[MUSIC] Thank you for taking the Coursera course on Introduction to Systematic Reviews and Meta-analysis. You have done a lot and you have learned a lot. Here are a list of the topics we have covered. We started by introducing to you what is a systematic review and meta-analysis. Then you learn how to formulate a research question, and how to search for studies, how to collect data from included studies. How to assess the risk or bias of clinical trials. And then how to synthesize all the information you have gathered and avoid bias in this process. And lastly, we talked about meta-analysis, and you should be able to read the results of a meta-analysis by now. You have achieved a lot and you have learned that systematic review uses pre-specified and rigorous methods to identify, appraise, and synthesize all evidence on a given topic. As a systematic reviewer, you should try to minimize bias and errors at each step of doing a systematic review. The most important step in doing a systematic review is to frame an answerable question. We talked about how to use PICO, the population, intervention, comparative intervention, and outcome to help you to frame an answerable question. You have listened to the lecture on searching for studies and that process should be as comprehensive as possible. Because, you want to draw your conclusions on your question based on the totality of the evidence. You want to identify everything that is out there, that meet you eligibility criteria. Risk of bias assessment is also important. It helps determine how trustworthy the evidence is. You can always conduct a qualitative synthesis of the studies created in your systematic reviews. When appropriate, you could design it to conduct a meta-analysis, which is the qualitative combination of the results from individual studies. If you would like to continue your journey in actually doing a systematic review here is how you can get started. The step one is to gather your research team. Again you need content expert as well as methods expert. Then you're going to develop your protocol for your systematic review. And the protocol will outline your research question, your eligibility criteria, and the methods you're going to use for doing your systematic review. And then you're going to collect the data. That means you're going to search the literature, locate eligible studies, screen them and collect data. As a step four, you will abstract data and appraise risk of bias in the individual studies. The next step, is to synthesize the findings, interpret, and assess the overall body of evidence. You will lastly write a report and when appropriate, or needed, update your systematic review. Our course covered the basics, or the introductory steps in doing a systematic reviews, but perhaps you will need more to actually do a systematic review. And the best of ways to get started is actually being part of a research team in conducting a systematic review. Yet, we hope you will be able to read a published systematic review critically after taking our course. And here I am going to show you the example that we started with in our very first lecture and we're going to read this review together. The review is on early versus late initiation of epidural for labor. And it was a review published in the Cochrane Library. What is the research question for this review? Well the population or the participants are pregnant term women requesting epidural analgesia in labor. And the intervention and the comparators are early versus late initiation of epidural. Here, early is defined as cervical dilation of less than 4 to 5 centimeters. The primary outcomes of the review include incidence of C-section, incidence of instrumental birth. Duration of first stage of labor and duration of the second stage of labor. The authors designed and conducted a comprehensive search for studies for inclusion. And this flow diagram shows you the results it got from each database, and how they screened the results, and how many were excluded or included at each stage. For this review, they included nine studies in the systematic review for qualitative synthesis as well as for their meta analysis. The next step in doing a systematic review is to abstract the data and appraise the risk of bias for each included study. Here each row is one study and each column is one of the domain on the risk of bias assessment that the authors use for their review. For example, the first column is on random sequence generation The second column allocation concealment. And as you can see most of the domains and for most of the studies ar rated as at low risk of bias, shown as green dots on the plot. And for blinding of participants and personnel, they're all rated as a high risk bias. After we have done this, you may find that the studies are similar enough, and then you can combine them together in your meta-analysis. Here is the meta-analysis for the primary outcome, and the incidence of C-section, again the authors included nine studies here. And each row represents the results from one study. Here we have, let's look at the very first study, for the early epidural group, there were 13 individuals had the event out of 74 randomized to that group. And for the late epidural group, there are 14 events out of 75 individuals randomized to that group. And the square in the states represent the point estimate here relative risk, or risk ratio, in the 95% confidence interval for that study. And the column to the right hand inside of the now or of the figures of each study you see the weight that each study's taking. And if you look at the sample size and the weight, Wong 2009 study takes about 84% of the total weight in the meta-analysis. And that's because that study has the largest sample size with many events. If you look down the force plot you will see the diamond and the point estimate for the diamond is relative race is 1.02 with a 95% confidence level that ranges from .96 to 1.08. Because the ninety five percent confidence interval covers the now value of one, there is no evidence showing that early vs. Late epidural has any impact on the incidence of c-section. On the bottom left of the forest plot, you will find estimates of the heterogeneity. The heterogeneity tau-squared is less than point zero one, and the chi-squared value is three point one eight, with a degrees of freedom, eight. And, i-squared value is almost zero percent, meaning that all these studies are very homogeneous. This also shows on the forest plot, because every study, if you look at them, are almost estimating the same effect. So that's why the heterogeneity is very low. So for this particular outcome, which is the incidence of C-Section, the authors did combine them in a meta-analysis. And the combined result is 1.02. That's how you would read the output from a meta-analysis, and the heterogeneity statistics for that meta analysis. As we discussed early not all systematic review needs to have a meta analysis. And now here is an example where the authors showed you the results from individual studies. But did not pull them together in your meta analysis. It's for the same review however this is the outcome on the duration of first stage. Again, each line and each stick shows you the results from one study. Here you don't see any diamond, because the authors felt that the studies are too heterogeneous to pull them together in a meta-analysis. So even within the same systematic review, you could do meta-analysis for certain outcomes and decide not to do meta-analysis for other outcomes. We look at the systematic review on a clinical question that we started in the first lecture which is early vs late initiation of epidural for labor. And we talked about how the authors formulated the research question, how they looked for studies, how they assessed the risk of bias, and how they pull the study results together. There are many more advanced topics on systematic reviews and meta-analysis that we did not cover. For example, we didn't discuss how to investigate heterogeneity. And there are also advanced statistical methods for doing meta-analysis. You may have heard of meta-regression, network meta-analysis, individual patient data meta-analysis, and multivariate meta analysis. We did not get a chance to talk about how to conduct a systematic review of observational studies or diagnostic test accuracy studies. And how to report a systematic review to follow the reporting guidances. These topics have been covered in a lot of the textbook on systematic reviews. Here we have for you a few suggested readings. For example, the standards for systematic reviews, put together by the Institute of Medicine, the Cochrane Handbook for Systematic Reviews of Interventions. As well more statistical oriented textbook such as Introduction to Meta-Analysis Methods for Meta-Analysis in Medical Research and Systematic Review in Healthcare. Thank you again for taking our course on Introduction to Systematic Review and Meta-Analysis, you have learned a lot on this topic but there are more to learn. We hope you will continue this journey and doing your own systematic reviews and integrating systematic reviews in your practice. Doesn't matter if it's medicine, public health, or policy, we hope you will base your decisions on evidence, on the totality of evidence summarized in systematic reviews. Thank you very much. [MUSIC]
