1
$\begingroup$

Rather odd question. I'm a surgical resident who is currently a research fellow in a lab studying fetal surgery. Our plan is to test a glue which we believe will help seal a uterus after we open it to operate. Our plan is to use rabbits, which have multiple pups gestating, separated in 2 "horns". The plan is to do same fetal surgery on both sides of the rabbit surgery, and only use the glue on one side.

I'm taking over for another researcher who has since left the lab, who came up with an estimate of 10 rabbits needed. He apparently used a z-test to calculate this, and to be honest I don't entirely follow his math or logic. During our renewal, IACUC (the animal ethics guys) apparently had a stats guy who got our renewal denied for using said z-test, because it shouldn't be used for small numbers. Of course, said stats guy did not explain what test we should use. Any help would be much appreciated-

Original text in the application (previously approved, now denied) is below.

To determine the number of animals required to demonstrate safety and efficacy of each formulation, we used the UCSF Clinical and Translational Science Institute (CTSI) sample size calculator for analytic studies (comparing the means of a continuous measurement, a z-statistic was used to approximate the t-statistic): http://www.sample-size.net/sample-size-means/. We used the standard type I error rate of 0.05 and type II error rate of 0.1 (for 90% power). Our expected effect size and standard deviation for input in the calculator were generated from our previous experimental rabbit model, with standard deviation 0.002 and effect size 0.003. These parameters generated a total necessary sample size of 20 fetuses for each formulation. Because the rabbit uterus is bicornuate, one side will serve as the control and the other will be used for application of the experimental glue. Rabbit litter size generally varies from 4-8 per gestation (2-4 fetuses per side). Thus, each rabbit may be used for experimental interventions on an average 3 fetuses, depending on the number of fetuses present, leaving an equal number of sacs as controls. Thus, 7 animals per group would be required. However, there is a known mortality associated with maternal laparotomy and hysterotomy from previous work in this field of 20% (Gratacos et al, 1999). Furthermore, there is the possibility that some rabbits in the groups may only have 2 fetuses available for intervention. Thus, we are requesting 10 animals per study group.

$\endgroup$
1
$\begingroup$

Since the same animal is used both for treatment and control, you could use a paired t test. I'm surprised the stats guy didn't mention the pairing ... A paired design will often give better power than an unpaired one, so that possibility of pairing was fortunate.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.