I'm working with a colleague on series of experiments treating mice with compound A to see if their condition improves. Baseline behavior tests were administered to determine the difference between WT and diseased mice, and then each genotype was to be separated, 50% for no treatment, 50% for treatment with compound A. My initial reaction was to randomly assign the mice to either treated or not treated based upon drawing mouse numbers out of a hat.
Instead, my colleague analyzed the baseline tests and decided which mice would be treated or not treated based upon these tests in an effort to "ensure accurate representation of all baseline levels". Now, ignoring the fact that I would have rather done this completely blind, and had someone else randomly assign them, is this an acceptable way of deciding on treated vs. untreated mice? I'm not 100% comfortable with it, so I'm curious what the scientific community at large feels.
Personally I believe that the main question that should be asked is, are the groups you are assessing representative of the population at large? If the mice were ranked in terms of baselines, and then alternately grouped (e.g. Group 1: 1,3,5,7, etc and Group 2: 2,4,6,8, etc) then I believe this would be a valid approach. Ideally you want the variation and means in Group 1 to be identical to Group 2. Random assignment is normally sufficient (and often the only method), but by its nature, especially with small N, will often produce one group which is more representative than the other.
Keep in mind that at the end of the experiment you could always compare low baseline to high baseline, as long as they are evenly distributed between treatment groups.
I'd be interested to know what others think on this matter.
The objective is to render the two groups/treatments equivalent (as if belonging to the same population). A priori pairing (based upon some validly measured baseline) followed by random assignment of paired individuals to the groups/treatments would insure their equivalence more than non a priori pairing and purely random assignment, unless the N is quite large.
Nathanael and Jacques, thank you both for your answers. I am used to working with larger populations of subjects, and as such have always utilized random selection. Given that this is a smaller sample size, it's reassuring to hear that there is validity to the method of assignment utilized by my colleague.
Randomization based on baseline is a good idea, but be aware that if you rank them and sort to groups of [group 1=(performance 1,3,5) and group 2=(performance 2,4,6)] then you have built in a bias whereby group 1 will likely start out with higher performance average than group 2. Check statistics of baseline groups before proceeding.
I feel you can do it two ways:
1. Randomly (blindly) assign the mice to treatment and non-treatment group. That assingment will be done by a third person. At the end of study, analysis based on whether treated or untreated mice was WT or diseased could be drawn. That would give results in categories as: WT mice when treated with drug A had this effect, or diseased mice when treated / not treated with drug showed that effect. But the major problem would be the unequal number distribution in each group at the time of analysis. If a particular group had significantly less number of mice, then drawing any statistical conclusion would be difficult, and the conclusions would not be that powerful.
2) Based on baseline data divide mice into two categories, and then randomly assign (blindly) equal number of mice into treated / non treated group in each category. This will reduce the chances of unequal distribution of samples.
All the best...
I see two possible approaches here.
1.Randomly assign the animals and then use the base line data as a covariate in a covariance analysis when you have collected your final data. That will test whether the baseline affects the final measurement and also correct for it.
2. Match the animals on the base line data, then assign matched animals to the treatments at random. For example, if you have two treatment groups (say control and treated) you should look for pairs of individual with the same base line value and assign them at random to the two treatments. This then becomes a randomised block experimental design, and it should be analysed by a two-way analysis of variance without interaction (I am assuming quantitative data). These designs are described more fully in www.3Rs-reduction.co.uk.
Michael
Hi Devin,
I am aware that some psyhcopharmacology studies have done that exactly; i.e. analyzed the effects of stimulant drugs on attentional performance as a function of baseline proficiency. The approach I believe is justified on the basis of functional neurochemistry and behavioral pharmacology whereby low performers typically derive a benefit from a drug whereas higher performers don't or even are impaired following administration of the same drug. A well known example can be found in Granon et al., 2000. See also Puumala et al., 1996.
Hope this help.
Stephane
Stephane and others,
I was just notified about your answer and references, which are quite similar to the study we're running now, which is hugely helpful. I was trained purely as a molecular biochemist, so some of these more behavioral questions are new to me, so I greatly appreciate everyone's input. Now I just have to hope that this particular treatment works in vivo.
Devin
ps - it just occurred to me that you could in practice test compound A according to both approaches discussed above. For instance, in the 1st stage you could give compound A to mice according to their baseline proficiency (use the median or mode to divide up the groups), whereas in a 2nd stage you could give them the same drug according to random allocation of mice to drug treatment AFTER A washout period of X number of days between stages.
Great idea Stephane, although since these mice are going to also be utilized to look at biochemical markers post treatment, I'll have to try it with the next set of mice.