Yes, meta-analyses on observational studies can well be performed. The technical meta-analytical issues are substantially similar with that used with randomised trials. For clinical question related to interventions the Cochrane Handbook (Chapt 13) suggests to use evidence from non-randomised studies only when randomised studies are lacking.

However, your specific case concerns dietary intake and cardiovascular disease and it depends how you construct you question. If you consider dietary intake as a “risk factor” then you can limit your assessment to observational studies that you can pool in a meta-analysis. Conversely, if you consider dietary intake as an intervention then you should limit your assessment to RCTs that you can include in a meta-analysis.

For reporting of meta-analysis of observational studies you should use the MOOSE guidelines (see https://www.equator-network.org; or JAMA. 2000; 283(15):2008-2012).

Hi Iosief, 

Thank-you very much for your response. 

I am not looking at dietary intake as a risk factor, or looking at dietary intake as an intervention (as the participants in the studies already have cardiovascular disease). 

The studies I have found have all been cross-sectional, so my research question is just to "describe their dietary intake". I want to find the average amount of carbohydrates, protein, fat consumed per day among the studies. 

Do you still think it would be feasible? 

Thank-you for your time, 

Emily 

If you have raw data and there is no potential heterogeneity you can still perform the meta-analysis. You have to pool then the rate/prevalence/average of the variables of interest. 

 Thank-you very much Iosief! 

Hi Emily. I know you can do it with the comprehensive meta analysis program very easily. They  have links on their website too if you want more information- www.meta-analysis.com or in the Borenstein et al., 2005 book.

Thanks, 

Mara

Hi Mara, 

Great! Thank-you so much :) 

Kindest, 

Emily