[MUSIC] Let's look first at the steps in a systematic review. You know this picture well. It shows where we are in discussing how to do a systematic review and meta-analysis. And we're at step nine and ten, developing the forms for assessing the 'risk of bias and extracting the data. And then, assessing the risk of bias itself. Number nine developing forms for assessing the risk of bias and extracting the data and number ten assessing the risk of bias. Let's begin by talking about why bias in the individual study is important to a systematic review and meta-analysis of a group of studies? What do we mean when we say risk of bias in systematic reviews? I think most of you would think ahead and say well, when we do a systematic review and meta-analysis, we only want to include high quality studies. Well that's what we mean. We dont want there to be bias in the methods that are used in the studies that we include in our systematic review of meta-analysis. That is we want a low risk of bias. We've talked about the fact that you want a low risk of bias in the studies themselves and you also want a low risk of bias in the systematic reviews. Today, we're just going to be talking about bias in the methods used in the included studies, which some of you might think of as a quality of the studies that are included. So, our objectives today are to define the sources of potential bias in studies included in the systematic review. To look at the potential impact of bias on the summary results and to discuss how to deal with potential bias in your meta-analysis. Now all of what I'm about to say wouldn't make any sense and we wouldn't need to worry about the risk of bias, if it had no impact on the results of the meta-analysis itself. That is when we summarize all the studies together, if there is no change in the quantitative findings from the meta-analysis, then we know that the risk of bias has had no impact on the results of your meta-analysis. On the other hand, we never know whether that's going to be true or not, and we have to assume that bias could have an impact on our results. And that's why we're so worried about it, and we want to make sure just to include what we would call high quality studies, or studies at a low risk of bias. So what is study quality and why aren't we using that term anyway? Why are we using a term like risk of bias which is convoluted and kind of hard to understand? So, I think what we mean when we're talking about study quality is we're talking about the methods used to conduct the study, the way the study was actually done, but we don't know that. All we know from reading a journal article or looking at clinicaltrials.gov, or wherever we get our data, is the way that the study was reported. That is the way people wrote up the paper or the way they entered the data into clinicaltrials.gov. So, we're never able to find out how well the study was done. We're just able to find out how well it was reported. And I will mention at the end and I'll mention now, that there are reporting guidelines to make sure that that study is reported well. The CONSORT statement for reporting clinical trials, and the STROBE statement for reporting observational studies. I'll talk about that again at the end. So some of the elements of study quality can be assessed by reading a study report, for the most part. That is we can asses the internal validity of the study. And as you've learned in your first year epidemiology class, this means minimization of bias. This means decreasing that risk of bias that I'm going to be talking about today. External validity is how well your study relates to the outside world. Some people say whether the study findings are generalizable. Some say whether the study findings were applicable. But that's what external validity means. We can usually tell that by reading a report. We can usually tell how relevant the study is to the questions that we have, how original it is compared to other studies that are done, and whether the study appears to have been done with ethical constraints that we all would accept. For example, has an institutional review board approved of this study being conducted. But there are parts of the study quality that we can assess. We have no idea usually what the protocol violations have been like in, for example, a randomized clinical trial. Usually, this isn't reported in the report from a randomized clinical trial. We don't know how well they kept records, we usually do not know what the forms look like. We don't know how often errors had to be corrected in data entry. What about study procedures? They may say that they measured, let's say visual acuity a certain way, but we don't whether visual acuity was actually measured that way. How much does the report, in fact, reflect the actual study? We have no way of knowing that. People have talked about that as being problematic, but we really don't have a way for accessing whether the report reflects the actual study and how it was done. And we certainly have very few ways of finding out whether the data were falsified or fabricated. That's something that every now and then we hear about. And we just hope and pray that the results we're reading are totally honestly reported. So there's a lot about a study that we can't tell from reading in the report. But what we're going to talk about here is what we can tell, those first elements of bias in the internal validity of the study. So the studies that you include in your systematic review and meta-analysis should be unbiased or at least you should minimize the bias in the studies that you select. Now what does bias mean when we're talking about intervention studies? There are three main types of bias that I'll mention. And I'm just going to mention the names of the bias themselves, but you have to understand what they mean. We can't just rattle off the names because these names are used generically to talk about bias. The first is selection bias, the second is information bias, and the third is a bias in the analysis. It's not really called analysis bias, but I'm trying to make it understandable, and I'll explain a little bit more as we go along. Now we're going to take a break, and when we return we're going to talk about minimizing bias in the included studies and specifically selection bias.
