2
$\begingroup$

Im doing a meta-analysis of randomized interventional trials to compare between two types of surgeries.

The follow up period at which outcomes were reported varied between 6 weeks to 24 months. Some studies reported their results at different follow up points, whereas others reported results at one follow up point.

I want to ask about the most appropriate way to pool the studies together in this situation. Should I pool the data of the longest follow up available or should I rather subgroup the studies each study data according to the follow up point and report the pooled effect of each subgroup?

I performed both analyses. However, they were not matching. The first analysis revealed a significant difference between the two arms, while the second showed no significant difference in all subgroups.

The I2 of both analyses was 0%.

Improve this question
$\endgroup$
1
  • $\begingroup$ If you have multiple effect sizes from some studies you might like to investigate multi-level meta-analysis which will let you include everything in one analysis. $\endgroup$
    – mdewey
    Commented Apr 20, 2023 at 12:32

1 Answer 1

Reset to default
1
$\begingroup$

Whether you pool the studies with differing follow-ups or perform a subgroup analysis is dependent on reasoning within the subject matter. If you believe that the outcome differs meaningfully between a short follow-up and a long follow-up, you shouldn't pool these studies together, but instead perform subgroup analyses. If you believe the effect of the exposure on the outcome is the same, regardless of your range of follow-ups, you could pool the results. What you believe (or experienced individuals in the field believe) happens to the outcome at different follow-ups impacts whether you pool or not.

You can also perform a meta-regression to determine the effect of follow-up on the variation in the estimates of the trials.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thanks! The time factor is important to determine the effect based on the literature. I decided to subgroup based on follow up at which outcomes were reported to have reliable results. Now If I want to explore other causes of heterogeneity such as age, what results should I subgroup? I thought I should be stick to the results of the follow up period that is most reliable to see the true effect. However, this was not pre-specified in the protocol? Is it a good approach? or is it better to use the results of the largest follow up in all studies? $\endgroup$
    – Ghassan Saeed
    Commented Apr 25, 2023 at 14:27
  • $\begingroup$ @GhassanSaeed Continuously subgrouping analyses may result in many strata. Instead, you could do a meta-regression to determine the effect of variables on the differences between studies. Regarding the results, I would suggest to try and reason what is most likely to represent the true effect. The most reliable period does not seem like a bad choice to me. Given that this was not pre-specified, it is not bad to mention in the final report that this was decided post-hoc. Choosing the largest follow-up is not always the best answer, as a long follow-up might introduce noise in the outcome. $\endgroup$
    – rjjanse
    Commented Apr 25, 2023 at 14:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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