[CANCER RESEARCH 39, 2204-2210, June 1979] 0008-5472/79/0039-0000$02.00 Adequacies and Inadequacies in Assessing Murine Toxicity Data with Antineoplastic Agents Anthony M. Guarino,1Marcel Rozencweig,Ira Kline, John S. Penta, John M. Venditti, Harris H. Lloyd, Donald A. Holzworth, and Franco M. Muggia Division of Cancer Treatment, National Cancer Institute, NIH, Bethesda, Maryland 20205 [A. M. G., M. R., I. K., J. S. P., J. M. V., F. M. M]; Southern Research Institute, Birmingham, Alabama 35205 (H. H. L.]; Toxicology Program Office, Battelle Columbus Laboratories, Vienna, Virginia 22 180 (0. A. H.] ABSTRACT regardless of the number of treatments. Finally, implicit and/or explicit assumptions were made regarding the reproducibility Previous retrospective analyses have suggested a very pos of these experimental results. itive correlation in toxic doses of antineoplastic agents between This retrospective study takes advantage of the accumulation mice and humans. Additional toxicological information has now of additional data to further assess problems associated with been accumulated and reveals a noticeable variability in the toxicological results in mice and their potential for use in dose existing data base. Nevertheless, it is likely that mouse toxi prediction in humans. cological studies will become a principal determinant for esti mating initial doses to be used in humans. Recognition of the factors responsible for differences in determinations of toxic MATERIALS AND METHODS dose levels in mice will enhance the proper utilization of this approach. The cytotoxic antitumor agents selected for analysis were confined to those for which clinical toxicological data were available for i.v. administration by at least one of the following INTRODUCTION schedules: high intermittent doses, weekly doses, and daily Antitumon agents are commonly administered in humans at administration for 5 or 7 days. These limitations were necessary or near their MTD.2 When a new agent is first introduced into because other routes of administration, particularly the p.o. clinical trials, its MTD is reached by progressive increments of route, may introduce further bias in correlation studies; the an initial dose that has been derived from toxicological data schedules chosen are those most frequently investigated in obtained in animals (4, 5, 9, 24). Significant interspecies dif humans. ferences in the rate of drug metabolism and excretion make it Toxicological studies in mice were performed under contract difficult to accurately extrapolate to humans the iatrogenic to the Laboratory of Toxicology and Drug Evaluation Branch of effects observed in animals (1, 6, 7). Nevertheless, based on the Division of Cancer Treatment in the National Cancer Insti the clinical experience with 37 drugs, Homan (20) has esti tute, according to previously described methods (13, 25). mated that an initial dose of one-third the MTD (in mg/sq m), Lethality was chosen as a nonspecific but easily measurable as determined in the most sensitive large animal species (bea toxicological end point. The lethal dose categories are LD10, gle dog or rhesus monkey), would be tolerated in humans in LD50,and LDso;in this study, this period ranged from 14 to 60 about 94% of the cases. Whether such a determination con days after the last dose administration, but more than 90% of stitutes the most efficient method for rapidly and safely reach all deaths actually occurred within 14 days. ing the effective dose in humans or whether rodent data may Statistical analyses were performed to assess true differ be used to advantage has been the subject of recent analysis ences between the related parameters of animal strain, injec (14). tion site, and drug-dissolving or suspending vehicles. Probit The value of combined animal toxicological data, including analyses and tests for parallelism of response were performed findings in rodents, for predictive purposes in humans has been wherever possible (3, 10, 11, 16). Slopes of log probit lines repeatedly advocated (12, 14, 26). This approach has been should be parallel if: (a) there is no statistically significant hampered by the availability of only limited rodent data not difference between the data (11); and (b) the cause of death expressly obtained for toxicological purposes. In fact, previous (i.e. , mechanism of toxicity) in the 2 treatment groups is the correlations had to be based on various assumptions. Specifi same, that is, both groups show the same ‘‘specificaction― cally, assumptions were made regarding the route of adminis (8). tration, considering i.p. and i.v. injections equivalent and ne glecting the effect of schedule. The toxicity pen course was RESULTS considered to be a function of the total dose administered, Clinical toxicity data were available for 58 drugs that have I To whom requests for reprints should be addressed, at Building 37, Room been administered i.v. in humans by at least one of the 3 5B22, Laboratory of Toxicology, National Cancer Institute, NIH, Bethesda, Md. 20205. schedules of interest (Table 1). The daily for 5 or 7 days 2 The abbreviations used are: MTD, maximum tolerated dose; LD10, drug dose schedule has been most widely used, for 49 (84%) of the that kills 10% of the non-tumor-bearing animals during the observation period; LD@,dose that kills 50% of the non-tumor-bearing animals during the observation drugs, whereas high-dose intermittent administration has been period; LD@,dose that kills 90% of the non-tumor-bearing animals during the studied for only 41 % of the drugs. observation period; 6-MP, 6-mercaptopurine; ara-C, 1-f@-n-arabinofuranosylcy Of the 3 categories of lethal dose, the mouse LD50,estimated tosine; BCNU, 1,3-bis(2-chioroethyi)-1-nitrosourea; FU, 5-fluorouracil; TIC mus tard, lmidazole-4-carboxamide, 5-(3,3-bis(2-chloroethyl)-1 -triazeno). by the probit model, is statistically more reliable than is the Received November 27, 1978; accepted March 2, 1979. LD10or LDso (11, 15) and was most often available for our 2204 CANCER RESEARCH VOL. 39 Downloaded from cancerres.aacrjournals.org on September 28, 2021. © 1979 American Association for Cancer Research. Drug Toxicity in Rodents and Humans Table1 for FU. The slopes of the probit response lines were parallel for Availability of mouse lethality data for 58 drugs used i. v. in humans all of the drugs except daunomycin, where the LD50and LD10 on the most common schedules differed by 52 and 275%, respectively. — No. of drugs for which mouse Is The variation in the total equitoxic dose per course could be thecorrespondingthality data are available for schedule in humans studied for 6 drugs, each given on 1 equitoxic dose on Day 1 No. of drugs and 1 equitoxic dose on Days 1 to 5 schedules in a single LD,0Schedule given iv. in hu- LD@+ mouse strain using one vehicle and injection route. The L050or LD@Single mans LD@ and/or LD10is shown for each drug on both schedules in Table 5. The 13(54)Weeklydose 24 20 (83)a differences in total equitoxic dose per course apparently vary (6)Daily 34 11 (32) 2 (33)afor 5 (or 7) days 49 23 (47) 16 greatly according to the drug in question. Thus, the difference dataareNumbers in parentheses, number of drugs for which mouse lethality was modest for BCNU (7%), intermediate for ama-C(23%) and available for the corresponding schedule in humans. isophosphamide (28%), large for cis-diamminedichlono platinum (77%) and 6-MP (150%), and extreme for 2'-deoxy thioguanosine (1440%). study. LD50's were known for 83, 32, and 47% of the drugs Toxicity studies have been repeated, usually at the same tested at the single, weekly, and daily schedules, respectively laboratory, with 16 drugs under the same conditions with (Table 1). Knowledge of other lethal dose levels does provide regard to the animal species and strain, route of administration, additional information of obvious importance in the analysis of vehicle, and schedule. The data for the vast majority of these toxicological data. However, fully usable lethality data (LD50 agents show striking discrepancies (Table 6). The average plus LD10or LOso)for the 3 schedules were only available for difference between each pair of lethality studies at LOsowas 54, 6, and 33% of the drugs. 42%. The LD50was identical for 3 drugs (thiotepa, pseudourea, Variations in the lethal dose levels according to the mouse and BCNU) and differed by 20% or less for azasenine, actino strain relate to data obtained largely in Swiss and C57BL x mycin 0, cyclophosphamide, porfinomycin, and daunomycmn. DBA/2 F1 (hereafter called BD2F1) mice (Table 2). Occasion Differences between 21 and 50% were noted in the LD50of ally, results were available for other strains, i.e. , AKA, BALB/ methotrexate, mitomycin C, ama-C,and TIC mustard. The LD50 c X DBA/2 F1(hereafter called CD2F1), and DBA2. differed by 71% with nitrogen mustard, by 82% with vincnistine, At least one comparison could be made for each of 16 drugs by 85% with cis-diamminedichloroplatinum, and by 245% with that used identical routes, vehicles, and schedules. For a camptothecin. number of these agents, noticeable variations existed in one on It is also noteworthy that there was a lack of parallelism in more of the lethal dose levels (L010, LD50, and LDso). The probit slopes for more than one-half (9 of 16) of the compani average difference between the LD50obtained in a given pair sons. Occasionally, the slopes diverged so dramatically within of mouse strains receiving the same drug was 45% and ranged the LD10to LD@range that reversals of results were observed between 0% (fluorouracil deoxyriboside) and 145% (vincnis according to the lethal dose category considered. For example, tine). For 5 drugs (6-MP, cyclophosphamide, pseudourea, yin the LD50of cyclophosphamide (58 mg/kg) in one study is 1@% cnistine, and camptothecin), the LD50differed by more than smaller than that in the other (67 mg/kg).