MedAll App
Download the MedAll App
All your healthcare resources in one place
All your healthcare resources in one place

SFP Session 6: Interview prep 1: Critical appraisal and practices



This is a session on critical appraisal led by Wayne from New York Hospital and Robbie from New Castle sf. Through this session, medical professionals will learn how to approach critical appraisal while being presented with scientific papers. A basic framework will be discussed and the PICO framework introduced to help critically appraise and present findings in the interview. The session will include practical examples, including a paper from the Lancet, to help learn how to apply the concepts to real life situations.
Generated by MedBot


With the critical appraisal station being used in most of the interviews, it is arguably one of the most important stations as most students will not have had prior teaching on how to critically appraise a scientific paper. Join us this evening where we aim to cover the PICO framework which will help you structure your thoughts when critically appraising a paper.

Learning objectives

Learning Objectives: 1. Understand why critical appraisal is important in medicine 2. Be able to define the PICO framework and explain how it can be used to assess the validity of a research paper 3. Be able to describe the purpose and parameters of a critical appraisal interview 4. Explain how to use the PICO framework to evaluate the purpose, population, intervention, control and outcome of a research paper 5. Explain how research findings can be used to shape and inform clinical practice.
Generated by MedBot

Related content

Similar communities

View all

Similar events and on demand videos


Computer generated transcript

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

Um, welcome everyone. Um, for those who just joined, um, my name's Wayne, I'm one of the, um SFP doctors in New York Hospital. Um I'll be running the session today with Robbie who is a New Castle sf doctor. Um, fine. So our session today will be um on critical appraisal. Um I'm not sure what, you know, everyone's background um is in terms of research experience, but for me, um I was pretty new to research when I, um, you know, applied for the SFP. So I think when I'm in terms of, you know, approaching, you know, the critical appraisal station during interviews and that was one of the things that made me rather nervous. Um, you know, just because I didn't have that much background. Um, you know, with research and also I don't think med school in med school, we get taught, you know, properly in terms of, you know, looking, um, you know, on how to critically appraise a paper. Um, but basically critical appraisal is, you know, going through a paper systematically when you're presented with a scientific paper. Um, it's easy to just read through it and, you know, sort of, you know, note and understand the conclusions that the researchers have made. But, you know, how do we actually know, you know, whether the study was designed properly, whether, you know, they actually tested the hypothesis? Um You know, what's the outcome that they measured? You know, what the outcome that was supposed to measure if that makes sense? Um And I think mostly with research, you know, it does answer the, what we want to know is that, you know, with clinical research, will it have any impact on how we change our clinical practice? Well, what we are researching on um you know, improve patient um lives and improve patient health. Um So I think a lot of clinic, a, a lot of um clinical research is done um you know, in with, you know, the goal of maybe influencing and changing the current guidelines. Um So for the SFP um interview purpose, um so there are when we, when we have the critical appraisal station um for me personally, because I applied to East Midlands. Um so for the critical appraisal station, I was given a full paper which is about 10 pages long. You include um authors, acknowledgements and references. And so I was given a full paper a day before. Um And I, I was given about 24 hours to just look through the paper and write, you know, a report um like a critical my critical appraisal. Um the paper obviously the um interview format will differ between deaneries and differ between, you know, the hospitals that you apply to. Um So I've heard of people having um just 20 minutes right before the interview and they were just given a brief abstract. Um and they had to use the 20 minutes to come up, you know, with, um to explain the paper to the interview is um basically, um how was your experience like Robbie with ESP? So I had an interview at Northern, which didn't focus on critical appraisal at all. And I had an interview at London which did, and they gave you a, an abstract for a paper. I think mine was about fluids and itu um and resuscitation in a particular cohort. And I was given about 20 minutes to review that and a clinical scenario before the interview in kind of the waiting room and then to critically appraise and present it in about 10 minutes. And I think, I think, you know, as we've kind of said, we're going to talk about how to approach critical appraisal. A critical appraisal is a presentation at the end of the day and it's all about for the interview, how you can put across the significant points and how you can talk about it or talk about why whatever you are appraising matters and why it doesn't matter. So, yeah, that was my experience um quite intense does require preparation to have a framework effectively. But yeah. All right. So speaking of frameworks. Um So this is the basic framework that I used when I um you know, had the interview, obviously um everyone has got their own style, but I did find this framework. Um well, um yeah, I did, I did find this framework quite useful. Um So when we, so when I was given the paper um the frame, so this was the framework that I used um the introduction PCO which is um a short summary of the abstract, which I will go into later um conclusion of the study. And then, you know, after reading the um the full paper um mentioning what the strengths and weaknesses of the paper were. Um And then, um I touched a little on internal and external validity um which I'll be talking about later. Um And then I ended the whole presentation with um you know, whether that one paper, the results of that one paper that I was given would change my clinical practice in real life. Um So there's a link on the chart if everyone wants to click on it. So that's the article that we'll be using tonight. Um And I'll post the abstract on the slides, um the upcoming slides. So if we go through the framework, um if everyone has a quick glance through the paper, so this is a paper um which basically um want, which basically wants to determine if um morning dosing uh antihypertensive medication is better than evening dosing. So um this, yeah, so the introduction of this um introduction wise um when you, when you look at the paper, um when you want to introduce the paper to the interviewers, um I used what, why, who, so what kind of study is this? Why was the study done? And who did the study involve? Um So if you look, look at the paper in the abstract itself, it will tell you um that this is a randomized trial published in the Lancet in October 2022. Um And the aim of the study was to look at whether taking antihypertensives in the evening would improve cardiovascular outcomes compared to taking it in the morning um in patients with hypertension. So that's your um population, that's your study, um your, your sample, basically your population of interest. Um And I think this is a very, you know, nice brief way of um introducing the paper um in the critical appraisal station. So, moving on to the PCO framework, um this is um another framework which goes a bit deeper into, you know, the specifics of the paper. So PCO stands for um population well, intervention control and outcome. I've put a few question marks behind control and I'll get into that. Um So when we look at this paper, um the PECO framework basically tells us that, you know, when we, when we think about the peak of framework and we can basically see if it answers the question that needs to be tested. So basically, this is a paper that wants to see if um adults with hypertension um would do better if they have morning dosing versus evening dosing. So, using the Pica framework, we need to make sure that the population, the, the people that we're using in this study are people who are taking antihypertensives. The intervention would be, you know, comparing people with evening dosing of antihypertensives and morning dosing of hypertensives. Um We'll leave control out for now. An outcome um is basically, you know, seeing how well people did, um you know, in the morning dosing compared to the evening dosing. Um So if you look at the, if you look at the um PCO framework, so what I've written here is that this is, this is a study that looked at about 21,000 individuals aged 18 years and above who had hypertension and were already taking at least one hypertensive medication. So, from this statement itself, we can see that the, the study actually sort of identified the correct people that we want to because we need to have, you know, the correct population, which is basically people who are taking antihypertensives um in this study, um intervention. So the intervention um was to change the time the participants took their antihypertensive medication to either in the morning or in the evening. And again, um the intervention is um well, a valid intervention because you are comparing between two groups, morning dosing and evening dosing. Um And the primary outcome is also stated in the abstract that I've put on the slide. So the primary outcome was defined as vascular death or hospitalization for a nonfatal myocardial infarction or stroke. So, the authors basically used um you know, these endpoints to find to see if you know, patients would benefit more from morning dosing or evening dosing. Um So, going back to control. So if you think about this study, so usually control would be something that is already usual practice. So maybe, um so in this case, because we don't, we don't have a control. Uh As for now, I think there is no evidence that says, you know, if you have, if you take your antihypertensive medication in the morning, um you know, there is no evidence that says that it's better for you to take it in the evening a vis rosa. Um So in this case, we don't really have a control because we're not comparing anything to the standard practice. We're basically comparing between two groups and trying to see which one is better than the other. And I think, I think just, just for the purpose of hypertension or something like this, this is where the concept of something called clinical equipoise comes in. I we genuinely don't know which one is better by doing it in the morning or doing it in the evening. And most of the guidelines state the morning and that's due to the idea in hypertension, at least of your BP is probably highest throughout the day. So control it during the day, but there is no evidence supporting that really. And this is what this trial is asking. Um, so, so although it's not a kind of controlled trial in the same way, kind of around in the same way, like, you know, placebo versus something else is, um, it is still, you know, it has a valid question to answer. Ie does one is one better than the other. Ok. So I've included another, you know, PCO framework that I found online. Um And I think this is something that you can use in all of the, you know, all of the stations when you're asked to critically appraise a paper. So, you know, by looking at an abstract, you can use the PICO framework to actually see if, you know, the authors of this paper have actually answered the, the question that's been asked. Um So, you know, if you get a paper about, you know, um whether gabapentin is better than placebo in decreasing pain symptoms in patients, you could use a PICO framework um to, to sort of work through whether the study is valid or not. Um So, you know, population um again, would be middle age if you do get a paper, you know about this, it, it could be um um the population would be, you know, male amputees suffering from phantom limb pain, um, intervention would be gabapentin. So they, they want to compare gabapentin with, um, the control with which would be placebo. Um And the outcome would be, you know, would it be more effective in decreasing, um, the pain symptoms? Um, yeah. So I just wanted to quickly touch on um, the differences between a randomized trial and a randomized control trial. Um, because I think most of the time, um you know, we, most of the time people tend to see randomized control trial. Um So I think the main difference between randomized trials and randomized control trials is that for randomized control trials, you have some um you know, control um an existing practice that you want to measure the new intervention against. Whereas for randomized trials, you want to compare between two groups um to see which one is better if one is better than the other. Um Yeah, so moving on to the third part of the framework um conclusion. So after going through the PICO framework, um I usually just do a short um brief conclusion of the study and that can usually, you know, be easily found on the um abstract itself. Um And the conclusion from the study was that um it showed that the evening dose of antihypertensive medication was not different from morning dosing in terms of improving major out uh major cardiovascular outcomes. Um So, you know, regardless of what paper you get given to, I think, you know, having a short conclusion at the end of the PCO framework would be quite um nice and it would be, you know, quite a nice way to sort, sort of just wrap things up. Um Yup. Right. Moving on to you, Robbie. So, so we've spoken a bit about presenting the paper and how you would approach what the paper actually said within PCO. But the next level above that is, is what the paper has done actually valid and is what the team or the authors have done actually worth considering in future. So, so you then start to think about comparing it to either other potential trials or actually looking at how the trial is designed or made. And now, as we kind of said on the kind of structure slide, the next thing after PICO that we tend to use was sort of strength and the weaknesses of the study. So talking about strength of the study, what makes it a good study? What makes it work? And is it clear in what it was assessing? Was it retrospective? Was it prospective? I was it designed 10 years ago and is looking forward and, and, and how did it assess it? Was it using things like surrogate end points um or multiple endpoints? And is it valid? And is it and does it work for what it's trying to show? So say, for example, this study, it was looking at BP and BP readings at home and the control of BP readings at home for people in these two morning and evening groups. Now, how this study did it was with home BP monitors. Now, we know that there's a level of an accuracy with them and it was also done with self reporting. So people had to type in what their readings were every evening and, and, you know, I think the strength of this study is that they were actually quite harsh in their exclusion criteria and they ended up ex seeing quite a lot of people and monitoring them quite closely. But perhaps a slight weakness is sort of how data is recorded. Is there a bias in there and what you're looking for in a paper, if you're able to look into that much detail is how do they control that? How did they discuss that? And this study actually did some quite nice statistical controls around the people who were or weren't adherent and which was nice to see you. And it's kind of a question that you're looking to ask to see when the ship team has thought about that type of thing. We just move on to the next slide if that's OK. I think this is you again. Yeah. Yes, I just wanted to quickly touch on internal validity and external validity. Um So when you're talking about, you know, strengths and weaknesses in the study, I mean, to be fair, this light could have be put, put um lower down. But um you know, when you're talking about strengths and weaknesses of the study, it's important to think about, you know, um internal validity and external validity. So just as a brief summary, internal validity means, um you know, if it, it just basically want it, it asks the question if the study actually tested the hypothesis, which is, you know, in this case, is one, is one, you know, one is morning dosing better than evening dosing. Um And when we talk about external validity, um we ask a question if the results of the study can be applied to real life scenarios. So um thinking, thinking about internal validity, um things that would affect internal validity would be, you know, biases. So when, when, you know, researchers, when they are selecting people, um you know, are they selecting people, you know, do, do the, do the research is, you know, maybe somehow for whatever reason, want the morning dose of antihypertensives to be more effective than the evening dosing. And so they select people who sort of adhere to, you know, morning dosing a bit more effectively. Um you know, multiple know, like different types of biases um can affect, you can weaken the internal validity of the study. Um So usually, um you know, if, if we see that a study has used, you know, things like blinding. If we see that a study has you know, a predetermined protocol that the researchers adhere to and that these are things that increase the internal validity of the study. Because if you follow a predetermined protocol, it just means that, you know, the researchers won't be, you know, changing the way they do things halfway through the study, which might affect the way we measure the outcomes. And obviously, you know, it might, um, you know, affect the findings of the study. Um And then you, you also have, you know, um things like um reporting bias where, you know, um when you're reporting the outcome of the results, um you know, in this case, the patients had to sort of, you know, report the blood pressures that they measure it um every day. Um you know, there could be an element of bias because, you know, it's, you know, maybe pe different people sort of read it the wrong way or maybe they just weren't measuring it the correct way. So I think there are loads of things that could affect um the internal validity of the study. Um And I think these are things that you have to mention about when you're critically appraising a paper. Um So moving on to external validity. Um so these are things which basically, so this is basically, you know, can we use the findings of the study to apply it to the real life um scenario? Can we, you know, use the findings of this study? Um to change the way we treat patients. Um So, you know, things that affect, you know, external validity would be patient. Um The, the main thing that comes to mind would be, you know, the, the the patient, um the population, um population size and you know, the variety of the population. So if you look at the study, it was um mostly white males and it was a study that was done. Um if I'm not mistaken, well, mostly in the UK. Um hypertension exists um everywhere, not just in the UK. And, you know, so I think it just, it just makes you think, you know, these, these um findings were done based on a UK um population will the findings of this study be applicable to, you know, people elsewhere around the world? But there, you know, you know, with people of say, maybe different ethnicities would uh an evening dose be more effective than the morning dose probably. Um So, and also I think, you know, other things to think about would be um, you know, like patient adherence to um the medication, you know, do patients maybe take, um you know, um do patients maybe adhere to their medication regime, you know, more in the morning compared to the evening? Um So using this, you know, the results of this study, I think these are things that you have to think about in, in terms of determining whether it's, you know, applicable in real life. Perfect. And the next thing to think about just following on from that is weaknesses of the study and how that and analyzing what is, or isn't a weakness of the studies. Now, every trial paper that you'll see that's published at the end of it, there will be a limitation section normally at the end of the discussion and if you have the full paper read that because that's your cheat sheet. Um, and it's worth having a look at that. The things that are generally seen though throughout every study, there is there's always some form of bias somewhere and there's always ways that papers do or don't control for them if they can. So, so say, for example, in this study, um, it's at risk of something called the hall film effect, which is the idea of if you're being watched, you're probably going to change what you do. Um, and effectively the idea here is the people who are part of this study know that they are in the trial. If they know they're in a trial, looking at BP medications, it is not going to change whether or not they take their BP medications. Um, and that's a bit of a problem because one of the biggest problems in BP management and complex BP management is people not taking their medications and not being adherent to their treatment strategies. And now that to some level, by virtue of this being a trial probably is a limitation of it. And, you know, it would be worth comparing the adherence states they have here versus sort of national adherence state and seeing whether or not that's applicable. Um, as we've kind of said already, this trial does have quite a nice strength in the fact that there's 21,000 people involved that is huge. Um, um, and, you know, like, you're just not going to get trials that are bigger than that in, in sort of single countries, er, even for things like repression because those populations don't really exist elsewhere. But, you know, there are, there are problems with it, as we've kind of mentioned, it doesn't cover all populations internationally, so it probably has some applicability to UK maybe us, maybe European populations, but less applicability to places like, say, for example, Japanese populations which have um other hypertensive epidemics. And similarly, this study is at risk of things like recall burst. As we kind of mentioned earlier, people were having to input their own BP monitoring, which means there is a risk of people just putting in BP incorrectly. And one thing this paper doesn't talk about is how it managed to adjust for that for people putting in things incorrectly. And so, so that's, that's kind of a thought in here of when you're looking at a paper, what biases does it have? And does it talk about how it controls for them? And if it doesn't, does that affect what you're looking at and how it responds sample size is a very common one that people talk about. Um, and that's when the concept of things like power comes into play and it's probably worth having a quick read about if you're interested in trial design. Um, but most of the time sample size, most of the time isn't actually the biggest limitation of a particular trial. Um, because if, if a trial has gotten to the point where it can be published and funded fully, it's probably got enough people in it probably, um, that gets a bit different when you start thinking about cancer trials and things. But, but most of the time sizes a strength rather than the weakness for those things. Cool. So, and just kind of pick up one other thing. So as we kind of said, this trial is large, very large, but one thing to kind of be aware of in the evening dosing group. And this is some of the papers that you pick up on in the evening dosing group, 202 111 more people withdrew from the evening group than the morning group. Um Now that, that's interesting and they actually didn't do a statistical correction for that because they, because they'd already done censoring elsewhere for these patients in this study. So there is, there is a question in there. Well, did the people withdraw from the group because of you know, lack of tolerance or issues with adherence, they just don't know. Um So there is a subtle limitation in there. But as I said earlier, if you have time to read the whole paper, it's in the paper and cool. So, so moving on. Yeah. Um So this is a, a consult flow chart and I thought it would be good to just add this in because um usually the, it, it's a flow chart that most, you know, well, no, most, you know, when you're doing a randomized trial, it's good to know how you select the people and to do that, it's good to add a con con consult flow chart, which is, which, which is something that most papers would, you know, publish and because it shows very clearly how the participants were, you know, selected reasons why some were not selected. Could it be due to age or were they, you know, taking more than to antihypertensive medications or were they just not tolerating the antihypertensive medication? So a consult flow chart shows you very clearly how the participants of the study were selected. Um And you know, with, with the con, with the consult flow shot. I do think that it, it's sort of, you know, it, it um shows the internal validity of the study because it shows you, you know how the sample population was selected. Um So, you know, it just makes, you know, the fact that the researchers were testing the hypothesis better a lot more. Um Yeah. So how will this change your clinical practice? Um So this is the um the sixth part of the, well, in like, in terms of framework. So after going through, um you know, PCO strengths and weaknesses, I usually just um do a short um you know, wrap up with, you know, whether my, I would change my clinical practice based on this paper and based on the results of this paper. Um so usually with cli with, you know, for, for something to change your clinical practice, obviously, it can't just be based on one study. So most of the time your answer would be if you were, you know, presented the findings from a NC in the interview, I would say most of the time your answer would be no, because you can't just change your clinical practice based on the findings of one randomized control trial. Um due to the fact, you know, because of all the other factors that we discussed earlier on. So if you look at this pyramid, um this is something that I would strongly suggest that everyone look up and sort of, you know, do a bit of research on. Uh So this is a pyramid that shows um quality of evidence. And so you have different types of studies and I would strongly encourage everyone to look up on each type of study and sort of learn the strengths and weaknesses of each study. Um So at the bottom end of the pyramid, you have observational studies. So these are studies that, you know, collect data, collect observational data um without, you know, actually analyzing them. So these are just, you know, observations um and you know, written up into a clinical paper and then, you know, moving a step up, you've got experimental studies and this is where um you know, randomized control trials um falls into. So these are um your studies that you are comparing sort of, you know, between two groups, like the one that we um are using for this paper, you know, morning dosing and evening dosing or, you know, maybe comparing say gabapentin to placebo in terms of treating neuropathic pain. Um So, and then moving up from here, you have the more um well, the sort of um critically appraised ones um like media analysis, systematic reviews. Um So these are types of um studies that gather information from the observational studies and experimental studies and they analyze, you know, tons of these studies and, you know, sort of come to a conclusion as to whether there is evidence that, you know, treatment works better than the other or what. And so usually, um well, correct me if I'm wrong Robbie. But I think usually when things are influenced, the nice guidelines would be stuff like me analysis. Um because the um you, I mean, usually it won't be just a randomized controlled trial paper that you would say that, oh, we're making everyone, you know, take the morning dose of like antihypertensive medication because it would be, you would need basically loads and loads of evidence to prove that one is more effective than the other before we actually implement it in real life and encourage patients to well, take that, that medication. Yeah. I, I think, I think for most things, that's the case and it's the type of thing where for common, common disease is answering big, big big questions, say for example, is medication a better than nothing. And then, yeah, absolutely. You're looking for multiple trials and multiple populations with reproducible things. But the thing that's also worth thinking about is whether or not it would be practical or possible for there to be another trial showing the same thing that somebody has already done to make a metanalysis out of or multiple trials to make a meal out of. So for example, the only, the only bit I've actually got any experience in clinical guidelines for is in pediatric leukemias and you get a couple of 100 pediatric leukemias a year and you might get 10 people a year with a specific subtype of a, you're just not going to have internationally. There just aren't the numbers to make multiple clinical trials to do meta-analysis on. So in that type of scenario or say, for example, with this trial where you have 20,000 people, you've got probably enough or you would expect that the people looking at that would might have enough evidence to answer that question sufficiently to say yes, this is the crack in this population or no, this isn't the case. We need more data. So it kind of depends to some level on what the question you're asking, how big it is and whether or not it's practical to get other evidence or other data in other areas and whether or not you need it. And certainly for most things you would expect it um particularly for things like surgical interventions or other things, assessing outcome because there's lots of variability within that. But for large scale studies or national studies, some of the time that can be sufficient, but it really depends on what the study is looking at. Um I, I think the next we're going to talk about is how or kind of a little bit about guideline change. So it's kind of as we've already spoken about. Nice. Nice is kind of the gatekeeper in the UK for how clinical practice happens. They are the group that decides where money goes and what treatments will and might be approved and how things work in the UK and how, how services are commissioned to some level. Um Now that's fine for the UK. But something that's interesting about night is it takes a very, very long time for the vast majority of things to come into effect through nights. And I think often the lag time is usually about 8 to 10 years now. That's not just the slow cogs of, er, sort of, you know, a big an organization like NHS or noise making stuff happen. Part of that is because nice operates off the idea of evidence but then also practitioner experience. So the people on these nice evidence panels that make decisions about care, they will hopefully have experience in their particular areas. What that means is not all the time when you're looking for evidence or if something has or hasn't been approved, it will be on nice. And so it's always worth looking elsewhere. And from a quick Google effectively of whether or not this paper has been used in other guidelines. It's been used in Korea and it's been used in Japan and it's also been used in the European Society of Hypertensions Guidelines, which is on the next slide. Um and something that is useful. And if you get time to look at a paper beforehand, so if you are say, for example, given 24 hours to look at a particular paper, I think previously the recovery trial has been used within CO V ID. If you get time, have a look at where your papers been cited because this paper has been used in the European guidelines, which is like a 200 page guideline document. But on page 66 this is the entry for it. And there's two things I want to take away from it. One it talks about this is a critical appraisal. I'll mention that on the next slide. But two, they actually say whilst you can experiment with giving nighttime dose of antihypertensives, adherence to antihypertensive medication is better or is better in the morning and worse at bedtime says these other three papers. Now that's interesting because that's something that you wouldn't necessarily see from reading just this paper. So one thing that is worth doing is reading around the area a little bit, seeing who else has spoken about this paper. Um And also if we go to the next slide just briefly, um, here is your PICO, um, and your, and, and if you do have time to read around the area and look at where else your research has been used and what other people have been saying about it, you'll probably get most of your answers yourself. Um And it leaves in quite a place to say. So I suppose the take homes are make sure you understand what your paper is saying, make it simple and read around a bit if you have time to because it's all out there. And most of these, most of these groups will use well established papers with well-established outpoints and sort of not incorrect answers, but with conventional ways of thinking. Ok, so we're moving on to our last few slides. Um, I mean, due to the, um, well, in the interest time there were some things that we weren't able to cover on. Um So I would suggest um for the interview, um there are a few things to read up on. So study types um go back to the pyramid that was in one of the slides and the hierarchy of evidence. Um you know, just have a quick read through of what open label means and what control trials mean um parallel crossover and biases. So these would be um your things to read up on because you, when you're arguing, you know, on, in terms of the strengths and weaknesses of the study, you need to understand mention, you know, some form of bias that's been either addressed or not addressed by the researchers. So have a quick read through on, on what selection performance detection attrition biases and, and how we can overcome each type of these biases. Um And also something to mention um is um selective reporting outcome bias. Um So do read up on that because it's um it's basically, um you know, when researchers choose not to report a certain type, um a certain outcome which might make the results of the study seem a lot better or worse than it actually is. Um So do read up on that. Um me methods of randomization, um you know, have a quick read through of what 1 to 1 and 2 to 1 randomization mean. Um And types to um of how we can randomize participants, um analysis of results. Um So some basic stats and intention to treat or pro pro protocol analysis, complete case analysis, um complete case analysis um and basic research knowledge. So I actually had a few questions asked in the interviews in terms of um you know, P values confidence intervals, null hypothesis. And what does 95% confidence interval mean? What does risk ratio mean? Um So I, I would say do read up on that and yeah, we've touched on the hierarchy of evidence. Um Yeah, so useful resources. I personally use the um past test book, the Dark blue one and the right. Um Do you use the purple one, Robbie for the interview? I think you're on mute. Yeah. Yeah. No, it's a nice paper and talks through not just kind of like big trials, it also talks through things like qualitative and economic papers which you are probably fair game. So it's worth having a look. And also if anyone likes Twitter, Trish Grand is great on Twitter and does a lot of C and stuff. So follow her and it's free as well. Fine. Right. So I think that's the end of our presentation. Um Does anyone have any questions? Cool. Um So before you leave, um I was told to let everyone know that we're actually running um a mock interview session and it won't um if I'm not mistaken, it will be on a first come first serve basis. Um I'll pop the sign up, um, form in the chat. Um So these mock interviews will be running, um on the third of November and the fourth of November. And we've got about, um, just over 20 interviewers who signed up to do it. Um It will be, you know, a very good, you know, experience in terms of, you know, preparing for the interview. So I would strongly suggest that, um, you know, you sign up for it if you can. Um But yeah, um we've got a feedback form for this session um in the chart as well and I think we would really appreciate it if you could give us some feedback on the, on the session today. Um Just so, you know, we can think about how we can improve things in future sessions. Um Ask you, we can have access to the slides, please. Yes, of course. Um So if you, if you um fill out the feedback form and I'll um send you these slides. Um One once we have your email. Um Yeah, we can certainly um send you the slides. Um Any other questions every morning mock interview, sign up asks if we have been shortlisted mind. Yeah, you can still sign up. Um Yeah, we do understand that some, you know, some deaneries still have not released um interview offers. Um But you, we would still encourage you to sign up um because it will be good practice um for when you do get an interview, it's also worth saying that we're running multiple of these sessions. So basically throughout November, I think it was running in December as well. So, um, if you don't know yet, there will hopefully be a session that will cover you. Um, so yeah, just sign up for ones you think might be relevant and then we can kind of adjust. Cool. All right. Thank you so much everyone for joining us. Um And if everyone um if you do have an interview, um good luck, um you can just drop us an email or messages on Instagram if um you have any questions about the critical appraisal if you've got any questions on what we spoke about today. Um But yeah, thank you everyone for joining us and we'll see you for the next session, which will be the A two E scenario. I'll put my email address here. Um If you've got any questions. Right. Right. Thank you, Robbie. Um And thanks everyone for joining us. I'm just gonna do a polite hang around for a minute just in case anyone has any questions as willing in the door. I think I also potentially mentioned about talking through a graph. Do we still want to do that? Sorry, I just remembered. No. Ok, understand. Um So just for the people who are left, I've just got a message where you just to let her know. Um um Cool. So I'm gonna talk about the survival curve that's in that, uh, paper that we've spoken about. I've got too many tabs. Uh, there we go. Um, so I'm assuming people can say this because I have no way of checking if you can count. Um, and so this is the, this is sort of the survival curve and it's the curve of hazard ratios by time of the primary end points of, you know, severe cardiovascular compromise ie having an MRI stroke, that type of thing. Um, by time, now what my first question and I'll have a look in the chat, I'll have to share it is. What does this graph show? What, what does it tell us if people just want to put answers in the chat? Yeah, I'm just checking that I'm just gonna see if you guys, ok, back onto my screen just if somebody could put like a bullet point of what this does or doesn't show outcome. Yeah. So, so, um, so yeah, who, who, who, who is exposed to what end point with per particular time? Absolutely. Um, now from the graph that, um, does it show? And actually if we is back on the call, would you be able to read out what people are saying in the chat? Just because I can't say, um, how long in time is this data reliable for looking at the follow up years? And I really want my question is what's happening around years? 6789 and 10, what's going on around here? Someone said loss of follow up people dying due to age, et cetera. Yeah. So, so we know we know from the concert diagram that about 303 150 odd people died in each group. Um So the answer is really follow up time and it's something, it's something to always be aware because you might well get shown a survival curve in your critical appraisal. And it's always something to be aware of when you're looking at survival graphs is what you want to see is really smooth incremental changes. What you don't want to see is when you have things that are like big boxes and what each of these sort of little adjustments are up here is that either somebody dying or somebody dropping out of the study, um, and that, or, and, and affecting you, what that ends up meaning is, you know, you can probably trust this data up until about sort of five or six years. But then after that, it becomes a lot harder to assess. The other thing that I quite like to see on, on, on these types of graphs. What are effective these fiber curves is competence intervals. And now I think it's, now we've got this sort of unadjusted hazard ratio. Um It doesn't really say when it's put in place, but I think it's probably fairly safe to say it's here somewhere and it could be a mean comp as well, but, but it doesn't really say when I'm here. Um But it's probably safe to say that the competence will overlap here. So there isn't a, you know, there isn't a significant difference between these ones, but when you get to here, it becomes a bit harder to tell. Um And what you would see around sort of year six or seven is these comp inter being very, very broad. Um And it becoming a lot harder to work out what's going on in sort of the latter stages. The other thing that always, always, always when you have a um survival curve is make sure you've got the number, the number that you're looking at. So we can kind of see from this, that you know up and virtually everyone in the trial gets about 3.5, 4 years of follow up, then a few people start to drop off and then there's only about 500 people from only about 500 people and then about 200 people in each group from seven years onwards. So I would say that from a very quick look at this, the reliability of this data is probably up to about year five year six if that makes sense. So that, so that's, that's something that's useful to think about when you're looking at cyber curves is what does it look at? You're looking for it to be a nice smooth curve if it isn't worry a little bit and what is it really telling you? Um I do, I think I might have another one open if I want to talk through something else. Um Any other questions about that or any other questions about specific graphs or stuff, stuff like that, people want to talk about just happy to if we talk about or anything else. Um Wing wants to add to my three simplistic overview of survival gift. No, I think you did a really good job of explaining that. Um I think that was like that was something that I didn't even. Um well, I didn't take note of because I think um when you look at the, the morning and um I think I, I think the um the, the reliability of the data wasn't something that you all just mentioned. Yeah, I it talks very briefly about median follow ups and things like that, but it doesn't really talk about how long people followed up for. You can kind of see from that, that we know that not all of those people died because they did there. Um But yeah, so, so it's interesting, I'm just going to very briefly if I still have it opened, I don't have a date. So I'll talk about forest plots and I'm also going to talk about um but I'm just going to very briefly talk about a different survival curve from some other data which might be useful to talk about. Um and just to highlight um so that day to time. Ok, fine. Um OK. I'll talk about forest plots. Uh just somebody say some things whilst to try and find a forest plot that's useful and interesting. Uh doctor for analyzing graphs. Um I think my nurse asking about this. So what I was um what I was told um was that when we usually it's more for a CF level type of, of, of types of interview where they ask you to analyze the graphs. And although it's something, you know, that's really useful to know, I think um from what I've been told, usually for sfe interviews, they the most they expect you to know in terms of like, you know, graphs would be, you know, the type of graph. So you've got the copy my gras or forest plots, they might ask, you know, what, what kind of graph is this. Um And um you, you basically um expected to know the different types of new graphs in which um data can be um presented in. Yeah, absolutely. And I think um you might get presented with a forest plot or you might get presented with a paper with the forest plot in. And I, but you should have time to look at specifically what you're actually looking at. I think the thing that's important is they're not expecting to be an expert in whatever paper they give you if you are great. But if you're not that's fine. It's just, can you pick up the General INS one and two? Can you present? Which is actually probably the main thing they're assessing. Um. Right. I'm really struggling to find a good one. What's a good one? I trying to think. I can't think of any child that for, that's really bad of me, isn't it? Um, that be what? Oh, no, because that was a single. Uh, I don't, yeah, I'm really bad at this. Um. Ok. Right. Cool. Found one, right? Ok. When it says to lo it's not loading, correct? I took five. Yeah. Oh, that's a crap. Uh Sorry. Ok. Any other graft types, forest plots you want to talk through? Um, let me, um, just give me one second. Does anyone else have any other questions? Just while I'm frantically looking for something? We haven't got any more on the chart, uh, in case uh I've got one, I've got one that sort of looks. Ok. Yeah. Fine. I'll just go. Cool. Right. So this is a forest plot sort of it. It's not as pretty as the ones you get through that cock collaborations but it, it works. Um So this a very quick look is looking at is multiple forest plots stacked and, oh, it doesn't use different studies. Oh, ok. Fine. No. Hang on. This is a bad example. Ok. I can't find one under pressure. I'm sorry. Has way you manage to find one that isn't crap. Um, I just saw lots of lines and it said overall, I was like, oh, yeah, that looks good. Yeah. So I did look at that but I don't know what it's about. What is it about? Give us, give us a summary. Do you have, ah, there we go. E right. Ok. I think I found one on the study. I actually know a bit about so I can talk about that. Um, so this is, this is a forest plot. Um that's looked at vinCRIStine pulsing in pediatric acute blastic leukemia. Um And e effectively, the idea of this is so how leukemia treatment works basically internationally is that every child who gets leukemia is treated in a clinical trial and every trial lasts for about 5 to 10 years. So the one that's going on at the moment is the altogether trial. The one before that was the U A 2003 that started in 2003. The one before that was like, I think it was a 1997 99. Um So, so basically everyone in the UK and the US that gets leukemia gets treated under a clinical trial and that means we have their data. Um And what this is is a comparison of giving vinCRIStine, which is the chemotherapy quickly with a lot of steroids in the high pulse group or slowly with not very much steroids in the low pulse group. And now kind of what I was saying earlier is look at the years going through and then outcome of what happened to them and event free survival. So what you get in a forest plot is you get the hazard ratio or world ratio of whatever outcome you're looking at and in this case, it's event free sur valuable um from this one historic treatment, this one convention treatment. Um and that's your kind of ratio and then you get these whiskers um which are your, your normally you're 95% competent interval. Ie how confident are you that the real mean or the real hazard ratio you're looking at is between these two lines. Now, there's a few things to note within these trials. The first one is the ones with the widest competence schools in this case, have the smallest study populations. So, you know, there's some ju trial 47 in the Inn trial like 30 really wide competences really hard to draw significance. Whereas in CCG 161, they had, I think actually overall they had about 800 people in that one. So, so worth noting, um so 12, there's usually a mechanism for waiting and the idea of forest party is so you can compare and combine what different trials say. So there is usually a mechanism for weighting different trials based on normally how many people were in that trial. So say, for example, if we look at this one, er the 95 trial that had, you know, like 2, 2.5 1000 people in and that has a bigger square than say, for example, er, the one that I've never heard of the, that one, which has like 400 people in which has a little square. Er, I think it's also, again, appreciate, you can also appreciate the fact that there is quite a narrow comp for the big trial and a wider comp for the little trial. Um So that's, that's fine. Now, what you then do is some very fancy statistics of where you combine all of those findings and you combine those hazard ratios, do some fancy things with effect modeling and you get an overall an estimate of what the real hazard ratio for a population will be given all of the various study populations you've looked at. Um now in this cohort. So in the one above, we can see that the hazard ratio overall was naught 0.79. So, so you were less likely to get a bad outcome, more of it. You had a longer event free survival with uh high pulsing the versus low pulsing therapy here, competent interval 1.68 no 0.91. So still still a fairly wide competence interval but but but doesn't cross the one line. Whereas in the contemporary therapy, the kind of a modern day therapy which is what's done at the moment and it was no 0.96 has a ratio with a lower bound and ac I of 0.85 a higher band of 1.09. Basically ending up, meaning that in this case, it wouldn't be a statistically significant finding. Let see the P value here because you are not 95% sure that your finding is correct. That makes sense. So, so from the fact that it goes over to 1.09 it means that there is a level of doubt that actually your finding probably crosses over to the other side of that line. So that doesn't really tell you what you should or shouldn't be doing. Does that make any sense? Roughly the answer can be no? Cool. Right. So that's a bit, that's a bit more of an explanation of forest plots again, like, yeah, again, I they might present you with a forest plot. I doubt they would give you a meta analysis because I think it's probably, yeah, read the graph and says what it says, yeah, that's the point. I mean, the point of the paper is to get across information. You're not expected to be an expert off the back of it, read the graph and says what it says, it's really not that difficult. I don't think you'd be expected to interpret meta analysis. Um because that is a bit fancy and the statistics behind the meta analysis and other effect modeling is hard. Um Whereas you probably could be given a survival, survival plot and ask to interpret that. Don't, unless you're really interested in medical statistics don't go into like, ka versus hazards versus log testing because you will want to die. But you probably would be expected to understand that one line is lower than the other and the comp doesn't cross, therefore, it probably is a significant finding at that time point. Um, is that fair enough, do you think? Yeah. Cool. Right. Any other questions before I go and have the scan and cry? Because I'm on nights tomorrow. Uh Sad. Yeah, so someone just messaged me actually during the session be like, hello? Are we about, are we starting nights tonight or tomorrow? And I frantically wrote to checking because I was like, I'm pretty sure it's tomorrow but it's ok, it's tomorrow. Um Yeah, fine, cool. Any other questions for anyone else in the? We haven't got anything in the chart, but thank you Robbie and thank you everyone for joining us. Um Yeah, we would really appreciate feedback. Um Yeah, because it really would help us know in terms of thinking about what to improve for future sessions. We've got about two more sessions um running. So those will be in the, we'll post it up when they are. Um It will be on the a tec station and then the last one would be the in personal station where they ask you why you want to apply to our deary and you just go through and how to answer these questions. Um But yeah, it looks like we haven't got anything else in the chart. Um So I think we can probably leave the session. Cool. Thanks very much guys. Thank you.