Computer generated transcript

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

Uh, good evening everyone. Uh I hope that you're fine. Uh Some of you may know me, some may not. Uh I'm going to introduce myself a little bit, uh, rapidly. I'm a pharmacist and an epidemiologist. Uh, and I have a phd and a diploma of ability to direct research in uh epidemiology and public health. Uh Actually, uh I do a lot of, of research uh related to public health, related to epidemiology but also related to clinical, uh a a and, and medicine uh aspects. Uh and this is, uh because, uh, uh I work with you guys with, uh lots of uh uh medical students and lots of physicians and, and pharmacists. So, II have really a very wide, let's say range of, of research. Uh And uh I'm, I'm really happy to be with you. Uh Actually I was, uh uh I had this particular idea of having a research club in mind long time ago and we wanted to start with it and then the corona happened and we, we, we stopped it. So I'm really now very happy that you are back uh to the track of freely doing this kind of uh research uh activity, extracurricular activity because I think that research is very important for you. And if you want to really have better visibility, if you want to have access to better positions in your training. And then later on in your practice in your uh profession, I guess that research is important, this is the world we live in. So I will be with you for three sessions. Uh I will start today with this session uh with the first session regarding the research design. Now, some of you may know what I will be talking about uh today. Uh And some of you may not uh I would like you to really pay attention. And if you have any questions, I'm ready to answer these questions. Uh And uh I hope that you will enjoy it. And if there's something that you would like to ask about, don't hesitate. So, uh and this is it. So today, we will be talking about some small introduction to research and we will mainly be talking about study designs. Uh And tomorrow we will move to sampling sampling procedures, uh calculating minimal sample size and talking uh a little bit about data collection and uh uh some additional uh ideas that are important for the research that you may want to, to conduct. So our learning objective for today are numerous. We will try to define epidemiology and health research. Identify the different types of epidemiological studies. Distinguish observational from interventional studies, understand the difference between descriptive and analytical aims identify the aims measures data sources from different types of studies, associate evidence, qualitative study types, understand the value of review types. And finally, to understand the difference between association and causal relationship. So you can see that we have many learning objectives and we have really ambitious, let's say learning objectives. I will try to go over the uh concepts uh as much as possible. So I'm starting with health research. What is health research? It is an investigation of a human health issue to learn more about it? So very simply put in simple words, this is health research. Uh Usually it is funded by the government, private foundation, drug companies with the hope that the new information will be useful to patients, to the community and to other researchers. And it can be applied to animals and humans in clinical and in epidemiological settings. Now, of course, uh what we will be doing today will not be related to animals. We will mainly be talking about human research, whether in clinical or epidemiological setting because uh I'm not specialized in animal research. Uh You can ask a biologist uh to talk to you more about animal uh research. I'm specialized in everything related to humans. So as for epidemiology, you know that epidemiology in the modern sense. First, when you, when you hear the word epidemiology, you think it's about epidemics. We're out of the uh COVID pandemic So we think a la lot about this uh uh event. However, you should know that nowadays, epidemiology is a discipline that aims to study the distribution of health phenomena in a population. And it can be any health phenomena. It can be an infectious disease, it can be a chronic disease, it can be a trauma, it can be uh quality of life, wellbeing, uh uh mental health, it could be anything in addition to the factors that condition their frequency. And this means it's the factors that affect their appearance, either by increasing their appearance as risk factors or by decreasing their appearance. We and here, I mean uh protective factors or preventive factors. So when I say that I am an epidemiologist, it means that I do mainly research. And what I do is any type of research related to any disease or any concept related to health. But I apply it to a full population or to a sample in a population, but I don't apply to a single patient. So having a case report, for example, a case report is the, let's say the uh uh the discipline or the uh uh the specialty of one physician working with one patient. So this 1 to 1 a relationship will lead to writing a case report. While for us epidemiologist and people who would like to work on wider research, let's say we go to study research on samples. Now, of course, there are several activities and I guess those of you who uh uh uh have attended the uh fifth year course, they know that epidemiology would cover for access, disease surveillance, investigation, research and intervention evaluation. And let me tell you that disease surveillance, investigation and interventions evaluation. They are all the um specialty of, let's say health authorities such as the Ministry of Health. Uh There is, for example, a particular unit that is called uh the uh epidemiology Surveillance Unit. So they are specialized in disease surveillance. They might do from time to time, do some investigation, particularly whenever there is a problem such as a massive, let's say uh uh uh intoxication or uh uh some food uh toxicity uh due to an event uh or something like that. Uh they would also work sometimes on the evaluation of intervention. They develop, they uh apply or implement an intervention and then they try to evaluate the effect of this intervention. So all this is done by the Ministry of Public Health. But what about us, we as researchers? Well, we mainly use the third axis of epidemiology, which is we use epidemiological principles in order to conduct a research. And once we do the research, we then try to publish this research. And this is a very important aspect of epidemiology because it complements the practical professional epidemiology part that is done by the Ministry of Health. And at the same time, it will really be very important in order to decide on interventions and decide on uh uh strategies uh uh that will be applied by the Ministry of Health. They cannot decide on strategies unless they know about the diseases, their distribution, their risk factors in the population. So, uh epidemiology is also a basic science for public health. As we are saying, it is really a very important uh a pillar of public health and it uses statistics as a major working tool. And here you know that and uh I think that all of you are starting to understand the importance of statistics. You cannot do human research without using statistics. There is this is, this is it, it's there and there's no way you can get away from statistics. Even if you don't like math, you will have to use at least a minimum of statistical principles in your research. Even if you have a statistician, you will have to already understand what he's saying, know what you ask from him. And once he delivers the tables for you, the P values, you have to know what is their meaning. So in all cases, what we are trying to do is to quantify health states. And when we say quantify, we say numbers and we say statistics and this is what epidemiology tries to do. Now, one additional idea, there are several non quantitative methods, we call them qualitative methods. Uh they are used but they are mainly used in the uh human science. Yeah, we're talking about psychology, sociology, uh maybe economics. So in these uh types of uh sciences, sometimes we don't use quantity, we use quality. And here we use qualitative methods and we can say that they are complementary of quantitative methods. So sometimes the ideal is to mix quantitative and qualitative methods in order to come up with some answer to a question to a research question that you might ask. Now based on all this. Well, here is a table that tries to summarize all types of research. And I can tell you that this is a very, let's say complete table. Uh that includes really the most important uh basic research here if you look at the extreme uh uh left. So you have some theoretical and applied animal studies, cell studies, biochemistry, material development, genetic studies. This is mainly basic research done most of the times on animals or on cells or even in vitro using substances and uh uh chemicals and uh uh tissues. Uh And as we said, in addition to that, we have clinical research and clinical research is somewhere between basic research and epidemiological research. Sometimes we call clinical research, translational research. So it's as uh as if we are talking about a translation of basic research into the real word. However, it is not still not the real word, particularly the experimental part. So I'm talking, I don't know if you can see my arrow. I'm talking about clinical research, experimental studies and here we have clinical studies, phase 123 and four, 1 to 3. Mainly it's during the development of a medication or of a vaccine, for example, or even of some types of interventions. Then you have the phase four study, which is sometimes can be experimental. It's whenever you're trying to compare to medication after you, for example, you uh market these two medications. No. In addition to the experimental part, clinical research can also be observational. And here we can conduct observational study in clinical settings, either a therapy study or a prognostic diagnostic studies where we are trying to uh either improve diagnosis measures of a disease or to find out prognostic aspects of a disease. We can do some observational study with medications, uh some secondary data analysis related to a treatment, uh some case series also or even a single case report. These are part of clinical research but they are considered observational. No. In addition to clinical research, we go to the uh uh red part which is epidemiological research. And let me tell you that epidemiological research can also be either experimental or observational experimental. It's whenever you have some field study or group study that you can apply on groups or uh populations or it can be observation and the observational studies in epidemiology. Well, these are the ones that I think you might uh know you have the cohorts that could be prospective or historical. You can, you have the case control study, the cross sectional studies, we will take the details of these but you also have some ecological monitoring surveillance study or a description of some type of disease based on registry data. Now, all these are part of what we call primary research. It means that you have original data. You are working on data that you take immediately from patients or from participants within a population. But there is also something that is called secondary research. And here we are talking about a meta analysis or a review, a systematic review or a simple native review of what has been already published. So you take the primary research and you try to join several primary research articles, you put them together and you do your uh review scoping systematic or meta analysis. And let me tell you, we can add to this what we call bib uh big bibliometric analysis. It's also part of these types of review. So you can see that you can specialized anywhere in research. It's so huge. This is a huge word. Uh It's up to you to decide where you want to be. Now. For me, my specialty is epidemiology and I work a lot in on observational studies. But I also help in uh experimental studies. Sometimes 90% of the cases. I'm in observational research. But some I also work in clinical research, uh experimental clinical or experimental epidemiological research. And uh I do also some secondary uh research, uh the principles are the same. Uh But of course, the context is different. This is, let's say a more simplified uh figure where you can see that the types of study can be etiological or descriptive descriptive studies are not that important from the evidence based point of view. I mean by that, that the uh level of evidence that is provided by the information given by these types of study is not really that high. There are also the etiological study that are much better. Uh whenever they are observational, they are of acceptable levels of evidence uh particularly for cohorts a little bit less with guest control and a little bit less and less with cross section and studies and for interventional study. Well, here, the level of evidence becomes better. You have particularly good level of evidence with randomized placebo control, uh double or single blind studies and multicentric uh research. And you have also some good level of evidence with quasi experimental. So based on on these two tables and figures, we will take them one by one and try to understand what we mean by every type of study. Here's the case report. Now the case report, I guess that you all have tried or maybe have read about a case report. A case report is a report where you have uh a new disease or a particular uh extraordinary presentation of a case in a hospital. And here the uh the physician decides to um well talk about it to his peers, to his colleagues. So in general, it's a kind of innovation that is starting. But the level of evidence is very low and you cannot conclude anything based on a case report. Sometimes several case reports related to the same disease might appear. And here somebody might decide to come and take all these case reports and try to do what we call a case series. So writing a case series is interesting because it will start to uh uh uh uh let's say, uh give some hypothesis, some new and emerging hypothesis about either a new disease or an emerging disease or a developing disease. And I can tell you this is how we started to know about HIV. This is how we started to know about SARS. This is how we started to know about MERS. This is how we started to know about COVID. So having several cases and writing about these several cases, this will lead to really starting to s to to suspect that there is something ongoing and we have to know and see what is ongoing. So a case series report, as we said is a simple descriptive uh uh account for uh of interesting characteristics. Case series. It will report generally uh uh involves patients seen over a relatively short time. Now, in general case series, well, they don't include control subject and people do not have the disease or condition being described will not be involved. And this is why when you don't have a control group, you cannot really understand what is ongoing among people who have the disease. So having a control group is something of major importance. And this is why case series are are of low evidence because we don't have a control group. No. Uh sometimes some people will not include case series in a list of types of studies. So they would say that while this is not a type of study, in fact, yes, it is a type uh a type of study, but it is of low evidence. As we said, uh on occasion, some, some investigator might include some control subjects. And uh uh whenever you have a controlled subject, it is here where things start to get more interesting and get more robust in terms of conclusion that you can take. Now, this is an example of a new disease, as we're saying, uh a novel approach to treating COVID-19 using nutritional and oxidative therapy. And here you have study design. If you look at it, it is an observational case series consecutive. So yes, this is an important type of studies for new diseases. What are the advantages and disadvantages of case series? They are easy to write. The observation are useful as we said because they will allow me to try to understand what is ongoing, some causes, some explanations uh of the observation. But there are lots of biases, selection bias information bias. And particularly since you don't have a control group, well, you cannot really understand what is ongoing, but we can generate some hypothesis and generating hypothesis is important because you will go and try to conduct other types of studies in order to really prove or confirm your current hypothesis. So now we move to another type or level of studies that are a little bit more important. And here we're talking about cross sectional studies, case controls and cohorts. So these are the observational analytical epidemiological studies. Of course to classify these studies. It is important to look at time. Time is a very important concept that will help me to classify the studies in epidemiology. You have, we are here today if we do a survey now and we measure at the same time, the exposure and the disease. So we are conducting what we call a survey or now we call it a cross sectional study or a cross sectional survey. It all means the same thing. If we are looking at the past, either we ask questions about the past of people, the past exposure of people. So here we have a case control study or sometimes we might reconstitute the history of a generation of a group of participants. And this is what we call a historical cohort. And if we are just taking a group of people and waiting with time doing an observation, a follow up of 510 years ahead of, ahead of us. So here we are doing what we call a prospective cohort study. So you can see that time is very important and we it will allow us to classify the different types of studies. So let us take them one by one to understand them better cross sectional study. It's a snapshot of the reality. You take a representative sample of the population, you do a concomitant measure of exposure and disease and you try to link exposure and disease or you just describe exposure and disease. There is no temporality. You cannot be sure that exposure comes before the disease. And this is a major weakness of cross sectional studies. It can be descriptive or analytical descriptive. It means I'm just describing analytical is whenever I'm trying to link any exposure to the disease status, trying to find if any exposure is a risk factor or it is a protective factor of the disease and it can be repeated to assess evolution. So you might try to do several cross sectional studies, for example, every five years or every 10 years or even every year, you can do that. And this will allow you to have a prospective vision, let's say of what is ongoing. But if you do it only once, well, it is not the best type of study, particularly because you cannot be sure that exposure comes before the disease. Second cross sectional studies uh uh still about cross sectional studies. Uh We are uh uh as we said, trying to see if there is a possible association between a factor and a disease. And here, the word possible is important. I guess that I should have colored this with a, another color because even if you find a significant relationship between a factor and a disease, but you cannot be sure that a statistical significance or an association is also a causal relationship. So it's just to also suggest some hypothesis. Uh it's a little bit better confirmation than case series, but it's not the best. So in the hierarchy, let's say of studies, it comes just after case series, but it is below other types of studies. So since we cannot verify temporality, any observed association will need to be better verified, we will take time into account and taking time into account means that we will have to do either a case control or a cohort study. And this will allow us to confirm causality. And let me tell you that if you want a full confirmation of causality, it's only possible by experimental studies. So you might ask, and why don't we do an experimental study from the beginning? Why why bother ourselves and go through all these types of studies knowing that they are not really very uh uh interesting from the evidence level point of view. Actually, it's an ethical issue if the exposure that you are trying to assess is possibly or potentially harmful. So if you are looking for risk factors, you don't have the right to do experimental studies, you don't have the right to experiment some substances that you think are toxic. Some uh uh uh exposure that you think are harmful, some behaviors that you think are not really uh uh good for patient's health. No, you cannot do experimental studies. In this case, you only do experimental studies. If you think that this is a positive exposure for the patient, a medication, a vaccine, a a positive behavior. Yes, you might do experimental studies but everything that is negative, everything that you suspect being negative, you cannot go for experimental study. You have to do a uh an observational study and you will go to either cross sectional or case control or cohort. Now, why do we use cross section study in general if you want to diagnose a disease or stage a disease distribution in a population? For example, I want to know what is the prevalence of a certain disease, let's say co PD asthma, whatever. And then I want to see what is the stage distribution of this disease in the population. So here I might do cross sectional studies. See how many have a mild disease, moderate severe disease. What is the percentage? How are they are distributed all over the population? The subgroups, men, women, uh elderly, uh adults, Children, I want to know this disease distribution. Yes, I would go for a cross sectional study if I want to assess the usefulness of diagnostic procedures. So I might use different methods at the same times. And here I can know whether the I have a new, for example, diagnostic procedure is good or bad. Uh Is it really uh uh measuring what it is supposed to be measuring? So, all the validation studies, whenever you want to validate a new screening procedure or a diagnostic procedure, a new scale that will help you measure either an exposure or a disease. All these are really done due uh using cross sectional studies. If you want to establish norms, for example, know the range within which most patients fit. So if you want to know what is the normal value of any measure in a population? Uh of course, labs, they will establish and provide the normal limits based on working on samples on cross sectional samples. So they will be testing people who are known to uh uh uh uh to have uh to, to not have the disease where the disease is absent and they will give you normal values if you have uh uh a a survey that you want to conduct, to gain insight in a very complex topic uh to learn how people think or feel about an issue, things related to quality of life, things related to wellbeing related to mental health. Here you will go for also cross sectional studies. So you can see that cross sectional studies are really very, very useful even though they will not help me a lot in establishing causality between a risk factor or a protective factor and a disease. However, I always start with cross sectional studies and then I go and confirm more and more in case control studies and in cohort studies, case control studies. Well, we are here, we are at the right of the slide, we have cases and we have controls. So we have people that are known to have a disease and people that are known to be healthy, they don't have the disease, they are controlled. And then we ask them about previous exposure. So for example, we ask if they are really, they were exposed at a certain moment in their past, they were exposed to smoking, they were exposed to alcohol, they were exposed to uh unhealthy eating stress. Uh uh they took some medication in the past, whatever it depends on your study. So you ask question about a past exposure and you will try to link this past exposure to the current status of case versus controls. So it is a retrospective study and it will help me to measure the association between a disease and one or even more than one exposure. I will work as having two groups of comparison. You have cases that is a parallel group to the controls. You don't put them together, you don't analyze them together, you always analyze them separately. Comparatively in a parallel manner. Uh we uh use this type of study whenever you have a rare disease or whenever you, we want to test the association with febrile exposure factors at the same time. And I guess that this is really very interesting. It's one of the best, let's say, uh uh cost versus equality or cost effectiveness ratio as a type of study, it's not very expensive. And at the same time, it will allow me to get really very interesting information. Now. Controls yes. Uh for questions do you want to take them now or like at the end? How do you you can you can ask me now if you want? Yes. Ok you have a question in the chart. Uh what is the win ratio in studies? And the person apologize if he mistook the term or the win ratio? Mm are you sure about this? Are you talking about the ratio of um cases to controls? For example? Mm he didn't he didn't answer because it uh you can take 121 cases and controls you can take 2 to 13 to 1 and up to 4 to 1. You are still gaining in uh power but if you take more than four controls to to cases call us, you don't really uh uh uh gain any more power in your study and in general we don't put more than four controls to one case. I hope that this is what you uh uh where in case this wasn't the question at the end of the session, there is a feedback where they can put the questions they want to be answered at the beginning of next session. So I can put the question for next session and we list them for your next session, but we'll OK. OK. OK. Very good. So are you still seeing my screen? Yes. Now that start case control study? Ok. Ok. So uh one condition uh that is very important in case control studies. Control should be chosen from the same population as cases. And here is the difficulty. What do I mean by the same population? I mean by that control should be included. If, if they had been cases, they would be included in this study. So it's really difficult sometimes to decide on the type of controls. It's difficult to decide if we want controls that are healthy, totally healthy or we want controls to have another type of disease. So sometimes we choose controls who have been admitted to the same hospital. We choose controls who are from the same family of cases. But in both situation, there are drawbacks and we have to really think a lot about how to choose our controls in general. Uh There are some references that advise you to take more than one group of controls because if you take uh family members, there is a risk of over pairing because they come from the same culture, they come from the same uh uh family, they might have the same habits, lifestyle. So, if there is something that you want to see in cases, you will not see it anymore because it will also occur among controls. If you take people that have been admitted to the hospital for another problem, well, you might also have a problem of overbearing because uh uh maybe the same toxic exposure you are studying would have caused this other disease or the other problem. You might have a profile of cases that is totally different from the profiles of controls. So this is why most of the time using several types of control might be recommended. And in addition to that, and this is of course, something that you all know uh you cannot compare apples to oranges, you have to compare apples to apples. So what to do if you have different gender or age or uh educational distribution between cases and controls, you will have to go and do either what we call matching or you will apply the multivariable analysis method in order to adjust and, and make these two groups of cases and controls make them really alike. So the profile of cases and controls, if they are different, you will have to try to make them alike except for the presence of the disease. And this will allow you to study the exposure. Now here, what I have here is an example. Of, of the very well known case control study. This was the case control study that helped in, uh, finding the association between the thalidomide, which was a medication used during pregnancy in the past and for cilia. And you can see here these Children who are born with, uh, uh, uh, problems at the level of their, uh, hands and, uh, the, uh, you can see here they, they don't have fullfledged hands. Doctor. Yes. Uh Doctor Edie, a Edie sent a link and also specified document by one ratio. If you want to get back to that point approach will combat practical guidance based on previous experience. But this is something totally different. You, we are talking, you are talking about composite endpoint, which has nothing to do with what we are doing. So uh I if you want uh maybe for uh uh next time, II might prepare one quick slide about it, but it's not related at all to what we are saying. So it's for composite endpoint and here it's something that is totally different. This is for scales and for uh uh prognostic and diagnostic tools. So it has nothing to do with study design. OK. OK. OK. Thank you. Thank you for all these. I will uh put it aside for for now, let me, let me continue and we will talk about them later on. OK. So now we move to the cohort study and a cohort. Well, it's a group of people of, of subjects that are followed up over time. And at the end of the study, this will allow us to calculate a cumulative incidence or an incidence rate of a disease during the study period. It will allow us to compare two groups to compare the incidence rate of a disease among the two groups. And by that, it will allow me to check first the natural development of the disease among the two groups. And to compare this development and see if the exposed have a different development of the disease, a different incidence of the disease with time. And here it will allow me to be closer, let's say to the causality between an exposure and a disease. So a cohort is generally prospective. However, you should know that sometimes we might go back in time and reconstitute the data as if, as if we were starting, let's say 10 years ago. So we have, for example, access to the files of participants of subjects of people, uh medical files or a. Um it could be a registry, it could be in a uh in an occupational setting. For example, where I have all the files of these participants and I can reconstitute apo stereo AAA cohort. And this is what we call a retrospective cohort. So in general, it is used for studying rare exposure, exploring the effect of an exposure of sever disease. So one exposure you take these people and you wait in time for their diseases to appear. And this is a figure that will show you here, you have exposed people. And as time passes, you will compare the incidence of the disease among those who are exposed and those who are not exposed. And this takes time into account. And this is a very important point. We are not comparing prevalence, we are comparing incidents. So we are trying to see how much time is passing before the event is occurring. And by this all statistical analysis will change, we will not use the usual statistical analysis that we can use in cross sectional study nor the one that we use in case control study. Here, there are special statistical analysis, we call it the survival analysis and in survival analysis. Well, we will compare incident, we will compare the hazard of the problem and we will calculate what we call a hazard ratio. So we will later say what are the types of association that will be calculated from every type of study? Just know that it's a very specific and special type of analysis that is done whenever we are using prospective cohort studies. So the retrospective cohorts, if you remember, we said we are here, then we go back in time. For example, employee health record in a manufacturing company, retrospect, a uh a hospital record or a, I don't know uh a registry record. And then we go back in time, we can go go back 20 years, let's say 10 years. It depends. I will check how many were exposed at a certain moment in time. And then I will do the followup in time and I will look at the outcomes that have already occurred in some subjects. So it's really a very interesting and very uh let's say a time saving way of doing a cohort. But at the same time, there is a condition, you should have enough information in these records. If the records don't include enough information, well, you will not be able to reconstitute this cohort and you will not be able to see what is ongoing and what is happening. Unfortunately, this is how it is. So it's a very tempting idea to do a retrospective cohort, but it is not always possible because you are relying on what have been registered 20 years ago. Mm It might be good and it might not no similar to the cohort studies that are prospective and where you start by an exposure and you wait for the disease to appear. You should know that interventional studies they are similar in a lot of aspects to cohort. First. The similarity is the prospective nature. So you have people who are today exposed to something and you are following them up with that. The difference is that the exposure, it's not, it should not be something toxic. It is supposed to be protective or curative. There is no way it is something toxic. And you know that if by mistake, we try something and it comes up to be toxic, we stop the study. So if you're doing a clinical trial and within the trial, some side effects, for example, that were not expected, they just appear, well, you will stop the study even ahead of time. Even prematurely, we don't want to be harmful. You know, that ethically first do no harm. This is the first ethical principle that can be used and it should be used in research as in practice. So it's very important to have an exposure that is supposed to be protective or curative. This is 1st, 2nd, the exposure of people in experimental or interventional study. Well, it was imposed by the research. It was imposed. I mean that it's the researcher who decided to expose these people. They are not due to just take the hazard or the uh uh uh a chance exposure. Oh A um let's say it is not a uh an individual decision to, to smoke, for example, or an individual decision of the participant to uh to drink alcohol or whatever it is something that is imposed by the researcher. And this is why it should be something positive. Second, s if you want to really proof the causality, the relationship between this this protective or curative exposure and between the health status. Well, you have to do what we call a randomized controlled trial because randomization is the key it is the key to the causality proof if you don't have randomization. Of course, when we say randomization, it means that IP SA factor, we have a control group and we are randomizing or assigning exposure randomly randomly using some randomization procedures such as doing it on the uh computer. There are specific software or sometimes manually, we used to do it manually in the past, just trying to see uh uh taking uh drawing AAA paper where it's written in group A versus group B. It's as simple as that. But now we don't do it like this anymore. We have a randomization plan that is uh done on the computer. So you can see you are here, you randomly assign the people to be treated by, let's say treatment E versus uh treatment A versus treatment B. And then you do the follow up in time and you compare the incidence, the cyst this time. It could be the incidence of the disease. If we're talking about a vaccine, for example, or if we're talking about some vitamins or if we're talking about um I don't know uh something positive such as a positive behavior, eating better uh doing some physical activity. All these, it can be also a medication and where you would wait for the time until the person is cured or the time un until uh uh uh the disease progress. It can be also something uh let's say like an educational session. Uh for example, the time until a woman decides to stop breastfeeding. So any event can happen at a certain moment in time. And here you are trying to see what's the effect of this intervention on comparing the incidence of this event. As I told you, randomization is the key, you have to have a control group. It could be a placebo or a reference. Uh having a double blind uh measure is very important because it will decrease subjectivity and having a multicenter trial is also important because it will help me it uh to have a more representative population of the total population of the people who have this disease or who should be included in this interventional study. But as for cohort, we should take time into account. So we use the same statistical method for clinical trials interventions and for cohorts, they are different from the ones that are used for cross sectional studies and case control studies. So put in mind that statistical method will depend a lot on the type of study that you are conducting. The test that will be used might be similar. But the overall modeling. Well, they are different one word about quasi experimental study. Here. The investigator controls the factors to be compared but not person selection. So it will distribute the factors to be compared on groups of individual, not on individuals. For example, you have two villages in one village, you do the intervention in the other you don't do it, but you cannot be sure that all people that are within that village while they were exposed to the intervention. For example, you diffuse an educational session about a certain uh I don't know, lifestyle uh uh uh habit within that village. But you cannot be sure that everybody watched that film or that video. So you just will wait and see what will happen. You will compare the two samples. but the level of evidence is not a really very high level of evidence because we did not control for compounders, we did not do a randomization. We did the randomization at the group level, not at the individual le le level. So we don't know whether you have differences in age and gender and education. It's also economic factors among people who are exposed versus not exposed. So this is why quasi experimental studies are used sometimes but well, they are not the best type of studies doctor. Yes. Uh We have another question. Yes. What would be the difference between case control studies and retrospective cohort studies? If in both we are reconstructing the past? OK. Very, very interesting question. Let me tell you in case control study, you don't reconstruct, you don't reconstitute the database. You rely most of the times on the answers of the participants. Hm. And you just associate a current disease to a previous exposure, but you don't take time into account the way you analyze the data is different. And in general, you would rely on people's memory. And this is why we say we have a lot of recall bias. Whenever we are conducting a case control study. While in a retrospective cohort, you are, let me show it to you. You are going back first to some types of health report. You are not relying on people's memory. This is one first thing and second and this is also very important. You are reconstituting the data and you are taking again time into account. So when we say working on cohorts and taking time into account, we are talking about survival analysis, we are talking about hazard ratio, proportional hazard ratio configuration. So everything is different from a case control study where you do a logistic regression. So the uh the types of analysis are different and the uh uh form of the data that you have is also different and the source of the data is different and the types of biases that are involved are also different. So yes, there are lots of differences. The only thing that that is similar is the fact that we are talking about an exposure that comes before a disease. And let me tell you that this is something that has always to be present in order to have a good uh uh uh uh level of evidence. And as we said, the cross sectional study, whenever you cannot be sure what comes first, it's here where you say that? Well, I don't have enough temporality uh uh uh proof and I cannot know what comes first. And this is why the level of evidence is so low. But the fact that exposure comes before the disease, it has to be there. And the way or the approach might differ between retrospective cohort and case control studies. Now, one more difference, maybe I did not say it. But in general, in retrospective cohort, you divide your population according to exposed versus nonexposed and you observe them with time and you might study different outcomes. You might uh uh uh uh wait for different diseases to occur and you might study the occurrence of these different diseases. While in case control study, it's only one disease versus controls. And whenever you are going back in time and asking question, you would ask questions about several types of exposures. Uh So you can see that yes, it's a matter of exposure before disease. But the approaches are different. The methods are different. The sources of data are different. The uh statistical analysis is also different. Is it OK? Just waiting a second for him to reply but it was perfectly explained for me. Mm We can continue while he is typing if you, while he is like this. OK. Now what about reviews? Reviews rely on available literature and here you will try to use the available literature and the available evidence and try to come up with better answers. Let's say, uh you will try to homogenize and synthesize the evidence in the literature in order to come up with uh an answer to a question. Now, in general, if you can reach what we call a meta analysis, uh the word meta analysis means it's the last analysis that you are doing a meta analysis will give you more or less a final answer. And if you just check uh meta analysis in, in the literature, you might find, let's say uh uh in the conclusion of this meta analysis, while we know that this drug is effective, no more studies are necessary to confirm these findings. But before reaching a meta analysis, there are different types of reviews that can be done. You can do a narrative review. And here a narrative review, you have a broad question. Uh You don't specify your source. Uh Most of the times it's biased source, you choose the article that you want to show and you just let down the articles that you don't want to, to show, you have a message in mind that you want to uh uh uh deliver. So the synthesis is qualitative. Um sometimes we, we say it is evidence based, but most of the time a narrative review is not considered as really the best type of evidence. However, if you go for what we call a systematic literature review, so here you are a lot more rigorous, you have a specific question you do some comprehensive search strategy. Yeah. And your keywords, you go, you choose one or more database, you go to MEDLINE, you go or you go to Scopus, you go to Cochrane and, and you really apply your search strategy, you come up with a huge number of uh abstracts and then you remove the abstracts that are not relevant and you, you keep the ones that are relevant and then you do your really selection in a uniform way, you critically evaluate the articles that you obtain. And sometimes you even do what we call a quantitative synthesis. And here it's called a meta analysis. So it is a very good evidence on whatever questions you are asking. Uh uh uh you should know that sometimes it is possible to do the meta analysis. The quantitative aspect if you are able to find quantitative measures that are similar homogeneous between the different types of study, but sometimes you do the literature, the systematic literature review, but you cannot reach a meta analysis. It's OK. Still a systematic literature is something that is very well see. So what are the steps whether we're talking a meta analysis or whether we're talking systematic review? You start by i identification of studies by your search strategies that should be very well done. And in general, you will ask for the lib and help in order to define with you the search strategy because this is their specialty. They know within every type of database, what mesh terms and what specific terms to put in your search strategy so that you will be the most comprehensive possible. Then you do the selection based on inclusion exclusion criteria, quality criteria. You select first the abstract then well from the list of abstract that you selected, you will go to full text analysis and you take the full text article and you select them based on quality criteria. Third step, you do what we call an abstraction. I mean we extract the raw data from these uh uh uh studies. Uh we extract effect size, we extract the uh association measure like here, for example, on the extreme, right, you can see that we have 1234567899 studies from every study. They extracted the OS ratio with its confidence in German and they give different weight to every study because the weight of the study will depend on its size, it will depend on its quality. So you give a higher weight for bigger studies, better size uh and, and better quality and you give a lower uh uh weight for studies that are of lower quality and of lower sample size. And you come up with a weighted overall average. This kind of analysis is called meta analysis. And here we were able, for example, to, to say that 0.37 is the overall odds ratio between a treatment and a this is So this type of analysis is sometimes feasible. Sometimes not. If you don't find the same similar uh uh uh association me measure, you cannot come up with a meta analysis, you just do a systematic review. Yes, there is another question the child. So now, so we have why the narrative review has low evidence. It has a search strategy like systematic review. Wait, the narrative review most of the time doesn't have a search strategy. But if you do a certain strategy and you really uh apply it in a very rigorous manner and you reach all the steps except the quantitative steps. OK. So here you might say that it has an acceptable level of evidence and we call it a scoping review. So if you start your work very rigorously, you apply the search strategy, you do a good selection and you reach the point where you don't do this abstraction. So you don't extract the data, you don't assess the quality of every article, you just take them, let's say you obtain 20 articles, you take them, you summarize them in a table and you discuss them. So this is called a scoping review. A scoping review is acceptable. It's better than a narrative review, but it's not as good as a systematic review because the systematic review has something very important, which is it takes into account the quality of the articles, the quality, the level of evidence is taken into account in a scoping review. We don't assess the quality. We just look at the content. Did I answer your question? Mm We'll just wait a minute to. So for answers if you want uh I can inform you afterwards or do you wanna wait? Let's let's move on and then we will see you then. Yes. OK. So this is a very uh uh interesting summary of what we've been saying. You have the lowest level is the case report and case study. Then you have cross sectional case control cohort whether prospective or retrospective. This is the analytical observational study. Uh I also told you that the cross sectional study could be descriptive, just descriptive, not analytical. And here you have the best evidence, the nonrandomized trial, concurrent historical controls. Well, we don't love them a lot but they are there and the best would be the randomized controlled trial and the meta analysis of several randomized controlled trials. Now, is it possible to do a meta analysis with observational studies? Yes, it is possible. Does it mean that the meta analysis of observational studies is of first level of evidence? No, because a meta analysis of observational study will still include the possibility of biases that are related to every type of observational study. So no, II, if you want your meta analysis to be perfect and to be number one in the e of evidence, it has to be a meta analysis of randomized controlled trial. So it has to be a meta analysis of very well conducted studies. Second, if a meta analysis is done on really heterogeneous types of study, so this will be in that the meta-analysis will not be perfect. And this is why you will have to calculate the level of heterogeneity of studies within the meta analysis and the more heterogeneous the studies that are included the lowest or the lower the level of evidence provided by the meta analysis. Well, if you have a meta analysis with really very good and homogeneous studies, then the level of evidence will be a really excellent level. This is another way of seeing it. And here I would like to see that above all these, we have the clinical practice guidelines. So you as physician, sometimes you don't have the n the time to read all these, somebody has done the homework for you and the clinical practice. Uh guidelines are issued by scientific societies all over the world and they are based on the level of evidence. And you will see that the clinical practice guidelines they include within them, the level of evidence for every recommendation that they, that they do. And one more thing I would like you to see that a narrative review, expert opinion editorials, case report, case series are here below the case control and others. The animal and lab study are even below these. Why? Because what you find in animals cannot be extrapolated to humans directly and there are a lot, lot of concepts that were found in animals that were never applied to human, they could not apply them to human. So this is why the lowest evidence is that the level of animal and lab studies. Uh uh here it's the level and grades of recommendations. Whenever you have some clinical practice guidelines, you have the level of evidence. 12345. Here it's the best level of evidence because you have at least one large randomized placebo controlled trial of good quality with low potential for bias or a meta analysis of well conducted randomized trial time. In uh uh uh level two, you have small randomized trial or one large randomized trial. But with a suspicion of bias or a meta analysis of such lower level, let's say trials or you have AAA meta analysis of trials with demonstrated heterogeneity as I told you if you have only observational studies. So you're immediately at level three, prospective cohorts is at three retrospective cohorts and case controls are at level four. And finally, studies without control group case reports, expert opinion are level five. So the recommendation of the clinical guidelines, it comes here, you have the A is for a strong evidence. So strong recommendation b strong or moderate evidence, it's OK, we recommend it. Uh But here we have a limited clinical benefit. See we don't know, maybe yes, maybe no, maybe it should be given, maybe not. Uh uh insufficient evidence of efficacy. Uh but the benefit does not outweigh the risk. D. Now, here we are into adverse outcomes contraindication. So D is generally not recommended and e it's never recommended, it should not be given. So you can see that all these are based on this and this and all the research that we were talking about before and your practice as physician. While you are here, you should have an evidence based practice based on research evidence based on your clinical experience, of course. And, and this is a very important, it's not very old. It's relatively new where you should take the patient reference into account the patient quality of life, the patient improvement and the patient choice of being treated versus not being treated. Uh the utility of a treatment uh uh based on patient uh outcomes. And this is what we call. And now it's a relatively I'm saying relatively because it's been here for around 20 years, let's say what we call a relatively a new discipline uh which is the use of proms, patient reported outcome measures. So, in order to measure clinical efficacy, we don't only rely on physiological measures, but we also ask patients about how they are feeling about how is their quality of life, how is their wellbeing, how much in pain or how much suffering they they are? And finally, my uh uh uh final words are about the issue of association versus uh uh causality yes, you might find an association, a statistically significant association or correlation. But this does not mean that it is a causal relationship between exposure and outcome. And here this is the hot weather that causes at the same time a sunburn and it will increase ice cream sales. But even though you will find a correlation between the use of sunscreens and the uh uh uh uh the sales of ice cream, they are not causally related. They are correlated definitely because they increase at the same time, but they have nothing to do with each others. They are correlated, but there's no causality association between both. So the same could happen whenever you find a correlation, it could just be one common factor that is making them increase together or decrease together or go in different direction but also together. But then they might not be really causally correlated. So, the causality concept is the cause of a disease that can be either an event, a condition, a characteristic, a combination of these factors. And most of the time, it's a combination of the factors because most of the time, one disease or one event. Well, it doesn't occur if you don't have this association of factors of conditions of events being together. For example, you might have a genetic predisposition. But if you don't have the ee environmental exposure, you don't develop the disease, you might have a an immunosuppression. But if you are not exposed to the bacteria in question, you will not develop this infection and so on, so forth. So it's a matter of several factors together, that together they cause the disease. So a sufficient cause. It's a complete causal mechanism where you have environmental, biological, physical, social and a genetic or uh of course, that occur together and this will cause this causal mechanism. So it's important to put in mind this issue of multicausality of any disease. There is no disease with just one simple uh uh causal factor. So nevertheless, echoes need to be either necessary or sufficient for its removal to result in disease prevention. So this is what we were saying. They are there together. If you remove one well, the disease will not happen. Uh and you will be able to prevent a substantial amount of disease. So you need to identify every component to prevent some cases of the disease. And this is what you will tell the policymakers. For example, if we are able, if we are able to, let's say decrease, let's say smoking level by 20% we will be able to prevent so many cases of cancer and cardiovascular and COPD. So you just can do this type of calculation in order to uh uh convince policymakers about the importance of prevention. So you have sometimes some predisposing factors, mainly the susceptibility of the disease that you cannot change related to age, sex history, genetics and so on and so forth. You have enabling or disabling factors. These are modifiable factors such as low socioeconomic status, malnutrition, poor medical care. They can all promote a disease. You have precipitating factor. This is particularly infectious agent that whenever they are there, oh the the the disease will rapidly appear such as the virus and type one diabetes, a bacteria or a mycobacterium tuberculosis and the appearance of tuberculosis, uh COVID-19, uh exposure to the virus and the appearance of the disease and so on and so forth. And you have also reinforcing factors. I mean, you have repeated exposure to a toxic substance or a hard work or something that happens after the first step of the disease starts and this will just uh uh uh sustain, let's say the exposure to this reinforcing factor and cause the disease to be sustained and to remain. So all these are considered types of factors that can really be causally linked to the disease. And that's what I had to tell you for today. I would like to thank you. I hope it was, was clear enough and let us see, stop sharing. That's it. Uh All the aspect of the pascal that uh they can recording and also keep questions for next, next session uh there that aspect. So you can expect some questions in that manner. OK. Will you, will you send me the questions? Yes, of course. I will send you the rest before time. OK? Very good. So you can clean them and let you want. All right. Ok anyway next time is tomorrow yes, tomorrow at 630 I guess. Yes. Yeah. Uh thank you. And also if you want I can send you the link that uh regarding the ratio if you want, I can send you the links that. Yes. Yes. Yes. Yes. Thank you. Mhm. All right and uh have a nice evening. You too everyone thank you very much for your time. So, uh I will see you tomorrow. Ok. Bye-bye. You're welcome.