eSignature Lawfulness for Animal Science in United Kingdom: Simplify Your Document Processes with airSlate SignNow

How to eSign a document: eSignature lawfulness for Animal science in United Kingdom

and without further Ado it's wonderful to welcome you to this um last week star research some seminar of the semester I'm Georgia Mason I'm director of the Campbell center for the study of Animal Welfare um those of us here in Canada um acknowledge that we're on traditional traditional territories and treaty lines of indigenous people in Inuit people invading people and hearing wealth of course were based on the traditional territories of the mississaugus of the credit if you're not Canadian and you're online and you're thinking why Canadians always do this it's part of a Truth and Reconciliation movement to really acknowledge the kind of dark and murky past of Canada's Colonial history and so it's just a little tiny um recognition that the land we're on was not originally opinions both um if anyone wants to take this kind of a little bit further have a look at some of the websites about indigenous involvement in World War and one two and a Korean War because obviously November is a month of remembrance and there's some interesting things out there yep there's a comment on the chat about as well yeah oh thank you okay the owl is now under my nose is that better sound to people online foreign and he directs the division about all welfare there he's got a um long and distinguished academic history really ruining his understanding of animals welfare and behavior to bear on the use of animals in research and he has a sort of critical lens where he uses what we know of behavior and Welfare to ask first of all why do we always keep doing better cages we'll need those that makes them weird second question why are the cages all identical when the humans they're modeling is so heterogeneous third question Wellness is biomedical research failing so badly to produce reproducible or um translatable results and then its current question is hey maybe we shouldn't be in a glass house throwing Stone because Animal Welfare science itself is perfect so what can we learn about critiques of other disciplines that we can apply to our own work amongst as many Publications and accolades include winning the ISE creativity award and winning a center for it's the John Hopkins centers for the ultimate Alternatives that animal needs three hours award and he's also got some fantastic Publications um many of which you can see on Google Scholar but he's been involved in some really important policy pieces like the arrive guidelines and also the US's Guide to the parent use of the oratory animals okay without further Ado I'm going to pass you over to Hannah who's going to talk about the principles of good science as they apply to Animal Welfare science well thank you very much Georgia for this uh nice introduction thank you first of all for having me uh for this presentation I'm currently sitting in a rather miserable um hotel room I mean not miserable it's just hot and uh tiny uh because I'm at the conference in Germany and um so I'm trying to do my best uh to give a lively presentation here because of course it's always a bit difficult to talk to your own laptop when you would rather talk to people and I can only see you in the very uh tiny right corner of my laptop because otherwise you just cover up all my slides and I don't know what to talk about so um it's quite a bold title that I've chosen for this um presentation here principles of good animals welfare science welfare in Brackets because of course much of what I'm talking about is about science in general or animal science in general because there just isn't that much um meta research going on on Animal Welfare signs so that I could use all examples from Animal Welfare science but I think that many or much of what I'm going to talk about is applicable to Animal Welfare science and I'm going to give you some examples that on this court is to some extent so when the first freeze right okay so what do we mean by good Animal Welfare signs or what do we mean by the quality of Animal Welfare science so if you look up what quality means then that's how good or bad something is and so we can ask like what is good science what is good Animal Welfare science and there are obviously different answers to this question science that answers relevant question is that what we're looking for or is it science that uses rigorous methodology or is it science that is morally responsible and I would argue that um good Animal Welfare science cannot be just one of those you may ask relevant questions but if you don't use rigorous methodology your findings will be useless if you're not using rigor's methodology you may be wasting animals for irresponsible research and so I would argue that basically good Animal Welfare science is relevant rigorous and responsible it has to be all of those but who is deciding whether Animal Welfare science is good or bad now generally whether the questions that we're asking are relevant that's in the hands of the funding bodies or the experts that advise funding bodies on whom to get money and we're generally assuming that if if someone is ready to pay for a study then the question must to some extent be relevant at least for those who are funding that research whether the methodology's record is that is usually in the hands of our peers so peer review decides on whether or not we can publish our findings and the fact that we publish or the fact that we're allowed to publish and that our papers get accepted means that other scientists thought that what we did was write and was good sign so that it's worth being published and whether or not what we're doing is morally responsible that's often in the hands of ethical review board to review our study protocols and especially in animal research of course ethical review boards play an important role in deciding whether or not the benefits of the research are sufficient to outweigh the harms that may be imposed on these animals in the course of research now I'm going to focus on the rigor of methodology today um for one because there is a lot of research going on on this and there are some substantial problems out there and it's also I think the most interesting one because it's the scientific community that is responsible for the rigor of methodology so first of all it's the scientists who use the methodology and decide whether or not to use rigorous methodology and it's their peers why do you accept or don't accept their research findings depending on whether they consider them rigorous or not now that something may be wrong with um science in general um that is obvious when you look into the literature of meta research and even though there is not that much out there on Animal Welfare science if we look at science in general then um there is certainly there have certainly been um loud voices um voicing concerns over the rigor of of uh methodology and Science and John yanidis was not the first one to criticize a poor methodology or lack of scientific rigor but he was certainly one of the loudest and most prominent and that is certainly also due to the fact that he dared to publish a paper that was entitled why most published research findings are false now that's of course not what we want to hear neither the ethical review bodies want to hear that because if they approve of research on animals that turns out to be useless because the results are false then it cannot be responsible research now this wasn't just a catchy title he actually did go into details and this is all based on on on uh um a large stream of meta research that he and his group conducted when they came to the conclusion that most research findings are false for most research designs and for most fields and by this I imply that it may also affect Animal Welfare science now since then we're talking about something that is often referred to as a reproducibility crisis and this is data from a survey that Eternal nature conducted some years ago when they asked their readers whether they think that we do have a reproducibility crisis in science and 90 of those who participated in this survey they did think that there were issues with reproducibility and more than 50 percent uh thought that there was a significant uh crisis uh in terms of reproducibility now this term reproducibility crisis has certainly been fueled by spectacular cases of replication failure and these are certainly the most famous ones it's two pharmaceutical companies and they did in-house replications of studies that have been published in in prominent scientific journals so they picked the good ones the famous ones and they tried to replicate these studies in-house because often these uh studies that publish or that report on spectacular findings um they may pave the way for advances in in in medical development now what they found was that in one case in the study conducted by bioscientists of 67 studies that they attempted to replicate they could only replicate 36 percent of the findings while 64 two-thirds of the findings could not be successfully replicated and in the study by Amgen it was even worse that only 11 of 53 study outcomes could be um uh replicated in their own hands now what are potential causes of such poor reproducibility and there are of course many and that's what I'm going to talk you through in the course of this presentation is a couple of of or or some of the most um prominent or most likely causes of poor replicability all of which probably contribute to these issues some more uh some less and there is certainly variation across fields of research and I'm trying to figure out what uh what if which of these causes uh are uh most relevant for Animal Welfare science so to start with of course there may be issues with reproducibility and certainly with translatability or generalizability if the animal models are are not um properly validated or if they are essentially poor models for the constructs being studied now this is a major issue of course in in pre-clinical animal research where we often use um animal models that are very distant from human beings and in evolutionary terms but also in in the general biology including behavioral biology and he has often argued that mice may not be the best model to model diseases in humans but in addition to this besides actually the the species gap between humans and mice there are other issues and these have been discovered more recently namely that it's not just that these animals are different from humans but it's also that we breed and house them in in ways that are not very conductive to them modeling relevant States and human beings and one of the issues is that these laboratory animals these Liberty mice they are highly artificial creatures in most cases they are highly inbred they are almost completely homozygous that's one thing and the other thing is that they have been um of course derived from wild mice but the way in which they have been bred and housed resulted in animals that are essentially germ free and they are very far removed in their microbiota and the pathogens that they carry from wild mice and interestingly more recently researchers has to have discovered that this is a major issue given that these laboratory mice the immune system of these mice doesn't really develop beyond the state of of a newborn that they have started to actually rewild these animals by transferring embryos into white mice in order to restore the natural microbiota and pathogens thereby creating so-called wildling mice and there are already some successful studies showing that the translational um power of these models is much better than than the the highly hygienic standard laboratory mice foreign besides um the genetics and and the the General biology there are of course also issue with the types of housing conditions that these animals are housed under and this is a picture from uh my very early steps in in applied ethology when I did my PhD thesis on on stereotypes and laboratorymize basically discovering that most laboratory mice when kept under the disparent standard laboratory conditions develop abnormal repetitive behaviors and this is a picture that depicts two different types of stereotypes in two different strains of laboratory mice so from the beginning on because I was working on Liberty mice of course it was never only an issue of Animal Welfare but but it was always because these animals are used for uh science for research um finding that the health and well-being of these animals may be impaired of course also raised concerns about the scientific validity of these animals for producing um valid research and of course more recently uh Georgia Mason and her PhD student Jessica Kate and colleagues they found that we're not much further than that and even today conventional laboratory housing increases morbidity and mortality and research wrote and substantially and this of course is a major welfare issue but it's also a major concern in terms of the validity of these model organisms besides the animals themselves whether or not they're adequate model organisms of course there's also a question about the measures that are used in research and whether these measures are valid outcome measures for the constructs that AR meant to measure and one of those measures or test systems that have been under have have come under attack recently is um the fourth swim test that you can see here where a mouse is placed um in a in a in a water path and then the time is measured uh how long the animal struggles to find an escape before it gives up and this test has been used and is still being used a lot as a test for measures of depression in pre-clinical animal research and even though researchers neuroscientists have long demonstrated and argued that this test has lack actually essentially lacks construct validity for depression or as an outcome variable for depression it still continues to be used and of course by this it has become an easy target for animal rights groups who didn't even have to complain about whether or not this is stressful and and and painful for these animals they could simply refer to other scientists demonstrating that this test likes construct validity and therefore should be used um with animals now what about animal model validity and Animal Welfare Science Now compared to pre-clinical animal research of course we are often much closer to our Target populations and many of us actually work on target populations so whether I'm studying laboratory mice in the context of a of a research laboratory or others studying laying hands or um or um cows cattle on a farm generally Animal Welfare scientists study study populations that are representative of their target populations and often they also study these animals in settings that are much closer to the conditions under which these animals are kept on farms or in Laboratories so to test both the the populations uh the study population as well as the study settings are representative or certainly more representative of our Target populations than this is normally the case in pre-clinical animal research but when it comes to um the outcome variables or the test systems that we're using to measure Animal Welfare then things get slightly burst and of course one reason for this is that animal welfare is an extremely complex construct I'm a welfare scientists are still struggling with finding common ground on what they mean by animal welfare and while some strictly consider measures of health and biological function others use Behavior normal behavior and deviations from normal behavior to measure animal welfare and yet others argue that the only thing that is of interest to animal in terms of Animal Welfare is how animals feel and so we should use measures about what and how animals are feeling now of course even those who are using measures of health or behavior implicitly they use these measures also as proxy measures of animals feelings and trying um to use indirect measures of well-being by measuring things that can be objectively measured because of course we all know that subjective feelings can't really be measured objectively so we always need proxy measures and this is I think where a lot of the trouble with the measures that we're using comes from but recently uh Liz Paul Mike Mendel Christian Nichol and colleagues from Bristol they have made an effort to figure out whether different measures of welfare actually converge and so they uh studied Chicken on the different condition and measured a range of measures using measures of preference using measure measures of judgment bias and also using more traditional candidate welfare indicators such as measures of stress and then explored or assessed whether these masses converge and as you can read on the left an excerpt from their abstract that overall the different approaches did not converge to identify a precise state of Animal Welfare indicating that all those different measures that are used by Animal Welfare scientists to actually measure states of welfare they seem to measure different things and so they concluded that further work is needed to establish which alternative measures of effective State might be more appropriate indicators of Animal Welfare now one problem that may contribute to this is that animal welfare scientists often use rather subtle interventions probably because many of us shy away from torturing animals we don't want to be nasty to animals we want to be kind to animals that's why we're Animal Welfare scientists but this may actually cause problems now even if I don't know whether this actually was a problem in this particular study simply by showing you that the two conditions that were meant to induce um poor welfare and good welfare on the one hand the good welfare state which is conditions that the animals generally prefer um is for laying hens a relatively large pen with a proper dust bath with a proper nest for egg laying with elevated perches and enough space for these animals while the generally non-preferred condition was a smaller pen without the dust bath without an appropriate nest for nestling and with only a small perch however if you compare these conditions to the conditions even actually good conditions for for laying hands in in Aviary systems you can see that there are many more animals uh in these in these pens and many more animals per square meter and the conditions are probably uh much worse than uh in both of these conditions and so the question is well are we really able to model poor welfare states versus good welfare states if we're using conditions that are so far off to the good side of Animal Welfare compared to the conditions under which these animals are normally kept okay so to conclude from this I do think that we are in many ways better off compared to uh traditional pre-clinical laboratory animal research in that our study population and study settings are usually much more representative of our Target populations and Target conditions but still um we have issues with with our outcome variables um that may not always be properly validated and so it's difficult to assess how good research is if we do not know how good the measures that we're using to measure welfare are and a problem that the interventions that we're using um to validate these measures are often so subtle that it may Simply Be unlikely to find major differences if we do not use more discrete treatments a second major issue that there's certainly also an issue um in in animal research in general but also in in Animal Welfare science is that we're often using uh sample sizes that are too small and so lack statistical power here are data from a recent systematic review and meta-analysis across many studies in the neuroscientists and neuroscientists and and Metabolism research and they found that most animal experiments are severely underpowered so in this case they um analyzed almost 500 studies with more than 2 500 effect sizes and as you can see the median power achieved across all of these studies is just 18 percent so normally we're aiming for a statistical power of about 80 percent 18 means stand less than one in five studies was sufficiently powered to detect an effect um even if an effect was there and that's certainly not a good situation now one of the reasons for this is probably that um in most cases people don't really do proper power analysis or proper sample size calculations and this you can see here on the right side of the picture where um the numbers of the numbers of experiments are displayed and the how many of those experiments used how many animals um in their studies and as you can see the largest number of experiments used 10 animals per group so the median group size is 10 animals per treatment group and I think this this is probably um not due to the fact that they all that all these samp size calculations arrived at a sample size of 10 animals per group to achieve proper power but this is just by convenience and by tradition researchers tend to use groups of 10 animals maybe sometimes 12 maybe sometimes eight per group but this is just not this just does not reflect proper sample size calculations and therefore many studies may end up using too few animals um than they should so how is that in Animal Welfare science I mean of course as with all animal science there is a certain pressure to use as few animals as possible that's part of the three hours principles but still we're not meant to use too few animals because if we're using too few animals then our findings may be meaningless and these animals how few they ever are may have been spent or wasted on inclusive inconclusive research now this is a recent systematic review and meta-analysis on all the different studies that have been conducted using judgment bias as a measure of welfare in animals and as you can see here there were many studies across many different animal species and and this is uh all these results that you can see here are meta-analysis across studies um using these different species now you can see that all of the um um summary effect sizes of the species-specific meta-analysis they all include zero meaning that none of them found a significant summary effect size if you combine them all together and do a summary of calculate the summary effect size across all of those studies then you will find an effect size of 0.2 in terms of hsg which is a small effect size and it's barely significant now these authors concluded that in individual empirical studies comparing two means to achieve a power of 0.8 at an alpha of 0.05 it would be necessary to have sample sizes of at least 50 animals per group for detecting moderate effect sizes and by moderate they mean an effect size of 0.4 which is twice as large as the one that they found across all of these studies and this is certainly not the number of animals that are usually included in Animal Welfare science projects so either we need measures that are more sensitive to the treatments that we're using or we would have to use many more animals in order to achieve sufficient power another major issue that is channeled across research is the analytical flexibility that we have as you know in contrast to clinical research and clinical trials we do not have to pre-register our study protocols and so we can adjust our analysis and and the decisions which results to publish in the end we can postpone this until we have seen our results and can then decide on which results we find most interesting and of course this may lead to biases in in the literature and one of the problems is key hacking P hacking meaning that when you have multiple outcome measures and when you've when you can divide your study sample into multiple subgroups you often have many possibilities for doing tests statistical tests and the problem is that if you do not correct for multiple testing then with an increasing number of statistical tests the the the chance is to find at least one significant result by chance increases dramatically as you can see here in the graph on the left side with only about 10 to 15 um statistical tests you are already above 50 chances to find at least one significant result simply by chance another problem is harking when the hypothesis um so harking is short for hypothesizing after the results are known when you adjust your hypothesis at the results after you've seen the results and again this leads um to a bias uh in favor of of false positive findings um and we have been looking for evidence on harking and we haven't really found any empirical evidence on harking but then in the course of a systematic review that Mariana Ross a PhD student in our lab conducted on um behavioral tests of anxiety in mice we found some outcome measures that had really extremely wide distributions of of the outcome variable to both sides um as you can see here um an example of the effect of Diazepam an anxiolytic truck that does not only have anxiolytic properties but also sedative properties on Locomotion in an open field test so when we saw results like these we wanted well how can it be that um the outcome of a study is so widely spread and yet the authors tended to interpret their findings um either in terms of an anxiolytic effect or a sedative effect of the drug and so we looked into this a bit more closely and what we can see here is data based on a systematic review including 151 studies testing for the effect of Diazepam on Locomotion in a field test now all the effects that are colored in blue were reported as the drug had an anxiolytic effect so these are the reported effects found in the results section while those who found a decrease in Locomotion here the negative numbers those who found a decrease in Locomotion they interpreted the effect of diaspam in terms of a sensitive effect interestingly there were some that found that even an increase in Locomotion was evidence for a sedative effect while some um in even interpreted um a negative effect on Locomotion as an anxiolytic effect but what is much more interesting and puzzling and in our view is can only be explained by harking is the fact that these effects have been predicted by these scientists so when we looked at what the researchers predicted in the introduction or in the method sections of their papers those who found an anxiolytic effect so an increase in Locomotion in the open field they also predicted the drug to have an anxiolytic effect in most cases while those who found a depressive or a sedative effect they already predicted that in the introduction or method section and interestingly those who didn't predict anything they didn't find an effect on either side of the scale and I don't know if you have a better explanation but we really did not find one other than that these predictions have made have been made post-hoc after the researchers had seen these results and then of course there are many other risks of bias um we all know um the commands of good research practice what we learn as students in undergraduate courses when we design studies that we should randomize our study animals to the treatment groups that those who treat the animals and those who assess the outcomes should be blind to treatment that outcome variable inclusion exclusion criteria should be defined a priori rather than after you've seen the results so that's all standard knowledge and to scientists but there is some evidence that many scientists don't seem to adhere to these commands of good research practice it's only indirect evidence again based on meta research and this is an example of a systematic review on pre-clinical animal studies of stroke and what they did they basically just um analyze papers for whether or not these measures to prevent risks of bias were reported in the method sections or not and as you can see those who reported for example that they used blinding and that outcomes assessed outcome assessors were blind to treatment they found much smaller treatment effects on average than those who did not report blinding in the method section so even though we don't know whether they simply didn't report but actually did blind um This is highly unlikely given that those who did not report uh blinding did find such um larger effects um of of their drugs compared to those who did report that they use blinding so it seems that scientific rigor reduces um the effect size is discovered in science and of course this has implications um for the evidence landscape so whenever we do systematic reviews meta-analysis of the evidence then of course it is important to know under what conditions these studies were conducted and it's very likely that due to the fact that often these measures to prevent bias are ignored and not adhered to that the effects that we find in the literature are substantially overestimating the true treatment effects this is also the reason or what instigated um the nc3rs in the UK to start developing reporting guidelines for animal research because in so many papers you didn't find anything written about whether or not these measures against risk suppliers were used and so that's where the the arrive guidelines for animal research come from that educated or reminded scientists what they should report in their papers and of course the aim was not just that scientists will report these but they would actually take these into account when designing and conducting their experiments but I think um Adrian Smith was right um the head of norikopa in Norway that um in principle what we need are design guidelines rather than reporting guidelines if we want to make sure that study Protocols are up to speed and so they developed a so-called prepare guidelines which they now promote by stating that you should prepare before you arrive foreign now how about risks of bias in Animal Welfare science this really is an area where we don't have much empirical evidence but there is some and this paper um from 2014 a group of Belgian researchers um they studied Observer Bias in animal behavior research in particular where the blinding has an effect on um on the results that we find or ex or expectations if we know the treatments that animals have been exposed to if that affects um what we're what we're observing and they did some interest interesting studies and here is one result where they um shared the same videos twice to a group of people so they trained observers to score positive and negative um social behavior and then they showed or they gave them videos which um they gave them in two versions where they slightly changed the light and the illumination but it was essentially the same videos scored twice by the same Observers but one time they told them um that um one group were control um animals while the other group were animals that were bred for high social breeding value meaning that they are particularly um friendly animals amongst each other and as you can see here um that those videos that were labeled as animals with high social breeding value they got higher scores on the side of positive social behavior and lower scores on the side of negative social behavior compared to the control videos even though it was exactly the same videos that were scored by these people and I think this is a good example demonstrating that unknowingly and unconsciously we're just affected by knowing um what treatment animals have been exposed to and the expectations that we associate with this contained our perception or our tendency to score one or the other if in doubt now often in particular animal behavior scientists and also Animal Welfare scientists they often complain that blinding is difficult you know you use several persons to collaborate it's often complicated sometimes it's even impossible because you can see what treatments the animals are exposed to but at least in those cases where it is possible to be blinded I think we should all take the effort to do it and here is a very recent study that um is has only recently been published as a as an uncorrected proof in plus biology by Natasha carbon colleagues also in collaboration with the nc3rs in the UK where they did a qualitative study of the barriers to using blinding and in Vivo experiments and suggests improvements and all those who are skeptical towards blinding and whether that's feasible in the context of their research I recommend this publication highly and the last point that I want to um go into some detail is the generalizability that we achieve with our study population and this of course is a particular issue in laboratory animal science where the animals are usually bread and house to be highly standardized so they're usually genetically identical and they are identical in terms of the conditions under which they are bred and housed and this rigorous standardization may actually compromise the generalizability of the results and because independent replicate studies are always different to some extent in the conditions under which they are conducted this lack of generalizability may also compromise the replicability of study findings now the interesting thing here is that while all those other causes of poor reproducibility that I've been talking about so far are actually considered as or are known to be bad for scientific validity and replicability while on the other hand rigorous standardization is recommended in many textbooks as the thing to do when you want to achieve highly precise and replicable study findings so in order to convince you that standardization is actually a cause of rather than a cure for poor reproducibility I'll go back in history sometime to the birth of reproducibility as a key principle to establish scientific evidence and I was back in the 17th century when the Royal Society in England decided that from now on in order for an observation or a scientific finding to be accepted as scientific evidence it had to be replicated independently by an independent scientist now of course we've come a long way since then and nowadays we're not asking researchers to conduct independent replicate studies before they are allowed to publish their findings but what we do instead is that instead of just observing one animal or one cage of animals of course we're using a sample of animals assuming that this sample the individual animals within this sample that they constitute our independent replicates and therefore produce um findings that are reasonably robust and replicable but of course if you look at this picture it is evident that these animals are a lot less independent of each other than animals would be an independent replicate studies so these animals they usually share the same genotype they share the same environment they are handled by the same personnel and the experiment is conducted on them by the same people so they have a lot of shared genotype and shared environment and therefore um are much more similar to each other than animals from independent replicate studies and I think the first time that this became that scientists became aware of the issues with this was a study conducted by a group of the of Behavioral geneticists in three different Laboratories um where they compared different strains of laboratory mice in a range of different behavioral tests and here is just one behavioral outcome measure um from an elevated plus mace test the standard anxiety test and as you can see there are some really robust differences between some strains here as for example between this h a strain and the c57 black six strain where you can see that there is some variation across the three Laboratories but in all three laborities there is a substantial difference between these two strains on the other hand the ones that I've that I've colored here you can see that in one laboratory in Albany the difference goes in is significant in One Direction in Portland it was significant in the opposite direction and in Edmonton there was no significant difference between these two strains so three fundamentally different conclusions drawn from exactly the same experiment and these researchers they went to Great length to make sure that the conditions across these three Labs were as rigorously standardized as is possible how about anyway I put the word despite here in inverted commas and this was actually the point that I was making after I had seen this paper namely that the reason why they did find such laboratory specific effects was not despite the rigorous standardization but because of it because of course there are things that you can standardize across laborities but there are also many things that you simply cannot standardize like the Personnel that is interacting with the animals the different smells in the laboratory the noise cape and so forth so there are always things that will differ between Laboratories and the more you standardize the conditions within Laboratories the more these differences between Laboratories will affect the results that you get from such an experiment so that gave birth to the standardization fallacy which I first formulated in a in a little letter to the editor um pointing out that actually standardization is um maybe a cause of poor replicability rather than a cure and since then I've conducted with my lab and with my colleagues a lot of studies on this um we've recently published a statistical account of the standardization policy we've done simulation studies on real data from pre-clinical research showing that if you combine multiple studies into virtual multi-laboratory studies that the results of these studies are much more similar and much better replicable compared to single laboratory studies where you find huge variation across across studies and we've also conducted a workshop with experts from different fields of research and came up with a perspectives paper on um suggestions how um we might actually incorporate biological variation into our study design rather than um excluding it by standardization without needing to use more animals for our studies so how about standardization in Animal Welfare science is that really an issue um there's certainly good news that um our study animals are usually more diverse than the typical laboratory Mouse I will study statings are usually less standardized but still um there and everybody who has um studied animals in different pens in different barns barns on different Farms they know that there are benefits Barn effects farm and study effects and they can sometimes be very large and they certainly need to be taken into account and I completely agree with Christian navrot and Lawrence keegox who recently um um stated in a in a paper on replicability in or replicability issues in Animal Welfare science that studies in farm animal welfare science can take advantage of the heterogeneous Farm settings that can serve as the population for sampling and ingly follow a framework that improves external validity and thereby also increases replicability so it's a bit like with clinical trials clinical scientists who often were jealous of the laboratory animal scientists because Liberty animal scientists they are in control of everything and they can control all the conditions um of their animals as well as their genotype but actually it's an advantage if you want to obtain robust and and rigorous results then of course you have to address the biological variation that is there anyway and otherwise you will just um um produce case studies typical end of one studies where a single genotype is studied under a single environment condition and you cannot expect that this will produce robust and replicable findings so if you look at all these different causes of poor reproducibility that I've been talking about here we can actually group them into the three different types of validity that are important for the scientific validity of our research and that is construct validity internal validity and externability and from this I was tempted to um propose the so-called three weeks principle of course in analogy to the three R's because whenever we conduct a harm benefit analysis on on the response whether an animal experiment is responsible we of course have to apply the three hours principle in order to minimize the harm in posts on the animals before we actually assess whether the the benefit of the research outweighs the Harms and I would argue that we should also more systematically assess to what extent uh these three ways construct validity internal religion external validity um are guaranteed to maximize the scientific validity of the research and and avoid wasting animals for inconclusive research I'm very happy that Switzerland has taken this up and in our newest version of the application form for animal experiments researchers now have to detail why they think their study Protocols are suitable to answer the questions that they're asking by giving some thought to the construct validity internal validity and external validity as well as reproducibility of their expected findings but there is more uh to responsible animal research than simply um scientific validity and implementation of the three hours um you all know that we have entered an era of open science and I would strongly argue that only open animal research can be responsible animal research here are some data from comparing studies that have either been pre-registered so where the study protocol has been pre-registered before the study has been conducted in comparison with the traditional literature and what is shown here is the proportion of null findings so the proportion of results that were not statistically significant and as you can see in those studies that were pre-registered the proportion of null findings is much much higher than in the traditional literature suggesting that there is a massive publication bias against negative non-significant results in the non-pre-registered traditional literature and of course if the negative findings don't get published then we have a massive publication bias and and every systematic review meta-analysis will be massively biased towards positive findings rather than true findings besides pre-registration um some you of you may have heard about Fair data management plans make sure that your data are findable accessible interoperable and reusable by other scientists um use pre-prints pre-print service to get your results out as soon as possible follow reporting guidelines and study protocol guidelines and make sure that you publish Open Access because as I said I do believe that only open research can be can ultimately be responsible animal research so with this I would like to thank you for your attention and acknowledge some of my lab members who have contributed massively to those data from my own lab that I've presented and some of my long-standing collaborators with whom I have worked on these issues for many years by now and of course those who have given me money and thought that at least my research questions were relevant to be studied so I'm