Survival Analysis - how to deal with replicates

Buckyball

New Member
Hi,

I hope someone will be able to help me with my problem, which probably is plain simple.

I work on immunity experiment with fruit flies and have pools of 10 flies in a vial which are getting the same treatment with 4 replicates (plus another 4 batches as controls).

So I have 4 vials:

Group1, Group2, Group3 and Group4

and monitor the survival of each.

So far so good.

What would be the best way for the sruvial analysis (I'm using Graphpad Prism)? I can certainly do a survival analysis for each group and generating 4 sruvival curves but that seems kind of odd to me.

From my feeling I should combine all 4 groups into one. Would that be right approach.

Thanks for any input!

hlsmith

Omega Contributor

Also, can you better describe this, I am confused.

Buckyball

New Member
At the end the experiment ist divided into multiple subgroups with various differnt pathogen treatments. We want then compare survival and see if our hypothesis about certain immune responses is true.

Right now I'm only talking about one treatment, lets call it A, and generation of the survival curve. So we setup 4 vials with 10 flies in each vial and every of the 40 flies gets the same treatment with A and then I monitor mortallity.

To make it easy, lets assume there is only one dead event a day, so for two vials that could look like:

Vial 1: 2,3,5,6,8,9,10,11,15,22
Vial 2: 1,4,6,7,8,12,14,18,19,19

So I could make two survival curves or pool the data and generate just one survival curve. The latter seems to be more logical to me or is this wrong?

hlsmith

Omega Contributor
n = 40 flies
m = 4 vials, so 10 flies in each vial
Treatment the exact same for each vial

So no plausibly differences between any of the flies and vials, correct. Time, batch, sampling, etc. will not affect survival? You can pool all of the data together, sure. If there is any possibly reason for survival to be conditional on which vial, then you should investigate that, by entering all data into model along with a categorical variable (M, which vial). However, given the sample sizes, I would wonder if you may have statistical power concerns for vial comparisons.

Buckyball

New Member

So no plausibly differences between any of the flies and vials, correct. Time, batch, sampling, etc. will not affect survival?
All this is very well controlled and should not affect survival in any way.

hlsmith

Omega Contributor
You are probably fine grouping them. Otherwise it would have been kind of like a meta-analysis where you have to control for random-effects.

I don't know your content, but do you know the flies' life expectancy without treatment, just living in a vial.

Also, I would as an outsider wonder whether once a couple flies die it changes the dynamics (environment of the other flies) in the vial, some type of ecological effect.