Hi, welcome back. This is Kate Dickerson. We're doing Section D, Refining the Question for the framing the question for systematic reviews and meta analysis class. So, if we refine the question, how do we go about doing this? And I'm going to talk more about this and then we'll go through some examples. Again, you can go back to chapter five in the Cochrane Handbook and look at that box, box 52a I'm giving here as an example. I'll give another one a little later on. That says what factors you should consider when you're developing criteria for p, the types of participants or patients or people. P works for any of those. So how is the disease defined? What are the most important characteristics? So, the wonderful thing about this chapter is it gives you little crib notes. And you can just check through and make sure you've done all those things as you define your question. Similarly, box 53A covers the factors to consider when you're developing your criteria for types of intervention. And 54C is types of outcomes. Now a comparison group was probably under intervention. Often comparison group is hard to separate from the intervention itself. Now, I'll mention once more that Cochrane is really concerned with intervention studies. So, you won't see anything in the handbook about exposures. However, the same things apply for exposures as apply for interventions. You have to change intervention to exposure, but it's the same, basic idea. So, don't get too worried about the fact that Cochrane is just dealing with interventions. You're going to have to stretch your imagination a little bit. But we'll help you along if you find that difficult making that extrapolation from Cochrane. I said I was going to talk more about outcomes and I'm just going to have this slide. And I will say that these examples are courtesy of Ian Saldana who has been for many years a TA in this class. And he is a Doctoral student at Johns Hopkins or by the time you listen to this he may have graduated. And you can also read more about how to define an outcome. In a New England Journal of Medicine article from 2011 that Deborah Zarin wrote. She's the director of what is called clinicaltrials.gov. That's a trial registration website where trials are registered at inception. And they Critical factors, the PICO and some other factors are put into a database on clinicaltrials.gov. But I'm just going to be talking about outcomes. Now clinicaltrials.gov defines five elements to every outcome. And we think these are pretty good because if you come across an entry on, clinicaltrials.gov, and it has all five of these elements defined. It's pretty easy to decide what they did in their study. But you won't often find that people have provided you with all five elements. The elements I've mentioned briefly, but you might not have recognized that they fit into this particular context. The first is what's called the domain. And that's what we call an outcome. So for example, anxiety's an outcome. Heart attack's an outcome. Visual acuity is an outcome. It's what we talk about when we talk about outcomes. It's the domain or name of the outcome. Now you have to measure that outcome somehow if you're talking about anxiety. It can be measured using the Hamilton Anxiety Rating Scale. If you're talking about visual acuity, it can be measured using that E-Chart I .mentioned or the Snellen Visual Acuity, etc. So there are specific measurements and measurement scales that are used. And you need to specify that, when you specify the outcome in your particular clinical trial and your systematic review. Then specific metrics are used for measuring an outcome. That is, is it a value at a certain time point? That is, what is the hemoglobin A1C at six months? That might be a specific metric. What method of aggregation do you use for the data. So for example, if you're looking at 100 patients in each treatment group. You can't look at each of the hundred people's values at a specific time point one by one, instead, you're probably getting a mean. Overall across the 100 people. So it might be the mean change from baseline or might be the mean value at a particular time point of hemoglobin A1C. That's the measurement that you're doing. And the time point is when did you measure it? Is it one month? Three months? Six months? So, all five of those elements are actually the proper way to describe an outcome for a clinical trial, or an epidemiologic study. It's complicated, I know. And you'll probably have to return to this as a reference again and again. But this is the proper way, we believe at this time. Who knows what's going to change in the future to measure an outcome. So let's go back again to PI/ECO, patients or populations, interventions or exposures, comparison groups or outcome. And keep that in mind because it's so important for everything we do. I can't emphasize that enough. Now I just want to say that sometimes you'll see people calling this picots or picos. And I mentioned sort of off to the side some people put settingiIn here. Well they also put timing, and that's the T and the S. So timing might include how long is the minimum follow up you want in your systematic review. For example, we're looking at education of teenagers for prevention of sexually transmitted disease. What is our minimum duration of follow-up that we think makes sense for looking to see whether this education was effective? I've already said that I don't think two weeks makes much sense. Knowing that kids have changed their behavior at two weeks in time. That's fine, but that doesn't mean that's what they're going to continue to do. So I would say if I'm looking at condom use after an educational program. That I would want to probably follow those kids at least a year. Maybe I'd look at them at monthly intervals, and ask them what they're doing, but I'd probably want to follow them at least a year. Similarly, if I'm looking at the effects of patching in preschoolers who have amblyopia. I'd probably want to look at whether their amblyopia's reverted after they stopped the patching. Or whether this is a longterm fix that we can expect. I mentioned also, we might be interested just in looking at people. For example, in a community living situation. If we're talking about falls, we might not be interested in falls in nursing homes. We might only be interested in falls of people living on their own. We might be interested only in inpatients or only in outpatients. So often people say that PICO, PICO or PECO isn't enough. That we need to add timing and setting, and it's fine if you want to do that. We tend to just use PICO without the T and S and incorporate that somewhere else. We don't forget about it entirely, but if you want to make it PICOTS, that's fine with me. So, sometimes, as I mentioned, you have a combined effectiveness and harm question that you're addressing. And as I mentioned, effectiveness questions are best answered with an RCT. While harm questions are best answered with a randomized trial, but this often isn't practical. Because the follow up's too short or the studies are too small. And so you may need to use observational data. So I just throw that in as a reminder. Let's go back then to the refinement of our question. We make a big deal in this class about deciding ahead of time what you're going to do, and trying not to be influenced by the data. However, I will say that to formulate your answerable question, you generally need to know a little bit about what's out there. It really helps to know what outcomes people who've done trials think is important. So for example, you might believe that quality of life is the most important outcome. In post-surgery for breast cancer, but it may never have been measured. I think it has been measured in many studies, but it may not have been measured. So you probably should know what are the outcomes that people in a field believe are important. So in any cancer study, they're going to believe that death is important. And you want to know this, because you want to know what to measure as your outcomes. Because when we're talking about the O, the outcomes. You are concerned both with the question and the outcomes that were present in the clinical trials. That you're going to include in your systematic review. And the outcomes that you think are as important as the systematic reviewer. It's kind of tricky, but these may be different especially because the trials may have been a long time ago. Before a certain outcome was really recognized to be important. And quality of life is one of those outcomes that thank goodness we're now realizing is important how the patient feels, but it wasn't always measured. So on one hand you need to guard against testing a hypothesis that you formulated after you've seen the data. And, on the other hand, you probably, or someone on your team probably needs to know a little bit about the topic. So you can choose outcomes that makes sense. So that's the end of section D, and next we're going to give some examples of what I've been talking about. I think it often helps to have some examples, to try to make the whole scene look a little clearer.
