Virtual IACUC: Reduction vs. Refinement

April 21, 2009

A comment on a previous post alleged that the scientific enterprise has not taken the 3Rs (Reduction, Replacement and Refinement) seriously, leading to a failure to reduce the number of animals used in research. Subsequent comments from Paul Browne and Luigi provided links to actual data which refute this claim, however it remains an interesting question to explore.
One of the thornier problems in thinking about the justification of using animals is when two or more laudable goals call for opposing solutions. For today’s edition of virtual IACUC we will consider what to do when Refinement calls for the use of more animals, in obvious conflict with Reduction.

Although the concept of refining animal research designs can have many aspects, one of the more important parts is the use of techniques which minimize acute stress, distress, etc to the individual subject animal.
There is a very general operating principle which underlies this decision making, namely that within the scope of justified experimental procedures less is better than more. For example, while a single major *surgery (say to implant an intravenous catheter for drug self-administration) might be justified, a second major surgery for a given animal requires extraordinary justification and may very well be refused.
Now, let us suppose that there is a non-zero failure rate for catheters in the course of a typical drug self-administration experiment. Suppose that you have an implanted rat, trained to self-administer cocaine, participating in the study, otherwise healthy when….the catheter fails. Since the animal has not completed the study design it cannot participate in the data analysis. That’s bad enough but consider that appropriate groups sizes are necessary to generate the gold standard statistically reliable result- which is essentially a requirement for good interpretation of the outcome, publication, etc. If you can’t publish it, it didn’t happen. So it may be the case that the loss of one subject blows the entire experiment, potentially obviating the use of all of the other animals on the study.
Let us consider the decisions the investigator and her IACUC might make to ensure that the majority of animals’ participation is not wasted.
First, you might just say fine, take your chances with blowing the entire study. A loss of a single subject does not inevitably happen for each group and does not inevitably ruin the analysis. Just re-run the study with a new group when and if the loss of a subject or two because of catheter failure blows the whole thing. Of course, you might just say this is all about probabilities and numbers. How frequently is a study blown, what are the group sizes and how many extra animals are required in the course of a year’s worth of experiments? I am uncomfortable with this alternative, however, because by design it dictates that in some cases the full **participation of some subjects will be wasted, simply because another member of the group lost a catheter.
Second, the design may use a excess of animals at the start, relative to what is anticipated to be needed to satisfy the demands of statistical power in the final data analysis. So instead of the 10 that are needed, the investigator always starts with 12 and performs all the surgical and behavioral procedures in all of the animals. The drawback to this approach is, of course, that since a thing like catheter failure cannot be predicted specifically (and if it could, the investigator could probably identify systematic reasons and fix the procedures accordingly) it is likely that there will be a chronic excess of animals in all studies which may or may not be needed. Whether justified or not, animal numbers increase.
Third, it is possible in many cases to surgically repair or replace the catheter on an individual basis as-needed. This, however, requires a second major surgery on an individual which is considered to require a high threshold. This is where we really get down to the difficult call. How important is minimizing the stress to a single animal on experimental need versus systematically lowering the total number of animals required for a given experimental purpose? This is what I would like you to discuss in the comments, DearReader. How would you balance these factors? What decision would you favor if sitting on an IACUC?
How do we balance the permissible procedures in one animal against the need to use additional animals?
*I use this example because it is fairly easy to grasp why less would be better than more, not to be sensational. The niceties of a repeated-measures design versus a between-groups design, statistical power, group size are fascinating but perhaps a bit complicated for a general audience.
**It is not a surface consideration of IACUC decision making but I do think there should be a goal to make each subject’s participation in research count as much as possible. With participation in a published data figure as the gold standard of what counts. Yes, there will always be pilot studies and supporting studies that go unreported but I think that we should always strive for the published figure.

No Responses Yet to “Virtual IACUC: Reduction vs. Refinement”

  1. Luigi Says:

    As Spock said:
    “The good of the many outweighs the good of the few, or the one”
    I think this would also be the general view of any IACUC. Insufficient statistical power is a show stopper when it comes to animal protocols. As you say — if the results aren’t usable, all those animals will have been used for nothing. And that’s unacceptable. Suffering, in contrast, is a fuzzy concept that merely requires a reasonable effort toward minimization.
    I’d use the extra couple animals. And I’d advise anyone else to do the same.


  2. Paul Browne Says:

    I’m with Luigi on this, if a study isn’t worth doing properly it isn’t worth doing…god knows I’ve read enough underpowered studies (clinical and animal) in my time to have very little patience for them.
    I’ve also known colleagues who have had to repeat studies because an unexpected problem took the number below that required for a valid result, not an acceptable outcome in my book.


  3. Pain Man Says:

    I can only tell you what I do.
    First of all, the difference between n=10 and n=12 is trivial, whether you’re talking power analysis or, IMO, rodent suffering. You keep running new guys until you get an asterisk (or the power you want) or give up, is the law of the land.
    Second of all, everybody does second veins. I’ve worked in two separate IVSA labs that ran over 100 animals/day. Not doing a second vein is not really an option. It would be a waste of a perfectly good subject. Doing one surgery is ethical, but not two? What about dual IV and intracranial surgeries, are those unethical?
    If I were shooting for an n=10, I’d order 18 animals, and here’s what would happen to those 18:
    3 would fail to acquire operant self-administration for sucrose pellets.
    1 would die on the operating table.
    1 would die the day after surgery.
    2 would fail to acquire cocaine self-administration.
    1 would OD on coke.
    10 would complete the study IF it’s a relatively short study. For IVSA that’s something like less than 2 months. Longer than that and you might start losing catheter patency. Sure some people are lazy and just sac the animal, and some aren’t skilled enough to catheterize a collapsed vein. But I’m pretty sure recath is in our animal protocols. And if you need more animals (for n, for asterisk, for power, whatever), you order more animals and start from step one.


  4. Pain Man Says:

    For IVSA that’s something like


  5. danae Says:

    After working in Tox & PK studies, I know how easily expensive studies can be ruined by all the abovementioned failures. Then it’s a real shame to see animals wasted and to have to do it all over again. Sometimes it’s as simple as a batch of faulty VAPs, some kinked lines, inexperienced techs unable to deal with patency issues with fussy catheters, etc. You need to account for possible, and more importantly probable, issues that inevitably come with working with live organisms and technology/drugs.
    I also think some species deal with repeated use better than others. I’m all for letting controls serve again, or making the most of any critter that is in good shape as a candidate for reuse, but some I think are better off with a one time only usage. Rabbits, in particular, unless you treat them almost as house pets, just don’t adapt well to research life and display increasing distress and abnormal physiological data over time. I really do question their use as models overall, but specifically, they don’t approximate “normal” physiology in my book. I say, if it must be rabbits, once is enough. Rats seem to be much hardier & more readily adapatable, more robust, less fragile.
    I don’t think “reduce and refine” are goal in and of themselves if the quality of data and quality of animal models must decline. Better to do it once & have ample data then cut corners & get inconclusive, n = too small data pools.


  6. qaz Says:

    This debate is missing the point of reduce and refine.
    Reduce is about reducing the number of animals WHILE STILL DOING GOOD SCIENCE. If you had a better technology you might not lose half (8/18) of your animals. If you had a technology that allowed you to get twice as much data from each animal (think multi-electrode recording vs. single-cell-recording) or a technique that allows within-animal instead of between-animal statistics, then you could reduce the number of animals you need.
    Refining is about decreasing the pain and suffering of animals in your care. Surgery was once done without anesthesia (many many years ago). No IACUC would approve that today. In fact, today, we do surgery not only with sufficient anesthesia but also with sufficient analgesia. Refining is about providing emotional and intellectual stimulation to social animals that need it.
    Rather than argue defensively about why we can’t reduce and refine, let’s give specific examples in which we are reducing and refining. I think it would help the whole animal rights debate if scientists would state up front that they care about animals, that they ask themselves every time if its worth it, and that scientists who aren’t studying pain and stress don’t want an animal in pain or stress because that will wreck their data. (Scientists who are studying pain and stress need to keep that pain and stress under careful control and don’t want any additional pain and stress beyond what they’re studying.)


  7. Luigi Says:

    DM provided a pretty specific scenario, so I’m not sure we missed the point. But you’re right — there is a larger issue that could (and should) be addressed. And all your points are well made.
    For the vast middle ground who think some use of animals in biomedical research is acceptable but that use and suffering should be minimized, I offer what others have already mentioned many times:
    ‘Reduce and refine’ is GOOD SCIENCE. Even a complete psychopath who loves nothing more than tearing the heads off puppies should be interested in using fewer animals, and minimizing their suffering. Because it’s GOOD SCIENCE. Within-animal statistics (c.f. paired tests) are more powerful than between animal statistics, both mathematically and because the necessary experimental design removes many potential sources of variability. Reduced animal usage is also cost effective. Per diem animal care is expensive! Supplies and tech time is expensive! And, as already pointed out by qaz, refinement mostly means making sure you’re really studying the treatment you want, instead of the side-effects of the procedure, or the effects of the procedure under irrelevant circumstances. In another thread, someone mentioned how rabbits tend to suffer more than rats under the same conditions. Obviously, unless you want to inadvertently study the interactions between stress and your treatment, the answer to the scientific need is reduced suffering.
    So biomedical scientists and people with concern for animal welfare are by and large all on the same team. We’re all interested in the same things. None of us want to abuse animals, and none of us are even really enthusiastic about their use. I’d personally love to sit in front of a computer all day running simulations instead of using real animals. It’d be way cheaper and I’d definitely sleep better at night. But we’re not there yet, not even close. In fact, we’re so far away that I can’t even imagine it happening. Current computer models of physiology and cell lines are laughably irrelevant to real intact animal responses. Ironically, the only way we can fix our understanding and develop non-animal tools is through more animal research. Before we can get answers without animals, we first have to figure out how animals work. Biology is still a very young science.


  8. DrugMonkey Says:

    PainMan, thanks for the additional details. Especially in essentially pointing out that all three scenarios that I outlined may be on the table all at the same time. It is never an either/or decision for the whole institution, all the time. We may fill in extra subjects at times, start with larger groups at times and justify additional surgical procedures at times. I was certainly trying to simplify this complicated balance in my example, but you are right that reality is much more complicated.
    I do have some comments on a few of your points, however.
    Doing one surgery is ethical, but not two? What about dual IV and intracranial surgeries, are those unethical?
    Sure some people are lazy and just sac the animal, and some aren’t skilled enough to catheterize a collapsed vein. But I’m pretty sure recath is in our animal protocols.
    The point that needs to be fully recognized is that we are not talking about some absolute decision on ethical/non-ethical procedures. The way the IACUC process works is that each use of animals has to be justified and just because, say, a multiple-surgery protocol is approved for one purpose that it will be approved for all purposes.
    In large labs such as you describe, this leads to no end of problems if everyone from techs to trainees to PIs do not know what is on each and every protocol. You very likely are drawing your experience from studies conducted under more than one protocol, probably several of them concurrently. Have you read them?
    It may very well be the case that the “lazy” people are actually following the requirements of the protocol under which their work is covered. Or are following the spirit of the overall process even if they are technically permitted to do a second surgery. For example it may be that an IACUC may permit the catheter repair if it is relatively rare. If they notice in their oversight role that 80% of the subjects are undergoing a second surgery this changes their decision making and for sure make the PI look deceptive if the original protocol estimated 5% or fewer. So the PI may set her own limits on how frequently additional surgical procedures should be used.
    A given researcher not being skilled enough on a catheter repair or replacement is actually an important point. Another part of the ethical use of animals (and an explicit part of IACUC oversight activities) is to make sure that staff conducting procedures are appropriately trained to do so. If they are not sufficiently trained or skilled, it is unethical for them to do the procedure.


  9. Luigi Says:

    It’s important to realize that your description sounds horrifying to animal welfare advocates. Your username doesn’t help much.
    I’m not saying we need to whitewash any truths. But it’s important to remember, as DM says, that your description is not exactly a model example for animal use.
    Technically, you can’t even take an animal out of the cage to see what color it’s belly is without a protocol in place. And if your approved protocol says you can do it at noon, then 12:01 is too late — you’ll have to wait until tomorrow. And officially there is no excuse for laziness or imperfect training. So being ‘pretty sure’ and allowing for mistakes isn’t really cool.
    That said, the level of regulation and expertise I imply above is simply unrealistic, even silly. Thus, certain ‘procedures’ have blanket permission, and protocols tend to generally be a bit vague for good reason. Ultimately it’s up to the researchers and IACUC to use good judgement. That’s an important trust that needs to be recognized and respected. We can’t rely on rules, either as guidance or as an excuse. There needs to be a moral compass somewhere in there.
    So at the end of your post, it might have been worthwhile to note that, before ordering more animals, you all first sit down and discuss best ideas for avoiding the 3 animals deaths or whether any existing data can be salvaged.
    I actually applaud animal rights advocates, and think what they do is important. Of course, blowing up cars, threatening researchers with harm, and other hijinks along those lines are a different matter. At the very least, those criminals need to remember that people are animals too.
    But reminders to always try to do better are good.


  10. DrugMonkey Says:

    PainMan… your description is not exactly a model example for animal use.
    It isn’t necessarily accurate either. The reasons for losing subjects from experiments are all valid and happen regularly. They just don’t all happen at once to each and every experiment. So the notion that you start with 18 each time to end up with 10 is a bit overdrawn.
    It is also most emphatically the case that we do not rest on our laurels, accept a certain rate of error and add animals accordingly. If problems arise, most scientists try to determine if there are systematic causes and adjust their procedures (and designs in many cases) accordingly to minimize subject loss.


  11. @UCLAMouseMan Says:

    One issue that is often overlooked is refinement for reduction purposes. Generally this is not a big issue in areas other than mice and rats, but it merits mentioning. Some people have a habit of running ahead with procedures at which they have not become fully proficient. Due to the difficulty of tracking individual proficiencies, the IACUC tends to not pay attention beyond what classes someone has taken in the past and there’s little to no attention paid to how well they conduct their procedures after. The end result is that someone causing distress to or loss of animals for lack of proficiency with their procedure of choice is often not noticed unless things start going disasterously wrong. I’ve even had one investigator become indignant when I said she was doing something wrong (and teaching someone else the wrong technique in the process) and offered to help her do it right.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: