For the past decade, biomedical researchers have worried that too many animal experiments can’t be repeated in other labs, or that a potential treatment that doesn’t work when tested in humans turns out to be based on flawed animal research. To address those problems, last week a working group of federal and academic experts advising the National Institutes of Health released ideas about how to shore up the rigor, transparency, and clinical relevance of NIH-funded animal research.
The group’s recommendations take aim at studies using vertebrates and cephalopods, such as squid and octopuses. They include having NIH add a page to its current 12-page grant application form that would ask researchers to describe study details, such as the number of animals they will use and plans for data analysis. Peer reviewers with statistics expertise would then assess the plan.
The panel’s 11 June report also recommends funding statistics training for animal researchers, asking scientists to explain their choice of animal model, boosting support for research on large animals, and encouraging researchers to record and study how factors in animal care—such as temperature, light levels, and an animal’s gut microbes—can sway experimental results.
The report delves gingerly into the issue of preregistration, which involves posting plans in an online registry before animal experiments begin. As with ClinicalTrials.gov for human studies, proponents of preregistration argue it will ensure that researchers don’t change their study design to get a better result or hide negative data. (Another version of this idea, called registered reports, has peer reviewers evaluate the plan.)
But scientific groups warn preregistration could stifle creativity in basic research, tip off competitors to ideas, and make researchers vulnerable to attacks by animal rights activists. Instead of requiring preregistration, the report recommends that NIH start with education about the concept and pilot studies of it.
ScienceInsider recently spoke with working group co-chairs Lawrence Tabak, NIH’s principal deputy director, and California Institute of Technology molecular biologist Barbara Wold. (Their remarks have been edited for brevity and clarity.)
Q: NIH launched a reproducibility effort 7 years ago. Why did you need to look at this again?
Lawrence Tabak: Because of the complexity of dealing with animals in research, it deserves its own special look. But many of the general principles during the initial effort carried through. We’ve amplified certain points and elaborated on others.
Q: Did you begin with the idea of mandating preregistration?
Barbara Wold: We certainly considered that. Where it can matter the most is in the late preclinical stages of animal studies. And even there, we recognized that it’s not clear exactly how well preregistration is going to work, and that this user community isn’t actually educated in even what registered reports are and what preregistration is.
Q: How would pilot studies of preregistration work?
L.T.: Don’t hold us to this, but, for example, if there’s an RFA [request for applications] for a particular program that might lend itself to a pilot, people will know that if they apply for the RFA, they are also signing up to be engaged in this pilot.
Q: Will you have to wait until those projects are published to know whether preregistration improved a study?
B.W.: I don’t know that we would have to wait that long. We would hope to be learning some things about registered reports and preregistration on the scale of 18 months, 2 years from now. Some fields already do this. And I think people have discovered that it’s not quite as onerous as they thought it might be.
Q: Will there be a point at which NIH could require preregistration for certain studies?
B.W.: We’re charging NIH with getting feedback. This really is intended to be a collaborative enterprise between the research community and NIH, and then you decide to go forward in a big way or in a more modest way.
Q: How will you know whether these efforts are making a difference? Would cancer experiments in animals be more reproducible? Would clinical trials have a higher success rate?
B.W.: It would be all of these things. This isn’t flipping a binary switch. This is actually a very complicated affair, where we feel there are numerous places where we’ll get improvement, but we don’t know which approaches are going to give us the biggest improvements and what the costs and savings are going to be.
Q: One implication of these rigor requirements is that researchers will use more animals, right?
B.W.: The necessity of achieving statistical power may call for more animals, but there are also places where we expect to gain real efficiencies. So we’ll probably have less doing little studies that have to be redone.