Hi Marcus,

The Agency for Healthcare Research and Quality has quite a bit of learning material on this topic. Let's begin with an AHRQ Methods Guide piece entitled "Selecting Observatioanl Studies for Medical Interventions (http://effectivehealthcare.ahrq.gov/ehc/products/196/454/MethodsGuideNorris_06042010.pdf)  and "Developing a Protocol for Observational Comparative
Effectiveness Research: A User’s Guide" (http://www.effectivehealthcare.ahrq.gov/index.cfm/search-for-guides-reviews-and-reports/?pageaction=displayproduct&productid=1166). The second has a series of powerpoint presentations also available (http://effectivehealthcare.ahrq.gov/tools-and-resources/slide-library/#ocerprotocol).

From the Cochrane Collaboration, Chapter 13 (Including non-randomized studies - http://www.mrc-bsu.cam.ac.uk/cochrane/handbook/chapter_13/13_including_non_randomized_studies.htm) of the Cochrane Handbook for Systematic Reviewers (provides guidance from the Cochrane Non-Randomized Studies Methods Group). Additionally, individual groups have given their own advice (e.g. the Cochrane EffectivePractice and Organization of Care Group (EPOC) - http://epoc.cochrane.org/).

Hope this helps.

Ahmed

P.S. Short answer, use crude data if you can and meta-analyze using mean (SE) using inverse variance in RevMan. Often you'll only get adjusted data so document what you can find.