Usually as long as the data is all anonymised, so that there is no chance for a person's private medical information to leak, the ethical issues are strictly methodological. Your meta-analysis needs to be rigorous, use all the available good-quality data, answer the most important question(s) about the topic, present a sound statistical analysis, present the results clearly and lead to beneficial outcomes. This is not substantially different to the methodological requirements for ethical experimental science.

Depending on your institution you might or might not still need permission, but I imagine that in most cases permission would be almost automatic once you applied for it.

The only other issue that comes immediately to mind is that in some cases we have strong reason to believe that a given piece of research is suspect. For example it could be a clear statistical outlier, or it could be from a source or  research area where we have good reason to believe that bias or fraud is a real problem. In such cases I tend to think that suspect research should be excluded from meta-analyses where possible or at the very least you should make it clear in your publication that the data set may be tainted. For example research by alternative medicine practitioners on alternative medicine is frequently of very poor quality, and aggregating a large number of seriously flawed studies is not going to produce useful information.

Usually as long as the data is all anonymised, so that there is no chance for a person's private medical information to leak, the ethical issues are strictly methodological. Your meta-analysis needs to be rigorous, use all the available good-quality data, answer the most important question(s) about the topic, present a sound statistical analysis, present the results clearly and lead to beneficial outcomes. This is not substantially different to the methodological requirements for ethical experimental science.

Depending on your institution you might or might not still need permission, but I imagine that in most cases permission would be almost automatic once you applied for it.

The only other issue that comes immediately to mind is that in some cases we have strong reason to believe that a given piece of research is suspect. For example it could be a clear statistical outlier, or it could be from a source or  research area where we have good reason to believe that bias or fraud is a real problem. In such cases I tend to think that suspect research should be excluded from meta-analyses where possible or at the very least you should make it clear in your publication that the data set may be tainted. For example research by alternative medicine practitioners on alternative medicine is frequently of very poor quality, and aggregating a large number of seriously flawed studies is not going to produce useful information.

I believe, ethical clearance is not necessary for meta analysis since we are dealing with anonymous data.

I believe that ethical clerance always should be prefered when dealing with human data. This goes especially for studies which You intend to publish since an ethical permission for Your study will be treated as a major demand from a serious publisher. 

Best regards, Pontus

I would just make sure that you acknowledge no ethical considerations when you submit your paper.  Also, if you contacted authors for data and they provided it, make sure it is de-identified and that you report in your writeup that you contacted authors (like contacting a participant in a primary study).  Goodluck!

Research that consists of a meta-analysis should be exempt from the need for IRB review.  However, I suggest that you contact the IRB with you proposal, as it is helpful to have an IRB letter stating that the research is exempt.  As noted above, it can be helpful for publishers, but it is also sometimes important in seeking grants.

The answer may depend on the country where you are conducting the research. The rules can vary significantly from jurisdiction to jurisdiction and, needless to say, the researcher is bound by the rules overseeing research in the country of the researcher's employer (affiliated research institution) is located and possibly at the data collection site as well if in a different country! All that to say, there's no black and white answer to your quesiont You have to check your local rules.

In general, the meta-analysis study is exempt from ethics approval as the study authors will be collecting and synthesizing data from previous clinical trials in which informed consent has already been obtained by the trial investigators.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4113204/

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4420933/

https://www.dovepress.com/author_guidelines.php?content_id=3400

Thank you all of you for giving your valuable answers.

A late response that I hope is helpful. At Rutgers (my institution), IRB (Internal Review Board) approval is required for all research that involves humans subjects. However, exemptions are are approved for research that meets specific conditions. But researchers cannot assume exemptions in lieu of submitting IRB applications. It's important to go through the formal process required by your institution.

Greg

If you use exclusively published data (not raw data provided by the authors), I do not think you really need an ethical approval. I have never asked for it and it is not even reported in the PRISMA steatments. Follow PRISMA steatmens and download article from Liberati et al. on explantion

As the data are in the public domain and if the study is not for degree purposes, it is unlikely that full ethical review will be required. The best thing to do is write a letter to your HREC chair, state that it is a metaanalysis of anonymised data in the public domain, not for degree purposes, is there anything else they require before approval.?