NIH policy on SABV and realistic review

May 06 2016 Published by under NIH, NIH funding

My prediction is that the grants that will do best in the next few rounds are those that successfully excuse themselves from including both male and female subjects. 

The grants that try to respond to the spirit of the new NIH SABV initiative will get comparatively hammered in review. 

29 responses so far

  • NeuroMaverick says:

    So is there a best way to successfully excuse myself from including both male and female subjects?

  • drugmonkey says:

    It's going to be section by section, and model system by model system, in terms of what they see as a good excuse.

  • NeuroMaverick says:

    I was afraid you were going to say that!

  • drugmonkey says:

    I only speak the truth.

  • Dave says:

    Best excuse is that prelim data shows male and females are the same, so only need to study one or the other. Riiiiight?????

  • Draino says:

    My transgene is on the Y-chromosome, therefore I will only study male mice.

  • Dave says:

    But what if you put it on the X chromosome?

  • Emaderton3 says:

    Just switch to breast cancer research . . . or maybe prostate cancer :)~

  • Grumble says:

    If you're worried about this, find someone who actually examines SABV as a main focus of their research, and ask them what they think would be acceptable for a grant in your field. They have been asked this a lot, have thought about it already, and (time permitting) can help you formulate a plan for the grant that won't get it automatically shot down.

    If you need to identify stages of the estrous cycle in rodents, find someone who knows how to do this to write a letter saying they'll help you.

  • drugmonkey says:

    So SABV means "turn it into a sex-differences grant" to you Grumble?

  • Dr Becca says:

    Best excuse is that prelim data shows male and females are the same, so only need to study one or the other. Riiiiight?????

    If your prelim data already show that males and females are the same, there's no reason not to include both in your subject pool, with no further need to power to detect sex differences.

  • Grumble says:

    Is there any way to avoid at least some element of that? (In cases where you don't have a good excuse, like studying breast cancer or whatever.)

  • drugmonkey says:

    Is there any way to avoid at least some element of that?

    I don't believe so. Unless there are very explicit instructions otherwise. There are not any such explicit instructions from NIH that I have seen. OTOH, I have seen weasel wordage that says "this policy doesn't necessarily turn everything into a sex-differences study". I think NIH is creating a lot of chaos here that could be avoided.

  • drugmonkey says:

    If your prelim data already show that males and females are the same, there's no reason not to include both in your subject pool, with no further need to power to detect sex differences.

    So. Let's suppose I run a preliminary study on the main variable of interest with full-N male and female groups. And I find no statistical difference. Does this mean that I run all subsequent studies as half-N M and half-N F?

  • drugmonkey says:

    Follow up question: Suppose your preliminary data show no sex differences but this sort of disagrees with the few papers that have been published showing a sex difference on exactly that experiment...or a closely related one. Maybe those prior results were publication bias and all of the failures to find sex differences weren't published. Or maybe you are introducing some weird factor because of your lack of experience with the other sex, how is the typical study section person to know?

  • Grumble says:

    "And I find no statistical difference."

    You might find the F subjects have more variability if they experiments were done on random days of the estrous cycle. That wouldn't necessarily show up in a simple test comparing M and F.

    "Does this mean that I run all subsequent studies as half-N M and half-N F?"

    Or can you choose all M or all F, because there's no difference? It's probably safer to SAY it will be 50/50 M/F and then DO whatever the hell you want.

    "maybe you are introducing some weird factor because of your lack of experience with the other sex,"

    Hence my suggestion to include an expert collaborator. Diffuses a lot of this kind of bullshit criticism. *Especially* if your preliminary data (which, presumably, your collaborator helped you get) goes against established results.

  • Dr Becca says:

    @DM - Yes, that's exactly right. Half and half. Why is that hard? The point is that in the end, you can say your effect is true for both males and females because you did your experiment in both. At its most basic, the goal of SABV is to get female subjects represented in experimental outcomes.

    For your second question, how is this different from any other instance of data that go against something that was previously published? This happens all the time, in every field. Do what you would normally do if you got a contradictory result - speculate, name some ways you could follow up to flesh it out, and move on.

  • Dr Becca says:

    @Grumble, the widely-held idea that female rodents are more variable in experimental outcomes because of the estrous cycle is not accurate. Monitoring it still might be useful and interesting in some cases, but overall, variability is not greater in females than in males.

    http://www.sciencedirect.com/science/article/pii/S0149763414000049

  • drugmonkey says:

    I cannot think of any half and half studies in off the shelf laboratory rodents in my fields of interest, Dr becca. You are blowing my mind with this.

  • Amboceptor says:

    If your prelim data already show that males and females are the same, there's no reason not to include both in your subject pool, with no further need to power to detect sex differences.

    There is sort of a reason. Throughout my career we usually used female mice, because they are smaller, so it's cheaper to give smaller mice a drug measured mg/kg, and also because they were easier to handle, less biting the researcher and less biting each other and messing up each other's health. And if you are using males and females of the same age, you're averaging mice that weigh 18g with mice that weigh 24g or so.... the weight loss curves are not the same and your statistics will not look as good. If labs have to use as many male mice as females, it will be inconvenient.

  • jmz4 says:

    "I cannot think of any half and half studies in off the shelf laboratory rodents in my fields of interest, Dr becca."
    Lifespan assays in mice usually use both. Although this is probably cause female mice generally get a better boost from longevity treatments and so they make the data look good.
    So all proposals going forward need to account for this? That'll be fun to watch people wiggle around.

  • Grumble says:

    " the widely-held idea that female rodents are more variable in experimental outcomes because of the estrous cycle is not accurate. "

    It's not accurate when examining a whole boatload of diverse studies that looked at different dependent variables, but that doesn't mean that YOUR particular variable won't be affected. To take the argument to the logical extreme, you wouldn't use the literature analysis study you linked to to make the argument that estrogen levels are no more variable in females than males. Until you do an actual experiment, you won't KNOW if your female subjects are more variable.

    That's not an excuse for not using female subjects, but it is a reason to pay attention to variability in females vs males.

  • drugmonkey says:

    Females were less variable in my first SABV compliant experiment. Maybe this will continue, maybe not. Is this now a second hypothesis to test? About variability as well as any effect on main variables of interest?

  • drugmonkey says:

    Jmz4- that isn't a half and half question. That is a full power for each sex issue.

  • Dr Becca says:

    DM, see my recently accepted paper.

  • drugmonkey says:

    Did you get any static from reviewers?

  • Dr Becca says:

    None at all. We also show in the figures which data points are male and which are female so readers can see for themselves.

  • Dr Becca says:

    That's not an excuse for not using female subjects, but it is a reason to pay attention to variability in females vs males.

    I never said not to pay attention to it, just that the risk of it is overstated and often used as an excuse not to study females. You should always be looking at the variability in your experimental groups regardless of their sex, in the event that you might need to use non-parametric tests in your stats.

    Is this now a second hypothesis to test? About variability as well as any effect on main variables of interest?

    Only if it interests you. I happen to find within group variability extremely interesting, and my lab has done well by diving a little deeper into these "hmmm" data sets.

    The mean ain't everything, folks.

  • Cynthia McMurray, PhD says:

    It is a good idea to test males and females, but NIH imposes this requirement without doubling the funds needed to compete such as analysis with power. Investigators are put in an impossible position that does not meet the spirit of the NIH new requirement.

    Unless there is prior knowledge, I would suggest either allowing a single sex analysis as a beginning and transition to sex 2 in a supplement, or increasing the budget allowable for a well-powered analysis

Leave a Reply