Does this compound extend the model organism's lifespan?


I am testing an experimental drug on snails in a small pilot study. The compound has the desired effect in vivo that we are looking for. As an interesting side note, I have noticed that the only snails still living now, four months later, all received the experimental compound. All snails that received a different compound have died, as well as controls which received nothing. Snails only have a lifespan of 12-16 months when kept at room temperature, and since they were full grown when we got them I assume the snails that have died have done so normally, dying of old age.

I have no background in probability and wanted to see if the experts here can tell me how to determine what the chances are that this is an authentic phenomenon rather than coincidence. I apologize for the small sample size in this experiment.

The experiment was as follows-

Snail 1 Injected with saline
Snail 2 Injected with 2mg/kg compound X
Snail 3 Injected with 10mg/kg compound X
Snail 4 Injected with 20mg/kg compound X

Snail 5 Injected with saline
Snail 6 Injected with 2mg/kg compound Y
Snail 7 Injected with 10mg/kg compound Y
Snail 8 Injected with 20mg/kg compound Y

Snail 9 No injection

Snails 2,3,4 are still living and healthy. All others have died. Thank you so much for any assistance!
Thanks for your response! The snails were separated randomly into treatment groups by weight. They were all approx. the same size, so I assume they were of similar ages, but that is just speculation.

To get an incredibly rough estimate, could I consider the outcome, living or dying, as a sort of heads or tails scenario? It seems the chances that I would get what I desire, heads (living) on only the ones that I want and all tails (dead) for all the other treatments could be calculated at (1/2)^9 or 1 in 512. I know an organism living or dying isn't such a heads or tails case, but I'd just like to know how rare it is for something like this to occur in general.


New Member
Subtle Guest,

It is always useful to brainstorm on all possible interpretations of data, often it make you think of things you were not expecting:

Some things to think about, instead of compound X helping could Y be killing the snails, could saline be killing the snails....

I guess when you say you randomized them I guess you mean each group had about the same average weight?

One way you could look at it:

If you hypothesize that all the treatments have insignificant affect on the avg lifespan (but 3 lived significantly longer because, even with snails, life is sometimes like that) then you ask how unusual is it for the the snails that lived significantly longer to happen to be the ones who got preparation X. I think you can model this with asking what is the probability of picking 3 blue balls out of a barrel of 3 blue balls and 6 white balls (randomized of course): this is: 3/9 X 2/8 X 1/7 = 0.01 or 1 out of 100; this is pretty unusual (3/9 bec you have 3 chances out of 9 on your first draw, 2/8 bec you only have 2 blue balls left in the remaining 8 balls, etc). But, as you say, the sample size is small and like I poked at, there maybe other things going on.

So how long do you think I will live if I eat snails given preparation X? :)



New Member
Another way to analyze it

Subtle Guest,

Here is another way to analyze the problem:

The data can be put into a contingency table and analyzed with chi-square. Each category, however, would have to have an expected value of 5 or greater; which you do not have, even if you combine some categories. However, the rule does not apply to 2X2 contingency tables; therefore, if the data is combined to make a 2X2 cont. table then it can be analyzed.

Therefore, a cont. table can be made as follows:

(Comp. X) (Non-Comp. X)
Lived long 3 0
Lived short 0 6

Sorry, I'm not able to make a nice table in this email; hopefully you see the 2X2 table above.

Now, there is a formula that gives the exact probablity that these two distributions (comp. X and Non-comp. X) are the same.

I plugged the numbers in and got a P of 0.01 (ie the same as the probablity calculation I gave above; they should give the same number, and it is nice to know that when you do it 2 different ways you get the same number).

I got the formula from: Practical Statistics by Russell Langley.

The nice thing about analyzing the data with chi-square is that this is the same way you would analyze it, if in the future you had more data so you could use more than 2 categories (actually, you would use chi-square but it would not be the formula used above).