search this site.

080309P - WRITING A META-ANALYSIS PAPER

Print Friendly and PDFPrint Friendly

Presented at the Scientific Writing Workshop held at the Kulliyah of Medicine UIA Kuantan on 08-09th March 2008 by Professor Omar Hasan Kasule MB ChB (MUK), MPH (Harvard), DrPH (Harvard) Professor of Epidemiology at the Institute of Medicine University of Brunei and Visiting Professor of Epidemiology at the University of Malaya


ABSTRACT
The results of meta-analysis are presented in the form of a paper. The paper should have the same sections as a normal scientific paper: an abstract giving the salient and most important findings and conclusions of the review, an introduction giving the background and objectives of the review, a methods section explaining the selection of research and the methods of analysis, and a discussion and conclusion section. Tables and references are also attached.

ABSTRACT
The abstract should be about 400 words. Although brevity is necessary, important and significant information should not be excluded. The abstract should be written in such a way that it can be read meaningfully on its own without access to the full meta-analytic report. The sub-sections of the abstract are: background, objectives, search methods, selection criteria, quality criteria, methods of data collection, the results, discussion and conclusion.

INTRODUCTION
The introduction should provide a background to the present review which includes previous studies and reviews and state of the art knowledge in the field of concern. The motivation and significance or importance of the review should be discussed.

METHODS
The methods section should give details of the search strategies (electronic and non-electronic) including dates and data-bases used. Selection criteria should be mentioned. Types of studies and types of participants should be mentioned. Approaches in assessing the methodological quality of the studies included must be given. Information must also be provided on sub group analyses and how the heterogeneity of the selected studies was assessed.

The main outcome measure(s) used should be defined explicitly a distinction being made between primary and secondary outcome measures. Data collection and data analysis methods should be explained in sufficient detail to enable an independent reader interpret the results. The following information should also be provided to enable assessment of bias: handling of missing data, existence of biases (publication bias, reporting bias, etc). It should also be mentioned whether the meta-analytic method was fixed effect or random effects. 

RESULTS
The results section should give the total number of studies analyzed as well as the aggregate number of participants. Summary characteristics of included studies must be provided including design, sample sizes, interventions, and outcomes. An analysis must be given of the relative distribution of potential confounding variables among the different studies. Summary characteristics of excluded studies should be given to enable assessment of the potential for bias. The overall methodological quality of the included studies should be assessed. The results should preferably be expressed quantitatively with statistics for comparison groups being given as means or proportions with 95% confidence intervals. Overall summary effect statistics should be provided with 95% confidence intervals as measures of precision. Results of sensitivity analysis should be provided to assess the robustness of the conclusions.

CONCLUSION and DISCUSSION
The conclusion should be based on the data and not any personal opinions. The conclusion should restate the main results. The discussion should include factors that help interpret the main findings and possible causes of bias. The discussion should be structured. It should cover completeness of the evidence, the quality of the evidence, and the possibility of bias. The author(s) should discuss implications of the results to practice and further research.

Sensitivity analyses can be used to assess the robustness of the meta-analysis by observing changes in the pooled effect estimator when changes are made in the way the meta-analysis is carried out. Changes can be using different inclusion and exclusion criteria, using a different statistical approach etc.

Publication bias should be investigated because it can compromise the whole meta-analysis exercise. A funnel plot is used to assess publication bias by plotting the effect measure or its logarithm against sample size. If there is no bias a symmetrical inverted funnel plot should result.

READING RESULTS OF META ANALYSIS
The following questions should be used as a quality and methodology check-list after writing a meta analysis paper.: are the methods clearly stated?, was the search for articles comprehensive enough?,  were the criteria for selecting articles for review stated?, were the criteria objective and were they adhered to?, was there a possibility of bias in the selection of those articles?,  was the methodologic quality of each article assessed?, were differences between studies explained or were they just glossed over?, was the combination of results from the primary studies appropriate?, and were the conclusions of the reviewer supported by data?