1
$\begingroup$

i have data from a "negative" clinical trial and would like to run an analysis to see if there is a certain group of patients (theoretically) that respond to the drug that was otherwise overlooked. What is the best way to go about doing this? The sample size is not too big (109) so power will likely be an issue.

$\endgroup$
  • 1
    $\begingroup$ Can you edit your post to clarify whether this is a group identified from theory or by fishing in the data? $\endgroup$ – mdewey Mar 7 '17 at 16:34
  • $\begingroup$ Just go ahead but make it clear in the writeup that this is a secondary analysis. $\endgroup$ – mdewey Mar 7 '17 at 21:32
1
$\begingroup$

Your reference to "a certain group of patients" [emphasis mine] suggests to me that maybe you are starting with a definite hypothesis that is informed by pathophysiology or maybe observational research. In this case, your approach should not attempt to "see if there is" evidence for the hypothesized response, but rather to attempt to falsify your hypothesis.

Thus, you ought to consider all the ways this data set (and others!) might falsify, constrain or reshape your hypothesis. If you think of yourself as a blacksmith, you ought to be using the data to hammer on your hypothesis, and not vice versa!

$\endgroup$
  • $\begingroup$ Thank you for your response.Would an interaction between the treatment and the subgroup of interest (yes or no) be appropriate? $\endgroup$ – Lucy Mar 8 '17 at 0:41
  • $\begingroup$ Very little good advice can be offered to you absent a more specific description of your theory and the data you've collected. By what kind of measures do you identify your 'subgroup'? Are these clinical measurements that have the usual, continuous character of most such measures? If so, then in focusing on a 'subgroup', you may have implicitly dichotomized a continuous variable – a great sin! (I ask only because 'dichotomania' is pandemic throughout the medical literature, in spite of the sustained efforts of Frank Harrell and others to get this disease under control.) $\endgroup$ – David C. Norris Mar 8 '17 at 1:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.