Computer generated transcript

Warning!
The following transcript was generated automatically from the content and has not been checked or corrected manually.

And right. So we're about to restart again. Um We're going to start again with Mrs, welcome are one of the consultants here in Cambridge and he'll start talking about to study a design for a few more books. Come to the room that will make a start. Uh Good morning all. So, uh you've had a lot of information in the last hour and a half. So I'll try and keep it very simple and we'll go through some basics of study designs and how do you ask the question and uh get it, you know, complete the project and published. Uh You heard from mcdonnell, it's getting research done and published is important when you come up to the consultant jobs. Yeah. So how many of you have done research and published? Quick hand up? Yeah. Uh CTS. Yeah. Two of them. Case studies, case reports. Yeah, a lot of us. Yeah. Variety. So there's something for everyone. Okay. And if you look at it, you know, a study design is, is very important in terms of how you organize your study so that you can get the best ever, you know, uh information from your research question and come to meaningful conclusion. No point in doing a study where you can't actually get a meaningful conclusion, then it's, it's a pointless study. Okay. So that's why you need to know what are the different basic study designs and what questions do you need to ask? And how do you ask questions? So that's what we will go through the next uh few minutes. Uh This is you've seen this slide in one of the few talks before. It's similar. And basically, if you look at all studies, they can be descriptive studies or analytical as orthopedic surgeons. Most of what we do is analytical. So it comes with numbers, you know, uh where you try and juggle numbers, you know, you, you learn about statistics, so and so on and so forth. So that's the topic analytical and they can be observational or experimental. We'll deal with it. I'll talk about it in the next few minutes. And then if you look at it, you know, descriptive studies surveys, you know, case reports or K cvs. And what you can't see here is uh qualitative which proof mcdonnell mentioned earlier on as well. So this is very important and this is becoming more and more important uh for MHR you know, we can do whatever study we want, but these are one actually patient's want or how do they feel? So uh I'll spend a bit more time on qualitative research studies. Okay. So if you call the top the observation studies. So what are observational studies? So we were looking at, you know what the variables, you know the different variables, variables, meaning is an exposure, uh you know, something intervention that you're doing for the patient. And what happens is the outcome. This is uh an appropriate study design. You can get some valid results and these are three different ones across sexual study, cohort study and case control. So what do they cross sectional? Uh It gives you a snapshot of the population. You say you take roof people here and say how many of these have our tractors. Um you know, and so you can give a snapshot that way and then you have a strict inclusion exclusion criteria. Okay, between 16 to 58 16 to 50. Okay. And amongst these, how many of arthritis and you can have an exposure or did they have any childhood septic arthritis? So you can say amongst this people have a childhood get septic arthritis. How many went don't have arthritis? Yeah. So that's kind of the associate gives you a snapshot at one point in time and it will give you an uh an assessment of the prevalence of the disease. Uh You can get what is called art ratio in terms of quite how many of them are exposed and what are the odds of that person is exposed to that particular uh intervention or an event and what they have in the outcome. So the next one, but it's difficult to drive cause in relationship, you don't know uh what you can just say, okay, this is a snapshot but you actually don't know what other factors were in world in them getting Delta. And they gave me some bias in terms of uh how do you assess cored studies? Uh again, so these are you take a population and then follow them for a period of time? Okay. So, and this is a longitudinal studies, it can be prospecting. So you start from the beginning, you know how many of these patient's who accepted arthritis as a child went on to develop arthritis later on in life. So you follow them from, you know, say age of five up to 50 then see what happens to them. So that can be query, time consuming, but it's got an outcome of interest. Um but it's the outcome should be something that you see often and not, you know, good for rare outcomes. Okay. Uh we can be quite costly um and time consuming. But there are lots of studies especially look at some kind of heart studies they they've done uh in uh in the US where they looked up heart diseases which start, you know, in a particular group of patients and follow them all 50 years. You know, the first generation, second generation, third generation, that's been one of the biggest studies. And, you know, I think there are some cause that is happening in orthopedics as well in the boss is looking at what happened to the natural history of uh childhood Sophia and parties and things and so on. So it's one of the studies. Uh you can do case control. Uh our studies do this kind of case control. What it means is that you select the participant based on the outcome. So, okay, all of them ended up having arthritis. Yeah. And then look at what is the relationship between the exposure? Okay, if these patient's who are arthritis, how many of them had septic arthritis? Okay. Uh as a child and you, you have a similar group of similar group of patients who don't have our tractors, but they are matched with agent sex and then look at how many accepted that that's as a child, but you need a lot, a lot of numbers. Um but they're good for rare outcome. See, there's something very rare you find so you can identify, okay, these are rare uh outcomes in the future and then you pick them up and then you follow them and see what exposure they had. Yeah, two particular uh incident. Um Exactly the outcome takes, you know, there's a bit of a lag between the exposure and outcome and you can correlate the risks of a particular exposure to what happens later on in life. And that is kind of the case control study. Yeah, but you can't tell you what is the incidence of problems? And this is the drawback of this kind of study. So this is broadly, if you know what's closure is, you want to look at the outcome, then you get a port study. Are you are looking at the outcome? And you want to see what is the relationship between the exposure, you need your face control and you want a snapshot of the relationship between the exposure and outcome. We do a cross sectional study. Okay. So these are basically and clinical trials, I think you can do to talk about clinical trials. Uh And these are experimental studies. It's a gold standard. Uh It tells you what is the relationship between the cause and effect. That means a variable that has happened or you know, treatment or intervention and what is the effect of the treatment? You can have randomized control trial for non random wise. Um And here you can also put in lab based studies but I didn't talk about it. But uh you can have lab based studies as well. So in randomize controlled trials, it is for animus because you randomly allocate participants to the treatment or intervention. And there are different types of randomization uh parallel and crossing groups are two common. And there are, you know, there are many other groups how you randomized, but it's mainly to reduce the past in terms of treatment allocation, the study study apartments that there are two or more groups, uh one group has an intervention or you can have two or three groups having different intervention and the other group has no intervention or has an intervention, which is a normal practice. So you're comparing, what is the difference between intervention be that you're trying to assess to intervention? A which is already a common practice. And again, when you're doing these, you have to calculate the sample size, which was talked about in the previous talk, there are various calculations you can do it. So you know how many patient's you want arguments you want in the study, uh so that you get meaningful results. Okay. Again, and you can have blinded, blinded, meaning a person doesn't know what they're having either single blinded, double blind or triple blinded uh and so on. And this mainly reduce the past in terms of uh research uh in terms of the assessment of uh intervention and, and this is what you need to know. Uh more studies have what is called not a hypothesis, which is already been discussed. I'm just going to be in touch with that. Here, we say that the intervention or a treatment, your testing is no better than the previous one. Okay. So you're not saying it is better, but it says it's not better. And then you disprove the hypothesis and there are different phases of clinical trials. Phase one is, you know, healthy volunteers. Uh that's what happened in North Dakota. Then phase two is you try on a small group of patient's Faysali is when you do the our cities, when it comes to our cities, you know that okay, this is going to work or not. You try the effectiveness. And phase four is once you've done the our cities, they need morning two later on and then check. Okay. Is that definitely going to be working or not? And it's uh two ways you can think of in the trials, you know, there are efficacy trials and effectiveness, efficacy's of you're doing an intervention, you think it's going to be affected, you know, and then this is going to have proof of concept. You sure that there is an efficacy that it can work. And then that is very important. When you go for bigger funding within hr you want uh 2 million ground, then you can say you've done a small study, you can sure it works and then you do the effectiveness trial where uh you can see the intervention works. So the draft do the pro for all of them. They are pragmatic trials where you have real life scenarios. So they're much bigger trials, but it's a smaller trial first to show that the bigger one works okay. And again, these are ways different hours workout interests of time, I'll just go quickly. So if you look at it okay, if you're reading a paper and then I think okay, what type of study is this? Yeah, first if you look at it but you do the investigate a sign and exposure. If it is. No, then it's an observation study. Is there a comparison group? Know it's a descriptive, it's a comparison group. Yes. Then you can have COVID study case control a cross section. Uh if it's an experimental at the randomize, controlled to have non randomized and this gives you a basic idea of the different kind of studies. Thank you. And and this is also commonly used a systematic reviews where you combine all these identical studies and to produce information, you know, uh with the largest amount of data, some studies have only 100 patient's another 1, 200 another 1, 300. So you combine them suddenly you've got a study of 600 patients' and you can pull the data and get meaningful conclusions. And then for that you need, the outcomes are measuring to be similar. Okay. That's why if you follow the systematic review, you know, then according to these Christmas guidelines, you can report, you assess the same kind of outcomes and you come to uh meaningful conclusion that you know, this kind of intervention works so very useful, but sometimes it can get boring. Okay. Uh And these finally descriptive studies, it can be services uh cases, case reports, we all know. But the most important the politics of studies. Yeah. So there are limitations of quantitative studies as we know because we're only looking at, okay. Does the intervention work or not? And we have some objective measurements, uh the scores are better but the patient doesn't feel that they got better or the other issues in life, you know, where in their major activities that they can't do, that's not explored in terms of all the prompts measures that we do. And if you want to uh look at some implementing new interventions as well and then health care research uh qualitative study is important. Okay. Um That's a patient experiences. You look at, you know, uh managing healthcare and you look at how services and organization work. And if you're involving patient's and public uh for the study, uh they, they can be done in terms of political research methodology. There's no hypothesis and you can't say uh because of uh intervention a uh this happened. Uh that is, and you can't predict with political research and all that you, you can say is okay. I'm gonna look at, you know, what is the mental health status of patient's coming to my young other clinic? You know, how is the status and does the mental health affect their out? And all that you can do is you can assess by interviewing the patient or you know, various other methods. Uh So these are different types of data collection. So you can do patient, 1 to 1 interviews or have what we call focus groups where in you can have a group of people coming in and discussing about a topic, observations, video, recording, online discussion's and so on and so forth. Um So interviews. So income beyond structure, that means you just, they'll have, you know, it's unstructured, you just have a chat with them and then discuss about the issues. That's a bad way to do it because you don't, uh when it's a bad, it's, it's, it's very difficult because you don't know what information you want to gather from the patient. Uh It can be very lengthy uh structured is when you have a set of questions you wanna ask the patient's uh they uh you know, deviating from it and that's not good as well. So most commonly used is a semi structure where you have say 10 sort of questions needs that you start. Okay. How is your hip? And then he said my hip was fine and you just don't go on to the next, but you explore okay. What do you say your hip was fine? What has happened? Uh And they said, okay, I, you know, I didn't have any pain today but you know, yesterday was, was difficult. Then you ask again. So why, what was difficult yesterday? So you explore in depth about the questions. So, but you have a kind of structure in terms of how you're dealing with it. Um Similarly a focus group, again, there's a moderator who introduces a topic, you know, what do you think about hip pain and it's five or six people will discuss about, you know, the hip a what issues they have and the moderator just, you know, they kind of, if they're deviating from the topic that brings them back and then engraved the discussion from all this, you get a lot of transcript, it's very tedious process. All the discussion has three, you know, typed out and then you look for patterns and you called data, you summarize data and then you get some conclusions from the interviews. Uh You don't need a lot of, a lot of, you know, patient's, you need about 10, 15 for most uh analysis. And these are different ways you can uh analyze the data. So briefly touch upon all these. But when you're doing a study, uh you as a researcher, you want to test and compare and analyse. So uh you're looking at okay, what is your population intervention? What are you gonna compare on the outcome? Um When you're asking the question, make it relevant to take somebody that you can measure and you can show that it is is significant uh study. So this is the right study for you. If there's something new that has happened, you know, like the albuterol from Metal metal on metal hip reso facing the issue reports were, you know, one case here and there are two cases. So you're putting a new finding. So case reports, a case series, uh the right kind of study design. If you want to associate some variables, you know, some intervention that has happened and exposure, uh you can do a cross sectional cohort or case control study. And if you want to compare one intervention to another or one intervention with no intervention, then you can have our city or one randomized control trial. And if you wanna combine the the results of all these trials, then you can do a systematic analysis, uh systematic review and meta analysis. Okay. But if you're looking at more kind of patient experience in healthcare cells, you can you need to do a qualitative research. But this is just a snapshot of uh research, the difference kind of study designs and then what kind of questions you can ask? Thank you. Sure. It's not, it's not most people. Yeah, you can't. Thank you. Right. Okay. Thank you. Uh We're now going to move on to Honda's talk, which is just uh okay. Yeah, that's historic. Contentious. Okay. She is. Yeah, brilliant. Thanks George. Okay. So you speak about levels of evidence, but it looked a bit about as to what exists a bit of history, um understand a bit more about the levels themselves, uh what they're useful and recommendations and considerations. So why do they exist? Well, conditions and research, busy people as we all know. Uh And if you imagine as for the Oxford uh Center for Evidence Based Medicine website, you have the schematic and they say, choose what type of research you're gonna use to answer your questions or where do you begin? That could be for any question that you're looking into. So you can get 2000 plus articles, seven RCTS, 300 cohort studies. But what is the best thing for your question? Uh As prof candies you mentioned this morning, um Evidence based medicine is the cornerstone of our practice. Uh The definition is right there and ultimately, it's about providing a hierarchal system of classifying evidence. So we can make good decisions for our patient's, it's a bit of history. Um The Canadian task force in 1979 wanted to characterize the level of evidence underlying health care recommendations and the strength of recommendations. And so they came up with this initial um organization of levels, as you can see our CT at the top expert opinions at level three. So look shorter than what we're used to. Nowadays, Sackett then subsequently expanded upon this when looking for anti thrombotic agents. Uh and he kind of increased it to 12345. And as you can see our cities at the top and case series at the bottom, and this is simply based on the study design and then theoretically the amount of bias as per what profit you mentioned this morning in terms of the different studies, having different types of bias levels. However, different questions exist, different medics of different questions in different ways. And so it depends on what you're asking and what you're studying and looking into. So, treatment research is very different to looking at prognostic research or diagnostic research and even economic analyses. And so this is where modifications were made. And so if you look at the top left prognostic studies, uh you know, the highest level for them, you could argue is a cohort study which is prospective uh as opposed to on the bottom, right, which is for therapeutic studies. And that's something most of us probably familiar with from medical school. Um You'll see the triangle versi, but that is basically the triangle that we're all familiar with where you have one A BC, 23, etcetera. And as you can see for therapeutic studies, a systematic review of our CTS or meta analysis is probably a lot more useful as we probably would all agree. So the triangle, as I promised now, there are loads of triangles. When you Google this, there's so many different triangles you sit there and think well, which one should I use. But essentially they're always at the top, you have met analyses and systematic reviews going all the way down towards uh expert opinion. And it's to do with the design of the study I level of bias. But also the fact that the top studies are more filtered as someone's going through the data analyzed it thoroughly as opposed to the bottom ones which are considered unfiltered. The other ways of summarizing it as Mr Kumar mentioned before, experimental studies, observational studies, etcetera. But that is what we're familiar with. But you see this only tells you the study design and surely the design of the study by itself cannot just be enough. You have to look at the study itself. I'm not going to go through these slides because in the interest of time, but essentially is what Mr Kumar and Prop can use. And everyone has kind of mentioned the benefits and negative negative points for each type of study. This is probably key to understand, to understand the level of evidence. Um but as I said, Mr Kumar is already gone through it. So it's all about bias and understanding the pros and cons of these studies. Oh okay. So the Oxford Center for Evidence based Medicine is probably the most useful one that we all are familiar with. So first developed in sort of 2000. September subsequently has been revised many original evidence ranking systems as I showed you before from seventies and eighties were essentially just looking at therapy and prevention. However, a team from Oxford decided we need to look into this and identify what research should be looked at for prognostic studies, uh diagnostic studies, etcetera. And so they came up with this is the most recent table 2011 1. So let's go through this. So if you look at it instead of a triangle, they essentially created a table which is probably a lot more useful. Um So if you look at the left hand column under question, they've imagine the clinician has a condition in front of them and they want to understand the epidemiology all the way to treatment uh and beyond. So if you look the first question is how common is a problem? It makes a lot of logical sense, doesn't it? And then you work your way down on the left. So that from the next column onwards on the left is the highest level of evidence for that type of question type, moving all the way to the right, which is the least amount of uh least uh evidence. Um If you also notice that the topic says level 12345, not one A BC. So from 2009 to 2011, what they did was they revamp this and they made it simplified because they thought all these subdivisions just gets a bit confusing for clinicians. So that's one of the benefits um going forward. We're looking at the literature. They suggest that actually, if you look at the treatment question because that's probably the question most of us will look at and be interested in. It's not just um what will happen if we do not add a therapy or sorry, apologies. Um What does this intervention help? It should be more? Does this intervention help and compared to its alternatives, what are the risk and benefits of this as well as the alternatives. And that is the next sort of iteration that might include that in terms of the table and that will evolve things such as um brother reviews, where you take all the evidence and really compare everything against everything, which is a much more holistic and overview, a bigger overview you could say. Of note, this is not just simply the study design and these guys are moving more towards quality of the study as well. So if you look at the bottom left, uh sorry, but the bottom with the single asterix, it says level maybe ground graded down on the basis of study quality imprecision in directness, etcetera. So how the person is conducted study masses as well as what type of study. So on the quality of evidence, essentially you've got your RCT and we all think that must be great that actually you got to look at the type of errors, the power. So a low powered study might miss a significant difference between to um sort of interventions of medications or surgery. So they're our scales out there such as Jada Deskee scale, etcetera, uh which assess all CTS and tell you how good they are. And therefore consideration when you're looking at papers, etcetera should also include, you know, the blinding process, random ization, um confidence intervals, etcetera, things you do when you uh analyze the paper and journal club, for example. So a couple of studies that just kind of looked into this. So Vandalia al this is from 2000 to looked at the RCTS and the JBJS and out of 2468 studies, 72 were considered our CTS. They're to investigators who use a scale with 14 parameters on it and the average score was 68. Now, an excellent score is above 75. And at that period of time, between 1988 to 2000, these two assessors who had a good inter observer availability, they basically found that 60% of studies had a score less than 75. So that's quite a considerable number of studies are not considered, you know, excellently done. And the conclusion was it was mainly due to lack of appropriate randomization, blinding, etcetera. And they suggested using a much more standardized approach in reporting are CTS. And as mentioned earlier this morning, the consort statement is something that's used nowadays. So this is where, you know, assessing the quality can make improvements. Similarly, Poo Manal looked at the evidence rating as documented by the authors and the Cochran reporting quality school. Between 2003, 2004 and 32 studies were found to be our CTS out of over 900 to investigators. And what they essentially summarize was that you should not assume that a level one is truly a level one and then a level one and level two that there's that much difference between the two. So it's worth again looking at the quality, not just the type of study itself. So how are these levels of study evidence use? So, you know, you've got them, we essentially get told, okay, look at your systematic reviews, look at your mental analysis, that's fine. But in terms of, you know, in the bigger picture for healthcare delivery, um it's all about recommendations and providing clinical guidelines. So the Center for Evidence Based Medicine, they've got a recommendation system where because they look at the quality of the evidence as well, they give you a BCD and that's based upon the type of levels. Uh So for example, consistent level one studies which obviously are considered level one truly and good uh you know, the recommendations given a a and therefore, as conditions, it's a lot easier than to go, you know, well, I'll use this type of treatment because there's a good amount of uh evidence behind it. Similarly, there's an alternative called Great, which is grading of recommendations, assessment development and evaluations. The BMJ um summarize this is being a transparent framework for developing and presenting summaries of evidence and provide a systematic approach for making clinical practice recommendations again, and RCTS assumed once again to be the highest level of evidence and observation studies are considered lower. But as you can see from this uh table there, um thing you know, studies can go up and down. So again, it's all about looking at the quality and not just assuming that just because it's an RCT, it will definitely be worth following what they find. And as I think, Profit mcdonald mentioned earlier, if you've got one study that's an RCT, you know, is that enough? Surely you need to have a few to really confirm findings. And this is what they essentially do. They take that data and then they say give a certainty recommendation as to how certain they are, that the authors have confidence that the true effect is um similar to the estimated effect. So go back to triangles which you know is what we assume when we think about levels of evidence. So top left a is what we've probably been taught all our lives. And what we still get taught be is probably what Morada I'll suggest is the great system I we take away the systematic reviews, which is what they use by the way to grade the the findings. But the wavy lines essentially represent the fact that things can move up and down. This whole world is fluid, it's not, it's not stagnant. And then see is what they suggest that maybe we should be taking the systematic reviews, but then taking it and using a magnifying girls to really uh look at the quality in more detail between these studies and allow things to move up and down. So finally, considerations, well, Oxford Group can suggest these, you should not assume obviously that the levels provide a definitive judgment about the quality of evidence need to look into yourself. And rumor that Pullman paper where, you know, a lot of level ones were not true. Level ones. Uh the type of question availability of research need to be considered, you need to make sure you're asking the right question back to Mr Kumar's uh presentation. You know, just because the cinematic review answers a question does not mean it's your question and you must consider your patient suitability, the treatment itself and alternatives and what's best for that patient to turn the risk and benefit um profile. So final site um I would say this is something I read when I was looking at the Oxford website and it comes like a disclaimer, they say just before you read or use their evidence levels, they say no evidence wrapping system or decision to can be used without a healthy dose of judgment and thought. And that's key your experience, your understanding of the condition as well as your understanding of the quality of data matches as much as just looking and saying, oh, that's a level, you know, to study. It must be great. Thank you.