QUANTITATIVE COMPARISON OF TOXICITY OF ANTICANCER AGENTS
`IN MOUSE, RAT, HAMSTER, DOG, MONKEY, AND MAN 1
`
`2
`
`•
`
`
`
`
`
`
`
`
`Emil J Freireich,3 Edmund A. Gehan,4 David P. Rall,5 Leon H. Schmidt,6 and Howard E. Skipper 7
`
`SUMMARY
`
`( 1)
`
`Toxicity data from small animals (mouse, rat, and hamster), large
`ani~als (dog and monkey), and humans were gathered, placed on a rea(cid:173)
`sonably similar basis, and compared quantitatively. Each animal species
`and all species combined were used tb predict the toxic doses in man (based
`on mg/m2 of surface area). Two models were assumed for the relationship
`between the maximum tolerated dose (MTD) in man and the approximate
`LDlO in each animal system:
`(dose in man.) == (dose in animal system i)
`and
`(dose in man) = Ai X (.dose in animal system i), ( i == 1 , . · .. , 6)
`(2)
`where A, is the fraction of the dose in animals used to predict the dose in
`humans (assumed different for each animal system, ie, i = 1 , ... , 6). It
`was found that when animal systems other than the rat were used the very
`simple model (1) was remarkably good for predicting the MTD in humans,
`though model (2) leads to slightly better predictions. Based on model (2),
`the animal systems are ranked in order of predictive ability: rhesus mon(cid:173)
`key, Swiss mouse, rat, BDF1 mouse, dog, and hamster. The best estimate
`of the MTD in man is made by weighting the estimates from the various
`animal species. Dose on an mg/m2 basis is approximately related to dose
`on an mg/kg basis by the formula
`(i = 1 , ... , 7)
`) = (km)i X (dose in mg/kg),
`(dose in mg/m2
`where (km)i is the appropriate factor for converting doses from mg/kg to
`mg/m2 surface area for each species. When the (km)i factors are known,
`equally good predictiQns of MTD in man can be made by either dose unit.
`On an mg/m2 basis, the M'TD in man is about the same as that in each
`animal species. On an mg/kg basis, the MTD in man is about Yi 2 the LDlO
`
`in mice, % the LDlO in hamsters, * the LDlO in rats, 1/3 the MTD in
`
`rhesus monkeys, and 1;2 the MTD in dogs. In each case the ratio is the
`(km) factor in the animal system to that in man. Hence relationships
`among the various animal species and man are somewhat simpler and
`more direct on an mg/m2 basis. These results support the conclusion that
`the experimental test systems used to evaluate the toxicities of potential
`anticancer drugs correl~te remarkably closely with the results in man.
`
`1 Received Dec 29, 1965; revised Jan 17, 1966.
`2 Study done under the auspices of the Acute Leu(cid:173)
`kemia Task Force of the National Cancer Institute by
`the Subhuman Subcommittee.
`3 M. D. Anderson Hospital, Houston, Tex.
`4 Biometry Branch, National Cancer Inst, Public
`Health Service, Bethesda, Md.
`
`5 Laboratory of Chemical Pharmacology, National
`Cancer Inst, Public Health Service, Bethesda, Md.
`Please address requests for reprints to Dr. Rall.
`5 National Center for Primate Biology, Univ of
`California at Davis.
`7 Kettering-Meyer Laboratory of Southern Research
`Inst, Birmingham, Ala.
`
`CANCER CHEMOTHERAPY REPORTS VOL. 50, NO. 4, MAY 1966
`
`219
`
`West-Ward Exhibit 1068
`Freireich 1966 - Page 001
`
`

`

`D
`wer
`nie11
`O .a·
`\VILS
`top<
`intE
`1don
`ttra .
`;
`
`f\ I mil
`l ~e
`
`per
`
`me
`1 ace
`
`\
`
`ith1
`
` !i
`
`\I
`
`icj
`I in
`~di
`lw
`I CE
`
`l~ 1 a:
`ii:
`0
`1v
`i\ ~
`le
`:,1· ~
`" (
`l. I
`ii:
`'~ (
`
`-.;.:
`
`No attempt was made to relate therapeutic
`doses in the various mammalian species. In the
`future this correlation should be attempted
`since the therapeutic target in the host is not
`the same as the toxicity target. However if an
`agent has therapeutic properties in an experi-
`. mental system, it is well to know the dose level
`for patients. Since there is some justification
`for using MTD's in cancer therapy, these dose
`levels were studied.
`·
`The plan of this retrospective study was to
`examine conside1·able toxicologic data obtained
`in (a) small animals, used in primary -screen(cid:173)
`ing and quantitative · secondary drug evalua(cid:173)
`tion; (b) larger animals, dogs and·monkeys, for
`the quantitative and qualitative aspects of tox(cid:173)
`icity at sublethal and lethal levels; and ( c) man,
`the target species. The goal was to determine
`what relationship exists, if any, between cer(cid:173)
`tain comroonly used toxicologic end points in
`the various animal species and man for a num(cid:173)
`ber of anticancer agents.
`Nothing in this report is intended to suggest
`or imply that short cuts are allowable in pre(cid:173)
`clinical or clinical toxicologic studies. Dose(cid:173)
`limi ting and serious toxic effects in man are not
`always apparent from even the most carefully
`done toxicologic investigations in animals ( 1).
`It is emphasized and should be clearly under(cid:173)
`stood that it is dangerous to attempt to ex(cid:173)
`trcipolate directly from anirna l toxfoity data.
`to ·mcixiniU?n tolerated doses in nia.n! New drugs
`can be introduced safely into clinical trial only
`through careful toxicologic and pharmacologic
`study in animals and then very cautious study
`in man, starting with much lower dosages
`than those which appear to be tolerated by the
`animals.
`
`APPROACHES AND ASSUMPTIONS IN THIS STUDY
`The published and unpublished data which
`form the basis for this analysis were obtained
`by numerous investigators using different pro(cid:173)
`tocols and end points. We used consistent and
`·reasonable general assumptions so that the data
`were comparable. The biologic end points, pro(cid:173)
`tocols, assumptions, and correctiOns necesssary
`to make the results more comparable ar~ de(cid:173)
`scribed briefly.
`Toxicologic End Points (See Appendix I)
`Mouse, rat, or hamster: Lethality-the dose which
`when administe:red by a certain route- and schedule
`killed a selected percentage (10%, ie, the LDlO) dur(cid:173)
`ing a specified observation period; 50 to more than
`100 animals were used in a typical determination.
`
`CANCER CHEMOTHERAPY REPORTS
`
`The biologic aspect of a drug development
`program
`to discover compounds effective
`against any clinical disease is generally an ex(cid:173)
`ercise in comparative pharamacology. In the
`typical program, compounds are screened in
`small animals against some easily produced and
`reproduced pathologic condition. A close rela(cid:173)
`tionship must exist between the screening sys(cid:173)
`tem and the ultimate clinical condition for the
`program to have the potential for success. Thus
`examination of this relationship is highly im(cid:173)
`portant. In cancer chemotherapy the similari(cid:173)
`ties and differences have often been considered
`among
`transplantable tumors, virus-induced
`tumors, carcinogen-induced tumors, and spon(cid:173)
`taneous tumors in animals, and between animal
`tumors and the various cancers and leukemias
`in man. However the similarities and differ(cid:173)
`ences between mice, rats, hamsters, dogs, mon(cid:173)
`keys, and man have been considered less often
`in terms of quantitative and qualitative aspects
`of the toxic effects of drugs. The consistency
`of the action of therapeutic agents among vari(cid:173)
`ous mammalian species is a keystone of most
`drug development programs, yet only rarely has
`this been studied in a quantitative manner.
`Classically comparative pharmacology and
`physiology have been concerned with differences
`which permit analytic studies of specific bio(cid:173)
`logic systems, and these studies have yielded
`valuable information. But it is equally impor(cid:173)
`tant to consider the much more frequent simi(cid:173)
`larities; we have tried to do this in the present
`analysis.
`Of all the toxicologic end points, lethal toxic(cid:173)
`ity is the easiest to measure with reasonable
`precision. Therefore we considered the lethal
`dose of certain cancer chemotherapeutic agents
`in various laboratory animals. For man the
`end point was the maximum tolerated dose
`(MTD). Hopefully two benefits might accrue
`from this evaluation: (1) If there is reasonable
`consistency in the reactions of various mam(cid:173)
`malian species, the toxicologic component of
`cancer chemotherapy screening will be shown
`to have a rational basis. (2) If such consistency
`is found, the problems of introducing highly
`toxic therapeutic agents into man might be
`approached more confidently. If major incon(cid:173)
`sistencies are discovered frequently, this would
`highlight the deficienCies in present screening
`systems and raise serious questions about the
`utility of these schemes for safe introduction of
`new drugs into man.
`
`220
`
`West-Ward Exhibit 1068
`Freireich 1966 - Page 002
`
`

`

`•
`
`unit. Only a simple transformation is required
`to change mg /kg to mg /m2
`; therefore the rela(cid:173)
`tionships developed are equivalent whichever
`unit is used. · The quantitative relationships
`were simpler when expressed in mg/m2
`A conversion factor (km) was used to trans(cid:173)
`form mg/kg to mg/m2 by the equation mg/
`kg X (km) == mg/m2
`; (~) factors for ani(cid:173)
`mals, given their weight, a1·e presented in table
`1 (Appendix II), and table 2 (Appendix II)
`pres.ents a way of transforming doses in mg/
`kg to mg/m2 for man, given height and body
`weight. Chart 1 (Appendix II) is a diagram
`·for determining surface area in man, given
`height and weight.
`Calculations based on units of body surf ace
`area have no intrinsic merit per se. Very likely
`some other basis such as surface area of the
`site of action of the drug, lean body mass, or
`some fractional power of body weight, possi(cid:173)
`bly related to length or some organ-membrane
`surface area, would be as appropriate or more
`appropriate. However the body surface area has
`been used to relate many physiologic param(cid:173)
`eters among species and means of transform(cid:173)
`ing the data are readily available. Further, in
`our clinical studies we routinely use body sur(cid:173)
`face area to adjust d1"ug dose for patients of
`different size and weight .
`RESULTS
`The first step in analyzing the data was to
`correct the daily dosage schedules for man and
`for animals, when necessary, to a uniform
`schedule of qd 1-5 days. Thus if an LDlO for
`mice, or lVITD for man, was obtained by a
`schedule of qd 1-10 days, we calculated that
`the LDlO (or MTD) for a schedule of qd 1-5
`days was twice that value. The next step was to
`convert doses (LDlO's or MTD's) from mg/kg
`• This was accomplished by the ap(cid:173)
`to mg/m2
`proximate formula
`(km), X (mg/kg), U=l, ... , 7)
`) =
`(mg/m2
`where the (km) i factor differs according to the
`species and also according to body weight with(cid:173)
`in each species. In the analysis an average
`(krn.)i factor was us~d, assuming that individ(cid:173)
`uals in each species were of average height-to(cid:173)
`body-weight ratios. The (km)i factors were
`derived from standard relationships b.etween
`weight and surface area as given in Spector
`( 40) and Sendroy and Cecchini (39). Details
`and other information on relating drug doses
`in mg/kg to doses in mg/m2 are given in ·
`Appendix II.
`
`221
`
`..
`
`poO' or monkey: {a) MTD; typically 2-4 animals
`.
`re eused at each dose level~ spaced by 2-fold incre-
`·lC
`le . "1:nts. In all instances tndividual doses which killed
`ld f;~ and lOO?o were used. The highest dose killing 0%
`·; i as considered the ~TD. ( b} Dose-related, hema-
`o_t F « oietic effects; localize? hemorr~ages .of th~ gastro(cid:173)
`lh t:~festinal tract; generahzed hemorrhagic lesions (ab(cid:173)
`'t dominal and thoracic viscera) ; stimulation of the cen-
`i
`: tral nervous system (CNS}; others.
`-
`el { Man: (a) 'MTD for a fixed schedule (dose causing
`>n
`1 miid to moderate sublethal toxic effects in a significant
`3e percent of patients) ; ( b) l\tITD for a variable sched-
`· 1 ule, ca:Iculated from. t~e daily d?se and median p~riod
`to toxic effects requiring cessation of drug; the Judg-
`ment of many clinical investigators was necessarily
`~
`... o , accepted in making this estimation.
`id I Because of the nature of the available data,
`1
`- "the toxicologic end points in the various ani(cid:173)
`1 mal species were related to the MTD in man.
`l-
`1r Although it was necessary to assume that the
`~-
`dosages resulted in the. same percentage of tox-
`.1 icity in each species, the results do not depend,
`'
`! in a major way, on this assumption. For the
`ie
`r- ~drugs in this study, the dose-toxicity curves
`.n
`1 were relatively steep E?O that if the true per-
`
`1- I centage of toxicity for a given dosage was,
`
`1'
`
`say, between 5% and 15%, the actual dosage
`;t , used would not differ very much from the dos(cid:173)
`age that should have been used.
`~-
`It was necessary to use toxicologic data ob(cid:173)
`~-
`•t
`tained by various routes of drug administra-
`tion, ie, intraperitoneal (ip) for small animals,
`Y
`··oral for small animals and man, and intra(cid:173)
`~.
`.. _ venous (iv) for large animals and man. In mice
`and rats the LDlO's obtained by the ip and iv
`;-
`a ~ routes are usually comparable.
`:s 1 Another variable for which some reasonable
`correction must be made is the dosage schedule
`Y
`.'including the total dose. 1vVe assumed that the
`c
`~~
` toxicity of anticancer agents is cumulative.
`Griswold et al. (3) reported that when the
`1 LDlO's in BDF1 mice cif 70 agents, including
`(the major classes of anticancer agents, were
`J compared for two schedules, qd 1-7 days and
`1 qd 1-11 days,8 the mean ratio (qd 1-7 days/
`
`e
`
`h I qd 1-11. days) was 1.56. This is very close to
`
`that which might be expected from direct cumu(cid:173)
`d
`lative drug toxicity (11 days/7 days = 1.57).
`,_
`d J Pinkel (2) and other investigators pointed
`out that the usual doses of certain drugs in
`a
`•- ~ various animal species and man were compara(cid:173)
`V I ble when the dose was measured on the basis
`: of mg/m2 of surface area. Consequently most
`:-
`I of the results are presented in mg/m2
`• However
`' since mg/kg is a commonly used unit of drug
`dosage, so~e results are also pres.ented in this
`
`h
`e
`
`!l
`
`s
`
`8 qd = drug given once daily for as many days as
`indicated.
`·
`
`VOL. 50, NO. 4, MAY 1966
`
`West-Ward Exhibit 1068
`Freireich 1966 - Page 003
`
`

`

`CHART 2
`
`0
`
`0
`
`CHART 3
`
`O o
`
`Comparison of toxicity data on anticancer
`agents 'tor the hamster and man (on a
`MG/M2 bnsis l
`
`O
`
`o Antimetabolites
`6 Alkyloting agents
`e Others
`
`0
`
`A
`~
`
`1 ~
`l ~
`t ~ ..
`N~::~
`
`\:(\.
`
`_
`
`...
`
`(!
`
`l~ :
`
`.1
`I
`. I
`~·
`
`~
`
`Cl
`LU
`
`L&J
`
`~ a:
`~
`1-
`::E
`:::>
`;2§
`
`I .0
`
`<(
`
`\
`i
`<:
`z
`~~·
`o4
`~ 0.1 ----~....__ ___ _._ ___ ____. __ _ ~ ,
`10.0
`0.1
`1.0
`100
`1000
`.
`HAMSTER: LDIO (MG/M 2:oo 1-5 Doy schedule)
`
`animal systems. We wish to describe the rela(cid:173)
`tionship between the dose-toxicity curve for
`man and that for each of the animal systems.
`Two models are considered:
`(dose in animal system. i)
`(dose in man)
`(i = 1, ... '6)
`(1)
`
`and
`(dose in man) = A, X (.dose in animal sys-
`tem i),
`\'.
`(2)
`(i =.1, ... , 6).
`Model
`( 1)
`is a si)ecial case of model
`( 2) i ·.:..
`since they are the same when Ai = 1. Model ·, I
`
`! 0
`
`q; IOOO
`
`:
`
`
`Nll">o~
`~ * IO.O
`
`I 00
`
`......
`
`::E
`~
`
`Q
`
`The basic data used in this study are given
`in table 1. Doses of 18 drugs 9 are presented
`in mg/kg and mg/m2 for the 6 species, along
`with source information and other pertinent
`data. An average dose (LDlO or MTD) of each
`drug was calculated from the multiple studies,
`if done, on each species. The average doses for
`the 6 animal systems and man are given in
`mg/kg in table 2, and in mg/m2 in table 3.
`Charts 1-6 indicate the closeness of the rela(cid:173)
`tionship between the logarithm of the LDlO, or
`MTD, in the various animal systems and in man
`when the dose is measured in mg/m2
`• Chart 7
`indicates the close· relationship between 12
`times the LDlO in the BDF1 mouse and the
`MTD in man when the dose is measured in
`mg/kg. The ratio of the (km) factors for an
`average man and a mouse is 37 /3 = 12.3. It
`will be shown later that relationships between
`systems on an mg /kg basis are the same as
`those on an mg/m2 basis if the ratio of (km)
`factors is considered.
`To examine further the relationship of dos(cid:173)
`, between the animal systems and
`age, in mg/m2
`.man, consider the following: For each animal
`system and man, there is a dose-toxicity curve.
`The basic data for each drug consist of esti(cid:173)
`mates of a single point, the approximate LDlO,
`on the dose-toxicity curves for man and the 6
`
`0
`
`CHART I
`
`Comparison of ioxir:ity data on anticancer
`agents for the Swiss mouse and mon
`(on o MG/M 2 basis)
`o Antimetobolites
`6 Alkyloting agents
`o Others
`
`Cl
`0
`
`10
`
`1.0
`
`0.l ..__ ___ .....__ ___ __._ ___ - . J ' - - - - - - - - '
`0.1
`l.O
`100
`10
`1000
`SWISS MOUSE: ~010 !MG/M 2: QO 1-5 Doy schedule)
`
`9 Chemical Abstracts' nomenclature and NSC num(cid:173)
`bers for the agents are given on page 243.
`
`2.22
`
`West-Ward Exhibit 1068
`Freireich 1966 - Page 004
`
`

`

`CHART 6
`
`Comparison of tollicity data on anticancer agents for
`the dog and man (on a MG /M 2 basis)
`O Anti!ilelobolites
`£::. Alkylating agents
`O Others
`
`0
`
`Qj 1000
`=:;
`-0
`Q)
`..&::.
`
`<.> "' ,.,
`
`100
`
`0
`Cl
`U"')
`I
`c
`0 ..
`
`N
`
`::E -t.:>
`
`E!
`
`L.&J
`(,/')
`0
`Cl
`
`10
`
`0
`
`0
`
`0
`
`CP.ART 4
`
`Comparison of toxicity data on anticancer
`agents for the rot and mon (on a MG/
`b.
`M2basis)
`o Antimetoboiiles
`!::::. Alky\ating agents
`e Others
`
`0
`
`1.0
`10
`100
`RAT: LD 10 lMG/ M2:QO 1-5 Day schedule)
`
`1000
`
`CHART 5
`
`Comparison of toxicity data on anticancer
`agents for the rhesus monkey and man
`(on a MG/M 2 basis)
`
`0
`
`O Antimetabolites
`6. Al kylating agents
`e Others
`
`/
`
`/
`
`~
`j
`
`ISi
`
`I~ 1000
`Cl> s: u
`"" 0
`0
`Lt">
`~
`0
`0
`N
`~ ......
`~ w
`
`100
`
`~
`
`(/)
`0
`0
`
`10
`
`~
`
`0
`LJJ
`I-
`<!
`a::
`LJJ
`_J
`0
`I-
`
`:::!:
`:::>
`
`::1: x
`<t
`:::;.
`:z
`<(
`:::;.
`
`J'
`·'
`)QQ·
`
`.a-1
`or
`lS.
`
`i) .
`L)
`
`"S-
`~)
`~)·
`el
`I
`I
`r~
`I
`
`0
`L.&J
`~
`r:r.
`L.&J
`....J
`0
`t-
`:::;.
`:::>
`~ x
`~ z
`
`<t'.
`
`<t'.
`::E
`
`a;-
`=:;
`"O
`Q)
`..&::.
`~
`>-
`0
`Cl
`\{)
`
`..!...
`0
`0
`;.::,
`
`~ -(.!:)
`
`~
`\..l.J
`V'l
`0
`Cl
`
`1.0
`
`0.1
`
`.....__ ___ _,_ ____ ....._ ____ .....__ ___ ~
`
`IOC
`1.0
`1000
`10
`0.1
`DOG: MAXIMUM TOLERATED DOSE {MGJM2:QD 1-5 Doy schedule)
`
`CHART 7
`Comparison of toxicity data on anticancer agents for
`the mouse and man (on a MG/KG basis)
`
`The 12:1 relationship showh on a MG/KG basis is equivalent
`to the \:1 relationship shown ona MG/M 2 basis lchart 2)
`The approximate 12:1 relationship (mouse:man)is in
`agreement with the ratio of the KM factors used
`for these species: i.e.,37:3 {man=mouse)=ca.12
`o Antimetobolites
`6. Al ky lo ting agents
`e Others
`
`0
`
`100
`
`10
`
`1.0
`
`0.1
`
`0.01 ~---__.._.....__ ___ __._ ___ __...._ ___ ~
`0.01
`0.1
`1.0
`10
`100
`BDF1 MOUSE: LD10 lMG/KG:QO l-5Day schedule)
`
`Cl w
`t-
`<(
`a:
`1.1..J
`....J
`0
`I-
`::!:
`:::>
`
`:!1 x
`<t'.
`::::;:
`:2:
`<t'.
`::1:
`
`10
`
`100
`RHESUS MONKE. Y:
`MAXIMUM TOLE.RATED DOSE (MG/M 2~QD 1-5 Doy schedule)
`
`1000
`
`(1) assumes that the dose in each animal sys(cid:173)
`tem gives a direct prediction of the dose in
`man. Model (2) assumes that the dose in man
`is a fraction (Ai) of the dose in the animal
`system and the fraction remains constant for
`the sample of drugs.
`A third model was considered:
`(dose in man) = A-i X (dose in animal sys-
`(~ = 1, ... ,6)
`tem i) Bi,
`where B, is the power to which the dose is
`
`VOL. 50, NO. 4, MAY 1966
`
`raised, assumed to be 1 in models ( 1) and (2).
`This model is a natural. generalization of (2).
`However, since the estimates of Bi were near
`1 for all animal systems, in fact within 1
`standard error (SE) limit, there is no advantage
`to using a more general model than (2).
`By these models, we wish to predict the dose
`in man from the dose in each animal system
`when both determinations are subject to samp(cid:173)
`ling variation (and other assumptions as men-
`
`223
`
`West-Ward Exhibit 1068
`Freireich 1966 - Page 005
`
`

`

`l
`CHART 8
`\l /:~
`Observed ond predicted dosoges(MG/Mi) in mo~
`using all animal systems (weighted estimates)
`A /! 000
`
`...
`~ 0
`
`·
`
`~
`
`~(~
`
`o Model I
`G> Model 2
`
`IOOO
`
`10
`
`0.1..___ ___ ,___ ___ ,___ __ ___. ___ __,
`10
`0.1
`1.0
`100
`1000
`MAN,OBSERVED DOSAGE (MGIM 2
`J
`
`0
`0
`Cl
`
`~
`~
`g:
`z
`~ 1.o
`
`:.to~d
`'.srsi
`:tab1
`uar
`ltlO"
`i lite
`1 ag(
`the
`Sv;
`do:
`I th:
`I de
`I 4.
`
`ll ~
`
`dE
`
`/:
`\ ~
`
`tioried) in the sample of drugs. The statistical
`considerations in fitting these models are given
`in Appendix III.
`Model ( 1) is the simplest possible model ; no
`parameters need to be estimated~ . Thus the
`doses in table 3 for each animal system are the ~ 100
`predicted values of the dose in man and charts
`;:;
`· ~
`1-6 indicate that these predictions are reason-··•
`~
`ably good. T.he standard. deviations, on a log
`scale, of a predicted value of log (dose in man)
`were calculated for each animal system. The
`systems are ranked in order of pred1cti ve
`ability in the top half of table 4: monkey, Swiss
`mice, BDF1 mouse, dog, rat, and hamster. A
`predicted value of the dose ·in man has been
`calculated by weighting the estimates from
`each animal system (see Appendix III) and
`the results are given in the last column of
`table 3. The standard deviation of a predicted
`value of Jog (dose in man) is 0.299, with multi-
`pliers of 0.50 and 2.0 for lower and upper
`standard deviation limits respectively. Thus the
`weighted estimate based on all systems is bet(cid:173)
`ter than the estimate from any single system.
`Assuming model (2), the estimates of Ai and
`A, + 2 SE are given in the bottom half of table
`4. Note that the approximate 95% confidence
`limits for the multiplying factor, Ai, include 1
`for .all animals systems except the rat. Thus for
`the other animal systems it is reasonable to
`accept the very simple model (1) as providing
`an adequate prediction of the dose in man.
`However when all systems are combined to ob(cid:173)
`tain an ·overall estimate of Ai (see Appendix
`III), the approximate 95% confidence limits do
`not include 1. Also, note from the bottom half
`of table 4 that the standard deviation of a pre-
`dicted value of log (dose in man) is 0.275, al-
`most a 10% reduction from that of model (1).
`Therefore model (2) is pref erred for fitting
`these data; however for future studies in ·which
`more precise estimates of LDlO are available,
`it may be that model ( 1) will be adequate.
`Using model ( 2), we can rank the animal
`systems in order of their predictive ability by
`considering the deviations of observed from
`predicted values of dose in man. These standard
`deviations are given in table 4. Thus the order
`is monkey, Swiss mouse, rat, BDF 1 mouse, dog,
`and hamster. The best predictions with model
`( 2) are obtained by weighting the estimates of
`the dose in man from all 6 animal systems (the
`method is explained in Appendix III). The pre(cid:173)
`dictions for the drugs in this study are given
`
`l::
`in table 5 and the weighted estimates based on
`all animal systems combined are plotted in I
`i"
`chart 8. The best estimates of dose in man, as I r
`~. 1
`indicated by the standard deviations in table 4,
`are given by weighting the individual estimates ~ l
`~~
`from each animal system.
`Another model was considered in which the
`) was related to doses in
`dose in man (mg/m2
`the animal species in a single equation:
`log (dose in man) == 0.284 + 0.847 log (dose in
`Swiss mouse)
`-1.064 log (dose in BDF1
`mouse)
`+ 0.539 log (dose in rat)
`+ 0.801 log (dose in mon(cid:173)
`key)
`- 0.175 log (dose in dog).
`
`This predicting equation leads to a slight im(cid:173)
`provement in the prediction of the dose in man;
`the deviations of observed from predicted dos(cid:173)
`ages were less (standard deviation of 0.249 on
`log scale compared to 0.275 by using weighted,
`. combined estimates). However a prediction of
`dosage in man cannot be made unless estimates
`of LDlO are available from all the animal sys(cid:173)
`tems mentioned; also the model does not provide
`any real insight into the relationship between
`the dos.e-toxicity curve in each animal system
`and that in man.
`·
`From considering charts 1-6, this question
`arose: Do the differences between the dose-
`
`224
`
`CANCER CHEMOTHERAPY REPORTS
`
`l\,,'
`
`i
`
`~· .. F.'
`
`J l\ H ll i ~ -~
`·~'. 1.r ·-•• ~.,·
`
`'\' 'i\·
`~4
`;1((.
`t{
`
`l !,
`
`West-Ward Exhibit 1068
`Freireich 1966 - Page 006
`
`

`

`·
`
`toxicity curves for man and for each animal
`system differ depending on whether an antime-(cid:173)
`tabolite or an alkylating agent was given? Us(cid:173)
`ually the animal species, except the rat and
`monkey, underpredict the doses of antimetabo-
`/ lites and ov~rpredict the doses of alkylating
`l agents for man. By a statistical test ( t test),
`~ there was some suggestion (P<0.10) that in
`I Swiss mice and BDF1 mice the predictions of
`·1· dosage in man were lower for antimetabolites
`than for alkylating agents. There was no evi(cid:173)
`l dence of a difference in the other species. Only
`4 antimetabolites and 8 alkylating agents were
`tested in all animal species. Consequently fur-
`1 ther study is needed to determine whether the
`difference between dose-toxicity curves really
`depends on the type of agent.
`,o 1 There is some value in comparing the rela-
`I tionships found on an mg/m2 basis with what
`~ would have been found on an mg/kg basis.
`n I Some indication of this has already been given
`n I in chart 7 which shows that there is a close
`s
`relationship between 12 times the LDlO in the
`~ BD Fi mouse and the MTD in man. Since the
`1,
`ii relationship between mg/kg and mg/m2 used is
`~s
`e 1 (mg/m2
`) = (km)i X (mg/kg), (i = 1, ... , 7),
`models (1) and (2~ become, in terms of mg/kg,
`n
`.
`(km)a
`(dose m man) = (km)n.
`
`~
`
`n
`,
`
`1
`
`and
`
`X (dose in animal system)
`
`(1)
`
`(km)a
`.
`(dose m man) = (km)m Ai
`x (dose in animal system)
`(2)
`where (km)a and (km)m refer to the (km) fac(cid:173)
`tor in the particular animal system and man
`respectively, and A. is exactly the same as
`stated before. Hence it should be clear that dose
`in man can be predicted equally well either on
`an mg/kg basis or on an mg/m2 basis. Thus by ._
`using the km factors and model ( 1) , the dose
`in man (mg/kg) i-s approximately Yi 2 the dose
`in mice, % the dose in hamsters, % the dose in
`rats, Ya the dose in rhesus monkeys, and 112 the
`dose in dogs.
`
`1
`
`' f
`
`DISCUSSION
`
`1
`
`Originality is not claimed or . implied for
`this analysis. We have confirmed and extended
`the general observations and conclusions of
`
`VOL. 50, NO. 4, MAY 1966
`
`Finkel (2) who confirmed and extended specific
`aspects of the basic observation of Rubner
`(36), made 80 years ago, and many other inves(cid:173)
`tigators later·.
`The availability of much more extensive
`toxicity data from the Cancer Chemotherapy
`National Service Center program, from certain
`other published sources, and from our own labo(cid:173)
`ratories seemed to make this present analysis
`timely. Also we believe it is important to use
`more definitive biologic end points .of toxicity.
`This analysis and study of data on toxicity to
`animals and humans of several types of anti(cid:173)
`cancer agents (tables 1, 3, and 5) lead us to
`conclude that the toxic do~e of an agent is
`similar among species when the dose is meas-.
`ured on the basis of surface area. The skin sur(cid:173)
`f ace area was used here though it is unlikely
`that the skin is the target area of action of any
`particular drug. More likely the skin surface is
`more or less proportional to the true target
`surface.
`To the extent that mammalian species are
`broadly similar and have corresponding organs
`and tissues, it is true that any surface area will
`increase approximately with the two-thirds
`power of weight (38). Thus the two-thirds
`power of body weight would have been a con(cid:173)
`venient unit of surface area to use and the re(cid:173)
`sults of the analysis would have been almost
`the same (see Appendix II).
`Pinkel (2) suggested that "cancer chemo(cid:173)
`therapists consider the applicability of body
`surface area as a criterion of drug dosages in
`their laboratory and clinical studies." We sug(cid:173)
`gest that a unit proportional to body surface
`area is sufficient and an appropriate unit is
`(weight)~.
`vVe have been concerned only with compari(cid:173)
`sons among -species, not within species, and
`with adult animals, not immature and adult
`animals. Also we have been concerned solely
`with anticancer drugs.
`Some of the toxicologic data tabulated may
`disagree \Vith unpublished and published obser(cid:173)
`vations of some experimentalists and clinicians.
`The Acute Leukemia Task Force of the Na(cid:173)
`tional Cancer Institute wishes to correct, up(cid:173)
`date, and extend this analysis at some future
`time. Those interested in seeing such correla(cid:173)
`tion efforts extended can help by providing ad-
`
`225
`
`West-Ward Exhibit 1068
`Freireich 1966 - Page 007
`
`

`

`ditional data, both clinical and experimental, in
`a form similar to that in table 1.
`The present study has emphasized the quan(cid:173)
`titative aspects of toxicity of anticancer drugs
`to animals and man. Regarding the ·prediction
`of the qualitative effects of anticancer drugs
`in man from laboratory animal studies, Owens
`( 1) suggested:
`Predictive value
`Good
`
`Questionable
`
`None
`
`Preclinical toxicity studies
`Bone marrow, gastrointestinal tract,
`liver, kidney
`Nervous system, including periph-
`eral neuropathy, extraocular pal(cid:173)
`sies, and CNS toxicity
`Skin and appendages, including skin
`rashes, dermatitis, and alopecia
`
`Of the 18 agents in this study, 17· produced
`limiting toxicity to the bone marrow (marrow
`depression: MD) and to the gastrointestinal
`(GI) tract. If the mg/m2 doses in man that are
`predicted by using the weighted combined esti(cid:173)
`mate are compared to the observed doses, then
`the largest ratio of predicted dose/ observed
`dose is 3, for thioTEP A. Consequently it would
`be reasonable to study preclinical toxic effects
`in the mouse, rat, dog, monkey, and hamster,
`to estimate the MTD (mg/m2
`in man, and to
`start clinical cancer chemotherapy .trials at
`a.bout one-third
`the predicted dose. This
`would have been a safe procedure for all 18
`drugs mentioned. Owens ( 1) suggested tltat it
`might be reasonable "to begin a human trial at
`one-tenth of the maximum tolerated dose in the
`most susceptible animal" (on an mg/kg basis).
`Since the most susceptible animal will ordi(cid:173)
`narily be the dog or rhesus monkey, Owens'
`rule of thumb on an mg/m2 basis becomes:
`begin trial in man at about one-third the dose
`for monkeys or one-fifth the dose for dogs.
`Thus there is reasonable agreement between
`the two recommendations. However if the ani-
`
`)
`
`/' l
`, .::·\
`. basis . ?:
`mal data are not placed on 'the ·mg/m2
`before using Owens' rule of thumb, any addi- ~·<
`tional know ledge which the small animals :.·
`',
`(mouse and rat) might contribute will be over-
`looked. Remember also that the toxicity values ,
`(LDlO's) for such small animals are often more
`reliable statistically because more animals are
`generally used.
`. (
`
`The ratios of animal/human toxicity (mg/m2 l
`
`basis) for the mouse, hamster, dog, a~d mon~
`key are remarkably dose to unity. Thus each
`species generally predicts for man. That this is
`true for the mouse is particularly pertinent to
`cancer chemotherapy. Extensive drug develop(cid:173)
`ment pro.grams which use mouse tumors seem
`to be on firmer ground than we had previously
`thought. In general the rat is more susceptible
`to these agents than the other species. The
`
`hamster is unusually resistant to amethopterin I
`
`and sensitive to the .fluorinated pyrimidines.
`The dog and monkey, long known to be reason-,
`ably good predictors of toxicity to humans, 1
`t.
`have shown up well in this analysis.
`We are nqt suggesting that it is wise to take ~
`mouse or rat LDlO's, convert the doses to I
`, and then start clinical trials at one- t
`mg/m2
`third this level (in mg/m2 for man). The addi- I
`tional safety provided by toxicity data from \
`multiple species is well established, as is the
`value of specific quaUtative knowledge on dose(cid:173)
`related sublethal toxicity and its reversibility. ~
`Finally it is suggested that the quantitative \
`relationships between toxicity to animals and
`to humans are simpler when compared on an
`mg/m2 basis than on an mg/kg basis. Broader
`use of a surface area unit, either mg/m2 or
`(weight) 2f.J, by experimental and clinical cancer
`chemotherapists, as well as biochemists and
`pharmacologists concerned with mechanism
`studies, might prove helpful in ma