A paragraph from the Book "Advances in Meta-Analysis" written by Pigott, Terri.

Another common question is: How many studies do I need to conduct a meta-analysis? Though my colleagues and I have often answered “two” (Valentine et al. 2010), the more complete answer lies in understanding the power of the statistical tests in meta-analysis. I take the approach in this book that power of tests in meta-analysis like power of any statistical test needs to be computed a priori, using assumptions about the size of an important effect in a given context, and the typical sample sizes used in a given field. Again, deep substantive knowledge of a research literature is critical for a reviewer in order to make reasonable assumptions about parameters needed for power.

Good Luck ...

https://www.researchgate.net/post/How_many_papers_are_acceptable_for_doing_systematic_review_and_meta-analysis?_tpcectx=profile_answers

https://www.researchgate.net/post/Is_there_any_minimum_number_for_studies_that_should_be_included_in_meta-analysis

http://www.springer.com/gp/book/9781461422778

I think that it depends on the scope of what you are analyzing.  If it is a broad topic, then a small number of papers may indicate bias in your analysis.

I am currently working on a meta-analysis of a broad field using 195 studies published over around 5 years.  I have seen other papers accepted for peer review with as few as 18 papers.  

If a thorough search does not produce more than 29 papers, and if the subject of your analysis is somewhat narrow, then 29 seems reasonable.  In your methodology section, document well the steps you took to find relevant studies.

I think Dr. Marek hit the nail on the head.  The key will be to document the process by which you identified the 29 studies but more importantly how you ended up excluding other potentially relevant studies that might not have met your pre-determined criteria. I know one particular journal that required the authors of meta-analyses to list all of the articles that were identified in the literature review and then to list each of them and identify the reasons each was excluded, if applicable.  

The more transparent the process, the less likely someone is going to push back on "only" having 29 studies if that number seems reasonable for the research question.  

Thanks a lot for your guidance Michael and Lance. Much appreciated.

I would argue its not the number of papers or studies that you find, but the number of patients or subjects in these studies. In my field, a meta-analysis can be done on 4-5 studies, if they have enough numbers, patients or events and can generally give us better statistical power (often the reason to do the meta-analysis), thus better help address the research question. Many well know bodies have made approvals for medicines and treatment advice on small meta-analysis. Good luck.

A paragraph from the Book "Advances in Meta-Analysis" written by Pigott, Terri.

Another common question is: How many studies do I need to conduct a meta-analysis? Though my colleagues and I have often answered “two” (Valentine et al. 2010), the more complete answer lies in understanding the power of the statistical tests in meta-analysis. I take the approach in this book that power of tests in meta-analysis like power of any statistical test needs to be computed a priori, using assumptions about the size of an important effect in a given context, and the typical sample sizes used in a given field. Again, deep substantive knowledge of a research literature is critical for a reviewer in order to make reasonable assumptions about parameters needed for power.

Good Luck ...

https://www.researchgate.net/post/How_many_papers_are_acceptable_for_doing_systematic_review_and_meta-analysis?_tpcectx=profile_answers

https://www.researchgate.net/post/Is_there_any_minimum_number_for_studies_that_should_be_included_in_meta-analysis

http://www.springer.com/gp/book/9781461422778

I think there are two pertinent aspects:

a) All existing studies must be included

b) The pooling method must estimate error correctly with the existing number of studies

In answer to a) there is no threshold

Thus the threshold depends on b). With newer models (IVhet and QE), error estimation deteriorates below 5 studies - so if existing studies are less than five, better to avoid a meta-analysis. With RE models, error estimation deteriorates as a function of both extent of heterogeneity and study numbers so the minimum is undefined but possibly larger than 5.