This is a clear case for repeated measures ANOVA or a mixed effects model. Volume is the outcome, timepoint is a within-subjects predictor (which should probably be treated as continuous or ordinal, rather than nominal), and group is a between-subjects predictor. With such a small sample, though, it is very likely that you won't be able to make very meaningful inferences. Personally I think, especially for a small dataset like this, simply plotting the full dataset tells you more than a p-value does:

A mixed model is a better approach to analyze such data. Use a mixed model. You can find associations even with such a small data with a mixed model.

Yes, mixed model, but not with a normal error structure. I would assume a gamma error structure.

Model formula: Conc ~ Time + (1|Mouse)

Error: Conc|Time,Mouse ~ Gamma(a,b)

Fixed effects: Time = continuous (in days)

Random effects: Mouse = categorical (12 levels)

Thank you all for your kind responses. I will look further in those methods that all of you proposed.

Thank you so much once again

I like mixed modle this is what i also think would most definitely represent the data sets, but also would show the difference with just simple old average and standard devation it may not look as significant but there is a clear difference.  Careful to mind mouse age and  sex, you may find varations with respects to drug response.. tumor size.. with respects to growth rate , mitotic division before and after treatment. Late or early stage cancer treatment... and dose responce levels of drug, and dont to forget to label your x and y axies with what time hours, days weeks.. and volume .. like, mm cubed..  

best wishes and good luck!