Hi,
What I am looking for is an experimental design for the scenario where a drug is given to randomized cohorts of patients and rather than measuring the average outcome for each case, I want to measure the percentage of patients who recovered compared to the percentage who recover in the control group. The rationale is that in a drug trial some patients with, for example, the right genetic make-up, will benefit from the drug while others without the genetic marker won't. If the percentage who benefit is low, the averaging of the results could mask the benefit to the few patients who can benefit. However, there is still a need to determine whether the difference in percentage of those who appear to benefit could have occurred by chance.
Any suggestions?
From a stats point of view:
If you have a "degree of response" measure then you can group the patients into 2 groups - no response or some response. Although this is a crude measure you can then use parametric stats to assess the degree of response while excluding the non-responders. You then need only work out why you get a response for some.
Best would be to use regression with multiple factors to identify which of those factors have a significant effect on the response. If a particular gene is required then there will be a low or strong negative factor for one genotype with a strong or positive effect for the other genotype.
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology
- Pharmaceuticals
- Drugs