Im currently working on a meta-analysis of randomized controlled trials (RCTs).

These RCTs have reported their outcomes at different time points.

The maximum follow-up points were as follows: 1 months, 3 months, 6 months, and 48 months.

To pool the studies without double counting, I have decided to include the outcome data of the longest follow-up period of each study only.

My problem is: one of the studies maximum follow-up period was 48 months, which is considered an outlier in this case. This study also reports outcomes at 6, 12, and 24 months.

My first question:

Can I include the data of the shortest follow up-period data of this study instead, which is 6 months? My intention is to only make the data homogenous.

My second question:

I decided to do subgrouping for the studies based on intervention techniques to test for subgroup differences. Is it okay if one subgroup contained a single study? (no study was double-counted)


1 Answer 1


Trials usually have different lengths of intervention and follow-up. The question on homogeneity in this case is a clinical one. For example, in a previous SR/ MA on hydroxyethyl starch, trials with shorter follow-ups showed sporadic results while trials with a longer follow-up showed almost identical results. This was theorized to be a result of the long-term harms associated with the intervention.

In your case, the outlier might be the most accurate trial since it presents the longest follow-up. So I wouldn't be quick to exclude the longest follow-up from the analysis just because it is the only trial reporting it.

As for having a single trial in the subgroup analysis, there is nothing wrong with that. In essence you are reporting the results of that trial since it is not being meta-analyzed with any other trials. Having said that, since it does provide data, it can be used to detect between subgroup differences. As such, again I would not exclude it for the noted reason (i.e. only one trial in the analysis).

  • $\begingroup$ Thank you Dr. Okay this solves my issue in this outcome. However, I have another outcome that contain only two studies. One study has a maximum follow up of 6 months, and the other one is that same study that report data at 48 months. There is no reason to choose the data of 6 months from the second study here as well right? $\endgroup$ Jun 29, 2022 at 10:56
  • $\begingroup$ Agree. Unless there is a reason to pick the 6 months over the 48 months, I would stick with the latter. $\endgroup$
    – abousetta
    Jun 29, 2022 at 11:21
  • $\begingroup$ Im really sorry to bother you even further, but I performed the two analyses. One choosing 6 months, and another one choosing 48 months. The latter I2 value was 95%, whereas the former was 39%. Additionally, there was a significant difference between the subgroups based on intervention technique in the second analysis. Is this a good reason to choose the first analysis over the second? $\endgroup$ Jun 29, 2022 at 11:36
  • 1
    $\begingroup$ @GhassanSaeed it is always much better to choose based on some theory rather than on the basis of some feature of the data. $\endgroup$
    – mdewey
    Jun 29, 2022 at 15:32

Your Answer

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

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