 Hello all. In this video, we are going to see about meta-analysis, the top most study design in health research. The objectives of this video, at the end of the video, you should be able to list out the steps involved in the meta-analysis and the systematic review to interpret the meta-analysis and systematic review with judicious interpretation and discrimination. You should be able to list the steps involved and also to review any meta-analysis and systematic review. Now, what is the need for meta-analysis? There are now more than 15 million citations present in Medline alone with 4,800 medical journals indexed in Medline with 10 to 20,000 new citations added every week. All these studies differs in their design, methodological quality, population which they study, the intervention, the test which they are using, conditions they are applying. So, as with the number of published studies has dramatically increased, life is busy and clinicians are unable to read all articles which are published in their topic of research interest. So, that is the background for need of meta-analysis. So, if there is a single study complying all the results giving as a single interpretation or single number, then it will be easy for the clinicians to read. So, there arises the need for meta-analysis and systematic review. Let us move on to the history of meta-analysis. The term meta-analysis was coined by Gene Weiglass who is an American statistician. Historically, a meta-analysis like research was conducted by Carl Person in 1904. He tried collating the data from several studies of Typhoid and aggregated the outcomes of the multiple clinical studies on Typhoid. We cannot separate Cochrane from meta-analysis because Cochrane has a huge role in pioneering and trending this study design of meta-analysis. In 1972, there was a first meta-analysis published by R. K. Cochrane, articles related to perinatal medicine from 1940 to 1970. The name Cochrane collaboration was named after the British healthcare researcher R. K. Cochrane. Later, the Cochrane Center was opened in 1992. This graph shows the number of reviews in the field of nephrology. So, there is an increase in number of reviews over a period of time. In March 2012, the number of Cochrane reviews in Cochrane Database reached more than 5000. Cochrane Database is the largest collection of records of randomized controlled trials in the world called The Central, which is published as the part of Cochrane Library. The motto of this Cochrane Library is to provide trusted evidence, informed decisions and better health. So, what you see here in Cochrane collaboration is the C, which stands for Cochrane and the mirror image of the C, which stands for the collaboration. And there is one forest plot of the meta-analysis present within this two C. So, that is the result of the seven RCT conducted on the efficacy of antenatal steroids for lung maturity. So, because of this meta-analysis, the role of antenatal steroids has been proven nearly 10 to 20 years before the usual point of finding the efficiency of antenatal steroids. So, because of that pioneering research in meta-analysis, the logo consists of the same forest plot of the meta-analysis. Now, moving on to the differences between narrative review, systematic review and meta-analysis. Narrative review is when content expert gives a broad overview of relevant information by his practical knowledge in a narrative format. Usually, the content expert writes about a particular field, condition or treatment. A quantitative summary of the literature is often absent in the narrative review. Systematic review is a process to identify comprehensively all the studies related to the topic for a specific focused question that too in a transparent way. Systematic review usually oppresses the methods of the studies, summarizes the results, it presents the key findings, identify reasons for differences across the studies, and cite limitations of the current knowledge. So, all happens with systematic review. The basic difference between the meta-analysis and systematic review is meta-analysis goes one step beyond where mathematically the results are combined, the process sometimes referred to a spooling of the results of all the studies trying to yield one particular number out of the studies combined are pooled. Term meta means analysis. So, it is analysis of analysis. Now, what is Cochrane review? Cochrane review is a systematic reviews which are published in Cochrane. We have aggregate level meta-analysis and individual level meta-analysis. In aggregate level meta-analysis, we perform a meta-analysis where a review team usually combines the aggregate level data in each primary study. The point and variance estimate of the summary measure will be combined. On the individual level meta-analysis, the review team will obtain all of the individual patient's data from each of the primary studies and they yield the results. Compared to aggregate level meta-analysis, individual level meta-analysis is very challenging to conduct. Now, we move on to the methodology of conducting a meta-analysis. We need to design a protocol first. Protocol is a plan or set of steps to be followed in the study. A protocol for systematic review should describe the rationale for the review, the objectives and the methods that will be used to locate, select and critically apply studies and to collect and analyze data from the included studies. If authors choose to conduct their review through the Cochrane collaboration, they will also be required to register their title to the appropriate review group, saves their spot for this topic and provides access to further Cochrane support. Then, second is the research question. We need to be clear about PICO, population, intervention, comparison groups and outcomes. Then, we should be clear about the inclusion and exclusion criteria for the studies to be included or excluded. Authors need to decide a priori on their population age range, conditions, outcomes and type of interventions, minimum number of participants in each group, published versus unpublished studies and language restrictions all need to be explained in eligibility criteria for a methodology of meta-analysis. We need to develop a search strategy and locate the studies. It can be from the Google Scholar, Scopus, M-Base, PubMed. So, all the databases can be used for extracting the articles. Then, the fifth is we need to select the study. Two reviewers will check for the quality of the study which meets the inclusion and exclusion criteria separately and the studies will be included for meta-analysis. Then, the sixth step is data abstraction of each of the studies. They have to contact the authors if any further information is needed. Then, the seventh and most important step is assessing the study quality. We have two common reporting guidelines for studying the quality of the randomized control trials. That is, Jadad scoring or Oxford quality scoring system which is a three-item five-point questionnaire. We can use that for studying the quality of RCT or the most commonly used consort guidelines that is consolidated standards of reporting trials which is a 25-item based question. We can use that also but for assessing the study quality of the meta-analysis we have quorum guidelines and moose guidelines. This is the quorum guidelines, the quality of reporting meta-analysis. This gives you the checklist-based approach for reporting guidelines in meta-analysis. Then, the next step in methodology is analyze and interpret the results. We have different softwares for this purpose but before using the softwares we need to understand what are all the parameters used for statistical analysis. Then, we have writing the findings or the recommendations. We need to find which interventions are efficacious for whom and under what conditions and what are all the areas, topic, interventions requires further research that we need to mention. Then, the next step is the dissemination of the findings. The reviews conducted through Cochrane collaboration get published in the online Cochrane database of systematic reviews. These Cochrane reviews are quite lengthy and detailed. If it is not with Cochrane collaboration then we need to publish in separate journals. We move on to the statistics involved in conducting a meta-analysis. So, we have two sections for this. We need to have the test for heterogenicity that is we are going to deal with different articles. The differences may be due to the differences in the sample size, different study design, different treatment protocols, different patient follow-up, different statistical analysis, different reporting, different patient response. We have statistical tests to quantify the heterogenicity which is expressed as Q statistic. This Q statistic will quantify the level of heterogenicity. If the studies which we are clubbing is too heterogeneous then we cannot pool the study results in this meta-analysis. So, this heterogenicity by Q statistic will help us in understanding the level of heterogenicity. Q statistic is similar in concept to an I-square test although a non-significant result by convention the p-value 0.1 is often taken to indicate that there is no substantial differences between the studies. The percentage variability between the studies that is present beyond what would be expected by chance. I-square statistic is also used for this purpose but it is expressed in percentage of variability. So, how each studies varies from other will be expressed in percentage. It is important to consider that this test is underpowered especially when the number of studies being pooled is small. When interpreting an I-square statistic the values are categorized between 0 to 30, 31 to 50 and more than 50 which represent mild moderate and marked differences between the studies respectively. The two commonly used statistical methods for combining data include Mantel-Hansel method which is based on the Fixed-FX theory and the Ders-Simonian-led method based on the Random-FX theory. In a Fixed-FX model we assume that FX sizes in our meta-analysis differ only because of the sampling error and they all share a common mean that is in Fixed-FX model we assume that FX size in our meta-analysis differ between studies because of sampling error and they all share a common mean. The analysis spot related to meta-analysis will be dealt later. Now we move on to the bias in meta-analysis. We have an English language bias which means FX size of articles presented in English language are different from other language and we have database bias and the most common bias which we encounter in meta-analysis is the publication bias. Then bias in reporting data is also there then bias due to the sample size that is the sampling error and the personal bias also will be present when we are studying an effect of an intervention or a treatment. How to identify this publication bias? To identify this we need to put up this funnel plot remember this is not for us plot this is funnel plot for this we need to put up the risk estimate in the x-axis and the sample size in the y-axis. So this point is the risk estimate of the study with maximum sample size if there is no publication bias then the studies will be equally divided into two sides that is on the either side of this maximum sample size study design you will get equally distributed on the other hand if publication bias is present only the studies which showed positive results with smaller size will publish here and the studies which has been done with smaller sample size which would have got the risk estimate lesser has not published here so in this case the funnel does not appear like this so the absence of the studies with smaller sample size indicates that there is a publication bias present here a funnel plot you can have this risk estimate here in the x-axis and sample size are standard error in the y-axis if we plot then naturally there will be a funnel shaped appearance and if publication bias is present then there will be absence of studies with the negative results so that's how we identify publication bias using funnel plot. Now how can we identify publication bias using statistical tests? Orvins failsafe n test is used to check how many studies should be added in order to reduce the test statistic to a trivial size. Eggers regression test tests whether the funnel plot is symmetrical or not. Trim and fill method imputes data if funnel plot is asymmetrical. Blinding in meta-analysis can be done while performing we can blame the name of the authors the institutions the name of the journals the source of funding acknowledgments all can be blended this will lead to more consistent results this is one software called comprehensive meta-analysis software where we can do analysis for meta-analysis so we can collect the data like this study name differences in the mean their confidence interval sample size of group A group B their confidence interval effect direction then standard difference in means standard error edges G standard error difference in mean standard error all can be collected can derive the forest plot funnel plot from this software the software is available for meta-analysis some of the free ones are epimeta epimeta revman this is a cochlear collaboration and meta package from R there are some non-free packages that is this comprehensive meta-analysis is non-free and meta module in strata is also non-free that is paid we all know forest plot is the important aspect of results of meta-analysis in any forest plot there will be a diamond shape here so this indicates the fx size if the sample size is small then the confidence interval will be long and also the fx size which it contributes to the final diamond will be small on the other hand if the sample size is big so this is the percentage weight so the weighting factor is given by 1 by standard error square standard error is nothing but your standard deviation divided by root of n fin is the sample size so sample size is inversely proportional to standard error and this standard error is again inversely proportional to this weighting factor so this weighting factor is directly proportional to sample size so percentage weight will be directly proportional to the sample size so as the percentage weight increases that indicates the sample size is higher so let me take the maximum percentage weight and the biggest square appears here and the smallest two lines indicating that the confidence interval is very low so when we club this the contribution by this weightage will be higher so based on this the forest plot will yield a diamond so this is the final point so after compilation or pooling of all the study results we get this pooled arts ratio so this is in case of aggregate level meta-analysis as i told earlier this individual level meta-analysis will be very challenging so what we can do is we can do a sensitivity analysis where we can get the same forest plot with the subgroup analysis so we can club this good quality results and out full trial quality results so we can separate and plot the results so sample size can be clubbed like that and the time period also can be separated and separate results can be obtained through this sensitivity analysis so that is about the forest plot and sensitivity analysis for the results in meta-analysis now we move on to the strengths of meta-analysis so with the help of meta-analysis physicians make better clinical decisions since it is the highest level of evidence then it is more transparent when compared to other studies because all the quality of the studies included for our cities will be available in literature usually comprehensive search strategy will be followed then it provide more precise estimate of effect since overall sample is increased then it can reveal the bias strengths and weakness of the existing studies the future direction of research also can be identified using meta-analysis then the delay between the research discoveries and implementation of effective diagnostic and therapeutic strategies may be reduced then review on qualitative studies have also been started we move on to the limitations of the meta-analysis the reliability usually depends upon the reliability of the primary studies then the most common bias in meta-analysis is the publication bias and selection bias even though we have forest plot and other statistical methods to eliminate this publication bias but this is a serious dirt present in meta-analysis meta-analysis tries to assess the methodological flaws present in the primary studies but it cannot rectify the flaws present the common criticism revolving around the meta-analysis study design is that one single number cannot summarize research field so arriving at a single number in meta-analysis by compiling all the research articles cannot be accepted by many any attempt to reduce results to a single value with confidence bounds is likely to lead to conclusions that are wrong perhaps seriously so in fact the goal of a meta-analysis should be to synthesize the effect sizes how consistent they are how they are dispersed the next criticism is the file drawer problem which invalidates meta-analysis the studies finding relatively high treatment effects are more likely to be published than studies finding the low treatment effects we must remember that publication bias is a problem for any kind of literature search the problem exists for the clinician who searches a database to locate primary studies about the utility of a treatment then the next important criticism is mixing the apple and oranges we need to remember that because of the differences in studies the pooling does not make sense then the next important criticism is the garbage in garbage out if the studies which are included for the meta-analysis are not of good quality then we are not going to get the correct study results so this gigo issue of garbage in garbage out will be there but this can be addressed with subgroup analysis and quality analysis the next most common criticism is sometimes important studies are ignored which may be intentionally or unintentionally because of the lack of access this can be addressed by proper protocol and inclusion criteria with transparency then the next important criticism is the meta-analysis can disagree with the randomized trials some still argue that randomized control trial with good study quality is always better than a meta-analysis and the last criticism is they are performed very poorly so to conclude like all types of researchers systematic review and meta-analysis have both potential strengths and weakness with the growth of the science an increasing number of these types of summary publications will certainly become available to researchers administrators and policymakers who seek knowledge on recent developments to maximize their advantages it is essential that future reviews should be conducted and reported properly with judicious interpretation by the discriminating reader so this is very important the reader should be equipped with proper judicious interpretation and discriminations of the goods and bats of a meta-analysis these are my references thank you very much for watching this video if you like this video please share it to your friends if you haven't subscribed to our channel please subscribe thanks again for watching