I need help with meta-analysis

Description- we have two types of medical treatments- A & B across multiple trials. I'm trying to calculate cumulative odds ratio of an adverse event happening after treatment A & B. However, with treatment A, the adverse event is HIGHER after treatment while with treatment B, the adverse event is LOWER after treatment. when I try using revman trying to calculate odds ratio, it would not obviously let me as there can be no negative number. FYI A & B are both treatments - no control.

Goal- to show that treatment A worsens the adverse events compared to treatment B in a meta- analysis/ANY PERTINENT STATISTICAL TEST

I would like to know if there is any other effect size/ measurement that I could calculate so that I can resolve this issue.

rephrasing the question- how to compare two interventions when one treatment improves an outcome while other worsens it?

Thanks in advance VP


enter image description here

  • $\begingroup$ It's not clear what you are asking. In what sense does one treatment improve and the other worsen the event? An odds ratio would seem to apply that all you have is event yes or no. In fact, it's not even clear what trial designs you are looking at. $\endgroup$ – Björn Mar 16 '19 at 8:53
  • $\begingroup$ thanks for the response; I edited the question with a sample data. Please let me know if this makes sense $\endgroup$ – Venkat P Mar 16 '19 at 15:37
  • $\begingroup$ What are those numbers numbers of patients with an event? Events in the same patient? The traditional way to deal with events in a previous time period would be to use a regression model (e.g. logistic regression for y/n outcomes, negative binomial for counts of events when multiple events can happen in a patient), where the pre-treatment number of events is put into the model as a covariate (possibly after a suitable transformation - e.g. logit for logistic regression, log for negative binomial with some number added to zeros to avoid taking the log of zero). $\endgroup$ – Björn Mar 16 '19 at 21:06
  • $\begingroup$ I share @Björn confusion here. If you are referring to odds ratios when you speak of negative numbers you must have made a slip somewhere as OR are always non-negative. Similarly for frequencies. $\endgroup$ – mdewey Mar 17 '19 at 16:22

Your Answer

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

Browse other questions tagged or ask your own question.