Dear Kulandaivelan 

Existence of heterogeneity is not a barrier to perform meta analysis but if there are some studies in common between SRs, performing meta analysis of meta analyses will duplicate the effect of those studies and will lead to error in results. therefore I recommend you to extract the results of all primary studies included in 10 SRs and conduct a meta analysis of primary study results

regards

Armin

Dear Kulandaivelan,

Why not? If you assume each meta-analysis as a "separate measurement" it is possible to perform a meta-analysis of meta-analyses.

-But, there must be some primary analyses in the set on intersection of those ten different meta-analysis. So, in your statistics there will be double, triple or more times impacts of some of the primary studies while some will have more or less impacts. This situation causes a type of bias for your results. 

-On the other hand 10 is assumed to be NOT enough sample for conventional statistics approaches.

Good luck!! 

If all the previous meta analyses addressed same research question, then there will be overlapping of the included trials there. So, if you again do a pooled analysis, duplication is inevitable. If you fell that a new meta analysis is required on the same research question, better to do a new meta-analysis.

My gut feeling is that you might struggle with at least three things with a meta-meta study:

1) Getting a representative sample size.

2) Unclear interdependencies between the underlying meta studies. If the underlying studies overlap significantly, then you may have a problem.

3) Aggregation biases. By distancing yourself from the source, then you loose control over implicit assumptions or worse, subjective biases in the meta studies.

I would personally have been cautious and avoid drawing any firm conclusions from a meta-meta study. It is therefore better to do a new meta-analysis as Souvik recommends.

Thank you all for your comments. As I expected over lapping will be the biggest problem. But my problem is being a researcher from developing country, it is not possible for me to full access to all eligible studies. Another biggest problem I am having with current set of data is some deal with specific age, gender, conditions like hypertension, diabetes, obesity etc. So now I change my question that is it possible to write review in this set of data? Please give me some expert opinion as early as possible. Thank you once again for all.

Lack of access to all eligible studies, as Kulandaveivelan mentions, is why Open Access is important and becomes increasingly more important in the future. Especially for meta-studies. Even my university library (from Norway) is not able to afford subscribing to all scientific publications, and need to focus on the ones considered most important, since the costs are prohibitive.

It is certainly possible to write a review on your data set. The question is whether you will be able to draw any firm statistical conclusions from it. You would need to propose a model of how the different factors influence the end result you are looking for, and then argue why or why not these factors can be included in your analysis. I am not a medical expert, so I cannot help you with how to model this.

My concern is, if there are already 10 meta analyses, what will your add to the body of evidence. If you have a good reason to do it (old meta analyses, flawed methodologies in the existing meta analyses), then I would suggest going back to the original studies rather than treating each exisiting meta analysis as an independent study

There are a number of methodological and statistical issues to consider- overlap, which several others brought up, is one as well as is the variation among the methods the reviewers used to search. select, and assess and quantitatively synthesize included studies. Colleagues and I just completed a methodological analysis of overviews in education- they can be highly problematic. There are good reasons to do an overview, so it can be worthwhile if you do it for the right reasons and if you do it well. Below are several sources- some of which if you search in Google Scholar you will find the full paper. The Cooper, Schmidt and Becker papers are a must read, in my opinion. The Pieper papers describe results of methodological analysis of overviews in health/public health.

Cooper, H., & Koenka, A. C. (2012). The overview of reviews: Unique challenges and opportunities when research syntheses are the principal elements of new integrative scholarship. American Psychologist, 67, 446-462. doi:10.1037/a0027119

Schmidt, F. L., & Oh, I. S. (2013). Methods for second order meta-analysis and illustrative applications. Organizational Behavior and Human Decision Processes, 121, 204-218. doi: 10.1016/j.obhdp.2013.03.002

Becker, L. A., & Oxman, A. D. (2008). Overviews of reviews. In Higgins, J. P. T., & Green, S. (Eds.), Cochrane Handbook for Systematic Reviews of Interventions: Cochrane Book Series. Chichester, England: John Wiley & Sons, Ltd.

Pieper, D., Antoine, S. L., Morfeld, J. C., Mathes, T., & Eikermann, M. (2014a). Methodological approaches in conducting overviews: Current state in HTA agencies. Research Synthesis Methods, 5, 187-199. doi:10.1002/jrsm.1107

Pieper, D., Antoine, S. L., Neugebauer, E. A., & Eikermann, M. (2014b). Up-to-dateness of reviews is often neglected in overviews: A systematic review. Journal of Clinical Epidemiology, 67, 1302-1308. doi:10.1016/j.jclinepi.2014.08.008

Pieper, D., Buechter, R., Jerinic, P., & Eikermann, M. (2012). Overviews of reviews often have limited rigor: A systematic review. Journal of Clinical Epidemiology, 65, 1267-1273. doi:10.1016/j.jclinepi.2012.06.015

Dear Kulandaivelan,

When I read your original questions, I thought: yes you can write a review and attempt to meta-analyze a set of meta-analyses. Especially if the set of meta-analyses concerns by and large the same research question so that the results of the respective meta-analyses depends on the applied selection criteria and methods, then you can study the influence of selection criteria and methods on meta-analysis results.

However, when I read your second post in this thread, it seems that you want to use the set of meta-analyses as shortcut to their underlying studies. The main reason you give for that is your lack of access to the original studies.

Then my question is, like Michael Edelstein: if there are already 10 meta-analyses, why do you want to do another one?

Another question: Have you considered contacting the authors of the original studies and ask them for reprints/pdf’s of their paper? I know that a few decades ago, that was common practice because a reprint showed better quality images than a photocopy from the library. I suspect that requests were also sent because some thought it was easier to ask than to go through their library and, more positively, to let the authors know of your interest in their work. I still send pdf’s to investigators who ask for them, especially if the requester would say that the paper is not in their library. Furthermore, when you contact the authors, you may be able to augment your meta-analysis with things that are not (clearly) described in their paper.

Rob

I agree with the Edelstein's and Boer's comments but would like to add this. Meta-analysis is not the only valid method of research synthesis. It is just one way of synthesising research, namely, pooling of outcomes. A descriptive (also known as narrative or non-quantitative) review can also be valuable. A synthesis of the meta-analyses might shed some new light on the topic. For example, a study of the quality of the meta-analyses might find that very few if any are good enough to draw conclusions (in which case you have a good reason for doing a new one!). 

I do not think this approach will add any value to the existing evidence body.

I agree with a new meta-analysis as indicated by other colleagues above based on the heterogeneity of the findings from the original studies, inconsistent results from previous meta-analysis, and somewhat outdated previous meta-analysis reports with a number of new studies being reported since last meta-analysis. In addition, in my collaborative experiences conducting meta-analysis projects, we also found additional justifications including coding errors and selection inconsistencies of the studies including subsidiary data from previous meta-analysis, and new innovations on approaching heterogeneity of data such as meta-prediction or meta-regression in addition to the conventional pooled analysis or meta-analysis. A systematic review is helpful in sorting out possible domains of contributing factors before a meta-analysis when there are a wide variety of domains and complex mechanisms at various micro to macro levels for the field. However, a meta-analysis ranks highest for the level of evidence; whereas, a systematic review ranks lower than the meta-analysis for the level of evidence in clinical decision support for evidence or science based practice.  

The short and correct answer for this question is: no.

In theory, yes. However, I would prefer to go to the primary studies.

Cheers,

Alejandro

Thank you very much for all researchers who provided input to my question. it is indeed a nice learning experience. let me tell my decision- I am going to write a review paper based on meta analysis studies as my intention was to share the current evidence with physicians of India in order to increase PA. thank you very much once again one and all.

I also sugest that you should go to primary studies and  conduct a new systematic review / MA

If you can gather the studies which used in these meta-analysis(MA)  and perform a new meta-analysis it is very better. in this cases our estimate ( pooled of mean, OR, RR or DR and ..) are valid. for exampe if we can assess to  primary data which used in studies ( Data bank in stata, spss or other )  it is not need to use meta-analysis. in this cases  we perform a IPD (individual patient data ) meta analysis because IPD is more powerfull than MA . About your question,  If you can access to all articles which uses in these meta-analysis, it is not suitable perform a meta-analysis on meta-analysis. meta-analysis have enough problems, MA on MA will have more than problems.

I agree with Saed that going to the underlying sources is better.

One valid reason to perform a meta-analysis based on other meta-analyses, is of course if you do research on the efficacy of meta analyses as a research method.