Before someone responds with the answer that I am violating statistical convention, here is my proposition for the use of a single group-under very strict conditions- that can logically be both the experimental and control at different time points in a single occasion treatment/outcome scheme. I have 108 subjects enrolled in a pain study. They will all have physical therapy between 9-10 am on a certain day and pain score assessed at start and end of PT. At 10:15 am same day they will all be exposed to a new therapy (pulsed EMF or PEMF). The exposure outcome (pain score assessment) will then be assessed again. We consider a 3/10 vas drop in pain score to be the desired outcome. We propose to use the same group but measuring two times-when they are unexposed (after PT) and after exposure to PEMF. It seems to me that the variation introduced by creating a control group made up of different patients is simply needlessly introducing error (variation in patient characteristics) into the calculations. Please comment with your thoughts...thank you
I would suggest you measure difference in pain score after PT on each patient ( ie post PT - pre PT), then difference in pain score after EMF ( ie post EMF - pre EMF) and compare the two means of differences. The patient is his/her own control.
you may also want to vary the order in which the PT and EMF are applied.
No need for an external control but you might also consider dummy PT and dummy EMF to control for placebo / investigator psychological effects.
Thanks for responding James-appreciate your input. We are measuring pain scores pre and post PT and using the post PT pain score as the baseline measure or pre-PEMF score. What is actually happening is that patients are getting some pain relief during their PT. Immediately after formal PT we apply the PEMF and we get further pain reductions and it is these pain reductions that we are interested in. In some ways the PEMF related pain reductions are arguably more difficult to achieve as a result of ceiling effects. In the end we want to capture an effect size related only due to PEMF and expressed as an Odds Ratio. It is not our intent to compare or contrast treatment effect sizes for PT vs PEMF. We would like to be able to make a statement such as " by using intervention A , one is 3.3 times more likely to achieve the outcome of interest" which in our case is a 3/10 VAS. I feel that communicating in terms of Odds Ratio's would have meaning to patients, providers and payers since probability is a concept many people are familiar with.
Why not just say x% of patients who have PEMF after PT achieve further pain reductions, while (100-x) do not; so odds of further improvement are x/(100-x). No need for odds ratio..
James
I think I am complicating this analysis unnecessarily. I like your suggestion. We will simply express the results in terms of odds and not odds ratio. Thank you again for helping with this!
- Badges
- Science topic
- Similar topics
- Biological Science
- Physiology
- Physiological Processes
- Stress