|
Sean_Collins -> Re: Lets review some literature. (February 23, 2006 3:32:00 AM)
|
While not in my area of clinical expertise (which happens to be cardiovascular and pulmonary pt, rehab, and physiology in general); I do also teach and have a lot of experience with research methods.
I found this article refreshing from a methodological perspective in many ways. Even in areas where issues can arise, the authors are perhaps even harder on themselves than most critiques might be (in regard to something like selection bias because they could not "randomly" sample). Random sampling, while the only way to be sure you are not obtaining a biased sample in some way is almost a foolish concept for clinical research, depending on the question being asked of course.
For example - this paper did not randomly sample from the full population of people meeting their inclusion and not meeting their exclusion criteria, but rather sampled from those at a particular hospital, during a particular time period, and of only those that actually presented to a hospital (not that sought other treatments for the same pain syndrome).
But, they are interested in specifically answering the question of which technique is better for the patients that do "self select" to visit the ER, so in many ways having a random sample from the full variance of people with neck pain would not be helpful since the question starts with "people that have the pain AND decide to visit the ER for it."
So, I guess I am saying they were able to do this with their design.
They comment on whether they could "blind the subjects" or the physicians providing care. I dont think such a methodological process is possible to answer this question. The only blinding that I think could have happened is that of analysis. So the analyst only knows that these two groups are different - but not how they are different.
In terms of placebo - would have been possible but certainly would have reduced statistical power given the small sample size. I do not agree with their assessment of why placebo is not possible with manual techniques - if there is a "powerful psychological pacebo effect of the laying of hands" then this is exactly what you are trying to rule out as an alternative cause of the intervention response.
The sample size is soooo important, and it is small because of the rigor of the inclusion and exclusion criteria for eliminating alternative explanations. The designers of this study knew their limitations in terms of sample size and designed an appropriate study based on that limitation - I commend them for this.
One thing I would add - which makes clinical trials weak (in my opinion) is that they are based entirely on this concept of "randomization" and "chance" that attempts to make clinicians ignore the clinical reasoning that is not removed from actual clinical practice. For example - since we have to randomize patients in a trial like this - we remove the possibliity that the physician might have decided that a particular patient would have been better off for their particular neck pain to recieve manipulation and others meds. The only good thing of this is that when you find a significant difference you are more accepting of an actual cause - effect since this "watering down" (that is created by having some people randomized into meds that might have been better off for some unaccounted for reason recieving manipulation, or vice versa) creates a bias to the null - in other words makes it harder to find a statistically significant effect.
Finally - the ONLY thing I would have changed is the analysis. In addition to the "stratification" approach utilized very effectively I think a regression or factorial anova approach that allows combining several of the questions they asked by dividing the sample into several different strata. By looking at a few strata at a time with a regression or factorial model they could have tested for interactions that might be very useful to know.
Before I write more - I guess I am interested in what other people thought.
Best -
Sean
|
|
|
|