Ethics of Research involving animals
Extrapolating the results of animal studies to humans: the scientific validity of animal research - continuation II
Critical evaluation of scientific validity
10.37 We have observed that, in principle, animal studies can be scientifically valid. Nevertheless, there is a need for continuing review of the scientific case for using animals in research and testing. It is axiomatic that any such use should be accompanied by active and critical reflection on the validity and relevance of the models and research studies.16 Although scientific claims in favour of the validity of animal research are not usually made in absolute terms, some public statements can over-generalise and tend towards the absolute.17 It is important, for a number of reasons, not to overstate the predictive value and transferability of animal research to humans, because:
- Critical reflections are a vital part of good scientific practice, having value in determining directions and priorities for future research, as well as in interpreting the results of particular studies and refining models.
- Better understanding of the differences between animal models and the human organism can in itself be instructive and can prompt beneficial lines of research (paragraph 7.10).
- It is possible that lack of critical evaluation of the validity of animal models can on occasion be misleading (paragraph 6.32).
- Over-emphasising the predictive value of animal tests can make acceptance of alternative approaches unnecessarily difficult. In toxicity testing, for example, existing animal methods have been validated by the OECD ‘by experience’ and have not been subject to the same formal validation processes as those now required for new non-animal Replacements (see paragraphs 9.4 and 11.24). ‘Claiming too much’ for the predictive value of existing animal methods can sometimes put unnecessary barriers in the way of regulatory acceptance of new in vitro methods.
10.38 It is clear that continuing critical evaluation of the scientific validity of animal models makes good scientific sense, and as our description in Chapters 5–9 shows, is usually a part of good scientific practice. For example, the majority of the scientific community takes the view that similarities between mouse and human genomes are sufficient to permit informative comparisons between GM mouse models of human diseases and the human clinical conditions in specific cases. Nevertheless, such models require careful analysis in order to assess their relevance and effects (see Box 10.2).
| Box 10.2: A recent retrospective study of the potential value of knock-out mouse models* in pharmaceutical discovery and development The study aimed to address ‘common and varied…questions concerning the value of mouse genetics for drug discovery’, including the following.
It might be argued that such a finding is not surprising since the knock-out mice were generated after the medicines were developed, when the mechanism of action of the medicines was already known. However, the authors also assert that more ‘prospective’ use of knockout mouse models is currently yielding benefits. A number of new pharmaceuticals are being developed against human biochemical targets the function of which has been determined using genetic research involving mice, including treatments for osteoporosis and obesity.‡ * That is, mice in which one or a few genes have been deleted, or otherwise disrupted, so as to prevent their expression. † Zambrowicz B and Sands A (2003) Knockouts model the 100 best-selling drugs – will they model the next 100? Nat Rev Drug Disc 2: 38–51. ‡ Zambrowicz B and Sands A (2003) Knockouts model the 100 best-selling drugs – will they model the next 100? Nat Rev Drug Disc 2: 38–51. |
10.39 The study described in Box 10.2 is an example of a systematic attempt to evaluate the scientific validity of using animals as models for humans, by directly comparing findings in animals with the results of corresponding clinical studies. There have also been two recent meta-analyses of such systematic reviews. One was conducted ‘to find out how animal research had informed ensuing clinical research’,19 the other to assess the value of pre-clinical animal studies in permitting safe and effective first-dose studies of potential new medicines in humans.20
10.40 The first paper, by Pound et al. (2004), examined six reviews, each of which compared animal and clinical findings in a specific and problematic therapeutic area (heart disease, stroke, wound healing). The authors concluded that these six reviews provide little evidence to support the view that animal research has contributed to the treatment of human disease. The study has been used to support claims that there is ‘no-evidence base for animal research’.21 But it has also been strongly criticised, in particular for its selectivity, given that other systematic reviews were identified by the team but were excluded from the analysis. Of the six reviews discussed in the paper, five were initiated following lack of success in clinical trials, which could have been predicted from better analysis of the relevant animal studies. The sixth was initiated because of difficulties in establishing an animal model of the relationship between social status and coronary heart disease. Nevertheless, the study has served to highlight cases in which there were some methodological problems in the animal studies and/or in which full analysis of the animal results available would have predicted the ineffectiveness of the treatment, had such an analysis been done before clinical work started.22
10.41 The second meta-analysis draws on the work of Olsen et al.,23 among others, and concluded that, although the relevant available data are ‘fragmentary’,24 the concordance between short-term toxic effects of new pharmaceuticals in animals and humans (during clinical trials) was 71 percent. This means that 71 percent of human acute toxicities resulting from compounds that entered clinical trials were predicted by pre-clinical safety pharmacology or toxicity studies in animals. It is noteworthy that this conclusion has been used as part of cases both ‘for’ and ‘against’ the predictive value of pre-clinical animal studies: thus while 70 percent of human toxicities were predicted, 30 percent were not, and the rodent tests alone predicted only 43 percent of human toxicities.25
10.42 It is also worth noting that the toxic events considered by Olsen et al. are likely to be at the more minor end of the spectrum of potential adverse effects. Compounds causing significant damage to animals would not have entered clinical trials. Reliable systematic data on compounds eliminated before human dosing because of major organ toxicity in animals are not available. It is therefore not possible to judge how many compounds were rejected because of their adverse effects in animals.26 As before, this observation could be used to support or contest the scientific validity of animal tests. On the one hand, it can be argued that actual concordance is greater than 70 percent, when the animal tests showing adverse effects too significant to proceed to human trials are taken into account. On the other, it might be argued that animal research may lead to the loss of potentially useful medicines for humans as compounds might be removed in the screening process because of significant toxicity in animals which would perhaps not occur in humans. However, those defending the use of animals would argue that the option of ‘losing’ some compounds in this way can be viewed as preferable to exposing humans to medicines that have not undergone prior testing.
10.43 Finally, it should be noted that the Olson study only considered toxic events observed in human clinical trials, i.e. short-term effects. Longer-term toxicities such as carcinogenicity and teratogenicity were not assessed. For these long-term toxicities it has been difficult to establish the validity of animal tests27 which have been criticised by toxicologists.28 Thus the concordance between animal and human long-term toxicities, if it could have been measured, may prove lower than found by Olson et al. for short-term toxicities. At the same time it needs to be acknowledged that assessment of long-term toxicity is a highly complex process. For example, while it may be straightforward to identify a number of people who have taken a certain medicine at some point in the past, it may be less straightforward to correlate possible negative states of health which occur, for example, a decade after the medicine has been used. Since people may have taken a range of other medicines in the meantime, and since factors such as lifestyle or exposure to chemicals in the workplace may also play a role, many factors need to be considered.
Footnotes16 This argument also applies to the use of animals in studies that are extrapolated to other animal species.
17 See Animal Procedures Committee (2003) Review of the cost-benefit assessment in the use of animals in research (London:
HO) for further discussion.
18 Some commentators claim that it is easier to achieve OECD approval for new animal, as compared to non-animal methods,
see: Written evidence submitted by Dr Gill Langley to the House of Lords Select Committee, page 100 based on references from the OECD.
19 Pound P, Ebrahim S, Sandercock P et al. (2004) Where is the evidence that animal research benefits humans? BMJ 328:
514–7.
20 Greaves P, Williams A and Eve M (2004) First dose of potential new medicines to humans: how animals can help Nat Rev Drug Disc 3: 226–36.
21 See, for example, rapid response letters to the British Medical Journal.
22 See for example Blakemore C and Peatfield T (2004) Missing evidence that animal research benefits humans BMJ 328: 1017-8
Box 10.2: A recent retrospective study of the potential value of knock-out mouse models* in pharmaceutical discovery and development The study aimed to address ‘common and varied…questions concerning the value of mouse genetics for drug discovery’, including the following.
What is the correlation between mouse and human physiology and hence the relevance of knock-out models in developing small-molecule drugs?
Does gene compensation (when the expression of another gene alters to compensate for the loss of another during development) prevent identification of the true function of the genes that have been
knocked out?
Since current technology means that the genes are usually knocked out very early in development, in what sense are the effects of the lack of a particular gene throughout development relevant to the
function of the gene in adult animals?
How far is the embryonic or neonatal death of some knock-out mouse lines likely to prevent the identification of many of the best drug targets in future?
In light of such questions, the study demonstrated that the 100 best-selling human pharmaceutical medicines between them have 43 human biochemical targets, the genes for 34 of which have now been knocked out in mice. A literature review revealed that, of these 34
knock-out models, 29 (85 percent) provide a direct correlation with the therapeutic effect of the relevant medicine. In the remaining five cases, early (e.g. embryonic or neonatal) lethality or unrelated abnormalities meant that the knock-out mice were not useful models for humans.†
It might be argued that such a finding is not surprising since the knock-out mice were generated after the medicines were developed, when the mechanism of action of the medicines was already known. However, the authors also assert that more ‘prospective’ use of knockout mouse models is currently yielding benefits. A number of new pharmaceuticals are being developed against human biochemical targets the function of which has been determined using genetic research involving mice, including treatments for osteoporosis and obesity.‡
* That is, mice in which one or a few genes have been deleted,
or otherwise disrupted, so as to prevent their expression.
† Zambrowicz B and Sands A (2003) Knockouts model the 100
best-selling drugs – will they model the next 100? Nat Rev
Drug Disc 2: 38–51.
‡ Zambrowicz B and Sands A (2003) Knockouts model the 100 best-selling drugs – will they model the next 100? Nat Rev Drug Disc 2: 38–51.
23 Olson H Betton G, Robinson D et al. (2000) Concordance of the toxicity of pharmaceuticals in humans and in animals Regulat Toxicol Pharmacol 32: 56–67.
24 The authors note that much of the relevant information is held by government regulatory authorities and pharmaceutical
companies and is not publicly available in the peer-reviewed scientific literature. The authors state that they can ‘only learn from experience and then only if we have access to information’.
25 Animal Procedures Committee (2003) Review of the cost-benefit assessment in the use of animals in research (London: HO).
26 Greaves P, Williams A and Eve M (2004) First dose of potential new medicines to humans: how animals can help. Nat Rev Drug Disc 3: 226–36.
27 Lo WY and Friedman JM (2002). Teratogenicity of recently introduced medications in human pregnancy. Obstet Gynecol 100:
465-73.
28 For example, Ennever FK and Lave LB (2003). Implications of the lack of accuracy of the lifetime rodent bioassay for predicting human carcinogenicity. Reg Toxicol Pharmacol 38:52-57; Johnson FM (2003) How many high production chemicals are rodent carcinogens? Why should we care? What do we need to do about it? Mutat Res 543:201-15; and Gottman E, Kramer S, Pfahringer B et al. (2001) Data quality in predictive toxicology: reproducibility of rodent carcinogenicity experiments Environ Health Perspect 109:509-14; Kennedy DL, Uhl K and Kweder SL (2004) Pregnancy exposure registries Drug Saf 27:215-28.