This post is concerned with all experiments that do not involve the use of living people.
In vitro experiments are literally those done “in glass” although in practice many experimental vessels these days are actually made of plastic. This category includes experiments done with micro-organisms, isolated cells, cell-free systems like broken cells or isolated enzymes and pieces of animal, human or even plant tissue.
In vivo experiments are those performed using living animals, people or plants. Plants are not often used as “subjects” in biomedical research but back in the 1850s and 1860s Mendel used experiments with pea plants to formulate his three laws of genetics.
In silico studies are those done using computer simulations and are not discussed here but are increasingly being used in some areas like drug design.
In an experiment, the effect of some intervention or constraint is measured such as:
- What effects does a substance or environmental change, like temperature or pH, have upon the activity of an enzyme, or the growth of bacteria or growth of tumour cells?
- What effect does a substance or environmental change have on animals’ growth rate, blood pressure, blood cholesterol level or death rate from a disease?
These experiments are designed to test a hypothesis such as:
- Cold exposure increases metabolic rate and therefore food intake in mice
- Adding purified cholesterol to rabbits’ diets increases blood cholesterol level
- Low doses of atropine increase heart rate in anaesthetised rats
- A potential antibiotic kills cultured bacteria or prevents them from multiplying
- A potential cholesterol-lowering drug inhibits a key enzyme in cholesterol synthesis.
In properly designed experiments one should compare the measured effect in control and experimental groups or situations. The control and experimental groups should be identical in all respects except for the intervention being tested. For example if one were testing the effects of insulin injections upon blood glucose in rats then one would start with two matched groups of rats. One group would be injected with insulin and the control group injected with just the solution in which the insulin was dissolved. Blood glucose would be measured before and say 1 hour after injection. One would perform all injections and samplings at the same time and in the same place. One might have a schedule of one rat from alternating groups injected every 5 minutes and the blood sample for analysis collected exactly one hour after injection. If one were testing the effects of substance X on the activity of an enzyme then one would measure enzyme activity in test tubes with identical contents except for the presence or absence of substance X. One can also design an experiment with control and test periods rather than have separate control and test groups i.e. record data before, during and after the intervention if its effects are only temporary.
We will see in post IV in this series that double blinding is an important feature of good clinical trials; neither subject nor experimenter should know which group the subject is in until the data has been collected. This is to allow for any placebo effect of a treatment and also to eliminate bias in the way the experiment is conducted and the outcomes measured. Likewise in non-human experiments, some degree of blinding is desirable especially if there are obvious ways in which the experimenter can consciously or unconsciously bias the results e.g. if skill and care is involved in carrying out the procedure or if there is any element of judgement or subjectivity in recording the outcome data. The same person should make the measurements in both control and experimental groups.
Role of animal and in vitro experiments
Animal and in vitro experiments are often the starting point of any major new advance in the biomedical sciences. It is, however, important to always remember that data from in vitro or animal experiments should only be used to generate hypotheses about would happen in people; if possible, these hypotheses must then be tested in people. Showing that a potential antibiotic kills bacteria in a test tube or even that it cures an infection in mice is some way from discovering a new human antibiotic but anything that did neither of these would be an unlikely candidate as a potential antibiotic. Showing that a substance kills or prevents the growth of cultured tumour cells or has beneficial effects upon an experimental animal cancer would be a promising start for a new cancer drug but a long way from having a new cure for human cancer.
In the protein gap post, I suggested that animal experiments had helped mislead nutritionists into exaggerating the protein needs of growing children and thus in creating the illusion of widespread protein deficiency. The false notion that some defect in the heat generating capacity of brown fat might be a major cause of human obesity was also largely generated using studies with rats and mice (see later). Poorly interpreted animal experiments do have the potential to mislead medical researchers. In the past, some medical researchers have been guilty of regarding mice as scale models of people without full regard for the differences between species.
Many of those who oppose the use of experimental animals, suggest that differences between animals and people make the results of such experiments of little value. There are indeed many reasons why laboratory animals’ responses to an experimental intervention may be a misleading guide to the likely human response and some of these are discussed later in this post. Nevertheless, most biomedical scientists believe that, for the time being, animal experiments are essential to maintain the pace of advances in medical understanding and treatment. Most of the alternatives to animal use, like experiments with cultured cells, isolated pieces of tissue, micro-organisms or even computer simulations are even less likely to unfailingly predict how a whole person would respond.
There are many examples of major advances in medical understanding or treatment where non-human experiments have played a central role:
- Much of our understanding of the inheritance of genetic diseases stems from work with peas and fruit flies
- Experiments with micro-organisms played a major role in our understanding of the molecular mechanisms involved in inheritance
- Experiments with giant axons (large nerve cells) in squid played a key role in our understanding of how nerve impulses are transmitted
- Experiments with dogs played a major role in the discovery of insulin and producing the first effective treatment for previously fatal type I (juvenile) diabetes
- The development of penicillin stemmed from studies using bacteria grown on agar plates.
Animal and in vitro experiments are often an essential first step on the path that leads to a major new advance in medical understanding or treatment.
The use of sentient, live animals is an emotive issue and an important element in the ethical case for their continued use is a convincing demonstration of their key role in advancing medical knowledge and treatment; without this, the ethical case fails. In 1992, I constructed table 1 which is an analysis of the species used by Nobel Prize winners in physiology or medicine over an eighty year period. Nobel prizes provide an unbiased sample of key landmark advances in human biology and medicine that have stood the test of time. Table 1 shows that 90% of all these prize winners used non-human species for at least part of their work. This analysis gives a less anecdotal indication of the central role of experiments with non-human species in medical understanding and treatment. When the analysis is split into before and after 1942 then we can see there is a trend towards greater use of species, like micro-organisms, that are more distant from us in evolutionary terms. This trend reflects the growth in molecular biology and as noted earlier, much of our current knowledge of molecular genetics comes from work with microorganisms. The figures in table 1 make it difficult to sustain the argument that studies with non-human organisms have not made a critical contribution to biomedical understanding and to the treatment and prevention of disease.
Table 1 “Subjects” (%) used by all 139 Nobel Prize winners in physiology or medicine 1901-1984
Humans Primates Other Other Microbes
22* 6 73 11 24
29 2 88 17 7
20 8 66 10 31
*About half of these also used other species
(Note that metazoans are multi-celled animals and so “other metazoan” would include molluscs, worms, crustaceans, insects and other arthropods. Vertebrates includes mammals, birds, fish and amphibians).
Animal use in UK experiments
In the UK, the Home Office regulates what experiments are permitted to be carried out with living vertebrates. Those performing such experiments must be licenced and they must be carried out in licensed premises which are subject to regular inspection. The Home Office publish annual statistics giving the total number of experimental procedures and a breakdown of the species used, the purpose of the procedure and the severity of the procedure. Table 2 shows a breakdown of experimental procedures by animal type. Just over 2 million procedures were carried out in 2016 (a further 1.9 million were involved in the creation/breeding of genetically altered animals without any experimental procedure). About three-quarters of all procedures used rodents.
Table 2 Procedures carried out on living animals in the UK in 2016 classified by animal type
Animal type Number of procedures %
All 2.02 million 100
Mice 1.26 million 61
Rats 258, 000 12
Fish 294,000 14
Birds 141,000 7
Other* 107,000 5
Specially protected 17,000 0.8%
*Includes guinea-pigs, hamsters, rabbits, farm animals, reptiles and amphibians
**These are horses, dogs, primates, and cats (no great apes have been used since 1987 and their use has been prohibited since 2013)
The rationale for using non-human species in medical research
The notion of evolution from common ancestors underpins and unifies the biological sciences. Each individual species has evolved with certain specific characteristics that increase its fitness to survive in its own environment. Nevertheless this theory also suggests that many processes and mechanisms will be common to many or even all species despite the modifications that evolution produces. Evolution modifies or improves a basic plan and this has produced a vast array of living organisms who share some characteristics but each is fitted to its own particular ecological niche.
As an example of this commonality, all independently viable organisms have their genetic characteristics encoded in their DNA and the genetic codes within DNA determine the amino acid sequences of the proteins (e.g. enzymes) that an organism produces and this ultimately determines the characteristics of the organism. These DNA codes are the same across all organisms and identification of these codes has been derived using in vitro studies. Substances that cause mutation in bacteria do so by causing changes in the DNA sequence and they are often carcinogenic because changes in the DNA sequence of human cells can cause them to become cancerous; hence the logic of initially screening chemicals for their carcinogenicity by testing how mutagenic they are to bacteria.
In vitro experiments
One might, for example, test whether a potential drug or a natural extract inhibited multiplication of cultured tumour cells or killed cultured bacteria. Such experiments might be a useful way of initially screening potential anti-cancer agents or agents that might have potential as antibiotics or antiseptics. One might test whether a potential drug inhibited an enzyme e.g. statins block a key enzyme in the normal synthetic pathway for cholesterol and they are now very widely used as cholesterol-lowering drugs. One might test whether a substance induces mutations in bacteria because substances that do this are potential carcinogens that could cause normal human cells to mutate into cancer cells. Drugs that relax smooth muscle in isolated artery walls or slow isolated an isolated heart may have potential as anti-hypertensive agents. These types of experiments can be particularly useful as a way of rapidly screening large numbers of compounds to see if they are worth studying further as potential therapeutic agents or for screening things to see if their presence in the environment or in food is likely to pose a potential hazard or benefit to human health.
Experiments with animals, especially small animals, have a number of very obvious advantages over human experiments. The experiments are relatively cheap because of the small size and rapid breeding of these animals. It is possible to do animal experiments of much higher technical quality than is usually possible with human experiments e.g. one can keep them all under identical, controlled environmental conditions and feed them on precisely controlled diets. By using highly inbred strains of mice one can essentially eliminate genetic variability between individual animals because all animals of the same sex are effectively genetically identical i.e. all the males or females of whole strains are like identical twins and tissue can be transplanted between them without risk of rejection. There are also fewer ethical limitations on the nature and range of experiments that can be undertaken although there are nowadays strict legal regulations and ethical guidelines governing animal experiments. One cannot normally set up a human experiment to test effects of an intervention expected to cause serious harm or even one where serious harm is a real possibility. If there is satisfactory justification, such interventions can be approved in animal experiments and this considerably extends the range of experiments that can be conducted.
One would expect that results from animal experiments will be relatively easy to repeat in different laboratories; these experiments should therefore have high repeatability or reliability. However there is always a question mark about the validity of applying the results of experiments done with say inbred mice under highly controlled laboratory conditions to genetically diverse, free-living people. If one did an experiment on a single person from an inbred, isolated and newly discovered tribe of people would one confidently expect the results to predict what would happen in people generally let alone predict the response of mice? Animal experiments are reliable but are they always valid when used to predict human responses? An expert rifle shot would expect to be able to achieve a very narrow clustering of results on the target (high reliability or repeatability) but if the sights of the rifle were poorly adjusted then this cluster might be a long way off from the central bull’s eye of the target (low accuracy or validity). Animal experiments have high reliability but their validity for humans should always be considered and, wherever possible, tested. There are sometimes fairly obvious reasons (see below) why the response of a mouse might not predict the response of a man. It is sometimes not even possible to even predict the experimental responses of one strain of mice from those seen in another mouse strain.
The potential of animal experiments to mislead human biologists
Misinterpretation of animal experiments helped cause two important scientific errors
In an earlier post I have suggested that misinterpretation of results from laboratory experiments with small mammals probably played an important role in causing two major scientific errors; the false belief in a crisis in world protein supply (the “protein gap”) and the belief that a defect in the heat generating mechanism (thermogenesis) in brown fat could be an important cause of human obesity. Despite these criticisms, I have also made a clear case in the above discussion that animal experiments have played a crucial role in advancing our understanding of human biology and in generating the consequent medical advances (as illustrated by table 1).
Animal studies have the potential to mislead human biologists if there is not careful consideration of how and whether the results from animal studies should be applied to people. As suggested earlier there may have been a tendency amongst some human biologists to subconsciously regard small laboratory animals as essentially scale models of people. This may be because the results of animal experiments have sometimes been hastily applied directly to humans without causing any obvious problems or misunderstandings. This means that scientists may become careless in applying their animal finding to people without always fully considering why the responses of a laboratory rat or mouse may not accurately predict the responses of free-living people. Despite the usual robustness of animal experiments in indicating how human beings would respond in a given situation there are clear biological reasons why what is observed in a laboratory mouse would not always be a good indication of what would happen in a person. For example, rapidly growing offspring of rats and mice require much more of the energy in their food to be in the form of protein than do slow-growing human babies and children. This is reflected in the big differences in the protein content of rat or mouse milk compared to human milk; human milk has only about a quarter as much of the energy as protein as compared to rat or mouse milk (see the “protein gap” case study.
Different strategies of mice and people during cold exposure
During cold exposure small mammals like mice must generate more heat to maintain their body temperature and energy expensive heat generation in a tissue called brown fat (non-shivering thermogenesis) seems to be the key to their survival in a cold environment. When people are exposed to cold their main physiological response is to conserve heat by reducing surface blood flow and allowing the surface to become cool and this slows heat losses. This heat conservation strategy would be ineffective in an animal as small as a mouse where no part of the body is more than a centimetre or two from the surface. When subjected to cold, people respond behaviourally by trying to cover themselves in more insulating material (clothes) and they respond physiologically by reducing blood flow to the skin and surface layers which reduces heat losses quite effectively. Any extra heat generation is largely achieved by increasing muscle movement e.g. by shivering. Mice can huddle together to reduce heat losses and have also been fairly recently been shown to be able to reduce their heat losses and conserve energy by becoming torpid i.e. by dropping their temperature down to close to normal room temperature (20-25o) and temporarily abandoning the energy expensive attempt to maintain their normal high body temperature (37o). Figure 1 shows continuous body temperature records in individual mice over a three-day period that included a 48h fast. Temperature was measured using temperature sensitive radio-telemetry emitters implanted in the abdominal cavity. These individual mice underwent two periods of deep torpor. At the end of these periods of torpor they aroused and warmed up spontaneously without any human interference and before their food supply was restored. Lowering of body temperature was therefore a normal adaptive response of these mice to starvation which enabled them to conserve energy when short of food.
Figure 1 Body temperature in two C57BL/6 mice recorded over a 3 day period that included an enforced 48h fast (after Webb, 1992)
This is more than just an interesting observation because it was a factor in misleading researchers about the cause of obesity. Genetically obese, ob/ob. mice have long been known to have a lower body temperature than their lean siblings. They are susceptible to the cold and may die if suddenly exposed to fridge temperatures (<40). In the late 1970s and 1980s, this was widely interpreted as a failure of their brown fat to generate sufficient heat because of genetically defective brown fat i.e. a failure of their heat-generating system. This led many obesity researchers, like Jean Himms-Hagen in the quote below, to suggest that defective heat generation in brown fat was a major cause of obesity.
“Recent studies on brown adipose tissue have shown that a defect in this tissue is one probable cause of obesity” J Himms-Hagen (1979).
However, knowing that mice can exhibit torpor when fasted, we suggested that their low body temperature might be an adaptive response to perceived starvation i.e. their brown fat is “switched off” to reduce heat production and prevent the “waste” of energy in keeping body temperature normal (370). Their low body temperature could be seen as a permanent semi-torpid state and their cold susceptibility due to a sort of disuse atrophy of their brown fat because it is permanently switched off. This alternative explanation seems to be in keeping with more recent understanding about the exact nature of the genetic defect in these mice and with the observation that people who have the same rare genetic defect have normal body temperature.
There are thus occasions, such as enforced fasting, when even the most commonly used experimental animals, mice, “abandon” the energy expensive struggle to maintain “normal” 370 body temperature for several hours at a time. Some other mammals like hamsters, hedgehogs, bats and dormice spend many winter weeks or months in hibernation where their metabolism and heart rate slows and their body temperature drops close to the environmental temperature. Anyone who has kept a garden pond will know that, unlike mammals, fish eat less in cold weather. The body temperature of fish and other non-mammalian vertebrates fluctuates with environmental temperature. In these so-called poikilotherms, metabolism is slower in the cold whereas non-hibernating, homeothermic mammals increase their metabolic rate in an attempt to maintain their high body temperature. It is obvious that fish would be a poor guide to the effect of environmental temperature on food intake in mammals but it seems that the less well-known ability of mice to become torpid also make some experiments with mice an unreliable guide to human responses.
The nutritional burden of pregnancy in mice and people
When a mouse becomes pregnant it produces, in about three weeks, a litter of pups that weighs around 40% of the mother’s original body weight. This new mouse mother then supplies milk to this litter that enables it to double its birth weight in around 4-5 days. A human mother takes 9 months from the time of conception to produce a baby that weighs just 6% of her body weight and if she exclusively breastfeeds this infant it may take as long as 6 months to double its birth weight. It seems unlikely that the mouse would be a good model for predicting the likely impact of pregnancy and lactation upon the nutritional needs of the human mother.
Species vary in the nutrients they require and their response to foreign chemicals
If a rat or mouse is deprived of vitamin C, it has no impact upon its health because these animals, like most mammals other than primates and guinea-pigs, can manufacture their own vitamin C. Human beings develop the deficiency disease scurvy within a few weeks of vitamin C deprivation. Penicillin is toxic to guinea pigs so if guinea pigs rather than mice had been used in the early testing of penicillin then it might have considerably delayed the development of this important class of antibiotics.
Even where there are not fundamental differences between the response of small animals and people in experimental situations, it may still be difficult to translate some quantitative findings like effective drug dosages or nutrient requirements of small mammals to large human beings because of variation due to size and inherent sensitivity. Scaling on a weight basis (e.g. mg of drug per kg of body weight) may be easy and convenient but is it based upon sound logic and practical evidence? Drug or nutrient doses may be more dependent upon the relative metabolic rates of the two species than relative body weight. Species may also vary in their fundamental sensitivity to a drug or their requirement for a nutrient that is not just due to size differences. Consider the problem of trying to decide upon the appropriate dose of LSD to use on an elephant. This example is based on a real piece of research in which three scientists in Oklahoma were testing whether LSD could induce a naturally occurring condition called musth in which male elephants become violent and uncontrollable. If the dose was estimated on a simple body weight basis from the dose effective in cats then this would give a dose estimate of 297mg because cats are quite resistant to the effects of LSD. If the dose effective in people was scaled up in this way, the estimated effective dose would be 8mg. If the dose was scaled according to the relative metabolic rate then the predicted effective dose would be 80mg if one relied on the cat or 3mg if one used the effective human dose. Of course the brain is the main site of action of LSD so if one predicted the effective elephant dose by comparing the brain sizes of people and elephants then the predicted effective dose would be just 0.4mg (after Knut Schmidt-Nielsen, 1972). Eventually the unfortunate elephant (a male Indian elephant named Tusko) was given a dose of 297mg and within minutes it went into convulsions and died. The paper describing this experiment has been the subject of considerable controversy and discussion and many people regard the injection of such a high and untested dose in a large, valuable and intelligent zoo animal as irresponsible. In an article in the UK Daily Mail in November 2007 Fiona Macrae uses it as the first example in her discussion of “the ten silliest experiments of all time”.
As another example of this problem of scaling, consider the problem of trying to model with laboratory mice, the situation of two human populations, one consuming 40kg sugar/person/year and another consuming 20kg per year. These sugar intakes represent somewhere around 20% and 10% respectively of the total calorie intake of these two human populations. If you scaled these sugar doses simply according to the difference in body weight of people and mice then the two mice populations would receive amounts of sugar that represented just 2.5% and 1.3% of their calories in the form of sugar.
Most of the differences between the physiology of mice (or rats) and men noted above are obvious and well known. Scientists working with these animals would generally have been aware of these differences when using mice as experimental models of people. However, do they always take such differences fully into account when applying their animal findings to human beings? Do they always choose appropriate species when trying to model the likely human response to any intervention? Clearly fish would be inappropriate for modelling the effect of cold upon metabolic rate and food consumption and the occurrence of torpor or hibernation would make some mammals less than ideal species under some circumstances.
Has the role of dietary cholesterol upon blood cholesterol and atherosclerosis in people been exaggerated by early experiments with rabbits? The blood cholesterol of rabbits rises and they develop atherosclerosis if cholesterol is added to their feed. Rabbits are herbivores and so their natural diet would not contain cholesterol and so one might expect them to be ill-equipped to deal with an artificial dietary cholesterol load. Omnivorous rats are much less sensitive to the effects of dietary cholesterol.
Loosely based on:
Webb, GP (1992) Viewpoint II: Small animals as models for studies on human nutrition. In Nutrition and the consumer. Ed. Walker, AF and Rolls, BA London: Elsevier Applied Science. Pp. 279-297.