Is there some recommended approach for extracting data for meta-analysis from diagnostic studies use derivation and validation cohort? Some practical advice? I failed to find anything useful.
There may be something useful published by Jonathan Deeks and the Cochrane collaboration. There is a working group on meta-analysis of diagnostic studies.
Darren pointed you to a good starting point. I never gave this much thought, but it is an excellent question. I could argue that it makes more sense to focus on the validation set statistics, since this would represent the case for generalizibility of the diagnostic tool. it also represents the lower end estimates, since the algorithm will inevitably perform worse on the validation sample than on the derivation sample.
Thank you. You always have valuable comments. I read Cochrane Handbook for DTA reviews before, but they are still incomplete with some chapters. I will see if I can find something from Jonathan Deek's publications. Right from the begging I use several of those sources: "Cochrane Handbook for DTA reviews," "Methods Guide for Medical Test Reviews (AHRQ)" and "CRD’s guidance for undertaking reviews in health care" but still I do not find anything on this topic.
Dear Ariel,
Your argument makes a lot of sense. I use derivation cohorts because of bigger samples, but I will try with data from validation cohort and test the differences. Now I have too much heterogeneity. :(
It may be a good idea to run your analysis on both sets. The derivation cohorts will represent the upper end estimates and the derivation cohorts will represent the lower end estimates. You could consider it a sensitivity analysis.