unusual sj-ndrumes in one j While slope. or more of a large number of but animals, albeit these mny have arisen established, spontaneously, will require admitting the possibility that they may bo y ordinarily o attributable to drug treatment, ! The indie: 3) How can protocol data be erfrapointed sajthjf Since it is only feasible ` nscons to use experimental protocols for the direct determination of relatively ^ it may be u high-risk dose levels, c.f}., 1 percent, which makes extrapolation methods j being made necessary, one must-consider that such extrapolation method", might yield ! jjrocediire p misleading results. Procedures exist which permit extrapolating the. ! curve.' mny ! results obtained at a number of test-agent levels to determine the dose j not loss tlin level corresponding to any desired degree of risk and to establish, with a , response en high level of assurance, a minimum hound on this dose level. These j Ic.-s than n methods, however, are based on the assumption that the relationship j course, nth. objcivcd between tumor occurrence and do,-e at the levels tested will r ext m point in continue to apply in the regions to which extrapolation is being made. { Once nns The validity of such an assumption cannot, be tested and, if it is false, may j "virtual $af lead to a sciious ovciestimate of the "safe" level. Such oicrcstimnlion * servative r would arise if the relationship of response to doso is less pronounced nl the l protocol da dose levels to which extrapolation is made than at the levels at which tests ! necessary r were performed. Another related source of difficulty is that tests are V sections. performed on relatively pure inbred strains of laboratory animals. Characteristically, such pure strains v ill show sleep dose-response relation ships, while the heterogeneous population to which it is intended to apply the results of testing may exhibit a shallow relationship. * ANA 1 To avoid the risk of ovcreslimnlion of ``sale" levels which may result from extrapolating with too steep ft slope, it is suggested here that a conservative result mny bo obtained by extrapolation with an arbitrarily low slope from the data nt hand. For example, quantal-rrsponse data, (bat is, flll-or-none response, frequently exhibit a somewhat linear re lationship when plotted on probability paper with a normal or probit V j* I In what 1/100 mill! tion will b< or tenfold > To illus'. prescribed , journal of the national cancer institute von. 2",-No BOR 014497 \ --------------- 4S A*- tendon is <lrawn to the possibility that this text may j copyrighted. The r-- > urged to, investigate its copyritrht status. ___ * * ` ( TESTING OF CARCINOGENS 459 scaling for (ho percent responding nnd a logarithmic soiling on dose, j According to tho kind of system being investigated the. dose-response I slopes observed may vary widely. Willi systemic poisons, rather steep dose-response slopes are generally obtained, reflecting (he narrow loga rithmic range between the lowest dose levels at which any toxic deaths occur and the levels which are lethal for all animals--such response | slopes are on the order of 10 to 50 or more probits per common loga| rithinic or tenfold dose increase (the technical definition of this slope is not given here; it permits one to perform any necessary extrapolations). The all-or-nonc response also arises in the study of the therapeutic effects of antibiotics. The response slopes in these instances arc generally i shallower, on the order of 3 probils per common logarithm. Lower slopes on (lie order of 2 do arise in virus-assay work, but. in these instances it may ho that use of a different response, curve, the single-hit or onepart if-'le curve, would be more appropriate, From such experience, it would appear that tho use for purposes of extrapolation of a slope ns t low as one prohit. per common logarithm is likely to be conservative. While slopes in the regions to which we wish to extrapolate cannot be established, the suggested slope of one is rather low compared with that, ordinarily obtained in the observable region. Tho iiuiiee.ted low slope is a key feature in the method to bo suggested ns conservative for the establishment of ``safe'' levels. Lor this reason , ii may be well to make clear just how weak or strong are the assumptions being made in its use. In fact the only assumption being made for the procedure to be conservative is that whatever form tho true response curve may tnko over the region of extrapolation, the average slope is nnt less than the assumed one. There is no requirement that tho true response curve he linear or even that the true slope should nowhere, be less than nr.mmc.d. The use of the indicated conservative slope, is, of course, arbitrary. Other values may lie specified and other scales for extiajiolat.ion may he employed. Once answers to the three questions arc provided, a defined level of "virtual safety," a prescribed level of statistical assurance, and a con servative rule for extrapolation, it becomes possible to determine, from protocol data, "safe" dose levels and to undertake the. planning of any / necessary experimental protocols. This is considered in the succeeding sections. j ANALYSIS OF RESULTS AT A SINGLE DOSE LEVEL . In what foliows we. will take the defined level of "virtual safety" to bo * 1/100 million, the statistical assurance level to be 99 percent. ExtrnpolaI tion will he on the basis of 1_ normal deviate or probit per common log ( or tenfold change in dose. I To illustrate how definitive "safe" levels arc obtained, consider that a \ prescribed dose of an agent has elicited no tumors in u group of 100 ex- j VOL. 27, NO. 2, AUGUST' 1PG1 I BOR 014498 t- lu.u uhs may ao copyrighted urged to, investigate its c ^yrlght_status. ______ 1... _________ _ . ................... The rciclc nM * i li .\ \ I t 460 MANTEL AND MITAN perimcntal animals. Wliilo tho observed rato of tumor occurrence is 0 percent, wo will lako, as an upper limit on tbo truo rate, that risk for which the probability for occurrence of as few ns zero tumors is 1 percent (100% less the assurance level of 09%). This is given by tho solution for P to the equation (1 -- P)1W> = 0.01. Solving, wo have 100 log (1 - P) = log 0.01 = - 2; log (1 - P) = -- 0.02 s= 9.98-10; 1 -- P = 0.955; P = 0.045 or 4.5 percent. We now know that tho observed outcome of no tumors in 100 animals is consistent with the possibility that tho true risk was, in fact, 4.5 percent. From tables of the normal probability function (G) wo determine that the normal deviate, Y, such that the integral from -- to Y equals 0.045 is --1.095, fora probit value of 2.305 - 5--1.095. However, the normal deviate, 3%, corresponding to a risk of 1/100 million can similarly bo de termined as --5.012, the probit being -- 0.012. The upper limit on tho risk for the dose employed is 3.917 = --1.095 -- (--5.012) normal deviates above, the desired safe risk, and, at a slope of one normal deviate per common log, it is necessary to reduce tho log dose by 3.917 logs to attain a "safe" level. The antilog of 3.917 being about- 8,300, it is determined that the "safe," dose, is 1/8,300 times that which had been tested. Table 1 shows the preceding lcsults together with those for several other hypothetical experiment rir.es in which no tumors were observed to occur in a group of treated mice. For each such group tho (able, shows the greatest risk consistent, at the 99 percent assurance level, with tho observed outcome. Also shown arc the corresponding conservative estimates, with (ho slope of one normal deviate, per tenfold dose increase, of the. "safe" (1/100 million) dose level expressed as a fraction of the dosage tested. Tho larger (ho experimental group among which no tumors occurred, tho greater is the value determined as tho "safo" dose. TAiu.r, ].-- Illustration of "safe" doses determined wlicn no risk was observed in sioglo groups Nmrilv-r of mice with tumors/ No. m-sted Upper limit on tumor risk at 0/10 0/50 0/100 0/500 0/1,000 level employed, 99 percent <op,urancr (]>orccnt) f Estimated "safe" dose (1/100 37 8. 8 4. 5 0. 92 0. 45 million) dose employed = 1 1/190,000 1/18,000 1/8,300 1/1,800 1/1,000 i I \ fi l I 4 f* 1 1 9 I I I t iii > f * Study of this table indicates that a control system can be established without the need for specifying a design protocol, though (.here might bo some merit in speeding a minimum size. For, whether the amount of evidence adduced to show that an agent is safe is great or small, it can be properly weighted to determine conservative safe limits. If tho promulgator of a drug wishes to have high tolerances established for his / JOUKNAL OP THE NATIONAL CANCEtl INSTITUTE n *i I 4499 ol compound, it jnents with a 1 is not t^pv'irc danger is dmut testing at rat! he lower. A for interpret in how much dal 1 The. method to the ca-e in used for illusl r In general it unfavorable re assurance level or The. solntim or fewer than solution, the c The prcced; tables of runti those of the O Let us con r =s 10 have Inferring to t- and that for I VOL. 27, no. mention is drawn to the possibility that this toxt may bo copyrighted, .urged to, investigate its co, status...... ..... *3'ho rcadc TESTIKO OF CAHC1NOGENB 4G1 compound, it would bo worth while for him to produce results of experi ments with a Inrgo number of animals, which would show that the agent h not especially dnngerous at certain dose levels. Where the lack of danger is demonstrated with n small number of animals or as a result of testing at rather low doses, tho doso levels determined rs "safo" will ]>o lower. A control system can bo constructed about tho possibility for interpreting tho data submitted, with no specification needed os to how much data should be obtained. THE CASE OF SOME OBSERVED RISK Tho method indicated for determining "safe" levels is not restricted to tho ease in which no observable danger was noted. That case was used for illustration because of the simpler mathematical solution involved. In general it will bo that an experiment testing n animals will yield r unfavorable results (tumors). The upper limit on risk at tho 99 percent assurance level is then the solution for P to \ Y>C,P`(1 -- P)n~` = 0.01 ("0 or 2 nCtP\\ - P)-< = 0.9D l-r-f] The solution for P is a value such that the chnncc of observing as few or fewer than r tumors is 1 percent. At values for P in excess of the solution, Urn chance for such an outcome is less than 1 percent. Tho preceding equation can be solved approximately by reference to tables of cumulative binomial probabilities. Such tables, as for examples those of the Ordnance Corps (7), show values of quantities such as ib nC.P'Cl - Py*. l-r + I Let us consider as a hypothetical outcome that, of n = 100 mico, r -- 10 have developed tumors, for an observed rat.o of 10 percent. Referring to these tables we see that for P = 0.19 I>c,P`(i- ry-` 0.9891 i-n and that for P = 0.20 ino jg nC,P`(l - P)"-* = 0.9943. von. 27, KO. 2, AUGUST 1901 ^4- I BOR 014500 nation is drawn to the possibility that this text may tio copyrigaau. i 4 * .arped to, investigate its cop. Ight status. __ ; : ; _ ; ' ' 1 i I^owwaw^giliii w miililVriihni r_--~J~~- ~ ' ' 462 MANTEL AND DHYAN We may take the solution for P ns approximately 0.102. The cal culating procedure follows ns before. The normal deviate corresponding to a 19:2 percent probability is --0.871 nnd, ftt the slope assumed, it will require a reduction in Jog dose of 4.741 = --O.S71-( --0.012) to obtain a risk of 1/100 million, 'rim "safe" dose is then determined ns 1/5!),000 of the flose which had becncmploycd, 55,000 being tho anti log of 4.74T. V------------------------ ---- -------------- -------- - USE OF CONTROL DATA In the methodology shown, it was assumed that the responso of interest, npponrancc of tumors, did not occur spontaneously. In general, however, it will be. desirable to use controls to check this and to allow data obtained for such controls to modify the determination of "safe" dose made. However, with the method already shown, failure to use controls or to take, control data into account will result in more conservative determina tions of the. "safe" dose. If, in fart, spontaneous rates arc rather low, they will have little cfTcct on the determination made. For this reason we, may adoptjtjjmcrdnvc which is somewhat more conservative than neeJssaYy for Inking control du_ta into account. This we can do~by taking, ns before;, the. upperliimt, for the risk in the, treated group ns tho solution for to z^op.'a - F,)"* 1-0 0.01 while the lower limit on the. control group risk is the solution for Pt to PtY*'-, = 0.01 i - r* where r, of n, treated animals nnd r( of nc control animals, respectively, sltowcd positive, response. These equations can be solved through the use of binomial tallies. At (his point Abbott's formula (8) can be used to obtain a modified value for the treated-group risk, this being computed ns PI = (P, ~ I\)l (1 -Pt). The computation follows as before, the normal deviate being obtained corresponding to P\* Since Pt cannot exceed P,, the use of control data cannot result in decreased values for the calculated "safe," dose. ANALYSIS OF RESULTS OBTAINED AT SEVER AT, DOSE LEVELS When an agent is tested, it is sometimes desirable to do so over a number, perhaps even a wide range, of dose levels. In such instance, all the avail able data should be considered in the determination of "safe" levels. 1 Tbo/wtldhussl.'Jlhthhn rpfiyphj^ri to tbc moderate) v rnuscrvntlvc procedure drperU^d here for dctrrtolclng F*. A more firorouj procedure would r-^ulre setting lltnlisoii the rath of binomial pniamctcrs. A Mould be flvco by reducing by unity the upper limit on the rath of controbto-trcBlmert nonrvspouse probabilities. journal of the national cancer institute As already indieat observed rc'-pmi'e six, variation of the fitt methods con-ci vativ large strnh si/,<-=. I (lie. result that extr; (be slope obtaining l si/.es or studios, the slope, taking aecoui may result in exit'1 lower ronfidcwo )ii (,han llmt whHi v is used foi extra pol An alternative d example, fitting tie determine the Inw< moderate percent a; with the observed 1 conservative value 1 percent dosage, o While (lie pmcei is based out lie vale results have drriv' But for binu.'Say p exactly a ['propria! nbly go''d job of g1 can yield I easonal ations being comp use of an invalid j of dose levels cor; In many situnti' in others quite s< large experiment: reflected in name, to risks in other one's estimate of is violated. Accordingly, v. wo will consider : single dose-level i The only assump tive and it is. of instances in vdiu is not mhlendiiit: generally appro, section, vol. no. 2, At BOR 0X4501 ;tent.ion is drawn to the possibility that this .-.urged to. investigate its c yrlght status. text tnay bo copyrighted. J r.c re M* .urn wiawMoi 1R 11 n 0 t I 3 1 l ! I J i t TESTING OF CAItCINOGENS 403 Parametric Procedures As already indicated, it may bo unwise to'extrapolate the data with the. observed response slope. Procedures for taking into account the statistical variation of (lie fitted slope will not suffice to make such extrapolation methods conservative. Tosco this, one need only consider the nscof quite large study slr.es. In this ease, statistical variation will lie negligible with (he result that extrapolation to low-risk levels will he substantially with (lie slope obtaining: in the observable range. (However, with small study sizes or studies, the designs of which are. inefficient for estimation of the slope, taking account of the statistical variation of the slope determination nmy result in extremely conservative estimates of "safe" levels. The lower confidence limit- on the "safe" dose in such instances may he. less (linn that which would he obtained when an arbitrarily low slope value is used for extrapolation, as suggested.) An alternative device rouhl lie. to employ parametric procedures, for example, fitting the maximum likelihood probil line to the data (/?), to determine the lower confidence limit on the dose corresponding to some moderate percentage risk, r.n,, 1 percent, to which risk-level extrapolation with the observed slope is considered reliable. Then, using an arbitral iiy eonservativi' value for the slope, and anchoring at the lower limit on the ] percent dosage, one could cxlrapolatc to the desired "safe" level. While the procedure jus! indicated i? straight forward, its employment is based on \ he. validity of (he pununet ric fund inn employed. Quite useful rrnilts have derived from r.uch parametric assumptions in bionsray work, But for biniissny purposes it is not essential that a parametric function he exardy appiopriate; it is; sufficient that, the function assumed do a reason ably good job of graduation. (Even soman bat innppiopriato curve forms can yield reasonably good relative potency estimates as long as the prepar ations being compared are tested over the Mime, regions of response.) The ir-e of no invalid parametric function can lead to inappropriate estimates of dose levels corresponding even to moderate risks. In many situations the estimates may be only moderately inappropriate, in others quite serious. With parametric procedures, the, use of rather large experimental groups in one region of the response curve can be. reflected in narrow-range, confidence intervals for dose levels corresponding to risks in other ranges. This can produce a false sense of security m one's estimate of the rnoderate-risk dose level when the parametric model is ; iolated. Accordingly, while ngreeing that paramedic procedures may be useful, we will consider the possibility for extending the method described for the single dose-level ease without the need for assuming any particular model. The only assumption is that the nrhitraiy low slope assumed is conserva tive ami it is, of course, implicit that the response curve is monotone. In instances in winch it ran be recognized that use of a parametric procedure is not misleading, workers rnny prefer this procedure rather than the inoro generally appropriate nonparamctric procedures described in the next section. VOL, 27, NO. 2, AUGUST 1961 : Jl w* ** w a* vw - 1' ^ '' urced to invcsticato its cop *ifrht staj.ur.. ^ - ---------- -..-1- ------- ~ii ~ ' ii mwiJj ,1iS.7i ,M. 464 MANTEL AND HRYAN Nonpnrnmetric Procedures In the preceding section it was suggested that the use of parnmelric methods, while straightforward, could lead to nonconsorvntivc results, There are no simple fixed rules for conservative estimates of the "safe" dose when sevcrnl dose levels arc employed and one is unwilling to male assumptions ahouL the dose-response curve in the region of observation. How estimates enn bo made in these circumstances can best he demon strated by illustration/ We will begin with some simple ideas. Suppose investigators at two laboratories independently test an agent, at n. level of 100 rng/kg. At the first, with 500 mice tested, no tumors are observed, and with the methods described previous^* the "safe" dose is estimated ns 100 mg/kg/1,800 = 0.056 mg/kg {fj, table 1). A somewhat lower "safe" dose of 0.012 mg/kg is obtained at, the second laboratory, based on the observation of no tumors among only 100 mice. It enn readily he recognized here that it would he inappropriate to reject the high "safe" level of the first laboratory just because of the low estimate obtained at, the second laboratory. Tho two sets of data arc consistent and in fact confirm each other. If any modification is to he made, it should he to consider that, with results combined, no tumors have occurred among 000 mice which would lead to a safe dose of about 0.005 mg/kg. Suppose that at still a third laboratory, tests are made at n dose of 50 mg/Jrg and, with no Ivmnr^ occurring among 500 mire, (be calculated "safe" dose, at that laboratory is 0.028 mu/kg. Here again we can sec that tho "safe" dose obtained at, the first laboratory should not he modified downward just bemuse a consistent result at (he third laboratory yielded a lower "safe" dose, if anything, it should be considered that the 500 mice not, responding at (he higher dose at laboratory 1 would not hnvo responded at the lower dose employed at laboratory 3. With these 500 mice treated as nonrcsponders at the low dose, there is then, including those at laboratory 3, a total of 1,000 mice not, responding at the low dose. (Laboratory 2 results are being ignored for this illustration.) This yields as a calculated "safe," dose 50 mg/kg/1,000 =- 0.050 me/kg. In the present case the "safe." dose based on the combined calculation is less than that for the data of laboratory 1 alone of 0,050 mg/kg and so the higher figure is retained. Had tho combined calculation led to a higher "safe" nose it would have, been correct to take that as the estimate. The point of these illustrations is that,, when the data obtained from a scries of doses arc. con- Istent v.itli each other, it is appropriate to take as tho calculated "safe," level the highest one pertaining to the results at any one close. Even a higher "safe" level may be taken when it can be ob tained through a justifiable combination of the results at the various doses used. What is meant hero by "justifiable" combinations can bo seen from the following example in which hypothetical results at four dose levels, low, middle, and high, are considered. * An allmjallvc method to tho one oboul to bo described Is itlvcn as mi oppendU. JOURNAL OF THE NATIONAL CANCER INSTITUTE PO'C i" Fi/R ml' r 1 2 3 A I p,,rrnl V "sido" do " hid* In the ah? able" combii succession, t' he cletc;mine combined re would luMiic Where da simple exam Dose in d.-e Ol'fbT 1 2 At the fin he larger if the resuhs stance in th in pan nth, lov or ir.cn! of the invo luted value two altern: nt. this do The sL-mfi value L no native tc'1 would vick parent In Occurs, t::: deuce at t "safe" du result 1/2' 1/100, v.Tr In pvur simpler lb haps two c vm,, rv. N -*r BOR 014503 TESTING OP CARCINOGENS 465 Pofc in size order Observed results Number of tumors/ number of mice "Justifiable" combined results Combined number of 1 umors/eoinbined number of mice 0/1 on 2 0/100 3 1/100 4 4/JOO i rarcMlhi-'M li.'llcnlo time this remit "site" due. (0/100);* 0/100; 1/100; 4 /lOO 0/200; 1/700; 5/200 1/200; 6/300 5/400 not be comeh'icil, ns tbr nrjt must yield a lilt'lu'r valuo for the In l!ic absence of inversions in the data, we can determine the "justifinhle" combined results at a dose by adding to the results at that dor.e, in succession, the results at still higher doses. A calculated "safe" dose can he determined for each dose used and for eacli of the various "justifiable" combined results corresponding to each dose. The over-all "safe" dose would he (lie highest of the various determinations. Where data show an inversion the procedure is altered. Consider a simple example: Dose in size order 1 2 Observed remits Number of tumor'!/ iiumbci of mice i/i on 0/100 "Justifiable" combined results Combined No. of turnors/combined No. of mice 0/100) ;l/700 [(0/100) ;l/iui>] At the first dose level it is clear that the cale.ulatcd "safe" dose would he larger if based on the combined vesnlls for both levels than if based on the results observed at this level alone; accordingly, as noted in one instmire in the preceding example, the result at the lower level alone is shown in purenfhoM's. At the higher dose level an inversion occurs; there is a lower incidence of tumor, even though the dose level is higher. In view of the ii;ver=inn, one would be less willing to accept as "safe" tire calcu lated value obtained on the lewis of remits for this dose level alone. The two alternative results show n in brackets at this dose level are the results at this do=o level and the contradictry remits nt the lower dose level. The signifies nee of the use of brackets here is that the calculated "safe" value is now to bo taken as the lesser of the \ nines suggested by the alter native results. In the present instance, the result at the lower dose 1/100 would yield the lower "safe" dose and so the alternative result is shown in parentheses, (it will not always bo necessarily true., when an inversion occurs, that the retained result will correspond to the higher tumor inci dence at the lower dose level.) In the present example the calculated "safe" dose, will he that corresponding to the first dose with combined result. 1/20(1 or that corresponding to (he second doso with retained result 1/100, whichever is the greater. Jn practice the application of the methods just indicated is much pimpler than the explanation would suggest. Ordinarily only one or pcr imps two of the combined results at a dose level will need to be considered. VOL. 27, NO. 2, AUCJUST IPC1 *0* 14504 Mention is drawn to the po.ssi' Mity that this toxt may bo copyn^ntcu. urced to. Investigate its cop,,JL(rht status- -- r .. -. -------------- _ * * ~ `~Tf ^, ._ _ r^____ ________ * i 466 MANTEL AND IJKYAN TJic results at some dose levels may immediately permit us to drop (Item ( i Tabu: 2.--Ill' from consideration. After only a limited nmmmt of experience it should be possible so do this rather rapidly. The calculations are actually simpler than those for the maximum likelihood probit method. [Tn fact, (he confidence limit procedures ordinarily employed in connection with the, prohit method arc not fully satisfactory. A more appropriate method is j described by Mantel and Pal-wary (10), but it could require a somewhat extravagant level of computational efTort.] And, while for completeness, Dos'1 me/ niou'i''. wo have indicated the need for considering the possibility of inversions, this will ordinarily not pose a problem. ; (I) I o nfini'M V 0 niHviTS An Illustrative Example 9 a (imtio (! OO.'i'.l An example from the. literature shows how the procedure just discussed \ o dots i can be applied. The data, from Bryan and Rhimkin (//), arc the results 1 0. 1)15(1 ti ir.it 2 j obtained after a single insertion of niclbyleliolnnllirene into mice, 12 dif- n. ni.'J5 ! fercnl dose levels being used in lhe study. The reader may refer to tlio 0 125 (I. 25 original article for details. No peculiarities arise in (his example. There are no inversions. At t 0. 50 1. 0 the four lowest levels no tumors occurred and the appropriate combined result is readily recognized in these instances. At (lie middle four levels it can be recognized that llmrc is no point in combining results and, finally, the four highest lex els can he. disregarded as these, all yielded 3 00 percent tumor occurrence. The procedure is illustrated in table. 2. The, first three columns show, 1 t / i data. At the maxm normal d< is indicate respectively, the dose, log dose, and the observed result. Column 4 allows eiieli combined result considered, and there should be a separate j line, for each such result. In the present instance only one combined result required to be considered at each dose. For each such result, column 5 shows the. calculated maximum risk at the 00 percent nssur- * nnce level. These were obtained from binomial tallies or calculated directly for the, ease of no tumors occurring. The normal deviate cor- ` responding to the maximum risk is obtained from tables of (he, normal distribution and is shown in column 0. Finally, column 7 shows the > calculated "safe" (1/100 million) log dose. The maximum for this, j 2.902-10, appears in the second line, and the over-all calculated "safe" dose is 9 X 10" mg per mouse. * Ono might, remark that this "safe" dose jR so low as to make impractical any use of it which may result, in its ingestion by humans. But we aro dcnling here with a rather indent carcinogen and if any compounds are, i to he Assigned tolerated levels which are virtually zero, this is one of 1 hem. 1 Text-figure 1 shows graphically the results and analysis of the expert- 1 Tr.XT-nc.' ment just considered. Normal deviates are shown on the vertical scale, while on the. horizontal scale the dose employed is shown as negative descending powers of 2. The points shown represent the outcomes at. each dose level; 0 and 100 percent outcomes are shown by arrow. The solid lino on the figuro is tko maximum likelihood probit lino fitted to (ho i j 1 |I nwthy! percent data data extrap- upper 1 JOURNAL OF THE NATIONAL CANCER INSTITUTE ,* VOL. 27, 1 Fillip WWP *J I* BOR 014505 -at ion is drawn to the possibility that this text way ou rcd to. investigate its popy^ ^ht status .____ ^ .. ',,. . )- .... -- TESTING OF CAnCINOOENS 4G7 Table 2---Illustration of methodology for determining the "safe" dose from results 4 at scvcrnl doRe levels; data from Bryan and Shiinkin (II) ild I Icr I Calcu he i ho i Result Combined result Maxi mum /J lated "safe" (1/100 iis i at i thee mg/ value No. of tumors No. of tumors 90 % assur Corre sponding normal million) log dose (2) -- (0) -'8 mntl.'iC Log dose No. of mice No. of mice ance deviate -- 5.012 ;5Si 0) (2) (3) (4) (5) (0 (7) n WI0214 0. 3SR-10 0/79 n nnw'75 tl. OTO-10 0/41 n no its 7. Ml-10 0/19 v now 7. 5112-10 0/19 eel o finis 7. 803-10 5/17 n ni:h; 8. 194-10 C/18 '(s * n. o:i 12 8. '('15-10 13/20 if- { n nr.-jj 8. 790-10 17/21 i (1 Pin 0. 0'17-ui 21/21 :j e n t'.j 0. Oils -10 21/21 |At i i (1. 50 1. 0 0. 099-10 10. 000-10 21/21 20/20 j eel \ 0/158 0/70 0/33 0/10 3/17 (i/lS 13/20 17/21 *-- -- 0. 0288 0. 0500 0. I 1 11 0. 2152 0, 4 SO 0. 729 0. 871 0. 058 -- -- ------ -1. 800 - 1. 58 1 -- 1. 205 -0. 7X9 -0 050 + 0. 010 + 1. 131 + 1. 72S -- -- """ 2. 075-10 2. UG>-10 2. 88 [-10 2. 709-10 2. 331-10 1. 972-10 1. 752-10 1. 450-10 -- -- -- i tin la. Above the first 8 data points the (riminles shown rorre.-pond to i ll>e maximum P values of table 1. Exlrnpobilion, with the. slope of one "t iij nonur.l deviate per ronmton log to the over-all calculated "safe" value, is indicated by u broken line. All triangles, other than the one from wliich -1 :o dI 1i d > r- f :sl `C 5i i d 00 i Trxr-rinunE -- F^timation of tho "site" dose from test results with a carcinogen ninlliylcholanthrenc, at rcvoral dose level:,. At each tort level both the observed perrenlagc response and an upper limit, 09 percent assurance, based on combined data are shown. Si>;td (me is the maximum likelihood probit line fitted to tho data. The "safe" level of 0 X 10~* mg per mouse ia in this instance estimated by extrapolation with the conservative slope of 1 normal deviate per log from tho upper limit on J' at the second dose level. i t VOL. 27, NO. 2, AUGUST 1961 I I I1 BOR 014506 i-urcecl to, investigate its copyright status. 46S MANTEL AND BRYAN extrapolation was made, should fall above the line. The difference in slopo between the solid and the broken lino may he noted. RELATION TO OTHER SYSTEMS While at various points in the preceding section the possibility for arbi trary selection of the n.uirnncc level, the level of safety desired, (he slope value used for extrapolation, ami even tho extrapolation curve (one could use some scale other limn normal deviates or prohits along which a risk probability may ho defined) have been emphasized, this should not be taken to mean that the, methods suggested arc. completely general. Rather, in each case the nature of the risk .munition should be considered. Some principles do earn' over, however. Tor example, one. may ho interested in trying to extinguish a bacterial or viral population by expo sure to increasing temperatures or by increasingly long exposure to bac tericidal or viricidal conditions. Determining the appropriate tempera ture or duration of exposure through the use of a conservatively shallow slope may he appropriate in stie.h cases. Consideration of tlm `'single-particle" or ''one-hit" problem (12) gives rise to a quiLo different answer than that developed in the. preceding sec tion. In this problem it is considered that a single particle can cause infection or death. If particles arc distributed at random in matnial being inoculated, so that if there is an average of in particles per inocula tion, the probability (ha! n particular inoculation will contain none, and so be mifc is the, probability that it will not be safe is then 1--c-171. The problem is to determine from test data what reduction in the inocu lum is necessary to ensure safety. One can inc directly the. methodology already given to set maximum values on the risk as a result of the outcome, of testing. This in turn establish!.-; a maximum value for m ns --logt(]-ma\. P value), the neces sary reduction to any desired risk level following directly. In this instance the inoculum corresponding to a risk of 1/100 million is approximately one millionth that, corresponding to a risk- of 1 percent. 'I'ljis contrasts sharply with the ratio of about, 1/2,000 when extrapolation is made bet ween'tlicso two risk levels with a slopo of one probit per com mon log. While it has been shown that in tho central region tho `'nne- purticlo" curve mimics the probit curve with a slope of 2 probits per common log (13), the contrast would suggest that tho comparatively steep slopo of 2 noted virtually disappears in the lower tail. Wliat further suggests itself ir, that the ''single-particle" curve, provides a most conservative rule for extrapolation. Where there is any suspicion that this curve, may apply, the procedure described in the text should not bo used. One might also visualize a two-hit or two-stago process, the low-level probabilities for each stage being approximately proportional to the dose used. A reduction in dose by a factor of 1,000 should rcduco tho joint probability by n factor of 1 million. In this case tho 1/100 million risk dosage would ho 1/1,000th that of tho 1 percent risk doso, JOURNAL OF THE NATIONAL CANCER INSTITUTE i I I I t / t i t An altern levels inav i evolved in < intuit ive'y In thb, rr one compel rv-ults at t given fur ! interfere!!"' response, w probit pros level and it known to b detenu in"! A drawl, not correc 11 level. Suc: response- a < and lifttv.-ar of-fit, xi- '' hood soli;- a positive n likelihood obtains a!--. Consider :b the rmixinr with the as at each do=Xi. for dcp-> based on t!, trial value, square wit 1, to a 00 % < At this ; just been g! hv the mw X,* based o heiuu dcriv To illudresnoud; th A P, and the Then VOL. 7, NO. BOR 03.4507 -Oil is drawn to the possibili that this text way bo c' yrighted. The reader d to. investigate its copyright status. _ * TESTING OF CARCINOGENS 4G9 APPENDIX Alternative Method An aileron live method (o Hint- described for handling data at several doso levels may be more appealing to the biomet ricnlly orient' d reader. It was evolved in discussions with a reviewer (J. Cornfield) and is objective and intuitively satisfying. In this method the data nro considered to consist of a series c' assuj-s: one comprising results at the lowest dose level only; another comprising results at the two lowest levels; still another in which the results aro given for those, at the three lowest love's, etc. (Levels where there is interference, as for example lethality interfering with a carcinogenic iespouse, should not be included.) Ju.r each assay, with appropriate juuhit procedures determine the maximum likelihood, 1 /100 million doso level and its lower limit, subject to the ies! rie.tion that the probit, slope is known to he unity. Of nil the lower limits on the 1/100 million dose level determined for t lie series of assays, the largest is selected ns the; ".safe" doso. A drawback to this procedure could he that, standard methodology does not corrcigly permit t he del ermimition of lower limits on tlie 1/100 million level. Such met hodology goes awry if, for example, then: are, no positives responses u( the throe lou C'l dose levels. A principle described by Mantel and Tutu-ary (10) permits c solution. Calculate, the chi-square guodmvsof-fil, xi- (nr log likelihood for those so inclined) for the maximum likeli hood solution. (This ehi square would he taken as zero if there aro no posilho le.-ponses at any of the dose, levels considered; the maximum likoiihuod 1/100 million love! is actually infinite. A zero chi square obtain? abo v.hen only results at the lowest dose level arc considered.) Consider nliornn'ivo tiial values for tho 1/iOQ million dose Je.vel le-s than the muMinut:: likelihood estimate. Each such trial value, in conjunction with the assumed slope of unity, provides an estimate of the, response, rate at ear); do=o level of the iw-ey. There is'thus, in turn, a chi-square value, Xi, for dip ntures of the observed responses fiom the ixtinmled responses based cm the trie! 1/100 million doso level, x? will e.xeerd xi and for some tried value X:~x? will equal 5.412, the OS percent, upper limit on a chi square with one degree of freedom (a 9S% limit on chi square corresponds to a 09% on the "safe'' level). At this point we introduce a modification which, for simplicity, has just been glossed over. Instead of computing x? based on weights implied by the maximum likelihood solution, ono should compute the quantity x!* based on departures from the maximum likelihood fit, the weightings being derived from the fit to tho trial '`safe" dose. To illustrate; Suppose that at tho -i'th doso level, r( of nt animals respond; that tne maximum likelihood estimate of tho responso rate is Pi and tho estimate corresponding to the trial value considered is P\, Then X| = (r, - n(PJ)*/n*Pr (1 - Pt) VOL. 27, NO. 2, AUGUST 1961 BOR 014508 .-urtfcd to investigate! its copyright status. 470 MANTEL AND BRYAN and Xl* = (r, ~ nS\)yn,h (1 ~ Pi). Tho trial value for which xl ~ xl* equals 5.412 is (lie lower limit on the safe dose for the assay considered. With k dose levels, there will he k sc(s of assay data mid k lower limits, the maximum of these limits being the one selected. (The likelihood ratio may be taken as an alternative criterion for setting limits; see 10.) !i [ 1 i i REFERENCES (7) Minim, G. H.: Tlie role of certain chemical and physical agents in the causation' of cancers. ]reminds before the Committee on Inter, (ale nrul Foreign Coin- mcice, 80th Con,arcss, 2d Session, on II.It. 7(124 and S. 2197, January 2G, l'.HiU, Washington, U.S. Covt. Flint. Office, J000, pp. '15-00. () Bum-:, C. I-, and C\tti:u, M.: Biological n-say. Ann. Jtev. Physiol. 5; 479- 5:i(.>, 19t:i. ' (5) Fiwr.v, L>. J.: 'Statistical Methods in Biological As=ay. New York, Jfafncr Publishing Co., 1952. (4) Burn, J, II,, Fis.mo, I), J, and Goodwin, L. G.: Biological Standardization, 2d ed. London, Oxford Univt r.-.il.v Press, 1950. (6) Bhxan, W. It.: Quantitative biological cxperimenlatioii in the virus and cancer fields. J. Nat. Cancer lost. 22: 129-159, 1959. (rr) l'l.in.iial Wouks Auunuy, Work Projects Administration ror. the City or New York: 'J allies of probability functions 2: 1912. (7) Oiidnanoi: Criur.i F.\ uru m;t OKDP 20-1: Tables of the cumulative binomial probabilities. Wnihinrton, Ordnance Corps, Septmubcr, 1952. )(S) Ani-orr, \V. S.: A mclhud of computing the effectiveness of an insecticide., i U J. Econ. Knt. 18: 205-2(17, 1925. (9) CraiM'ian, J., and Man run, X.: Some new aspects of the application of maxi mum likelihood to the calculation of the dosage response curve. J. Am. Htat. As.-oc. 45: 181-210, 195(1. (70) M anted, X,, and Patwary, K. M,: Interval estimation of single parametric functions. Proc. 37d Session of the International Statistical Institute, Tokyo, Japan, 19GU. In press. (77) Bryan, AY. It., ami Sui.v.cin, M. It.: Quantitative analysis of doac-rcsponsc data obtained with linen carcinogenic hydrocarbons in strain C3II male mice. J. Nat. Cancer Inst. 5: 503-531, 1943. (IS) Lea, D. E.: Actiuir- of Radiations on Living Cells. Cambridge, Cambridge Univ. Press, 10 SO (IS) Mr.vNLL:., G. G.: Inherently low precision of infeclivity titrations using a quantal response. Biometrics 13: 149-1G3, 1957. t 1 i ' > ' ` j ( j ' | I f I 1 stmtAir The i I" ft'It M.F. 'no 1. M re i II. o nifUjnf', as 1 i 4 a, 11 * -1"i! t -i ihc rlinPin l 1 ' of 'i bis ns f prc^M' *' r> L crlK. " cylr>j>^\ho ourrC'l in OELU'f for the nutibolie is bn-f'i liomolot:1 hoilif: if by iy:ii *' prnft' if Trnr.-'er within i! IT]10Ilei! lymph* m CVttrly -is lymphnr I Itnc. n : Thu in'*1' HrAlth. c? t:. 'lUpWIW BOR 014509 nation Is drawn to the po.ssit Uty that this toxt may' b- copyrighted 7ho ur-cd to, investigate its co^y^frht status. u c r^ ,,r)i.Y asp applied pharmacology 20. 419-438 (1971) Food and Drug Administration Advisory Committee on Protocols for Safety Evaluation: panel on Carcinogenesis Report on Cancer Testing in the Safety Evaluation of Food Additives and Pesticides1-2 August, 1970 Received April 26, 1971 INTRODUCTION -.report is concerned with an evaluation of the present status of testing for earcinor.iion of food additives and other chemicals, which come into contact with man s.illy through his diet. Although carcinogenicity is only one of the many mani,,,iis of toxicity. it has somehow assumed a role of considerable, possibly magni.poriance. It is the only manifestation of chemical intoxication that is dealt with by ,! legislation. For practical purposes, a carcinogen is a substance that when adminiIhy an appropriate route, causes an increased incidence of malignant tumors in iinviit.il animals as compared with a control sciics of untreated animals. The n,c to a carcinogenic substance may consist of (1) an increased incidence of tumors ,pe `ecu commonly in the control animals, (2) an occurrence of a type of tumor not it all in the control animals, or (3) a combination of the occurrence of a different >! tunuii ;md an increased incidence of one or several types of tumors seen in the _ \ir.n cvpiesxcil in llii'- rcpni t arc (lime oT (he Advisory Committee and are not lo lie regarded ' it u.UoMcuts of lltc U.S r nod and Drug Ad min mi .it ion. at Advisory Committee mi ProtiKok for Safety [-valuation. ii el th,- Committee: Norton Nelson, Ph D,, Director of the Institute of (environmental me, Nv.v York Univcisity Medical Center. New Yoik, New York 100Mi, Cliun man: Julius ii, M D . Professor and Head of Pliaimacutogy Dcpaitmeiil. Jclferson Medical College, ' iplu.i. Pennsylvania. 19107; Jerome Cornfield. O.S., Research Piofessor of Uiostalistics, i I'linhuiL-li Graduate School of Public Health, KASf-.tl Building 9t>50 Rockville Pike, Ilothcsda, .mu * Leo Friedman, Plt.D.. Professor of Nulriuon and Food Safety, Massachusetts 'c uf I cchnolngy, Cambridge, Massachusetts 02139 (Ptesent address: Duccior, Division of and 7 usicology. Bureau of Science, Food and Drug Administration, Washington, M.'i'li. ' Robert t. Gossclin, M.D., Ph.D., Professor and Chniiman of Department of imlop.v, Dartmouth Medical School, Hanover, New Hampshire 03755. Cabin M. Kunin, A.uiciair Professor of Preventive Medicine and Medicine, University of Virginia School of at. f'luilettesvillc. Virginia 22901; * Ted A. Loomis, M D , Ph D., Professor of Pharmacology, -ii> ol Washington School of Medicine, Seattle, Washington 90105; Philippe Shuhik, '11 (. Ii, R Pint . Director, Epplcy Institute for Research in Cancer. University of Nebraska College ".I; .nc, Omaha, Nebraska bij!05; James L. Wluttcnbcrger. M.D,, Professor of Physiology, aJ Fimeisity School of Public Health, Boston, Massachusetts,02115; JamesG. Wilson, M.D., Pinfessor of Research Pediatrics and Anatomy, Children's Hospital Research Foundation, . nun. Ohio 45229. vuc Secretary: James R. Cribhett, Committee Management Officer, Bureau of Science, Food and (Administration, Washington, D.C. 20204. 'Vn of the PaneI an Carcinogenesis: Dr. Philippe Shuhik. Chairman: * Dr. Leo Friedman, Dr. si Nelson. Dr. James L. Wluticnhcrgcr, Jerome Cornfield. ' Jtmt to the Panel on Carcinogenesis: William Ltjuisky, Ph D., Epplcy Institute for Research in 'ft. University or Nebraska College of Medicine, Omaha, Nebraska 68105. Xmtmtiip irnniniicd June 30, 1969, 419 / BOR 014510 x.cntioa ir; drawn to the possibility tliat this text may bo copyrighted. The render urged to.investigate its cop ^ight status. _ 420 IDA PAM.I. ON CARC INOGI NI SI.S controls. In some experiments the only manifestation of an effect consists of (4|.ln. , occurrence of tumors in the treated animals than in the controls, the incidence hi . same in both. In yet another variation the only effect seen may consist of (5) an m.. in the number of tumors per animal, the number of tumor beating animals Im; . same. Any one of these effects may suggest the presence of a carcinogenic b.t/.n.i occurrence of benign tumors only would not classify an agent as a carcinoren It is apparent that all elfccts which in one way or another influence the o,.u,:; of neoplasms cannot be precisely the same. The different situations cited h,i\c examined in detail by experimentalists interested in mechanisms of action ol\,;\ gens; however, from the standpoint of the toxicologist many different cflcctv nM, lumped together in an clfort to simplify a practical task. In a brief consideration of die basic research aspects of tfiis problem in terms practical uses made oT the determination of "carcinogenic action" in animals. ,i problems will be posed and discussed: 1. Are Tumors Eicr Induced That Do Nat Occur in Control Animals? The primary prohlcm faced in answering this question is that relatively small bers of animals must, of necessity, be used. When a rare tumor of man is coimda.1 say a tumor with an incidence of less than I " ' of all tumors occurring m man n : be remembered that it is only by examination of a very large population that Mich It arc discovered. There arc on record various examples in which an experiment,ilm 1, concluded that he lias induced a new lesion for the first lime only to discover ibe occurring in his untreated controls at a later date. Since many of 1 he enmnneem in the experimental laboratory are common environmental contaminants, allvu .u 1 concentrations, it is not surpusmg that continued observation of "untreated" o often yields examples of all the induced tumors seen in test animals. It is thus id-advised to conclude that a tumor seen in treated animals net rr ivo. the controls. The implication of this conclusion is that the occurrence of ononrvo.1 i [i lesiops, however rare, cannot be used with confidence to classify j^compminiLud 'carcinogen. '' 2. Arc Benign Tumors of Significance to flic Toxicologist ? In the first instance benign tumors may cause death in man and animals william < undergoing malignant transformation. The induction of a benign tumor is. itwll. U fore, an indication of a serious adverse reaction. There can be no doubt fiom a sun,, experimental studies that benign neoplasms arc often precursors of mahgn.mui", Under these circumstances, it would be wise to take serious note of the occur ren. benign neoplasms in experimental studies, although this alone is not sufficient Im a, elusion of carcinogenesis The occurrence of metastascs provides an unequivouhl. i stralion oT malignancy. There arc, however, many tumors induced experimental!' " are invasive and arc classified as malignant, that metastasize only rarely duiir-1 , average experiment. The absence of metastasis, in the view of most pathologist', d i not rule out the diagnosis of malignancy. Transplantation has been used by some investigators as additional proof of m.f nancy. This method is employed infrequently and has certain drawbacks. There ate.'' example, some tumors such as the mammary fibroadenoma of the rat that arc hemp: all respects and yet may be transplanted readily. ; w'hai D the Sigi lime goes on, mated with it ,,,i,e that have b :l' i Vi tain cxpcrim .,,,,,ng radiation' ,,.li may someth ., ,111 luiuors kni , ,|,c application .mphomas. howc \ aoc have been . uiiu'gcns artti ' f,atin and Wiscl. iiiMis hcluecn vi .i.iuoreiis may h ,unl by the fact 11 ,.,.11,1 io induce l , moated only w, . i,mu as the pol , -p.iuailly associ. am,,I state. 1 Ins ikd into new! \\ e ate. thereto ; mxicologic p; .oimgen in the ii.uismilted by mill be difficult .ihmi.mcy is kni liaie that leukc mu' of African 'll cl. I he suggesiivc iiiily of ways i, .'Mionuiantly c upoi lance. I low 'miles elucidate imled routine!; In (lie present uli the status c la uing for car iy he difficult 'mil as can be d 'iiinals. Advice uumals for bio; Vuional Acadci BOR 014511 -cation is drawn to the possi' lity that this toxt may b copyrighted, urced to. investigate its copyright status. .____ ________________ TUT |k ') Till in, . Tils I'i f . ( - Il.'l/.ir,! j i , tni>i\n 'C lilt IIM . -ll ll.C, c . n of i,n Tech in. i i U'niu . f ,, ils. ;i r r. V mu,PI - .'mi'i.L i urn ii i I Mil )l g ; III , l illl'l'll i" ` cd" .Hi- "IT I'l l I I ! : oi \i i, ipniunl I V i 11 ll 111 f I 11 veil, ! I .1 Mil i i i ;iik ii-i L` l H 11* * . . 1 It fill .1 , KM I ill"-1 cm.ill' ' diimii' i igivtv. -I- if of in i' ( ere me, f . . c I'ciiik'' ' CANCER TESTING 421 i ji Is the Significance of Viral Oncology to Testsfor Chemical Carcinogens ? iitinic goes on, more amLmorc tumors of laboratory animals have been found to be ,,C(j ivjtli identified siirnl agents. Most notable arc the lymphomas of the 'T]hJThavc been shown in some instances to be causally related to a specific * i.iin experimental models have demonstrated that lymphomas induced in mice by u radiations (Licberman and Kaplan. 1959) or by a chemical carcinogen (Toth, ,mav sometimes be transmitted hv a virus. Ollier studies have demonstrated that iiLmorsTnown to be transmitted by a virus in mice may be made to occur sooner explication of certain chemical carcinogens (Kiisdthaum e/a/., 1940); these same minas, however, arc not influenced by all carcinogens (Toth and Sluibik, 1967). ,, li.uc been many studies to determine the effect of combinations of chemical ifcns.'ind viruses which arc not necessarily carcinogenic. Some of these studies i .uni Wisely, 1963; Puran-Reynals, 1952) indicate that it is likely that inter- between viruses that are not usually considered oncogenic, and known chemical - .gens may be important in the occurrence of neoplasms. The problem is complip'vtlic fact (hat, in recent years, various common viruses (c.g., adeno 12) hat e been ) in induce tumors when injected into infant rodents. These viruses have been lid mil) with nonnconlaslic conditions in man. In addition, a ubiquitous virus, inns the polyoma virus, has been found to mlcct many mice although it is not -.fill' associated with the occurrence of neoplastic disease in these animals in the ! talc. Tins polyoma v irus cun. however, give rise to many varied neoplasms when ol imo newborn mice and hamsters. etc, therefore, faced with serious problems insofar as testing and interpretation 'wcologic purposes arc concerned. Is there any value in a lest of a chemical -ogen in the Akr mouse? This mouse has a high incidence of lymphoma known to nvniittcd by an identified vims. Ifan enhancement of this lymphoma occurred it T he diliicult to extrapolate such results to man since, up to this time, no human T.uicy is known to be transmitted by a virus. On the other hand, many scientists cihat leukemia may eventually prove to be a virus disease in man, and in the in-e of Afiican Lymphoma (Burkin tumor) evidence continues to accumulate to this t suggestive evidence that viruses and chemical carcinogens may interact in a ;> nf v.ays indicates that tests of chemical carcinogens with viruses administered muantly or demonstrated to be in the animal model may be of considerable `iar.ee. However, for the time being it would seem essential that until further basic ' a elucidate the significance of these phenomena, such studies cannot be rccom' vj routinely. "It present state of our knowledge it is important that the toxicologist be familiar 'lie status of viral oncology and that the viral background of animals used for rnig for carcinogens be established. Although standardization of animal colonies fc difficult, investigators performing carcinogenesis bioassay should know as A as can be determined about the viral, bacterial, and parasitic contamination of the `jls. Advice on studies directed toward the development of defined laboratory "Jb for bioassay can be provided by the Laboratory Animal Resources of the 'Wi! Academy of Scicnccs--National Research Council. / / i I Il T BOR 014512 mention is drawn to the possi lity that this text may ' copyrighted, Tnc reader urged to.investigate its copyright status.___ ^ ` 422 I DA PANLL ON CARCINOGI.NL51S 4. Is It Important to Take the Genetic Background of the Test Animals into . A common criticism of animal studies that reveal a carcinogen is that tin- Mr,', . undertaken in animals with a high tumor incidence and that the study thus appt,^ to sensitise animals. Conversely others allege that a study was not salisfuctorv K the particular animals used were notoriously insensitive to carcinogenn ... Cancer researchers arc extremely fortunate compared to many other biologists m t , available to them many inbred strains of mice with known incidences of mv;- disease. In many instances there is also a considerable background of inform,!- the sensitivity of these strains to (lie effects of known carcinogens. It is thus p,.s,,> . test unknown compounds that resemble known carcinogens chemically m rta,.. small groups of animals and to derive meaningful results. ' Shim kin eta/. (1966) has recently suggested that investigation of compound-, m i- A mice and gauging the response by the incidence of lung adenomas might come sensitive and accurate screening test for carcinogens. This contention requires experimental study before it can be recommended generally. However, for specific situations there is no doubt that many of the special strains available .c 1 some instances, specially bred hybrids, provide a much more homogeneous p.ipi . to study and obviate many of the problems in evaluation of effects that are i..- randomly bred animals. There arc, nevertheless, many investigators who prefer a randomly hml .im- the basis that the genetic heterogeneity of such animals is more akin to (lie situ.o. man. There is, of course, much to be said for this viewpoint provided it is nw: that often many more animals are required under these conditions Only in a very limited number of eases have the particular tumors naiim' - inbred animals (Akr lymphomas, C3H mammary tumors) been shown to hcas,,., with a viral transmission. In the remaining instances the precise mode of origin . tumors remains mysterious. The genetic influence lias been established .uul le studies arc needed to uncover the mechanisms concerned. Under these conduit'-., difficult to construct experiments that have clearly demonstrable relevance m p licular situations in man. Special strains can be used for special purposes and pa- incorporated into screening tests. 5. What Are Cocarcittogens and Promoting Agents? The term cocarcinogcn was coined bv Shear (1938) to describe an cnhanccniu" < the local carcinogenic effect of a weak solution of bcn/o[r;]pyrcnc on mouse shm > i noncarcinogenic solution of a fi action of creosote. Various other noncurtinogciw 1 already been shown to have a similar action. Notable among these was the m. vegetable mixture, cioton oil. Wb.cn mixed with a solution of bcn/.o[r/]p>icr.r below the concentration needed to induce skin tumors in mice, croton oil. tionc.u, genic alone, augmented the action of the carcinogen and gave rise to many imr Subsequently it was found that a single noncarcinogenic dose of the carcinogen o be followed by repeated applications of croton oil and that many tumors would n (Terracini ct al., 1960). Similar experiments in rabbits showed that carcinogcnm.i the skin could be begun by applications of coal tar or methylcholanlhrcnc and i pletcd by wound healing or applications of turpentine. This two-stage induction of" rs was the basis fot i " I he promoting . ,,-H'd: it has been fr. ,-r substances with ' ,s,i!lio!ti and Shu aonslrnled in carcin Inch a two-stage m v wliote much nior; .-l, can be said to has I a cioton oil or subst .at Although such n ,-wnt state of kimwl I v.treinogcniciiy or pi ii'it. testing should I Mixtures and the T studies on record n.mogens have, in fulion. Lacassagnc minis administered upciitivc inhibitioi . ued and extended lumsti.ited that ac: hepatomas in rat .unis investigators .luciion. Since then uiiungen.s hove bee. . oar lack of mfon riinogcns m anima '.nil miisi involve : ii.ncd in animal s there are many ), , 'cscnl in our cm iro dy known to be act '.'liuill, if not imp< nievi of the total I he polycyclic nr m".'.wi polycyclic ar im residues, tobai "d smoked foeJs a nil similar quanti 'i/.ud. However, 'postirc. In Ihc in .'pi ofcsimiating " pproach to the ass vJch an approach . BOR 014513 ;cDtion Is drawn to the possi*-lity that this toxt may urged to investigate its copyrightstatus. _ * copyrighted. The read \i 1 > n : i' id. . r i, . (, n! ,, *f i .- , , r' i * ' t Uf/- / ^ j (, riu * ' .<( i [l|. r * C ( ; i \lri ' i ` i 1C M ' i ' . / ri: .* MU f t V It.'" ' J.`v Ml ' .t : t fTM' I 1 rtflJ *f r l'f i* * CANCER TESTING 423 tmjuIic basis Tor a hypothesis describing the phases as "initialion"and "promo* pc promoting agent croton oil has proved to he the most potent so far dis< a lias been fractionated, and interesting substances with this action isolated, hstances with less activity on the skin have been found (Lijinsky and SafTiotli, ^;i',oiii and Sluihik, 1056). The two stage mechanism has not been convincingly 'rated in carcinogenesis in sites other than the skm. There arc many situations >,n mo-stage mechanism appears as an attractive explanation. We arc still at a ,lcrc much more study is needed before many of these concepts, valid for the n,k said to have general application. Front a practical standpoint it is apparent '.on oil or substances of similar activity would pose a considerable hazard to the Vilnuigh such interactions ntay pose an important public health problem, our .. ,|,,|C of knowledge does not make it practical to recommend routine tests for mrcnictly or promoting action; if circumstances suggest the existence of such ('.'ting should lake this into account. This area needs intensive study. 'Mci an<l the Total Em ironmaUal LootI of Carcinogens on record that are concerned with the effects of mixturcs-of chemical :en$ have, in the main, reported the occurrence of inhibition of tumor .'n. I.acassagnc a cil. (1945) repotted that polycyclic and heterocyclic hydro* ;n!minixlcicd together resulted in inhibition: this was thought to be due to ,\:nc inhibition foi similar receptor sites. Subsequently these studies were re* ,inj extended with essentially similar findings. Richardson ct ol. (1952) later cited that administration of mcthylcholanthrcnc could inhibit the induction -iit'in.is in rats bv p-dimcthylaminoazobcnzcnc. Many additional studies by investigators (Gclboin, 1967) demonstrated that this was due to enzyme .hi. Since then many instances of effects of enzyme induction on the action of mis have been icported. Many students of carcinogenesis have drawn attention -I,nk of information on the elfects of low doses of combinations of different ecus in animals. It has been said that the practical conditions of human exposure must mvolxc the presence of many carcinogens of different types has not been J in animal studies. e .ire many known carcinogens that are chemicals, physical agents and viruses 'in our environment. Some of these agents are known hazards toman; others are ''into be active in animals. It has become apparent to some researchers that it is uf not impossible, to assess the hazard from any single carcinogen out of the I i'f ihc toial environment. . p'luxclic aromatic hydrocarbon carcinogens illustrate the point. There are pnljcyclic aromatic hydrocarbon carcinogens present in coal tar, certain petro'.'iihics, tobicco smokes and tars, polluted urban atmospheres, (^laccaaliheoilcd -'fed foods and oilier products. Obviously the discovery of an additional source liTilTquaiititics of these compounds could be said to contribute to a general . 1 However, the assessment of the hazard must take into account the tola! -tc. In the instance of hazard from ionizing radiation a relatively arbitrary confestimating "hazard above background" has long been used. Eventually a similar -di to the assessment of hazard from chemical carcinogens would seem indicated, 'in approach is even more difficult in this instance because of the difficulties of i i i t I I III i BOR 014514 drawn to the possih "lity that this investigate its cop^.^nht status. toxt may b'' copyrighted, The reader 424 IDA PANPL ON f ARriNOCit.NI SIS quantifying the levels of the many different carcinogens present in the environ,,;,-. . the present time the committee on Quantitative Carcinogenesis of the Internr Union Against Cancer (IJICC) has set itself the task or providing a tabul.ii,,-!.1 known quantities of chemical carcinogens in the environment and listing an.ihmethods. This task is formidable but underlines the need for more information lv philosophical basis for the evaluation of hazard from carcinogens can be cvtahlo' a rigorous manner. 7. Arc There Special Teatarcs Associated with Hormonal Carcinogenesis? The refusal to establish a legal tolerance for the herbicide aminolriazolc ruin'd questions that have not been answered To the satisfaction of many low.,.: Aminolriazolc is one of a series of thyroid-stimulating agents that give rise it, p . hyperplasia, which in turn can lead to the occurrence of thyroid adenomas a*id v. ,, nomas. Additionally such effects arc intimately bound to the total hormonal mi'. the animal and may be accelerated or inhibited in an indirect manner. In Hie um., stilbcslrol, a substance with hormonal activity of great interest to the lowv.a interested m food additives, the situation is similar. Piolonged adminisiraiinn . t hyperplasia in the first instance, leading to the occurrence of benign tumms ,m,| carcinoma. However, in the instance both of certain of the ibyioid-siiimil.iti'u; moncs and of certain csuot'.cns the induction of hepatomas, not appaicnth nv, . directly with hormonal effects, can occur. In the main, bossescr, hormonal carcinogens arc distinguished from other i., nogens in that, the final result is preceded by an exaggerated physiological clUai The problem of assessment of hazard from hormonal carcinogens is. tlierd-m- less complex than any oilier problem in chemical carcinogenesis. The In/.it.I i exposure to hormones must he evaluated both in the context of the horimrn.il i.m and as chemical compounds When the tumors induced arc clearly related In p' ' logical ctfects it may be possible to assess dose response relationships and olltci v (ions of exposure on a reasoned basis. However, in the instance of many Imu other effects of a chrome nature may occur which have not, so far, been assou.iti.sl the known physiological iole of these agents. 8, The Role of Pathological Reactions to Inert Materinis Physical carcinogenesis in the subcutaneous site has been extensively studio! I well known that a vaiicty of chemically inert materials may he inlmthke.) * ' cutaneously in rodents and will give rise to sarcomas aftei a relatively long I.' period Most such studies demonstrate that the ability to induce tumors is tel.itevlnophysical nature of the mntet iul (hardness, perforations, configuration, etc.), and. v' the mechanism remains unknown. In certain studico with food additives the occurrence of bladder calculi siihw.|t associated with tumots of the bladder has been the subject of much discussion In' case of the compound polyoxyethylene monostearate (Myrj 45) bladder cakuh reported to occur at very high oral dose levels; some ofthese animals developed hT ' papillomas, ft has been widely assumed that the two events were related nltimui'h11 has not been proved, ft has been shown, however, that glass beads inserted uiio u urinary bladder could give rise to papillomas in rodents. 11 can, therefore, be rcasouf . that in ll,': ' _;||K.,kc of the ealcu . v in a secondary ma !:rconditions where i,cd and conclusion DISCUSS One of the pi "nary .^deration of tecon . c most of lI'-e wr ,necs have been m .,e been performed - oiopiiaicly incorpo' ..nsdons of the com. Ttuhisiry and the tu lhe Dcsirabiliti <>f tnhnals in Ron,'me j hoi c is agreement I he use of S'T , muls. as free from , ,,based SIT- amma ,e conditions the S , epiihlc to infecti< l Ik'ic o consider.i' ved wall eonventio - ..male the view that 'lor chronic loxi ' I he IShulJer Imp!. Several chemicals mlieant increase i: urol animals imp .: sli increase ihc i. ;-me. This assumpi '.inonstratc safety t ueil both with thi >ltle secure interp : an. by this assay `'aid he obtained 1 unetl only with 'cinogenicity. 1 s.iicC this report million of the VIU ".a, Geneva WAU, Ml'clh SVcishnrger, (, 1907. If. / t BOR 0145X5 mention is drawn to the poss-' ility that this toxt way . urccd to, investigate its cop/rip:ht status. copyrighted. iroill!"-' ' ' Intern 0 > i -uil:Hi"n ' i Mflj' .til I ! 7 CSlablr - ' - 7 " raise K'1 s Inxin'1 : rise tu r mux .11 I inn.it >i" the iii.i i - ' W lli'in ! .waium i ' mmi - '* t " 1 lOHit.iti"; ' 1 ,tly i'"'*- 1 n'Ui-t *1 ' it cnij1 , then h'r' ` ; lu/.it't ' nuut.it -i < tied tu r' 1 id nti.et i- ' any turn issue t;tt< i " ` y sluilnV t i troilun-.! i 1 |y li'M;' I; . s refill'd i t ), itiiil. i - i MihnH|. emsinn I- 1 .r cab i:h " : eloped U i ' i iililuniy1! 1 ' serted <>* ' ' , l>c rr;t'i>' * * CANCER TESTING 425 - -J that in the Myrj 45 finding the induction of the tumors was secondary to the net of the calculi and. therefore, that the compound manifests tumorigenicity , j secondary manner, ft would seem logical that tumors arising experimentally .nations where physical factors might play a significant role should be carefully and conclusions based on common sense evaluations. DISCUSSION OF UPDATING TESTING PROCEDURES .'filic primary tasks to which the committee has addressed itself has been a -.ration of recommended test procedures for carcinogenesis. It is over a decade -iosl of the written recommendations were niadc.J During that lime many ..sfi.i'c been made in biological knowledge and techniques; many experiments --,:cn pcrfoimcd in carcinogenesis. It was Tell that this new knowledge must be .;i,nely incorporated in test procedures. The following proposals emerge from ,;w of the committee and its consultants and after a meeting with representatives ary and 1 he academic community. , Dew-ability of Introducing Specific Pathogen-Free (SPF) Animals or Germfree <'ah in Routine Tests c is agreement that germfree animals are not required or desirable for routine II,c use of SPF animals is also not recommended us a requirement. Healthy A.asftce frmn infection as possible, are recommended. Some investigators have ,-ctl.SPF animals and used lhem in animal quarters with standard animals. Under mndihons the SPF animaK were, in fact, less satisfactory since they were more ruble to infection than standard animals. ar is considerable dispute as to the length of lime that SPF animals survive coin! *ith conventional animals. There docs not seem to be adequate evidence to subre the view that the introduction of SPF animals might lengthen the period avail' r chronic toxicity testing. 'Bladder Implantation Method of Assay for Carcinogenic Activity ;ut chemicals which are carcinogenic by other tests have been found to produce a -jnt increase in the incidence of bladder carcinomas above the incidence found in I animals implanted with the pure vehicle. It is assumed that other chemicals ` increase the incidence of bladder tumors with this technique arc also carcino- . 1 his assumption may be correct. However, at this time it is not adequate to wtr.itc safely of a chemical with this as the only test. The number of chemicals I both with this technique and with more conventional methods is inadequate to .secure interpretation at present. Hosvcver, a chemical which is clearly carcino-> this assay must be regarded wills considerable suspicion, and further data J he obtained with more conventional assays. More impouant, negative data *cd only with this method cannot be accepted as adequate evidence of lack of ogenicity. ' .r ihis report was prepared, "A Report of the Panel on Carcinogenicity of the Cancer Research "'wn of the UlCC," Carcinogenicity Testing (I. Bcrcnblum, cd.), International Union Against '.Geneva IQG9. has appeared. See also, "Tests for Chemical Carcinogens," by John H. and xdi Weisburgcr, in: Methods in Cancer Research (H. Busch, ed.), Vol, 1, pp. 307-398. Academic mi. \ i *i jij.ip' .'.P'y.irirw mf*m BOR 014516 tentioa is drawn to the pose. iXity that this toxt may urced to investigate its copyright status. copyrighted. The reader 426 FDA PANtX ON CAKClNOGf NFSIS 3. The Use of the Dog in Tests for Carcinogenicity potent human bladder carcinogen. Had this drug been tested properly satisfactory substitute. rel.ii,.) ll Oil! p him n ' li e d. 'I'conp t le.ii! ' prow : ., 4. The Use of Species So Tor Not Commonly Used in Such Tests There is general recognition that new species in addition to the common:, rodents arc needed for carcinogenicity testing. Increasing the number of sp., -, would greatly increase the margin of confidence with which safety can he pi. Up to this time there have heen no species Hint can be recommended with km 1 A variety of species show promise, but in all instances further study is needt I ' they catt be recommended for routine test purposes. The most discussal .id1 species arc primates. Recent studies show certain primates to be highly Mivq some carcinogens; howcscr. much more work is needed to justify thcnddiiiun.ii i >; and time that would be necessitated by the introduction of these animals in rone procedures. Among the other species that have been suggested arc fish, partiin'" . rainbow trout, and various avian species such as chickens, parakeets, ,a-! Japanese quail. At present there arc no specific recommendations that am ofspecies should be considered for routine testing. 5, General Eneironmentnl Conditions Although Committee attention has been directed specifically to airconlainuu' animal quarters in studies of chemical carcinogenicity, it is ci idem on rcllaiion i; is only one part of the environmental system that can be inadvertently cuni.i with the chemical under test or other carcinogens in the environment. Thus, a it. under test may enter the general laboratory atmosphere from faulty technique r ' preparation or by secondary acrosob/ation of animal wastes. Other example- .ur. lamination of diet, air, or contact surfaces by chemicals used for insect or w control, or from preceding tests. Although it would be inappropriate in a document of tin's kind to attempt l"1' specific requirements for control of environmental contamination by potcn'Mi i nogens, (lie Committee emphasizes the extreme importance of good amm.i! utc f tics in an)' long-term study of low grade toxicity or carcinogenesis. Details of Ur. ' caging, and environmental control may vary from one facility to another, but thec tial ingredient of good animal care is the personal component--the wcll-inun highly motivated caretaker staff who arc genuinely concerned about the ucli-lvc; the animals. Minimal standards for animal facilities and care are now specified by law. hen guidelines, including special requirements when toxic, infectious, or radioactive > ,.,^-vaie used, arc ^ ,,,,i.il facilities an . ;j.v Use of a Sra 1 iiciuue safety cv. pl.-s diets of pra . ..mis or, on occ ,v Concern with ilUl gt ow th, rep i paoiighout the , .,-i ms of the pro' . ; In 1953, Ta ,,K'i of constiti .udi because tl .-I., set undctcc , ,mogenic actis . .'Iv"' smcc then, mar ; t.is there is an ,... nts with intc 1 uiMis that in the uoid "contanv . ,!|s I in- occurrence ;k,il response h. i I ivtciing (190 |i is clear that a -lompoMlion O' t sti ingent qu i.c soil from on . instancy of i. many grow i .mg practices t "conlamina s he serious ( 'iaie siudiex i i .inimals app i croup did n - li (he presenc ' ulicrmorc. v optible stra 1 1 unants in ih I lie essential -rat and ihe > Mials on cbci as made upc SOJl 0lisl7 tcntiou is drawn to the doss' n t v fy,.,* 4-^.4 ^ *".lBVC*^gto 19 bop/.finht tatua" text may v, copyrighted. Thc v i I. it. I ' CANCER TESTING 427 jrc used, arc available in published documents such as thc Guide for Laboratory a facilities and Care (NRC 196S). - - Lie of a Standard Diet in Experiments for Carcinogenesis safety evaluation studies, including carcinogenesis tests, have been based on i diets of practical ingredients supplied commercially by laboratory feed manursor. on occasion, mixed from practical ingredients in thc inscstigator laborai ,-ijKorn with this basal diet has been limited to (I) meeting thc requirements for i growth, reproduction and longevity and (2) maintaining a constant tomposiaiichout the experiment. Usually thc constancy of composition is assured only v. of thc proximate analysis, vitamin content, and occasionally, amino acid conla |95J, Tannenbaum and Silverstone wrote: "...natural foods contain a .r of constituents which have been given little attention in nutrition and cancer h been use they arc apparently not dietary essentials. In addition, there must be , set undetected. Perhaps among these unregarded substances arc some with - conic activity; and others that potentiate or oppose thc action of carcinoi ((hen. many examples to prove thc truth of these words have accumulated. , (here is an awareness of the presence in foodstuffs of many nonnutrient com- with interesting biological activities, including carcinogenic activity. It is ilhst in the evaluation of carcinogenic potency, as in any bioassav, it is essential ;J "contamination" from environmental sources which may produce erroneous occurrence of nonnutrient components in foodstuffs and their influence on bio' response have recently been, reviewed by Friedman ct at. (1966), Schocnta! (1965), btcjing (1966) i dear that a complex diet of practical ingredients is difficult to maintain constant pavilion o\n a long period of time with regard to all ingredients. Even with thc tri.ipcnt quality control, one must expect variation in the same crop grown on the 'il from one growing season to another. How much more difficult it is to maintain '. uncy of composition with regard.to ingredients gathered from many sources uiy growing periods, different harvesting situations, different storage and pro:practices, different pesticu!. residues and residue levels, etc. Experience tells us wniamination" of the typ. d w ould significantly influence carcinogenic studies serious (Newberne, 196' racini ct at., 1967). Often the interpretation of .studies is extremely dill', because the incidence of malignant tumors in con corals approaches that ob . ed in the test groups, forcing a conclusion that thc "updid not differ signifies dy from thc conti of Such a conclusion is consistent presence of low level cmvinogenic activity in the test group or in both groups, '.emore, what can be thc significance of the incidence of "spontaneous" tumors in ihlc strains when one is not certain about the presence of carcinogenic contnis in thc diet on which thc animals have been maintained? '`essentia! nutrient requirements of many laboratory animals, particularly that of au*J the mouse, have been identified to the point where it is possible to maintain ft on chemically defined diets. Somewhat simpler and more economically feasible mdc up of purified ingredients, such as purified casein, sucrose, starch, salt mixes, BOR 014518 tcatioa is drawn to the posr 'illity that this urced to. investigate its copyright statu text may ''o copyrighted. Vho ro.v 428 FDA PANtL ON CARCINOGI.NrSlS and synthetic vitamins have been used by many investigators. Halvcnvu/ (p>(. used such diets in studies of hepatoma over a period of several years in r.nnlv., Wogan and Newberne (1967, 1968) have recently successfully completed an Is study in rats on the carcinogenic potency of aflatoxin H, using a diet \cry siiiii!,u used by Allison el a/. (1950, 1954). The feasibility oT using highly punt,:1 synthetic diets mixed in the investigator laboratory under his continuous corbeen demonstrated. The advantages of such a basal diet in carcinogenesis stud the-point of view of maintaining constancy of composition with regard to ,f ponents, and particularly with respect to elimination to the highest practical if,. all potential dietary carcinogens and other factors that may influence caretrv seem to be sufficiently compelling to recommend its adoption in all carutio' evaluation. The increase in cost of feed, although appreciable (from approxima'ely Hu, ,, pound to approximately 50-60 cents a pound) represents only a minor fraction , total cost of a carcinogenesis study. It must he emphasized that as much cau e be exercised in preparation of hasal diets as in the preparation of experiment!1 containing test substances. 7, Pathological Study It is of the utmost impoi lance that a complete and accurate pathological evum bo conducted on all animals used in carcinogenesis experiments. A trained pad; should he concerned with both the gross and microscopic diagnosis ol all iic.-p' found. The pathologist should have a part in designing and conducting the i\pw The autopsy should lie done under his general supervision. It is essential that the plogist be experienced in the diagnosis of animal neoplasms; knowledge of human t, alone is not a sufficient background for expertise in this field. It is desirable tfi.c gross and microscopic examinations be conducted without knowledge of the iri-.e of specific animals. Many pathologists employ a check list to ensure that a complete autopsy h.n ' < employed, and it is felt that this is a good procedure. S. Rapid Tests There is no rapid test that the Committee recommends for routine scimn ; 1 carcinogenic chemicals. 1 he most promising procedure is tissue culture. I his met' 1 used routinely fort he evaluation of oncogenic viruses, whet e "transformation" o! t cultures by viruses is used as proof of oncogenicity. In its inception, "transform ' was a phenomenon characterized by the dcsclopmcnt of unlimited growth poten' a culture on continuous transfer, and by loss of contact inhibition as could he so * the architectural pattern of growth. The proof of the changed characteristics <'1 culture was the induction of sarcomas on rcinoculation into an animal. Since tli.it' only the manifestations of lack of contact inhibition have been used as the indie,it-1 9 tire instance, of certain experiments conducted with well established lines of uiiimo experienced virologists, valuable comparative data on the characteristics uf ihlb" viruses and their mutants may he obtained. However, these findings cannot yet h - as an absolute diagnosis of carcinogenicity of an unknown in the absence of tdhUi" .. procedure including .iMt- that the possihdi In spite of these strict ,v in llcrssald and S. ,.,,ocultures. It must) pound m t iro. Th . arm. ( 10 of .Von horn sin I l,c newborn moms. Mimgcnic chemicals . ,,ise was more serh.itt - adult. T his was cor .n.iicmogen in lho ;n mmiiine response u . ior may also play ; l ei lain compounds . ,\inngen may be le~n ,>u-e has limited ap; nances, it may be i n of materials nia i '.nuls can then he i cm only a few mont iten the period oT ,i.uiccs on record in : laieinogcns at bin bull cl ai, 1964) an nomas by DM HA ,,asin in rats (Htroi nr) by urethan in i u newborn anim.i' I sis might be a satirecommended as 1 the Combir.r.no' I lie evidence is st: ,!rms or of seu-ra 1 weaning. and ob:i r detection of cars ' mkovitch et(d. (I tlie life of animal Hirer indicates th. 1 frets for careinor In the context of i primary concern . ' complete a sc'i i BOR 014519 "teation is drawn to the poss ility that this text way ' > copyrighted rn,, jjrced^to ievostigate its coK/rifrht status. u ' ;nc ri e i i 11 U * . ' CANCER TESTING 429 vcjufc including rcinoculation. Even when this is done, efforts must be made to .. iliai the possibility of "spontaneous transformation" has been excluded, . .-lie of these strictures more effort should be made to follow up recent invcsliga. ^ ii^-wald and Sachs (1963), Hcidclbcrgcr ct al. (1968), and others on the use of ,, cirlwrcs. ft must be cautioned that many carcinogens arc metabolized to the active .mnd h> The activity of such compounds might be missed in an in vitro a fn( Newborn Animals newborn mouse and hamster have been used widely for the study of both - ifciuc chemicals and viruses. In some studies, it was found that the newborn was more sensitive to the rapid induction of lymphomas and lung adenomas than This was correlated with a longer persistence of the unchanged molecule of jfcinogcn in the newborn than in the adult. In (lie ease of viral agents the lower level line response in the newborn is considered of major importance, and the same may also play a role in the instance or the chemicals. .sun compounds such as fluorcnylacctamidc Hint arc metabolized to a proximate i'jicn may be less active in the newborn than in the adult. Therefore, the newborn , c has limited applicability as an overall lest animal for carcinogenicity, fn some ,\cc it may he convenient to use newborn animals since relatively small qunnM'f materials may be administered on'y once or a limited number of times. The -can then be maintained without further treatment and tumors be observed -a few months in the case of very active materials. However, this test docs not ,nthc period of total observation required for a negative result. Indeed thcic ate -.won record in w hich tumors have occurred very late in life after administration i .inogens at birth. These include the induction of (a) hepatomas by DM N in mice mini, 1904) and by urethun in mice (Delhi Porta ct al., 1967), (b) subcutaneous us hy DM HA in rats (Totli and Shubik, 1963), (c) kidney and fiver tumors by m in rats (Hitono ct a!., 1968), and fd) neurogenic tumors (brain and peripheral vtby urcthait in rats (Vcsselinovitch and Mihailovich, 1968). It has been suggested r.avhorn animals might be (tented at birth and then followed for only one year. aright be a satisfactory screening procedure under limited conditions but cannot commended as a routine test procedure. t v Combination of the Chronic Toxicity and Reproduction Study '; cvitlcrtcc is striking that sometimes administration of a single dose to I-day-old tv or of several doses beginning on Die firs! da v of life and at intervals up to time `..ming, and observation up to one year, constitutes a very sensitive assay system for `.tection of carcinogenic potential, The knowledge, for example, from the work of which ct al. (1966), and Druckrey ct al. (1967), that tumors may be produced late '.`life of animals after administration of a caicinogcn during the pregnancy of the Af indicates that the embryo and fetus may sometimes be particularly vulnerable ct for carcinogenic substances. the context of a safety evaluation protocol in which subtle chronic effects are ' vary concern and difficult to demonstrate, and in which the purpose is to assure `ntpletc a screening of as many potential deleterious effects as possible, this BOR 014520 .nt1^ is drawn to the .possibility that this text may bo copyrighted The urged to. investicate its cor .ight status. `y Utcc1, jhcro-'GC; 430 VDA PANTL ON CARCINOC.LNI.SIS information on the susceptibility of perinatal animals suggests some mojiiuy in the protocols of chronic toxicity studies. The chronic toxicity studies may be carried out independently or coordinated . . the reproduction studies, at the convenience of the laboratory. At present, itstudies, including carcinogenesis, begin with weanling animals derived from a n. stock colony, maintained on adequate practical diets. These animals are ilwn . laincd at desired dose levels for a "lifetime," usually a 2-ycar period for rats.'Ilies' received no perinatal exposure. Since one of die important purposes of ilie iLtoxicity tests is the detection of carcinogenic potential, it would seem dcsiiahle io!,:: the exposure as early in life, i.e., as close to conception as possible, because one ... reasons Tor a reproduction study is the detection of effects in the offspring, fuming of the parents before and during conception and gestation, it would seem prolu.i' ; combine the first stages of the reproduction study xvith the cluonic toxicity mi r the initial animals started in the experiment at weaning would be bred and the nth; - . either used solely for the 2-year chronic toxicity test or continued for the remain ' the 2-xcar period to serve as chronic toxicity test animals as well as the I', cniuy after producing the second generation of the reproduction study. This would ;ul! advantages of perinatal exposure for carcinogenesis studies, and the chronic tnviun carried out in this manner may prove to be an adequate lest for carcinogenic ha/aid 11. Statistical Aspects of Safely Evaluation There arc two questions that may be asked about an agent under test: a. For a given set of experimental conditions (c.g,, total dose, fractionation, rm.'r administration), is there any elevation above control levels in the incidence of c.r.ri.e., is the agent carcinogenic? b. If yes, can a dosage be found below which there is no elevation, all oilier tip-- mental conditions remaining the same; i.e., even though carcinogenic, can a safe be found ? Although we are concerned with carcinogenicity, it is clear that the formal argmu- largely applies to any other adverse effect of interest as well. Extrapolation of lone t." /chronic effects from animal to man introduces more uncertainty that extrapolation > yj acute effects. We consider statistical aspects of each question in turn. Is the Agent Carcinogenic ? a / / Although a positive answer to the question as posed can be given in some parti. ^ v/ instances, no unqualified negative answer is ever possible. Thus, if none out nf n animals exposed to the agent develop cancer, and n is a large, number, the clcvaii. - any, is certainly small, hut no matter how large the value of n. it is not logically pc to conclude on the basis of such evidence that no elevation has occurred. Hie 1 might in fact induce one cancer in every 2n animals, and the experiment, being toiiaz, to detect it, would be incapable of providing cogent evidence for the absence nf aelevation whatsoever. This is true no matter how large is the vahie ofjn A negative experiment cannot therefore.be rgmirdort :Uwn;.ihmvn that noth : lion^exists. It can, however, supply an upjjerJi.mil to the possible carcinogenClearly, the larger the value of n, tlnTcToserlhis upper limit will be to zero, but there n* ..,,,ch that will set it cqi ..mpulc l!'is linPcr llm ' :lK|, cancer will bcindi proportion may ha one animal will dev. iiall i/ test animal-, wi ,l,,c of P is the probab is easily obtained ^ | ^ 0.011 ". The .known value of I'. V, Jo\v. Urrr.a Comic Wmdfi i 10 100 1000 lliesc upper limits a aid using only 90co ., 2 3 cancers per 1000 iiman population for -1 more than 2 tumor allion with confident ..are than three mill flic situation is not iMc. We show below /'<- where the so pmportion of animal equal for treated and for the case in whic Ui-rF.R 95" 6 Experiments Same Inc. n 10 25 100 500 1000 ` The approximation * : <, -- i nifcufin*"* pr* BO* Ol*521 .S'; j T T ' *' <"3 rcntioa is drawn to the poss ility that this text may urged to investigate its^copyright_status. ___ 1... _ t copyrighted. The reader _______ r cj " safe i' : r >' U::* `J 'll,* !- ' i!.-||h>*i r Cl! * 1 no '>' i)t' n!. i' ' ncte > * CANCER TESTING 431 nthat will set it equal to zero. An elementary probability argument shows how to --uic this upper limit in (he simplest ease. Denote the proportion of animals in mccr will be induced by the agent under the given experimental conditions by .proportion may have any value from zero to unity inclusive. The probability that .., nc animal will develop cancer is A and that it will nojjs (I - P). The probability j!l test animals will fail to develop cancer is then (1 - P)n. We then ask: For what (Pf /> is the probability of the observed outcome of zero out of n small, say 0.01 ? , is easily obtained by setting (l -- P)" equal to 0.01 and solving for P to obtain , | - 0.01l/". The interval 0 to Pm,, then provides a 99% confidence limit on the riiwn value of P. Values of Pmtt for various n and confidence coefficients arc shown Upper Confidence Limit tor Possible Incidence for Experiments IN Which All n Animals Tested Ahe UNAFrECtED, by n and CoNriDENce Coefficient n 90% 95 % 99% 99.9% A U\Q ~ oi * Vp - v- D -T. Mvs to too 1000 0.21 0.023 0.0023 Pm.i 0.26 0.030 0.0030 0.37 0.045 0.0046 0.50 0.067 0.0069 ;*.vc upper hmils are uncomfortably large. Even with as many as 1000 test animals : MngordyVu^TonTTcTencc limits, the upper limit yielded by a negative experiment Tunrers per 1000 test animals. No one would wish to introduce an agent into a an population fojj^iig[ij\oji]jiBLLO!jId.bc said.! IlLLOJl llil l it, wo a kljsroJiahh'jipod uce "re thui. 2 tumors pci 1030. To reduce tltc upper limit of risk to Tjumors per one niuili confidence coefficient 0.999 would require ajicgaU\'c_rcsult in somewhat ^ m1i.hi three million test animals. ' c situation is not improved when concurrent observations on controls are availr We show below the approximate upper 93% confidence limit on the difference fi where the subscripts stand for treatment and control, and P is, as before, the pillion of animals developing cancer. This limit is shown for various n (assumed ilfnr treated and controls with 2n equal to the total number tested) and values of '.'fthccase in which treated and control groups develop the same number of tumors. Upper 95% Confidence Limit for Possible Ellvation Incidence for Experiments in Which n Treated and n Control Animals Yield the Saml Incidence, be* n and Control Incidence (Cornfield, 1954) n 10 25 too 500 1000 0.01 0.10* 0.065" 0.033* 0.015 0.010 0.10 0.31* 0.19 0.09S 0.044 0.030 0.25 0.45 0.28 0.14 0.063 0.044 0.40 0.51 0.32 0.16 0.072 0.051 Nipproximanon used introduces some error in this range. BOJ? 0l*S22 '..ij i. icntion is drawn to the possi lity that this toxt may urged to, investigate its copyright status. copyrighted. The reader 432 FDA PANEL ON CARCINOGENESIS Thus, even with 1000 treated and 1000 control animals, each developing 10 inn,,.. can he 05% sure only that the elevation in incidence rate does not exceed 1 10 the agent has probably induced at most 1 additional tumor per 100 test anmuU |> be observed that the practical difTicultics in obtaining a completely uncxpuwd... arc many. As has been noted in other parts of this report, the environment ofo,-f,.. mental animal (diet, water, and air) arc often uncontrolled and may, at times, inti,.1 low levels of the test substance itself or other carcinogens. Considerable additional safely is of course assured by testing doses well aim: . actual use level. But this leads to the question of a safe dose, consideration otv . seen to be an unavoidable consequence of the astronomically large number of ,r required to reduce the upper limit of possible carcinogenicity to an acceptable k,; / Can a Safe Dose Be Found? The above calculations make clear that the only practicable basis for cstimatim* j dose is by extrapolation downward fiom results obtained at some level well ah. . actual use level. But this extrapolation introduces serious uncertainties, which i>. , rceogni7cd if rational methods of safely evaluation aie to be developed. 'I he haw,; lem is that extrapolation outside the range of observation must be based mi pet';unvcrifiablc. assumption about the mathematical natmc of the dose-response ic1 r ship near zero dosage. The variety of dose-response curves possible in di!1 carcinogenesis model experiments further complicates the problem. It might be thought that the basis for such an extrapolation could be prnn,!. ` observations in the observable range. To show how far from being the ease this au . is, we give below three different dose-response curves, mathematically defined mu1 dosage range of 256-fold. All three have the same TO,,, and TPI(,.4 The first is a p: curve, the second a logistic curve, and the third the so-called one-particle curve Expected Percent of Animaes with Tumors Actual dose (TDi0) 16 8 4 2 1 1/2 1/4 1/8 1/16 J'robit curve (%> 98 93 84 69 50 31 16 7 2 Logistic curve (%> 96 92 84 70 50 30 16 8 4 One-part ndc curve (%) 100 99a94 75 50 29 16 8 4 It will be noted that below theTD50 the three curves differ by little and that in any eg ment of practicable si?e (fewer than several thousand animals), it would not be pm to conclude from the actual observations which one of the three best described tlir i * TD50 - dose which results in tumors in 50% of the animals; TD,6 = tumor dose for If'*. animals. uqu,wn below, however.; , (l|,e hundred million do irkl'tHy. FxrR A rot MU) V Al l't.` CllkVIS IXSCRUUN TD. TDo i 1 Do h`n>l TO U.CVfOhOt TD, TDo ood v>i il:C one in cme-luindra ,imully" safe dose is o a ouxandth using the loci- ^cn with ;m experiment, virtually" safe dose is o ! imi.OOO depending on i I lie extreme unrebabili ,.o the basic source of b , icvinc (Cornfield a a!.. ..Riivation time had sc a.i uliing to an cxtrapol. inleeular basis a> Buvle i (early extrapolation fr< jVvilics and impon.lcr. i npiudem to pluceexces ii:c dose-response curve.. p'lvxieal or chemical ba Nothing that has beei vv hot docs seem clear i. (wieologic dosc-respon n the question 1 ihnlines oj a Brood Gc 1 1 he impiaetieabl;. la;l /xi-.dblc public health ' I p.lalion. set limits to 1 K-vimg program. Draw' 1 road outlines of a fc.t I. Testing should he maximum tumor incid I ' nlc above the actual i| BOR 014523 - # ` - nation is drawn to tho po.ssib >ty that this text may bo 'opyrichted ttp r^- . , *. ; ,, 1 1 iyriuntca* Jh0 rc-^or CANCER TESTING 433 'iiwn below, however, the TD0 0not (ne >n a million dose) and the TD0i(,<,0n0, (one - hundred million dose) obtained by cxtinpolation of these three curves differ .i.JIy. I viR/n'ot-ATED Values or "SArF." Dosts roa Three Drrn.utNr Dose-Response CuRvrs Descrhunc! OnsrRvro Responsis in the 2% to 50% Response Range Eqhally Well Probit curve Logistic curve One-particle curve , TD, TIV, TPn,0OOl TDo fioaofl! TT>i _ ^ \ Dfl,OWN'D! 0.040 0.015 5 0.001 36 0.000 412 100.0 0.022 0.003 15 0.000 009 8 0.000 000 16 100,000.0 0.014 4 0 001 44 0.000 001 44 0.000 000 014 1,000,000.0 j ,,r,c in one-hundred million dose, which Mantel and Bryan (1961) call the I!-," safe dose is one-hundredth the TDt using the probit curse, onc-Jmndrcd ^''^TmTudne the logistic, and one-one millionth using the one-pa rticlc^curvc. Thus, ..ah ,m experimentally well-determined TO,, or dose at which Pnwl - (5.01, the itfv" safe dose is obtained by using a safety factor which can vary from 100 to L' !<!depending on the curve selected. 'TuTcine uTTrehaKiTitv of extrapolations outside the observable experimental range c Rive soiiice of the failure of the early safety evaluation program for the Salk l-(CornHeld ct a/.. 1956), even though t he observed ctu vc connecting log liter and a,,.man tune had some theoretical phitbcul chemical basis Nor is the uncertainty ne to tin extrapolated value surpiisiug, since even a relationship with as firm a .nl.ir basis as Boyle's law breaks down ;d extremes of pressure and temperature, y. extrapolation front the observable range to a safe dose has many of_t_hCLper- and imponderables ofjyxJ/jiiyolntion from animal tojmm, and it would be ,t;ni to place cxce.v.Txc ichancc on mathematical sleight of hand, particularly when ' ^-response curves lucdarcTargcly empmea! dcsci ijptions. lacking any thcoretjcal L' d iv chemical basis. x tl.inj that lias been said bears on whether a threshold dose docs or does not exist. ' Tvs seem clear is that niore fundnj.nenuj_cxpcrini.;y(ation than that of the usual akcic dose-response investigation in intact animals is necessary to shed much light ; question. ~~ an of a Broad General Strategy af levin;: . impracticably large numbers of animals required to detect carcinogenic effects of v public health importance, together with the unccrjamties of downward exiraset limits to the amount of protection that c.m be achieved will; a routine ';program. Drawing largely on the previous work of Mantel and Bryan (1961), the -ratlines of a feasible cancel testing program might appear to be as follows: Testing should be done at doses anti under experimental conditions likely to yield "an; tumor incidence. This would mean the use of doses sevejaLorders of magni-hovc the actual use level. Although it is a simple exercise to calculate the number / B0R 0usu teation is drawn to the poss ^. utirrrc'/eiHd t-n to i<n.--vostiptawtotnCits"p^OcSSoiyillicth;yt sthta:xttusth. is te xt .may M copyr^ightecdu` T1 haeL r""c`c *^ 434 FDA PANLL ON CARCINOGCNCS15 of test animals required to achieve ideal protection, this number is likely to cw-i bounds of practicability, and some practical compromise is called for. Huo f,,.. special ease in which (he control incidence is zero, negative results on 100 (cm ,y.at a high dose level would do no more than establish with 95 % confidence that ii v- dcnce docs not exceed 3 in 100 at that dose level. With an incidence or 10 p, , animals m hoTfi controlFand lest animals one is 95% sure only that the elex.iu. incidence with test docs not exceed 10 per 100 test animals at that dose. Clearh v-1 and fast rule can be laid down. Positive evidence of carcinogenicity will be supplied when the lower nv.-- limit on the elevation in incidence (Pr - Pc) is above zero. Expression of eonfidcncc limits rather than as a test of significance is to'be preferred, since cu-n ,, Die lower confidence is below zero and no positive evidence exists, tire upper lun.i i j be well above zero and this will serve as a constant reminder that failure m n-\ j positive evidence of carcinogenicity is not the same as a positive demonstration ot h carcinogenicity. 3. For compounds judged carcinogenic at tcsHcvcls, a virBgdlyxafc dose o- / principle, bcjjstiniatcd bv downward extrapolation using styme-arbitrai ily sHlscJ > ' conservative doxc-rcsaflQSc curve. Mantel and Bryan (1961) suggest the use of ,u with a probil slope of 1 per J0-fold dilution on the grounds that all dowc-ii-;curses with carcinogens arc in fact steeper. But as they dearly recognize, use .g , , rather than probits would be even more conservative and has an equal jioiils < (Mantel, 1963; Druchrcy, 1967). A reasonable upper limit to carcinogenicity ,u use levels would be provided by use ofn straight line connecting the origin ami '6;; corresponding to the TD[. This is equivalent to saying that the expected proper, tumors at some fraction,/, of the TD, is/>; 0.01 arid would also follow front use >1 one-particle curve. Any dose for which this expected proportion is less Ilian some r'rarity low level, say 1CT8, would be judged safe. This would lead to few conflicts wr'-1 / results of applying the Delaney clause. Unless thejest level wcjrjiMcast to1, tor. / / proposed use level, ajjnding of_c:ireinoenicity_nt JhrAcstJcvel would jcaijjn vo t t . 54lc useTcvel as to amount to a hndmg o/jitonacceplnbilily. It is clear that the r j J tainlies involved in extrapolating downward from test to use levels are not restis I ' carcinogenesis and may apply to oilier toxic substances as well. 4. For agents not judged carcinogenic, such an estimation of safe dose won1! > logical, hpt would he so low as virtually to excludcfrom use agents for which liie;, , no positive evidence of carcinogenicity. For such agents an alternative is a sc.tlm;'.1 < Vv ward from the test level bv a more modest factor, say 100-fold. This is cqun.T ,;1 downward extrapolation using a probit curve with a slope of about 2 per infold It should be clear that no absolute guarantee of the safety of such a use level could N; i ' It would be desirable to compute loutincly the upper limit of possible elevation m tv dence at that use level by downward extrapolation, using !hc^nmc_onser,vejiy_:ceclurc recommended for agents found to be carcinogenic, j.c,, by multiply-1 '' upper limTTTo excess elevation at that level by/, the fraction of the use to the leu '< ; Thus, if 100 controls and 100 animals tested at 100 times the usejcvcl. cadi >iTF tumor-bearing animals, the upper 95% confidence limit to the elevation at tin11 level is 0.09S (Mantel, 1963, Table, p. 38) and at the use IcvJl 0.0098 (0.098 -a l<x*l. ' about 1 per 1000. j^.mvnvwlerHonr and Conrtus the committee recommend1 ... 0.limited for carcinogenic! .Mini nr.vtion to be developed .ancillary information aver the decisions should be cut I, At least two species arc I two rodent species be 'Imogens on reeord amour .Yciurds in consideration i ,1s is considered inadequa extend this tequiremen; r - .miivc limlmg. No other i ecus, to lie recommended ..wild have to be evaluated Uaudomly bred anim , ;,mis should he reserved ! t. Knowledge of the me ,!.v of comparative metalmv enable selection of ic ;u.iie, such information s' 1, It is essential that cor a siillicient to rely on untii.iK. Positive controls ` rmicailv analogous in ; .Mii.iminaiion of the test <> For animal studies. -oiled which should h c.'mcs `nch as addqiv gx 'uipoimd. It is essentia' 1 a To to determine the ] 7. Jesting should I>c numnn tumor indueN lest numbers ofar i'mI would not be accc: <>f future testing to ` as. recognizing that f a Although it is pc's ii-iiies involved in eh [ximissihle levels that 10. 1 he potentiatin' o.iy he an important ; if impractical to reec ' lion be conducted. ' /i i BOR 0ll<525 nation is drawn to the possibility that this ur^cd to investigate its copy^ght_status. text ; may bo copyrighted. t tt `I 1 ^ r ,. i i* l . f t* jr! IM' -i t ` < |.`V I II'V' ' ic I* - ,!,! * c ** i t * * -!l i ii i I in it- r kv .-M i 1 C XU v' CANCER TESTING 435 -mrmlannns and Conclusions -lommiuce recommends that the I-DA require that food additives and pesticides 'jjfcd For carcinogenicity. FDA scientists must exercise judgment as to the type nation to he developed. The requirement may vary with each ease, depending on .ilUrv information available. : decisions should he subject jo periodic review bv an independent advisory body. (TlcaM two speclesTire recommended for all carcinogenicity tests. It is suggested rodent species he used, 'there arc many examples of species variation to pens on record among rodents. The use of more than one species raids additional jrdsin consideration of extrapolation to man. The use of Ihc.d.oj'. for a two-year nconsidered inadequate/oijlie slticbiolrflcciitogcnicity. and it is felt unrealistic ,-.nd tins rcqiuremcnt routinely to the seven or more years needed to establish a _,c finding. No other nonrodents have been sufficiently investigated with card* -v to he recommended for routine tests at this time. Any new species or strains I h.ixc to be evaluated with a range of known carcinogens. K.uulomly bred animals arc teconimended for routine use. The use of inbred ^ dm;Id. be reserved for `pcci.il tests. r" knowledge of the ij'&idmlicjvMlway of a test substance and particularly know- ..(comparative metabolisms (quantitative as well as qualitative) in different species :r,,iHc selection of test species to be made on a logical basis, and, where appro* .-.such information should be developed his essential that concurrent negative controls always be included in the tests; it is wtlioent to rely on data of tumor incidence from prior studies with untreated mis. 1 p..'Mise controls arc desirable when it is known that the test compound is analogous to known carcinogens. Ihccautions should he taken to prevent -lunation of the test animals by the known carcinogens. f or animal studies, a scimo. nthctic diet or a_dicLo/JcJeo^'au;aJitpo.sition is rccom- :d which should he ns free as possible (iom incidental discrete chemical sob ers such as additives in individual ingredients, pesticides, and particularly the test v'und. It i> essentia! that, if fieedom from additives cannot be assured, analyses be to dctcimine the level of additives present. Ic'tii'.g should be done with several doses including one_jcyel likely_to yield a -.mi tumor incidence. 1 led mini Hers of animals currently used are capable only of determining incidences not be acceptable, if applied to human populations. It should he the objecTfuture testing to'movc as rapidly as possible to more nearly adequate test num iccognizing that the number of test animals required to nchicvcjdcal protection is j in cvcced the bounds of practicability. Mikouch it is possible in principle to estimate "safe" levels of a carcinogen, uncer-11 "cv involved in downward extrapolation from test levels will usually result in hi "iwible levels that arc the pracucaLequivalcnt of zero. b* ' Fhc potentiation ofcaTcinogenic activity by dietary cocarcinogcns or promoters h an important public health problem. Incomplete knowledge at this time makes .'Tracheal to recommend that routine tests for cocarcinogcniciiy or promoting '!> be conducted. Where circumstances suggest such a possibility, appropriate tests j i I BOR 014526 '-urncd to invcstirrato its cor -irrht status. T 436 FDA I'ANt.L ON CARCINOGENESIS t should be carried out. Meanwhile, research should he undertaken aimed at deterr - whether such tests arc needed and, if so, how they should be conducted. 11. The use or animals with a substantial tumor incidence of known viral m;..... example, Akr lymphoma) is not recommended for routine carcinogenicity vtu.i this time. A viral profile of the test animals should be obtained. ! 2. Oral administration should be used for all routine testing of food addm-^, . special cases, other routes of administration may also be needed. 13. Every .effort should be made to assure as "clean" an environment as po'Sp ; , the animals under study. M. Ideally carcinogenesis tests should be begun prior to conception and K ,, . tinned in the offspring. 15. Newborn animals should be reserved fer special tests. 16. At the present time there is not enough information available to provide j h for recommending any rapid test for carcinogenicity. Implementation of these recommendations will require additional funding frmn >. , industry and government for facilities and personnel considerably in excess id u currently available. REFERENCES Allison, J. ft., Wanni-mac-htr, R. W,, Hili, R., Micili'aresf, J. F,, and Crosshv, m i (1954). Diclary piotcin and tumor-host relationships in the rat. A Nntr. 54. 59), Allison, j, B., Wash, A. \V,, Ijathem, J. H,, and Wainio, W. (1950). Some die.n ( 2-acc(ylnminofluorenc on the dog. Cancer Re.r. 10, 266, Berwald, Y., and Sachs, L. (1963). In vitro transformation with chemical carcinogen* Y (London) 200, 1182-1 184. CoRNiiiLii, .1. (1954). Measurement and comparison of loxicilies: the quantal rciivi.' h Statistics am! Mathematics m Ihotopv. (O. Kempthome, T. A. Hancroft, J. W. (iimc- w, J. L. Lush, cd;.), pp. 327-344. Iowa, State College Press. Ames, Iowa. Corni ii lh, J., Hai pi.rin, M., and Moort, F. (1956). Some statistical aspects of s.ifd) ic.< of Salk poliomyelitis vaccine. Rah. Health, Rep. 71, 1045-1056. Dflla Pori a, G., Capivanu, J., Parsii, I.., and r. Colnaohi, I. (1967). Cancernro o urctano in itipi ncoiiuli, lattanti c adulti, del ceppi C57UL, C3H, BC3EI, C3IIf I sai Tumori 53, 81-102. Druclrcv. ft. (1967). Quantiiativc aspeeLs in chemical carcinogenesis. U.I.C.C. Marter-' No. 7, Potential Carcinogenic Hazards From Drugs, pp. 60-77. Drucekiy, H., Pri.ussmann, R., I vankovic, S., and Schmaiil, D (1967). Organoimi't * nogenc W/rkungcr) he/ 65 vcrschicdcncn A-NilrosoVcrbindungcn an DD-Hv Z. Krehsforach. 69, 103-201. Dhran-Ki ynals, F. (1952). Studies on the combined effect of fowl pox virus and iv` cholanthrcnc in chickens. Ann. N. Y. Acad. Set. 54, 977-991. Friliiman, L., Mickelsen, O,, Yang, M. G., Wugan, G. N., Ca.mpiif.ll, J. A., and Mo"' A. B. (1966). Nutrition Socielv Symposium, "Nutritional Significance of the Non-S'. Components of Food." Fed. Prac. Fed. Amcr. Sac. Exp. Bin. 25, 102-144. Gn hoin, H, V. (1967). Carcinogens, enzyme induction, and gene action. Adean. Cunrn t 10, 1-81. Halver, J. et a!. (1965). Afiatn\in.induced liver tumours in rainbow trout (Salmo g.nni < In: Mvcotoxins in Foodsndis (Wogan, G. N., ed.), pp. 209-234. MIT Press, Camt: Massachusetts. HEiDrLDLRGER, C., Iype, P. T., Roller, M.-R,, and Cur.N.T.T. (1968). Studies on IiuIuem*. carcinogenesis in organ and cell culture. In: Proliferation and Spread of Ncophai: t The University of Texas M.D. Anderson Hospital and Tumor Institute, at Hum' " pp. 137-156. Williams and Wilkins, Baltimore, Maryland. ,, ,,.V, 1., f.AOL'iLK, g. L. (i.h'ine-Mciulel rats bv a si ill 10. (VMI'ICII.S'., DULCEKf.Y, f-f .Mi Nachkommcn tizcJi cir .......... v 1IV p,, and Wisr LV. A. V, ( i,i nillticn/a virus and aero* l ,, v.vu.Nf, A., Ilrc-I In;, an jea-d by a weakly active cu ,.i,|vn. But. J. Cep. Pathol * uikmvn. M., and Kaiu an, .i.luccd lymphoid minors c ; .isvev, W., and Sai i phi i, ( Miviiy among hydi tigenate li.it Delimit'd. Clin S/ier, vIxmii, N. (1963). Tlic cone IUI 109. vl ,mi i. N., and Drv an, W. [ 27,455-470. vuihrm:, P. M. (I9r,s) c l/tintoxiny in Foods! utt\. j I. ii king, H. G. (1966). Fee' t H 324. ii vriaon, H. E , Si n r, A uni adrenal hislologiv respi .'0 ineltiylcliolanltueuc uei sciun ii. U-. and Shi hie. V iki mal carcinogenesis. A ir.l 969. v mm ni al, R. (1965) Toxic n ii sit, M. J. (1938). Studies Inii'i. J Cancer 33, 4`>9 .s S'iihein, M. U., Wi ism. r,,i i I1'/.6), Bioassay of 29 alk, J. "sat. Cancer law. 26, 9' 1 mi niiaum, A., and Sn 11 /.'n (.451. I*11 a \( ini, B., SminiE. P . a: mmivc uuh single upplie. < urn it Res. 20. I -1 s I > m: vcini, B,, Magi i , p ' dietary levels of dimethG bail, I!. (1963), Dcvclopim i/a'iiucalfy induced mou ban. R., and Shguie, P. ' 4 me i hy I ben? [u Jam hr, ban, R , and Siiciiin, P r l'||p>renc and dimcthylm ban, R.. Mac.ee, P, N , a iimnc administered orall (Mr. 24, 1712-1721. V i| > F'f i^4ii|jjA #iim UJ! p M** BOR 014527 i nqlxoh i l> ciinwii to Clio pobsitxii iy likeL i-uxs toxL mixy oo cojpyjrx^iitco, fj no i*c^oc 'urged to, investigate its cop- tght status. ? CANCER TESTING 437 t Laoufur. G. L., and Sratz, M. (I96S). Tumor induction in Fischer and -V.c-Mcndcl rats by a single administration of cycasin. J. Nat. Cancer Inst. 41, 1003-- akii. S., Druckrfv, H., PumssMANN, R. (19G6>. Erzeugung neurogener Tumoren bci \jJkommen nach einmaligcr Inicktion von AthylnitroshnLQsiotT an sehwungeren B 53, 410. u. A.. Strong, L. C., and G aronfr, W, U. (1940). Innucnccof mcthylcholanthrene .. mt-idenee of leukemia in several strains of mice. Proc. Sac. Fxp. Bin/. Med. 45, 287- .. p and 'Vrst ly, A. V. (1963). Production of lung cancer in mice by inhalation exposure .'tixa/a virus and aerosols of hydrocarbons. Proer. Fxp. Tumor Res. 3, 186-215. mi.-.f. A.. Huu-llot, and Ruoali, Y. (1945). Inhibition of the carcinogenic aclion pro-,t hva weakly active carcinogenic hydrocarbon on a highly active carcinogenic hydro- . n. Brit. J. I-\p Pathol. 26, 5-12. ,.,,Si M,, and Kaim.an. H. S. (1959). Lcukcmogcnic activity of filtrates from radiation-- ol hmphoid tumors of mice. Science 130, 387-358. W-. and Sai i ion i, U. (1965). Relationships between structure and skin tumorigcnic , t) among hydrogenated derivatives of several polycyclic aromatic hydrocarbons. Ann. 1 Dermatol. Clin. Spec. 19, 34-44. __ ^ a u. N. (1963). The concept of tbicshold in carcinogenesis. Clin. Pharmacol. Tlterap. 4,\ vl/ i IW. , i, N., and IIrvan, W. R. (1961). "Safely" testing of carcinogenic agents. J. Nat. Cancer 27. 455-470. t nsr, P. M. (1965). Carcinogenicity oT Aflatoxin-Contaminatcd Peanut Meals. In: ,/.iibcr in Poadstnlls. pp. I87-20S. MIT Press, Cambridge, Massachusetts. i'm,, H. Ci (1966). Feed and foods us sources of carcinogenic factors. Note. Rcr. 24, N. .iiivis, II I... Si it r, A. R., and Horsos-Naciit-m art., E. (1952). Liver tumor inhibition : jilienal histologic responses in rats to which 3-methyl-4-dimclhyhiminoa/.obcn/.cne and - .iSik/holantluene were simultaneously administered. Cancer lies. 12, 356-361. in, l'., and SiumiK, P. (1956). 3 lie effects of low concentrations of carcinogen in epi,il carcinogenesis. A comparison with promoting agents. J. Nat. Cancer lint. 16, 4 ' %>, m u. R. (1965). Toxicology oT natural products. Food Cos/net. Toxicol. 3, 609-620. .i. M J. (1938) Studies in carcinogenesis. V. Methyl derivatives of 1,2-hcnzanthraccnc. * r J. Cancel 33, 499-537. is, M. B., WusrsuRGt a, J. M., Wi'isrorger, E. K., Gurareit, N., and Sunizfff, V. iV M Dioassav of 29 alkylating chemicals by the pulmonary tumor response in strain Amice. ',,/ Cancer hot. 26, 915-935. spm'm. A., and Silvtrstone, H. (1953). Nutrition in relation to cancer. Advan. Cancer ' I 4SI. "i isi, n,, Shurik, P., and Della Porta, G. (I960). A Study of skin carcinogenesis in the * iiilh single applications of 9,10-dimcthyl-l,2-bcn7.anthraccnc at different dosages. . a Res. 20, 1538-1541. mi isi, B., Macij:, P. N., and Barnfs, J. M. (1967) Hepatic pathology in rats on low ' u? levels of dimcthylmtrosamine. Brit. J. Cancer 21, 559-565. II (1903). Development of malignant lymphomas by cell-free filtrates prepared from a 'is.illv induced mouse lymphoma. Proc. Soc. Fxp. Ihot. Med. 1 12, 875-S75. (1963). lt. and Shuiuk, P, (1963). Carcinogenesis in Lewis rats injected at birth with 7,12rtli)lbenz[ij]anthraccnc. Brit. J. Cancer 17, 540 545. ' R. and Shurik, P. (1967), Carcinogenesis in AKR mice injected at birth with benzoi .'Hcnc and dimcthylnitrosamine. Cancer Res. 27, 43-51. s. B, Magee, P. N., and SituniK:, P. (1964). Catemogenesis study with dimcthylnitros'nc administered orally to adult and subcutaneously to newborn BALB/c mice. Cancer M, 1712-1721. / f v 1 T l 1 > i i \ \ ir BOR 014528 t urged to. investigate its c-oyriRht status. -- *. v- % FDA PANEL ON CARCINOGENESIS Wogan, G. N,, and Newhirne, P, M. (1967). Dosc-rcsponsc characteristics or ail,,. carcinogenesis in lhe ral. Cancer Res. 27, 2370--2376. '' ' Wogan, G. N., and Newderne, P. M. (1968). Sequential morphologic changes in ,ifl,, , carcinogenesis in the rat. Cancer Res, 28, 770-781. 1 Vessei.inovitoi/, S. D., and Mihailovich, N. (1968). The development of neuroma plasms, embryonal kidney tumors, 1 lardcrian gland adenomas, Anitschkow cell mu,, the heart, and other neoplasms in urethan-treated newborn rats. Cancer Res. 28,8ss ,, v.,,.|IKiY AND Af't'LII D HU SIToxicity and Exert Ochratoxin Toxicity and I Oehratown A or 1 Monoii, C. A.. Taxicot. /\prl- -``I' pure oehrati**un < present in barley anorexia and hub of lho mtub bed . urine and feces, . ur kidneys. \ hi.iIiiam A ix a myci hi der Mcrwc el at. I `>1 Reports (Krogh , ,-,>ris id animals fed ei ik-so feeds is due to i nepliroio\ins citi ini -nations led us to i 'li pure oehratOMn A 'aulinin A-produein, .1 kidneys and its e,\c Hurley cnltuies B-! .. ia Walheek a a!., i1 ".moldy bariey ( \ V 'i,(toxin A mid ! 2 | a/. 1970) and oxalic ueliraloMii A For `i-.h.ited daily with ' irhonate and led N' Alii ad libitum or in l N.V11) T groups '.el oclirato.Mii A ,n "s) in 24-lir samples eer<55 ;45, v/v)extra -ti'i.inol (4 :1, \ \)cx; npounds. Routine I 'lunals nccropsicd on I lie average total a BOR 014529