Chambless and Hollon (1998) discussed some issues regarding follow-up treatment studies (see page 10). Retention seems to be particularly problematic for follow-up studies in that differential rates may bias results. In order to be retained in a sample for a follow-up study, participants must complete treatment and show improvement.
This raises several questions for me.
First: Why must a participant show improvement to be retained in the study? (It is possible that some important concept somewhere along the line just went flying way over my head. However…) Isn’t the goal of these ( at least initially) randomized trials to determine what treatments are effective and which ones are not? True, the primary aim may be to determine if Treatment A is more effective than Treatment B; however, I would think that if Treatment A was effective for X number of people but not for Y number of people, there may be important differences between these groups that would benefit from further study. Additionally, Chambless and Hollon (1998) report on “tantalizing” evidence that suggests that some therapies may have a delayed effect, not showing effects (positive, I presume) until after treatment has been completed. Therefore, if follow-up studies are retaining samples based on improvement, some sort of lag time should be considered before dismissing sub-group Y entirely. Similarly, if we (and by we, I mean some group of researches interested in the effects of Treatment A vs. B) are studying the efficacy of Treatment A (presumably because the efficacy has not yet been established), then how do we know at what point Treatment A is not “working” and will not – at any point – benefit Mr. Y?
Second: Follow-up studies tend to require that participants complete treatment. Sure. That seems reasonable. Why compare the effects of Treatment A with Treatment B if only 75% of Group A actually received Treatment A and only 64% of Group B actually received Treatment B? This is a hairy problem with a sticky answer.
However, failure to complete treatment may be particularly relevant in practice. Patient compliance with any given treatment may be more important than the overall efficacy of the treatment. No matter how well a given treatment works, if a patient does not or will not continue (or even begin) the treatment, little benefit can be gained. Therefore, it may be beneficial to track participant outcomes even after decisions to discontinue treatment. While this may be difficult due to attrition (e.g., participant does not complete treatment due to geographical constraints, diminished interest in study participation, etc.), the effort may prove worthwhile. And although statistical power is likely to be limited, preliminary data may give way to important information pointing to future study and of clinical interest.
1 comment:
Failure to complete treatment is indeed, as you suggest, an interesting and important variable of its own. I know of at least one study that compared therapies on that score. That is, it asked which therapies were more likely to scare clients away. It didn't find that any were more likely than any other, BUT it did find, I believe that patients who are "reactive" (don't like being told what to do) were, no surprise, more likely to drop out with more directive therapy approaches (e.g., CBT). Interesting!
Post a Comment