A basic introduction to fixed‐effect and random‐effects models for meta‐analysis

Research Synthesis Methods 11/2010; 1(2):97 - 111. DOI: 10.1002/jrsm.12

ABSTRACT There are two popular statistical models for meta-analysis, the fixed-effect model and the random-effects model. The fact that these two models employ similar sets of formulas to compute statistics, and sometimes yield similar estimates for the various parameters, may lead people to believe that the models are interchangeable. In fact, though, the models represent fundamentally different assumptions about the data. The selection of the appropriate model is important to ensure that the various statistics are estimated correctly. Additionally, and more fundamentally, the model serves to place the analysis in context. It provides a framework for the goals of the analysis as well as for the interpretation of the statistics.In this paper we explain the key assumptions of each model, and then outline the differences between the models. We conclude with a discussion of factors to consider when choosing between the two models. Copyright © 2010 John Wiley & Sons, Ltd.

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Malaria contributes significantly to the global disease burden. The World Health Organization recommended the use of artemisinin-based combination therapies (ACTs) for treatment of uncomplicated falciparum malaria a decade ago in response to problems of drug resistance. This review compared two of the ACTs-Dihydroartemisinin-Piperaquine (DP) and Artemether-Lumefantrine (AL) to provide evidence which one has the ability to offer superior posttreatment prophylaxis at 28 and 42 days posttreatment. Four databases (MEDLINE, EMBASE, Cochrane Database and Global Health) were searched on June 2, 2013 and a total of seven randomized controlled trials conducted in sub-Sahara Africa were included. Results involving 2, 340 participants indicates that reduction in risk for recurrent new falciparum infections (RNIs) was 79% at day 28 in favour of DP [RR, 0.21; 95% CI: 0.14 to 0.32, P < 0.001], and at day 42 was 44% favouring DP [RR, 0.56; 95% CI: 0.34 to 0.90; P = 0.02]. No significant difference was seen in treatment failure rates between the two drugs at days 28 and 42. It is concluded that use of DP offers superior posttreatment prophylaxis compared to AL in the study areas. Hence DP can help reduce malaria cases in such areas more than AL.
    Malaria Research and Treatment. 06/2014; 2014.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: 1. Grasslands cover ~30% of the Earth’s terrestrial surface and provide many ecosystem services. Many grasslands are heavily managed to maximize these services for human benefit, but the outcome of management is anticipated to be increasingly influenced by various aspects of climate change and elevated atmospheric CO2. The relative importance of global change vs. land management on grasslands is largely unknown. 2. A meta-analysis is used here to examine drivers at both scales primarily targeting services provided by grasslands relating to plant productivity (above- and below-ground biomass) and soil processes (nutrients and soil respiration) in 38 manipulative experiments published in the last decade. We specifically target effects of (i) single and combined land management practices (LMs), (ii) single and combined factors relating to broad-scale climate change and elevated CO2, and (iii) combined management practices and changes to climate and CO2. Collectively, this examines the general efficacy of global change models in predicting changes to grassland functioning. 3. We found that combinations of management practices had approximately double the explanatory power for variation in grassland services compared with individual or interactive effects of factors associated with climate change and CO2. These interacting management practices such as nutrient additions and defoliation predominantly influenced functions associated with productivity or biomass both below and above ground. The effects of interacting factors of climate and CO2 influenced a wider range of ecosystem functions, but the magnitude of these effects was relatively smaller. 4. Interactions between management practices or between climate change/CO2 factors always had higher explanatory power than any factor in isolation indicating that multivariate synergistic models of environmental change can better describe impacts on ecosystem function in plant communities (e.g. relative to univariate climate-based models). Given that the magnitude and direction (positive or negative) of the interactions varied widely, this also implies that the outcomes of these multivariate interactions can vary spatially, temporally or by immediate context (e.g. management prescriptions). 5. Synthesis. Although our work confirms how climate change and CO2 can affect many ecosystem based functional attributes, it suggests that combinations of LMs remain the dominant set of factors in determining the performance of grassland plant communities. Land management may thus be critical for influencing projected responses to future climate change and elevated CO2 in models of grassland function at least for factors relating to primary production.
    Journal of Ecology 06/2014; 102:896-904. · 5.43 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: We read the article 'When should physical rehabilitation commence after stroke: a systematic review' by Lynch and colleague with great zeal (1). Authors conducted an interesting review to determine the effects on mortality, function, and complications when physical rehabilitation commences 'early' (within seven-days of stroke) and to describe the effects of early transfer to rehabilitation wards/hospitals when sustained rehabilitation is unavailable in acute stroke units. They concluded that there is some evidence that rehabilitation should start within three-days postevent, but effect of commencing rehabilitation within 24 h is unclear. They also suggested that the functional outcomes of patient with stroke are better when the time inter-val between stroke onset and rehabilita-tion transfer is shorter. We strongly believe that these findings will certainly have an impact on stroke care and the transfer procedures. But there are certain points that we would like to highlight. Statistical methods and summary measure While it is a common practice to report odds ratio (OR) and relative risk (RR) as a summary measure, it is clinically more meaningful to report number needed to treat and absolute risk reduction (2,3). In case authors wish to use OR or RR, it is better to use RR as summary measure as it is statistically more stable. Besides, one should not speak of 'risk' when using OR. Authors in given review used OR for number of deaths and reported 'risk' , which is wrong. OR and RR cannot be interpreted similarly (4). In meta-analysis, fixed-effect model is often used when the treatment protocol used by different studies remains the same (5). Authors should have used random effect model while pooling the results as there was significant heterogeneity in intervention protocol among the studies. This does not change the result but does change the magnitude of summary measure used.
    International Journal of Stroke 06/2014; 9(6). · 2.75 Impact Factor


Available from
May 21, 2014