Skip to Content

The Mouse Trap

If you haven’t seen it, this series by Daniel Engber at Slate, on the use of the mouse as a laboratory workhorse, is excellent. (And I’m not just saying that because he references some of my disparaging comments about xenograft models, although that did give me a chance to teach my kids what the word “acerbic” means).
He has a lot of good points, which will resonate with people who do research (and inform those who don’t). For example, writing on the ubiquity of C57 black mice, he asks:

So one dark-brown lab mouse came to stand in for every other lab mouse, just as the inbred lab mouse came to stand in for every other rodent, and the rodent came to stand in for dogs and cats and rabbits and rhesus monkeys, the standard models that themselves stood in for all Animalia. But where is Black-6 taking us? How much can we learn from a single mouse?

A lot – but enough? That’s always the background question with animal models. My take has long been that they’re tricky, not always reliable, and still, infuriatingly, essential. The problem is that even things like xenograft models are terrible only on the absolute scale. On the relative scale – compared to all the other animal models for new oncology drugs – they’re pretty good. And compared to not putting your drugs into an animal at all before going to humans, well. . .

19 comments on “The Mouse Trap”

  1. johnnyboy says:

    could you reference some of those past disparaging comments of yours on xenografts ? Just curious.

  2. Anonymous says:

    I’m a cancer researcher who follows this stuff pretty closely. I have always wondered if alot of the “miracle drugs” which supposedly work in mice but not in humans might just be human-specific compounds. That is, they kill a human tumor but not the mouse it is growing in because it is human-specific, not because it is cancer specific. Historically there have not been many human non-cancer cell lines to test cancer-specificity against (though this is getting better now). This could be a common, yet artifactual, result.

  3. luysii says:

    Animal models of human disease are usually profoundly different. Take stroke. Stroke is the third leading cause of death in the USA. A useful therapy for it would be great (we don’t really have one — forget TPA if you’ve heard of it — the evidence for it is extremely unimpressive, academic cheerleading for it to the contrary). Funding for animal models for stroke and various therapies has been enormous. I treated tons of strokes as a neurologist and tried various therapies as they came out. None worked.
    The following article is old but authoritative. [ Stroke vol. 21 p. 1 – 3 ’90 ] Of 25 different compounds of proven efficacy for treating focal and global brain ischemia (e.g. stroke) in animal models over the past 10 years based on published articles in refereed journals, NONE has proven efficacious in clinical trials in man, nor are any in general use today.
    By ’95 I stopped reading the animal stroke literature. By the time I retired 11 years ago, the number of failed trials of treatments of stroke that had worked in animals was 65 (as I recall). If there were a great new treatment for stroke we’d have all heard of it by now.
    This really isn’t too surprising, as one of the largest differences between us an other animals is our brains.

  4. Anonymous says:

    #2 continued:
    And, if they are human specific, they would likely show toxicity in phase 1. Then, a low dose would be chosen for phase 2 which was not toxic, but was (of course) also totally ineffective.
    My guess – this has happened with >10 anticancer compounds that failed in the clinic.

  5. bacillus says:

    This is partly the result of 30 years of capitulation to the animal welfare movement. We have ended up using mice in favour of every other animal model for political reasons rather than for scientific reasons. Moreover, the same movement has driven animal care costs through the roof, so mice is all most of us can afford. It strikes me as obscene that we in the West spend more on the comfort of lab mice than we spend on the healthcare of kids in the developing world. Ironically, the latter are sometimes supposed to be the primary beneficiaries of our cozy and costly mouse research.

  6. TX says:

    Mouse models are OK as long as one is aware that the molecular research conducted on them is structural in its nature and has little if any predictive value (the supposed gold standard of any scientific endeavour) outside the lab where one has to take into account annoying aspects such as biological variation (both genetical and phenotypical) and differences in environment. This is an interesting question within toxicology, where people want to use transgenes specifically sensitive to carcinogenes etc. The moral choice to use a supersensitized animals for screening (anything that doesn’t pop up there must be safe…) or simply a way of priming your system to see the results you want to see?

  7. RKN says:

    Based on the first article in the series, the only one I’ve read, the conclusion is clear: mouse models are not predictive. Yes, there are many practical reasons for using mice. Yes, you can’t get your drug into humans w/o showing efficacy in the mouse. Yes, mouse and human genomes have a lot of overlap (tho I was pleased to read that any more this fact is more “rhetorical than scientific”). Yes, mouse experiments get you published, and may keep some academics funded.
    But in terms of the mouse being useful model of human disease, the gig is up, they’re not worthwhile, they don’t work.

  8. lynn says:

    Mouse models still OK for antibacterial drug efficacy and PK/PD modeling.

  9. Vince says:

    #RKN: Based on the first article in the series, the only one I’ve read, the conclusion is clear: mouse models are not predictive
    In my opinion, this is the wrong conclusion to walk away from the topic with.
    Mouse models are hugely beneficial. As properly implicated tools, they have advanced human health dramatically and are indispensable. Take this from someone who has/does work on a non-mouse animal that’s outlined in the piece.
    The problem is with the application of the model, of utilizing it right, of getting the right mapping between model and reality. If they diverge too greatly, your results can not be expected to cross-over. Unfortunately, this seems to be growing problem in the field.
    This is hardly a problem just in the biomedical fields, or just in mouse models of pathology. It’s seen across the sciences, from geology to climatology, to the financial world where just recently well designed risk and leverage models fell apart when applied to the dynamic economy, and beyond.

  10. Vince says:

    #3 luysii: Animal models of human disease are usually profoundly different. Take stroke.
    I’ve seen you make a similar post before and really enjoyed your perspective. I’m in full agreement about the lack of usable output we’ve created (tPA is shit), but wonder about the reason.
    We’ve either targeted vascular solutions or dove right for ionotropic solutions (ie. MK801). While the early chelator findings of calciums importance in the signalling cascade is clear; I do wonder if we’re overlooking a wide swath of the system space that we’re just *seeing* due to shitty models not covering it.

  11. ABBA says:

    You are spot on. Politics has royally screwed this up too. Vocal minorities will undo civilization, and that is what they want.
    I met a wacko last week who claimed that not only should animals have rights, but plants should too. If you push these people far enough, you realize that they hate themselves. They think the world would be better without humans.

  12. Anonymous says:

    All inbred animal models are hyper-sensitive – they are almost pre-destined to give a false positive result – and the clinical data over the last 20 years has unfortunately proved this.
    The only data that predicts positive clinical success for a new mechanism is human data. Animals are still useful (though not perfect) for PK/PD and safety studies – they help you design the best molecule for clinical testing. However, unless you have got quality human genetic, pharmacologic or human cell/tissue phenotypic data then you are going into a PhII study with no data that predicts a successful outcome and effectively taken a $25-50MM (depending on your full costs to get there) bet blind. In this case the chances of success are less than 1:100 and is one of reasons why the industry is in its current state.
    Get the human data before putting chemistry on a project.

  13. pete says:

    Jeepers. To read most of the comments above you’d think that all of the gene knockout results ever observed in mice are utterly irrelevant to human physiology. I beg to differ.

  14. NoMoreStatistics says:

    someone posted about stroke earlier, a simple disorder to avoid, both kinds. Now here comes the ‘crackpottery’. Most blood vessel leaks are caused by a lack of copper vitamin k and vitamin d, leading to improper repair and hardening through calcification. You can look up the profound effects of high dose k2 on calcification, the dangerous effects of copper deficiency, sytemically on the vascular system, including aneurysms and cardiac hypertrophy. Ischemic strokes are caused by related mechanism with ‘furring’ which loves to collect cholesterol ( there’s a multibillion dollar red herring), and lastly check out a recent study on 10mg folic acid supplements (if you know anything about folate and are marginally rational, it should alarm you) with rather unexpected results.
    Mouse models aren’t useless but check out some the differences in ld50’s for a start. They are literally and metaphorically a different animal.
    Mice dont’t produce cortisol the way higher mammals and primates do, that right there is going to skew the metabolism of so many compounds in so many ways both subtle and gross, how many other aspects are different? The VDR is different in function as well, how ubiquitous are those two things alone. Modelling (yes the kind that involves silicon,big sigh) is the eventual answer, until then keep on truckin’.
    I thought the genome was going to save the world, can we stop fumbling in the dark already and admit we need an ‘Appollo’ like interdisciplinary project even to get a basic handle on the sort of chaos we’re talking about.

  15. Virgil says:

    Interesting article but his numbers are off a bit (probably to make the article more populist)…
    “…a basic lab mouse can be bought for about $5 and maintained on a nickel a day…”
    Sorry, I’m gonna call BS on that one! A C57BL6 from Jax/Harlan/CRL costs closer to $20 when you include shipping and handling charges, and per-diems at most universities run from 40-50c upwards.
    Another big variable he really doesn’t get into is the quality of care between different institutions. Things like noise levels in the vivarium can have a huge effect on animal health. Go grab a handful of aging studies from PubMed, and look at the lifespan of the control mice – anywhere from 2.5 to 4.5 years depending on which institution the work was doen at.

  16. “infuriatingly essential”…how true. As someone who’s worked a lot with lab mice, I can really appreciate your comments in this post. Thanks Derek.

  17. zmil says:

    He’s not talking about BL6s. According the accompanying slide show, you can get a CD-1 mouse for $4.85. Of course BL6 is probably what most researches think of when they hear ‘basic lab mouse’. It’s all I’ve ever dealt with, at least, in my very limited experience.

  18. Morten G says:

    @8 lynn:
    Read the article. There’s pages worth on why mice are crap models for TB (which is a bacterial infection).
    It would seem prudent in drug development to check if the drugs for the disease that work in humans actually work in your mouse model.

Comments are closed.