I heard from colleagues that case studies and group studies are typically not grouped together in the same meta-analysis. Several papers discuss the topic of which type of studies are "combinable" (e.g. Egger's excellent series of methodological papers in BMJ), moreover acknowleding that the criteria according to which studies are vs aren't combinable are subjective; however, this particular aspect (combining group and case studies) is not one I found discussed.

I did find discussions of randomised controlled studies vs observational studies, which concluded that "the temptation to combine the results of studies seems to be hard to resist", and which claims RCTs should be meta-analysed separately from observational studies. But this of course only makes a distinction in the intervention type (there was or there wasn't one, resp.), while saying nothing about the sample size (group N>1, or case-study N=1) aspect of studies.

For the intervention I'm interested in, the literature consists of only few (randomised&controlled) group studies, and many single-case-studies (with N=1 or a little above 1). Is it advisable to combine all these studies together in the same meta-analysis, or are there grounds for making separate a meta-analysis for each study category?

In particular, insofar as (as Egger et al 1997 state)

a simple arithmetic average of the results from all the trials would give misleading results. (..) Methods used for meta-analysis use a weighted average of the results, in which the larger trials have more influence than the smaller ones.

, I all the more don’t understand why combining case- and group-studies is problematic, as long as the former get 'penalised', either on account on having a small N (although one can have a single group larger than each of the groups in a RCT!), or on account of "single group" leading to a decreased study-quality score - right?!

  • 1
    $\begingroup$ Pooling together RCTs, non-RCTs and case reports is hardly justifiable from a clinical perspective. Yet, you may consider compiling all case reports into a single dataset and then analyzing it as a non-RCT. Then it can be pooled together with other non-RCTs, but remember to use multivariable adjusted effect estimates and not crude ones. $\endgroup$ – Joe_74 Feb 20 at 10:36
  • 1
    $\begingroup$ Thanks for the suggestion! Would clinical reports with 2 patients fall under the "case report/case study" rubric as well, or should I only compile into a single dataset reports with strictly N=1? $\endgroup$ – z8080 Feb 20 at 11:06
  • 1
    $\begingroup$ Any case series (even of 10 patients) in which you have individual patient data can be used to create a per-patient dataset. Remember though to add a column to specify the study ID. You will end up doing a part aggregate/part patient-level analysis. $\endgroup$ – Joe_74 Feb 21 at 10:55
  • 1
    $\begingroup$ Thanks, that's helpful. One more question if I may. How do you distinguish between studies that are not RTCs (i.e., don't have a control group), and "case reports"? Since we know from the onset that these will all be intervention studies (i.e. not observational), to me the sole question is whether there was a single group of treated patients (non-RCT) or whether patients were divided in groups (RCT). $\endgroup$ – z8080 Feb 21 at 17:06
  • 1
    $\begingroup$ It is from those very resources that I formed the impression I expressed in my last comment. It seemed though that you had a different definition, that's why I asked how you distinguish between those things. $\endgroup$ – z8080 Feb 22 at 9:50

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.