Use of Animals to Predict Human Toxicities

| Home | | Pharmacovigilance |

Chapter: Pharmacovigilance: Non-Clinical Safety Evaluation and Adverse Events in Phase I Trials

Animals are not perfect models for humans but there is, currently, no alternative means of assessing the effects of product administration on the whole organism.


Animals are not perfect models for humans but there is, currently, no alternative means of assessing the effects of product administration on the whole organism. Some non-clinical safety assessments are performed using in vitro methods, for example potential for genotox-icity is partly assessed using a bacterial cell muta-genicity assay and a chromosomal aberration assay in mammalian cells. Many of the parameters examined in safety pharmacology, metabolism and toxicity studies are, however, functions of the whole animal.

The use of animals in safety studies is necessi-tated by regulatory requirements and assumes that animal toxicities are generally predictive of hazard to humans. This assumption is the result of experience, which indicates that toxicology studies in laboratory animals yield data that are predictive of human toxi-cities. It is, however, essential to review this funda-mental assumption for both scientific and political reasons. Only if we can be confident that animal models yield data that is predictive of human toxicities can we be confident that safety assessments are useful and justify the test systems. The consequences of poor prediction include inappropriate use of animals, unforeseen toxicities and unwarranted restrictions on potentially useful drugs, which may limit their thera-peutic benefits.

The concordance of the toxicities of pharmaceuti-cals in animals and humans is, then, fundamental to the use of animal study data in safety assessments prior to human administration. Commercial confiden-tiality limits the availability of data on this subject, but there are some literature reports on the subject. In a survey of 139 drugs approved in Japan from 1987 to 1991 (Igarashi, 1994), animal toxicity data were drawn from 468 repeated dose studies, mainly in rats and dogs but with a few in mice and monkeys. Forty-three percent of clinical toxicities from 69 marketed drugs were not predicted from animal studies. The best predictability was for cardiovascular events and the poorest was for skin and hypersensitivity reactions. More recently, a multinational survey of 12 pharma-ceutical companies was reported, in which data from 150 compounds that produced human toxicity events were reviewed, and the human toxicities related to the non-clinical findings (Olson et al., 2000). When toxi-cities in rodent and non-rodent species were examined together, there was a concordance rate of 71% with the human toxicities. The concordance rate for non-rodent species was 63%, whilst for rodents alone it was 43%. Ninety four percent of these concordances were first observed in studies of 1-month duration or less. The human toxicities that showed the highest concordance with non-clinical data were haematolog-ical, gastrointestinal and cardiovascular effects. The lowest concordance rate was in cutaneous toxicities.

The two reviews both indicate that cardiovascu-lar toxicities observed in clinical studies are likely to have been observed first in animals and that cutaneous toxicities also seem to be less apparent in non-clinical than in clinical studies. On the surface, it appears that the Japanese data from marketed drugs and the multinational data from products in clinical trials had similar rates of concordance between non-clinical and clinical findings. The Japanese data indicated that 43% of toxicities in marketed drugs could not have been predicted from the non-clinical results, whilst the rate of concordance for drugs where human toxicities were observed during development was 71%. Presum-ably 29% of toxicities could not have been predicted from the non-clinical data. Caution should be exer-cised in correlating these figures because a number of those products which caused toxicities in clinical trials will not have reached the market. In addition, some rare toxicities may not be detected in clinical trials and may only be revealed when the product is on the market and used by a much larger and more mixed population. Olson et al. cited reviews of clin-ical toxicity resulting in withdrawal from marketing; only 4 of 24 cases and 6 of 114 cases could have been predicted from animals. This poor rate of prediction is considered to be unsurprising because late-onset toxic effects are usually idiosyncratic and therefore inher-ently of low incidence, are not dose related and are not related to the drug’s pharmacology.

It may be impossible to improve the rate of predic-tion of rare and idiosyncratic human toxicities from non-clinical studies, but care should be taken to maximise the rate of prediction for those toxicities related to the metabolism of the test material or to its pharmacological actions. The choice of animal model is very important in this; inter-species differences in metabolism influence the metabolite profile, the route and rate of clearance of xenobiotics, whilst differ-ences in the specificity and/or distribution of receptors give rise to differences in pharmacological responses to a given pharmaceutical. In order to maximise the usefulness of the non-clinical data in safety assess-ment, the species used for toxicity testing should be chosen based on their similarity to humans with regard to pharmacokinetic profile. Additionally, the chosen species should be responsive to the primary pharma-codynamic effect of the substance wherever possible, and in some cases studies in disease models may be warranted (EMEA, 2000b).

Apart from species differences, there are several other factors that may increase the rate of incorrect predic-tions of toxicity when moving from animals to man. These include differences in the way the toxicity is observed and recorded (eliciting verbal accounts of symptoms is not possible in animals), the presence or absence of concomitant medication, pharmacokinetic and metabolic differences, age (animals are young and humans may be old), state of health (animals are free from disease), the small numbers and homogeneity of the animals studied compared with the heterogeneity of the humans, dose differences, housing and nutrition (optimal in animal studies) as well as timing differences.

Overall, the published data suggest that between one-half and two-thirds of pharmaceutical toxicities in humans can be predicted from non-clinical data, thus supporting the use of in vivo toxicology data in assessing the potential for human toxicity. Recently, however, the importance of choosing the most appro-priate animal models and tests and of applying their results in the most appropriate way during safety assessments has come into sharp focus. This has been highlighted by the severe, life-threatening side effects suffered by six healthy volunteers in a Phase I, first-in-human, clinical study in the UK. The medic-inal product administered, TGN1412, is a mono-clonal antibody that was being developed as an immunomodulator and is one of a new generation of medicinal products which are being developed as technology allows the identification and targeting of more complex biological systems.

Repeated dose testing of the antibody had been performed in cynomolgus monkeys and it had been well tolerated by the animals following repeated dosing at doses of up to 50 mg/kg/week for 4 consecutive weeks. This dose level was therefore taken as the no observed adverse effect level (NOAEL) and was used as the basis for calculating the starting dose for the clinical study. The method of calculation followed draft guidelines that are frequently applied to inves-tigational products (FDA, 2002; see also following section) whose mechanisms of action and secondary effects may be better understood than those of the new generation products such as TGN1412. Having applied safety factors and allometric correction factors to scale between the monkey and humans, the human starting dose was selected at 500 times less than the monkey NOAEL. The devastating effects caused in the volunteer subjects clearly demonstrates that use of the no observed adverse effect level obtained from the repeated dose cynomolgus monkey study did not provide a sufficient margin of safety for human dosing of this product.

Examination of this case and of the wider implications for safety assessment of medicinal products – with specific reference to (1) biological molecules with novel mechanisms of action; (2) new agents with a highly species-specific action; (3) new drugs directed towards immune system targets – is ongoing (Duff et al., 2006). Further comment here is therefore inappropriate.

The investigations into this case seem likely to result in new guidance on the safety testing and assess-ment of such products. Whatever the outcome, the case has highlighted the fact that, whilst standard toxi-cological testing has had a good record in predicting the safety of new chemical entities and biologicals whose activities and targets are well understood, new strategies are required to assess the newer genera-tion of products that are designed to modulate more complex biological systems. For more conventional products, however, current toxicology testing strate-gies and safety assessments remain useful although it is pertinent to examine their role and success in predicting human toxicities. This chapter therefore focuses on these types of assessment.

Contact Us, Privacy Policy, Terms and Compliant, DMCA Policy and Compliant

TH 2019 - 2022; Developed by Therithal info.