I suggest that you speak with Randy Carter, Department of Biostatistics at your University or one of the other biostatistics faculty at U. of Buffalo. Solid department there and they can advise. Take your data they will need to see the measurements. Also, keep in mind that mouse models for vaccine protection do not always correlate with vaccine candidates for human use. Different MHC background thus different epitopes presented.
Currently we want to test the protective capabilities of new antigens using a pneumococcal sepsis model (IP injected).
I suggest that you speak with Randy Carter, Department of Biostatistics at your University or one of the other biostatistics faculty at U. of Buffalo. Solid department there and they can advise. Take your data they will need to see the measurements. Also, keep in mind that mouse models for vaccine protection do not always correlate with vaccine candidates for human use. Different MHC background thus different epitopes presented.
Mouse vaccination results can be quite different than human or non-human primate results. That said, I think that you are needing to design the initial experiment, rather than dealing with data already collected. You will need to defend the number to IACUC.
Simple T test values give decent results for 4-6 animals, and 6 is a not uncommon number to use in primate studies. But that only works if everything goes right. If an animal, for instance, escapes, or if an injection is not done correctly and one or two animals don't respond, your study is junk. I've seen that in a primate study, which was quite sad. A lot of money, time and effort down the drain, to say nothing of problems with getting further grants.
Generally, more is better, and mice are cheap. You should acquaint yourself with statistics. Take a class, it's well worth it. You won't be able to have a proper conversation with your biostats folks without basic background in significance.
This should have registered right away, but ti just percolated through that you are using a septic model.
It is VERY important that you always keep in mind that rodent model of sepsis does NOT apply to humans. Humans are unique. Our siglec-13, which is the control siglec for TLR4, is mutated out. It doesn't work. As a consequence, humans are 3,000 times more sensitive to LPS than mice. The LPS LD50 for an average human (whole body, not per kg) is only double the LD50 for a 25 gram mouse.
Monkeys and chimpanzees have working siglec-13. So they are also highly resistant to sepsis. They aren't as tolerant as rodents, but compared to us, it hardly matters. Deadly human doses of LPS won't cause a flicker in a chimpanzee or a rhesus macaque.
So very careful about any animal results for sepsis when going to humans. It has killed patients because the animal lab scientists didn't communicate with the physicians, didn't know, and so the septic effect wasn't expected. Many physicians are similarly unaware, and most physicians will NOT know how to treat sterile-induced sepsis. (That is sepsis induced by injection of LPS (aka endotoxin) that is not infectious.)
Based on type of test and statistical analysis the minimum number of animals required for the test are fixed. Generally in potency tests by challenge methods they recommend 16-20 animals. For other potency tests like antibody estimation the number is less .For safety the requirement is different.
It also depends on the genetic background of your mice. There tends to be less variation between individuals when using inbred mice like BALB/c. Outbreak mice like MF1 tend to vary more in their vaccine response. You will need to do a power calculation to work out group size, and so you will need an idea of the standard deviation of the result you will be measuring. A statistician will be able to help you do this. I'd also recommend a mixed cage model so that your experimental unit is each mouse, rather than each cage.
Kirsty that is an interesting approach, would you mind expanding on it?
A mixed cage model involves caging (randomly selected) mice together and giving each one a different treatment from amongst those on offer as part of your experiment. The 'experimental unit' is defined as "the smallest experimental unit to which the treatment can be independently applied ". In a mixed cage of 5 mice, each one would get a different treatment so the individual mice become the experimental unit. If you group the mice by treatment instead into cages of 5 then they will be subject to pseudoreplication. The experimental unit would then be the cage rather than each individual mouse and you'd be left with n=1 rather than 5. I normally repeat the model twice. Hope this helps!