Mouse longevity studies typically use 100 male and 100 female controls per site per year, and 50 male and 50 female treated mice per drug per site.

“if you have a drug that's authentically slowing aging one of the things you most want to see is that the very oldest animals in this drug-treated group are living longer than the very oldest animals in the untreated control group”

“if we have a good drug um we might have of the mice in that pool population if 80% of them are treated and only 20% are control we know that we have a drug that will uh give you a much better chance of being alive when you're really really old”

“the test statistic we use is we wait until 90% of the mice are dead in both populations the control and the treated population and then on the date when the 90% Mouse dies we say what fraction of the mice are in the treated group of the ones that are alive and what fraction of the mice are in the control group”

“we Define arbitrarily 90th we're going to do it at the 90th percentile all the time yeah so in other words it's sort of pre-specified yeah yeah oh ABS right it's pre-specified it's written down before we do any test in fact it was written down 20 years ago”

“the further you push it up 95 99 99.9 The fewer mice you have to work with and so after a while the statistical power drops you'll miss a lot of stuff if you only have like one or two mice alive at that age we've picked the 90th percentile because there'll still be a few dozen mice alive at that age at the 90th percentile”

“if if we're lucky if we have a good drug um we might have of the mice in that pool population if 80% of them are treated and only 20% or control we know that we have a drug that will ex uh give you a much better chance of being alive when you're really really old”

“at each site we are using 100 male controls and 100 female controls so pooling that's 300 males and 300 females in the control group each year and for each drug we use 50 males and 50 females that is half as many as the controls again at each site”

“we oversample the controls that is we double the size of the control group because we're going to compare controls to drug a controls to drug B controls to drug C so in a sense we're reusing we're getting our best mileage out of the controls since uh we can increase statistical power for every one of those pairwise comparisons by adding extra mice to the control group”

“what we do is a better way which is was worked out by David Allison about 20 years ago um the test statistic we use is we wait until 90% of the mice are dead in both populations the control and the treated population and then on the date when the 90th per Mouse dies we say what fraction of the mice are in the treated group of the ones that are alive and what fraction of the mice are in the control group”