Computer generated transcript

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

I so girl let me some liver injuries. No. OK. Sorry to check for. So Helga, you can start sharing your screen so long. OK. HGA. I think profs might be busy in theater. Um I think he will join as soon as as possible. I think we can, we can start our meeting so long. I don't think he will mind. So. Good afternoon, everybody and welcome. Uh today, Doctor Hil House uh register is going to talk about common res research language. Um uh Doctor Man, our consultant has guided her to prepare the talk and we are expecting Professor Sad Sheikh, president of the College of pediatric Surgeon to join us as uh invited guest. So Hilda, you can start your talk. So now thank you. Good afternoon everyone. My talk, as you can see is on common research language. Over the time that I've tried to prepare this talk, I realized that it's a giant talk and I've tried to keep it quite summarized. There are a few aspects that I haven't covered in this talk. But yeah, let me show you what we will cover and um talk through some um aspects of common research language that we come across when we read articles or prepare our MS. Um So first we get a whole bunch of different types of studies. There's observational studies which are well observational. We just see we what happens. Um I'll go through in detail those four that are listed there. Then we have experimental studies that always involves uh intervention for us. They're usually clinical trials and they're usually action and control groups. But we'll also go through a bit more detail in those. Um Then we have meta analysis and then systematic reviews and I will go through each one a little bit more in detail observational studies as you can see has basic four groups. They all don't involve any intervention at all. It's just a purely observation of what is actually happening in case series. We just report on whatever cases or maybe even multiple cases that we've seen. Um sometimes audits fall or audits fall under the same category, but a little bit bigger where we actually just check what are we doing at this point in time and they often the start of research. So the very first point to see what is the status quo and where are areas that we struggle with that we need improvement on um then do an intervention and then follow up on that case controls of um also retrospective reviews usually involve two groups and we compare those two, but it's retrospective, there can be bias in those groups. Um Yeah, not is, is the next step of trying to see where we are, then cross sectional studies are different. They, we usually come across them as surveys where from all sorts of areas get, send a survey, please fill this out. How do you rate this or that on a scale of 1 to 5 or 1 to 10, whichever one they send us. It's a snapshot of what is happening in time. But a single snapshot doesn't um compare anything over time and then cohorts of um a group or sometimes groups that share a common trait that's followed over time. They can be prospective or retrospective, prospective. Um For example, would be if we have appendicitis patients that we admit to hospital and we follow them up and see which group develops complications and try and see what is the traits of each group that predisposes them to complications or no complications. Retrospective for cohort then would be the same two groups, but we identify the problem later. So we identify, let's say for the appendicitis patients, they have developed complications and we now do a review on their notes and see what are the differences between the two. Why did one group develop complications? One didn't, um these are usually cohorts on databases. And since we have started a database for our department, that's probably a lot of the research that we are going to do. How do you know what is observational study or sometimes it's pretty obvious as you can see enteric duplication in Children and case series. It tells you exactly what it is. Sometimes it's not quite as clear from the heading and we actually have to read through the methods of what was done in this study on upper gi study on operative management of reflux disease. In the methods aspect, you see it's a retrospective analysis. Um and it's a retrospective cohort. If you read through the full article, that's the easy way to see what is the one, what is the other? Um And you can start grouping, which um what article you're reading? What study does it fall under? We'll go as we go through the talk, I'll explain more why that becomes important. Then we come to experimental studies. Like I said earlier, there's always a intervention involved and they are usually clinical trials in our settings. There's a few aspects that form part of an experimental study. One is the controls. Um It's usually two groups can be multiple groups but often two groups, one with a new intervention and then one with a sham intervention like a placebo drug or compared to a normal routine intervention. If we compare it with a new antibiotic treatment or something, then you can have multiple different controls. One would be a self-control where you use the same group of um patients and you uh review them before and after each intervention or you can have a crossover, which is two groups of patients. The one gets the intervention, intervention A first, the other one, intervention B first. Um and then you cross them, they do their own self-control and then you cross them over and do the same study with the other group again. Sometimes we see that if they try and see. Um well, the study I came across with had um anti cholesterol treatment and A compared group A group B, group A had the old treatment. Group B had the new treatment and then controlled them over a period of time and then crossed them over and did the same study. Again, it's relatively easy to get more numbers into a study. If it's um limited in terms of numbers, then there can be external controls. Um Often the historic controls, for example, if we come up with a new treatment that has never existed before, we then just control against what has been there previously on previous records, we can also have uncontrolled experimental studies. However, the a little bit frowned upon as you can't rule out bias and then the gold standard is a double blinded randomized study. So what does blinding mean? Well, we can blind patients, we can blind their caregivers or we can blind the observers or we can blind multiple of those. So does the patient know which treatment are they getting? Do the caregivers know which part are they giving or do the observers know what's happening. And a double blinded trial would be that patients and observers are blinded and only the caregivers are the ones that know which intervention who is getting this is done to try and minimize, minimize bias during the study and get more objective results. And then randomization um is required to try and get an equal spread of patients that represent the actual population more accurately. There's different ways to try and randomize and pick patients. I'll go through that in a second. This again is done to try and decrease bias um and to get the more accurate result that will then reflect, reflect the actual population's parameters. But always keep in mind if you read any studies, what are the biases are, they're often hidden and not completely um easily seen when you read it through the first time. But if you read through any study, look what could be potential biases. Um so that you can keep that in mind when you interpret the results, then meta analysis is considered the next um higher level of evidence. This is um done through a very wide and all-inclusive literature research of a specific question that we might want answers for and then look what other studies have answered and combined that data. This is often done. If there's multiple small studies you've been done throughout the world to try and combine their data, to try and see if they become si statistically significant, they have to be rather specific in their inclusion and exclusion criteria so that we can actually be sure that apples are compared with apples. Um These are the specific inclusion and exclusion criterias make or break the quality of the paper. And then when we do statistical analysis, we again have to be careful what and how the variables were compared um to see whether that is actually a true reflection of what's happening in a population. I've quoted here one article on laparoscopic anterior versus posterior fundoplication for gastroesophageal reflux disease. It's a systematic review and a meta analysis of randomized trials. But the one big critic to this article is that they didn't differentiate between the different, especially the different Posterior Fund applications that are available and they just grouped it anterior posterior and not the various versions that are available. And therefore there might be bias in their conclusion. And that's just to highlight that we have to read these articles um with the investigative mind and think about what they have written and how does it include and exclude certain aspects. Then on systematic reviews, they are often considered the highest standard um of evidence. It combines qualitative and quantitative research. It is considered the best summary of the topic at that specific point in time. It also includes an extensive research through the literature, but it's not just analysis, it also includes a discussion and a recommendation at the usually at the end of the um articles and it's relatively easy to pick them out. Like this example here, surgical management of Crohn's Disease. A state of the art review will basically says what it is and we know that this is a systematic review. Then just come to some more definitions. We have descriptive statistics. Um They basically just give us a summary of what we've seen in our study. So these includes averages modes. So what is most often or number of boys or girls or number of appendicitis patients that we have seen? But we can't do any real comparison between the uh yeah, between the param or the statistics that we have. Um So we just describe what we see in Neos current emet. That's basically what she's doing. She's describing what she's found. Then we have inferential statistics, they there to um we have a sample, we analyze it and we then want to infer and see if that is true for the entire population. Um Just some definitions to that population with whatever um aspect we study is called a parameter in a sample, which is a sample of the population which um whatever data we um study and then do mathematics on what's called a statistic. Keep in mind that the sample, the accuracy is linked to how well it represents the population, which is sometimes quite tricky because often we don't know what is our population's um default and we're trying to study exactly that. Then variable is the group norm for any data value that we're looking at. And then a data point is the single value of a variable that we're looking at. Just to understand what is going on. If you read through your article, what do these different things mean when they talk about it then coming to some more data types, they're basically categorized in two big groups, categorical and numerical categorical. Well, the name says it, it's basically categories. In those two, we have two categorical ones, nominal ones. It's basically things that we can't really quantify any other way. There's, it's gender, it's colors, it's diseases, we can count them, we can describe them, but we can't really do any more analysis on them. Then we get ordinal categorical data um which is also still things but they have some order to it like pain scores or like a style um questionnaires where we are, we are asked to rate something from 1 to 5. Um these have orders so they can have um can be rated a little bit more than just normal, plain nominal cate data. The numerical data is the one that we can actually do some maths on. They grouped basically an two and out of those 22, again, the interval ones um uh numbers that have meaningful increments of difference like temperatures where you, it's one degree and two degrees higher. And you can actually say from one up to your next up is the same degree going up. So the defined increments as things grow up, they're not percentages of each other of 1% 2% going up. Then racial and numerical um data has a true zero. Um where baseline can be established between our data value and the absolute zero. This is for example, age or white cell count, you, you can't be minus you. There is zero age, there is zero white cell count. If there is a minus white cell count on a um blood result form, I will question what the lab has been doing with the blood. Then two other numerical um data types are discrete. So one that has um definitive minimum size or binominal values like yes or no, you can make yes a one and no a zero and then count those or um if you roll a dice, 123456 on a dice, those numbers are discrete values and then continues is anything that can be subdivided further and further. So if we had better analysis is if we could take her BP more accurately than what we can do in the ward at the moment, you can actually get not just ABP of 100 and 20/18, but you can get 1 25/82 or you can get 125 point 3/82 0.3. So those are continuous data types. White cell counts. Um H BS are also considered continuous numerical data types purely because there are so many, it's a white cell count of 10 to the power of nine that practically you can subdivide them into tiny little pieces. These are important because different data types need different analysis. And we have to have a bit of an understanding of what the differences are. And if you dig deeper into statistics, understand which test you apply to, which to give you actually a meaningful answer. And if the wrong test was applied to the wrong data, you can't trust the result. So what happens? Um Well, with our descriptive statistics, just a few words on um what the different things mean? I take it that most people have at least heard of me medians and most uh mode somewhere along, at least in their school career, hopefully somewhere in undergrad training as well. But they're basically just used to describe statistics. Um Mean is an average, very easy. Um Everyone knows how to calculate average median is the middle value of the data that we're looking at. That means you count how many different ones you have and you take divided by two and the middle value of that is the median. So it's not the average, but it's the median value. The mean and median may or may not be the same if you look at it. And it's quite helpful in our own data. If we look at the mean and the median, they tell us whether they are outliers, especially if the mean and medium have quite a um difference in their values. And then we have to look through our data and see what outliers are they? And are they significant, what happened there? And whether or not we include them in our data? And then most a mode is the one where you count, which answer, especially in like at schools has been there the most often. Then from there, we actually wanna describe it a bit more. So we want to give a range. What is the minimum, maximum value quarts and percentiles? Um Kartel or percentiles are just um more accurate or with finer detail than Kels. But Kels, as you can see here on the picture, um your median will be your second quartile and in between all the data that you have, you again count all the different values you have and you take the middle value between the maximum and the median value and that will be your third quart. And then on the bottom, you take the minimum value and count the amount of data you have data points you have. And halfway between those two is the first quart. And then standard deviation is the distance between data points and means and they are um worked out again, relatively complicated. So I haven't included it here but it to see what would be a normal range and who would fall outside it. We use standard deviation these days to plot the weight and height for our Children. Um and see whether or not they fall into a normal range. And if you fall in the third or um standard deviation, you know that the weight for height is rather low or high. And this is a outlier of the normal range of data, but there's a lot more to it. I just don't have the time to go through that. Then coming to methods of obtaining data, um we often can't use the entire population there is. So we try and get a sample as a representation of that population, the larger the samples, the less likely that the the more likely that they will actually represent the population. But then we have to look for bias in the sampling area. If it was a sampling area, how was the sampling done? Um In observational studies, as I've said earlier, there's often bias. So we have to look for it. In observational studies, we often look backwards, we have to deal with the data there and we can't try and move bias out of it. And we, when we write up article, we actually have to mention that and say that in this article, there could be virus for this, this and that. So how do we sample um I was uh population so that we can get a sample that we can work with. So there's simple random selection that requires that you know exactly who or is part of the population. And you pick at random 50 people out of it, for example. But it requires that you need to know everyone. It's relatively simple if you want to do a study on the hospital staff, because hr should have a list of all the people who are employed at um re and you take 50 people out of those. And that would be a simple random selection. You can also do a systematic sampling where again, you need to know exactly who all is the entire population. Again, if we use fre a staff, for example, we would have a master list. We know who's employed at frea. But we can say we can go through the list and say we only take every 10th person who's on that list and put them into the group that we're going to study that's called systematic sampling. Then cluster random sampling is often done if we don't have a master list of who is um in the population. So we wanna study, let's say hypertension or I don't know he, yeah, that's a hypertension. Um and who comes to clinics. Um We can pick a few clinics and then from there, say every 10th patient that we want to see and ask them some questions and then stratified random sampling, you have two mutually exclusive groups. So we could say gastroschisis patients and we stratify them into male and females because gender is mutually exclusive and then um look at the different data points between those groups. Um Again, what's important for the methods of sa sampling is if we read through our article is to see whether the sample that was taken in someone else's study is representative of our own context. So, again, read through that section carefully and keep in mind the population, we see um often if we have a first world study compared to our third world second setting, it's very clear that it's not the same population. Therefore, we cannot necessarily use all their data for us then coming to hypothesis testing, which is used um And well, trying to see what is the statistical significance of things. So any study we start, we need to make a statement. Um girls do better uh with or have a better outcome in with gastroschisis than boys just as an example for our statement. And we then have to define our null hypothesis. And alternate hypothesis, null hypothesis state that there's no difference between the two groups. Um We often work out on AP value and we'll come back to AP value in a second. Um We work out the P value and if the P value in medicine is more than 0.05 we cannot reject our null hypothesis. Therefore, there might not be a difference between the two groups. The alternative hypothesis is that there is a difference and we reject the null hypothesis and accept our alternative hypothesis if our P value is less than 0.05. Um These are worked out in graphs and normal distribution graphs, we work out what is the 5% on one end or the 2.5% on either end, depending on whether we do a one tailed or two tailed test. Um But I'll come back to P values in the next slide to explain a bit more about it. Um Again, sometimes not mentioned in the studies or articles that we read. But we actually have to figure out what was the null hypothesis? What was the alternate hypothesis? And whether they're supposed to use or whether they wanted a one t answer, a twot answer, one told would be um for our gastroschisis patients. It is, there is a difference. So it's just uh yeah, there is a difference but it doesn't matter which side it is whether boys do better or girls do better. That would be a two tell test. And if uh a two tell hypothesis, if we say the girls do better than the boys, that would be a one tailed answer. And the statistical analysis that we do on the data that we have depends on whether our hypothesis. Uh alternate hypothesis is one tilt or two tilt and we therefore have to pick the correct tests then coming to P value in every single study that we see. There will be some mention of AP value. It's actually quite complex to try and understand where it comes from and how does it get calculated. But it is the probability of an event occurring. And it's done through the calculation of a geometrical area. That's why we keep working on these graphs and highlight the one area. It's used to check if the probability to find the same value in both groups is low enough that therefore there are two significantly different groups. Um We, like I said earlier in medicine, we kind of decided on um 5%. Therefore, ap value of less than 0.05 a significant AP value, more than 0.05 means that the likelihood that the two values of, of the same group is not specifically different um other parts of studies in the w or scientists in the world from biologists or microbiologists depending on what they're studying to actually define their P value differently. And we just have to actually read through what is the P value and how did they get there? So again, for discrete values where we have 1234, no commerce whatsoever, it's very, very easy to calculate AP value. You just draw a graph and you um pretty much can read of the P value of the graph on continuous data types. It becomes much more complex and we can actually only work with probabilities of a range of values. There's no specific values. That's why we talk about P values greater or lesser than a certain value that we pick. Um And then we work out through complex calculations how to gi what is the P value for our data? And then compare where does it fit? So just to illustrate it a bit graphically, we have a normal distribution graph, we decide that our 95 statistical significance threshold. Um It's 95%. We draw the line there and then we check, where does our observed result fall? If it falls to the lower end of the graph on the tail end, it's not supposed to happen on the tail end of the graph. We know that it's statistically significant or at least we hope so. Um if it falls into under the big area of the groove, we assume that there is no significant um stati statistical difference. However, the 5% is a value that in most medicine we have agreed on, it doesn't necessarily mean that there is 100% statistical significance. Um But if we make it less than 95% especially in medicine, we find that a lot of our studies end up being non statistically significant. And therefore we have settled on a 95% statistical significance. Then just to highlight some of the types of tests that can be done to work out P values, they roughly get grouped into numerical testing and categorical testing depending on the data you have. And then paramedic metric. And nonparametric tests. The to go through each one of them is too complex and is going to take us weeks to get everyone to understand what they're all about. But I just wanted to highlight some common ones, names that you'll come across if you read studies and ideally should try and read up around it to see what do they mean and where do they come from? So, under param parametric testing, we have students t test. You see it in quite a lot of ST studies. The Nova one compares more than three groups, but then you have to do further individual testing between the groups to get actual P values, linear regression. And then under the nonparamedic metric tests, um we have also a whole bunch of them that kind of correlate to a lot of the parametric tests. The Mann Whitney U test comes under multiple different names, but man Whitney use is the most common one used and it's quite a commonly used test that um I've come across in a whole bunch of studies, the same as Crystal Vus and then the Wilcoxon and Spearman rank. I haven't seen that frequently. And then categorical testing has a child care square test and a fish exact test. Those are fairly common ones that we come across. Unfortunately, it will really take a lot of time to try and explain where do they come from? How do they get to their? Um So I just settled on mentioning them that you know, that they exist, but it takes a lot more time to understand them. Hello. Are you coming to the end of your talk now? Sorry, almost, almost. OK. Then as last one, confidence intervals that we come across also relatively often. Um again, it works um with a bell shaped curve, it's the assurance that the population parameter falls closer to the study mean. And we again use customary in medicine a 95% confidence interval. It is not the percentage or the probability that it will fall under. But it means if I repeat a study 100 times 95 times, the mean of the actual population will lie within the um set limits for whatever data we're working with. Um And the graph here is just to illustrate how that kind of works. If we pick 68 we have a lot of data that falls out of it. 95% we already include a lot of it, but you have a 99.7 confidence interval. Um It's very difficult for in medicine. Therefore, we settle for 95 confidence intervals. So in conclusion, statistics is a whole degree on its own. And there's a lot more to understanding statistics than what I can share in 20 minutes, half an hour. Um We need to understand statistics to be able to read more than just the introduction and the summary of any article. And actually understand a bit more what is going on in the article so that we can um compare it to our setting and see whether it's useful to us or not. Also, if we write our M ES, we need to understand a lot more about the language used to be able to do a proper literature review and to be able to do our own me. And it's important to read critically through studies and understand what they are talking about and what their limits are so that we can actually um make them work for us or reject them if it doesn't work for us. Thank you. So now I have it there. Uh Thank you, Hlga. Uh It was, it was a nice presentation. Uh She could join about 10 minutes late, but he has joined. So I will invite him to, to come in now and I'll, I would like him to actually uh tell us what is the significance of all this knowledge in our actual pediatric surgical practice. And how do we incorporate this knowledge while we are studying or practicing and while we are conducting any research so bad? Thank you. Thank you, me. Yeah, sorry, I got stuck in theater so I missed the first bit of the talk but uh I will try and, and sort of just uh fill in the bits and pieces that I think will help answer the question you've just asked uh overall um your registrar given a fairly good cover. But the way I look at statistics when it comes to uh a daily application or an application of it, in terms of my regular practice, I look at statistics as falling into two separate areas. The one area is how do statistics apply to a research project that I am interested in conducting or how can I apply research to answer a question that I have? And then the second aspect of statistics is how does it help me interpret the data that is presented in a particular paper or how the paper is actually reviewed? So just very quickly, if we look at uh a research project that we would have, it would be a question that I would want to ask as to how many of my Children with necrotizing enterocolitis survive uh as a simple example. So I would look at that very quickly define a population. It's all my patients who present with necro necrotizing enteric colitis. It's fairly easy. And I would uh ask my question, which would be my hypothesis and my hypothesis would be that 20% of Children with necrotizing enterocolitis divide. Once I do that, then I know I'm going to have to collect the data and my data will fall into two categories. It will either fall into numerical type of data or the socalled categorical type of data that was discussed earlier. Once you know what type of data you have, it's very easy then to apply the statistical test that you will use to reach a conclusion that will either prove or disprove your hypothesis. So if we have categorical data, it's very simple. Are there two or three paired groups? And you know which test you use? If you have numerical data, then you've got to decide is that data likely to be normally distributed or not, which is the bell shaped curve that was spoken about if the data is going to be normally distributed, which means I have a large population and that's not going to apply to my study because my study is most likely going to be nonparametric. It's my very own patients. It's from a select group, it's a small number of patients. So the actual data is not going to be representative of the general population of follow up, be c so that will be nonparametric data. So if you have the parametric data, which is from a large population, all the medical schools in the country and all the N EC, then we will apply very specific tests. Those are the pet T test un PT test. And if it is nonparametric, we know which test to apply. So actually doing it from a, a paper point of view or a research question point of view, it is relatively simple. If you break down your research project into a simple question with a straightforward hypothesis or two or three questions need to be answered and then identifying the data that you will collect and the type of data you collect. So statistics from that point of view as applies to your research project is fairly uh straightforward on my side in terms of interpreting statistics from a paper that's a little bit more of an issue. And that would probably apply more to, to the question that you're asking as to how does it apply to daily pediatric surgery? And uh to answer that question, I think it's more important that you are able to interpret the terminology that is used so simple. Things like what are the differences between mean and medians are, are fairly easy but sometimes terms like standard deviations or standard error means or how a relative risk. What does a relative risk mean? What do the odds ratio mean? I think those are important specs to understand when it comes to into your uh daily sort of practice. So if you read an article on gastrosis and you want to try and see if that data is relevant or not and how it applies, then you would have to understand these terms. I think COVID has been very, very useful in this particular scenario because you look at the number of papers which have been retracted from very large journals from from journals, sorry, which have a large readership like the lancet and these are articles that were published there rather rapidly and found to have weaknesses in the statistics and the way they were reported. So it's always important to have some sort of an idea as to what the terms mean and how you can actually interpret them. So, I mean, simple things like uh the standard deviation we know will refer to where you have a continuous. So it's a very a numerical variable which uh will cover data of continuous type. And all it does is it measures how variable your actual data is. So or what the scatter is, are all the data focused around one point or is there lots of scatters? So you're getting figures from 0 to 100 rather than all around 50. Similarly, for things like the risk ratio, I think that's something important that people have to understand. Uh I'm I'm not sure whether it was already mentioned, but when it comes to interpreting percentages where the answers are sort of yes and no, it becomes a little bit more difficult. So that's when risk ratios or odds ratios become applicable. And a really useful term is the number needed to treat because for our type of practice, where we look at a particular sort of intervention, let's say octreotide for uh chylothorax, for example, knowing what the odds ratios are or what the uh relative risk is of the patient getting better, doesn't really mean too much to us. But if we are told that you need to treat 100 and 50 patients with high dose, uh somatostatin or simvastatin. In order to relieve one patient with uh a chylothorax, it tends to make more practical sense in, in terms of how we we work. So the odds ratios, uh relative risk, those are the ways the data are now reported. When it comes to the type of practice, we see PP values and things tend to be more important when you have continuous variables. But we should actually probably pay more attention to relative risk and odds ratios because our patients often have uh or we are usually looking for a simple yes or no answer in terms of an intervention that we are doing rather than how likely they are to benefit from a particular uh intervention. Um The, I think the other important uh factors are the type one and type two errors. Uh If that, that's sort of discussed, then that would be an important um aspect to understand what the difference between the two types of errors are and interpreting AP value I think is very important. So whenever you have AP value, it may, for example, carry a lot of weight in a randomized control uh trial, but it is not necessarily significant in a question that we would ask from a case control type of study that we often do. So the P values really only plays a role in trying to disprove the null hypothesis. And very few of the studies that we get in pediatric surgery actually follow the follow the idea or follow the principle of disproving a null hypothesis. Although P values are widely reported, they are, its specific use is supposed to be in disproving the null hypothesis that, that you actually proposed. So I think that's an important uh aspect if we can look at a sort of simple example, as the preoperative radiotherapy or chemotherapy versus surgery as compared to surgery alone for a tumor. For example, you have to look at the overall survival at your endpoint. But what you've got to then focus on is proving that your overall survival is not likely to be what you expect. So that's the null null hypothesis. And um II, I'm not gonna go into detail about it unless you really want me to. But they, it's important to understand what a statistically significant result is in terms of AP value as compared to a clinically significant result. So the P value is just telling you that the chances of the null hypothesis being correct is more or less than 5% in terms of the value that we have used. But the clinical significance of that result is not contained in that particular number, your odds ratio, your relative risk tend to give you a better idea of or the number needed to treat a better idea of what the clinical significance of a result is going to be as compared to the P value. Um Just I think uh a last point without having to go into anything else is just to come to review this whole evidence-based medicine uh concept. And this pyramid of evidence based medicine concept, which is right at the bottom, you have uh your own experience versus right on the top, the socalled uh mesa analysis, systematic reviews and the rest of it, what we've got to understand is that based on that type of uh evidence based medicine, a huge focus is placed on the randomized controlled trials and evidence that is derived from that. But the randomized controlled trial does not reflect real world experience. And real world experience is a concept that was introduced a long time ago, but has recently become far more important in terms of applicability and to a large extent, the courts of law now find real world experience to have the same level of uh of um sort of legal significance as does uh evidence based medicine and real world experience. Very simply put refers to the data that is collected from the type of trials that we often do. So we go and pull out a whole lot of files from a patient in the clinic, from patients in the clinic with gastroschisis. For example, we review that data and then we publish the uh statistics based around that data. Those statistics because they are case controlled or retrospective tend to have a lower degree of significance on this pyramid of evidence based medicine, but actually are far more applicable clinically to our patients than a randomized controlled trial on the management of gastrosis which or preselect the group of patients which are going to be analyzed. So a randomized controlled trial is not looking at the real world patient that presents to you, but it preselect the patients that we have a particular interest in. So I think real world experience is something that you, you have to know a little bit about and you have to understand that it, it it can sometimes be considered at the same level of evidence as randomized controlled trials uh are considered especially in a surgical discipline where our surgical experience is often derived from experience. Uh I think that that should cover most of the comments. Uh Melinda, I'm happy to answer anything specific. But uh just from a topic point of view, I think we have to apply what we can to our practice and interpret it from a relatively simplistic point of view. Then trying to understand why a man with me test you test is used rather than an analysis of variants or ano anova. For example, you know, some of that is that is perfect. I think what uh you have given uh is a very important advice uh for me. No, no, I think uh you, your advice has been perfect. I don't think we need anything more uh at this stage because you have actually given clear cut guidelines to the registrars when they select their this topic to keep it simple to have only one or maximum two or three questions. And, and that way it will be much easier to do the mini desert that's number one. And secondly, you have also brought of this very important concept of real world experience, which is what is more clinically important for us in our day to day practice. So, thank you for that. I think uh I'll invite uh Professor Hiley who also has huge amount of uh research experience uh to give his comments. Uh prof we know this is a very huge topic. Uh So just whatever you would like to advise us. Thank you pro Yeah. Yeah. Uh Good afternoon, everybody. Um I II, I'm pleased to see that this topic has come up uh for discussion. I don't wanna go into statistical tests largely because I don't understand them. Um But what I would like to emphasize is that as practicing pediatric surgeons, we have responsibilities and our first responsibility is to apply known science to our patient care. And that means that we've got to be able to read the literature sensibly, we've got to be able to interpret the literature and we've got to be able to apply what we've learned from the literature to our um to our practice. And, and secondly, we also have a responsibility to do research ourselves and, and, and, and it's, I mean, it, it, to me it would be terribly unfulfilling. If there was a AAA practical, a practicing pediatric surgeon, you ended up at the end of your career doing exactly what your boss told you to do at the start of your career, uh to, to manage patients. Um And so nothing has been, uh nothing has changed, nothing has been learned. So at, at this most basic level, we simply gotta record what we do and, and see if it works. Um And then if we find that it doesn't, we can try something else. So to start the, the research tradition at, at a school, I think it's important to start off with observational research and just pick basically your whole practice and look at it and see what's going on there and then um uh make your, your conclusions from uh from your own experience. And then we need to compare it to, to people who have populations like us. There is no absolutely no merit in comparing a series of gastroschisis patients, gastroschisis management in East London with what they are achieving in Boston Children's Hospital. That, that, that to me is unrewarding, unfulfilling and very often just depressing. What we should be doing is we should be looking at East London and saying, how are we doing relative to Cape Town or Harari or Nairobi? Um And uh and making sure that we are comparing apples with apples because then the um um the research bug catches on in surgery. We, we can't do sham operations. So there aren't going to be many um randomized prospective double blind controlled, the surgeon always knows what he's done. Um At least I hope he does. Um So we are stymied from uh um there we go. And there was a very important paper. I thought that uh perhaps you should look out. It was by our colleague um David Mucker who you will know uh Milan who said um evidence based medicine. Are we boiling the frog? Because when you think about it, these meta analyses are all based on randomized controlled clinical trials, all of which are of different quality, some of which are very strictly managed and others are not. So you can end up making a a decision on the meta analysis of bad data and call him that. So I think uh Professor Shaikh's point about real world experience is vital in uh in in the sphere of uh surgery in general and pediatric surgery in particular with that, I'm gonna shut up. Thank you for the opportunity. OK. Thank you pro uh Thank you very much. Uh Now I'll invite Doctor Manic Chen who is our consultant, pediatric surgeon and who is uh quite keenly interested in, in research. And actually, she has been instrumental in in starting an electronic uh entry of our operation uh notes and and then a new electronic uh database for outpatients. So, yeah, sure that this is your comments. Hi everyone. Yes, thanks. Rough. Um Thanks to how for that very difficult um, topic to cover. And she did well, I think to explain as much as she could in the time given. And uh, I think prof and Prof Hadley have just summarized everything pertinent to the pediatric surgeon. So, uh a lot of just two very quick points that I'll kind of summarize basically what they said, I think, um, is that we, we are informed by the research. So it's very important, valuable to um uh create research that we can all look at and be informed by them in order to make decisions about our patients. And I think they both kind of made that point very well. And then also that we have uh a paucity of data from the developed developing world. And we're looking at developed world research and we're trying to fit it into our um environment, which is, which is not very helpful for our patients. So we need to keep that in mind and we need to add to that database of the developing world. Thank you. Thank you. So that, that's, that's uh very important comments you have made. And um I think previously we uh did a lot of work. I mean, we treated a lot of patients, but we, our record keeping was poor, we access the files. And uh that uh was an important reason why uh we could not collect uh authentic data. But I think with our electronic record. Uh We are far better off than what we were before. So, thank you for that. I'll invite now, Doctor Selo Mataya, who is uh our other consultant also quite keen in research. In fact, he has recently uh enrolled for a phd. So he is the right person to give uh his opinion and comment. Sellar. Thanks for, I don't think I'm the right person yet. Um But yeah, I think um what someone told me once that um when you're actually reading a textbook, it's like basically looking back in time. So as much as a textbook is relevant for what you're doing, um it was ultimately printed now and the work that was actually done in the textbook was uh uh five if not even 10 years old. So that's why it's important I think for to have continuous um updates of what's happening. Hence why like Journal Club are very pertinent and important because other than teaching you or giving you exposure to what's recent about that specific topic, they also help you critique uh an article and help you kind of start dwelling or kind of looking into the research world and the research language because um unfortunately, researchers like a French woman or a French man um from afar, they look pleasant, but once you come to them close and try to communicate to them and you have no idea how to speak French, it becomes difficult, then it's not something that's so attractive anymore. So once you know the language, then you're able to communicate better and be able to actually enjoy um research. Um I kind of personally did not like research much. Um But I'm starting to get the hang of it um because I am slowly learning French. Um And it's a very interesting language to learn. Um Thanks, bro. So, thank you, sir. That was very nice simile. Thank you very much. Uh I don't see doctor uh Majola. Uh So I think uh we are coming to the end of the topic. I think I'll just invite prof she uh just to give a final comment before we close. Uh Sad, thanks Melin. Uh I think it's um having these kind of discussions very important because hopefully it will get people uh interested in, in research and doing this kind of stuff. But I think the most important thing is you don't have to spend way too much of time learning about statistics because you have people who can do this for you. So prof Aley would say since this year I mentioned it. Is that why bark if you can get a dog to bark for you? So the idea is to get the right question, collect the data and you do have people who know statistics really well to help guide you with the tests and the rest of it just understand how the test is applied to your clinical practice. Knowing the theory behind the tests and things is something that you will have somebody way more experience than we are to help with. Thanks Melin. Thank you so much. I think again, very wise words and same advice to the registrars. Keep your mini desert topic simple. And, and uh it will consume less energy, less time. It will be interesting for everybody to read it and uh it will save you a lot of time and energy. So, thank you very much. That was quite an important uh uninteresting topic, but very nicely covered by, by HGA. And uh we got nice advice. Uh Next week, it will be only A and M meeting only uh restricted to the staff of our department. But in two weeks time, we will have a talk about nuclear medicine in pediatric surgery. Doctor Botha, our medical officer will give this talk and we will have uh doctor Anita Brink who is a nuclear medicine physicist at Red Cross Children's Hospital. She will be the invited guest and we may have one more nuclear physicist who could also attend. So we will see the rest of you in two weeks time. Have a good evening. Bye-bye.