Cadmore media player playing audio Gordon Guyatt, MD, discusses composite end points.
Audio Player
Accessibility
Using forms mode, you can utilize keyboard shortcuts to make usability more efficient.
Space bar will play and pause the audio. Pressing escape will close all displays and pressing it again will jump to the end of the player.
You can press A to open the Keyboard Shortcut Dialog which also provides an audio description of the player and all of its features.
Transcript
Result Count
of
Search
Segments
Result Count
of
Search
Name:
Gordon Guyatt, MD, discusses composite end points.
Description:
Gordon Guyatt, MD, discusses composite end points.
Thumbnail URL:
https://cadmoremediastorage.blob.core.windows.net/11b95c16-281a-4050-9029-26fed7666c07/thumbnails/11b95c16-281a-4050-9029-26fed7666c07.jpg?sv=2019-02-02&sr=c&sig=RENtF1nmwBmh9%2BHbJTGqxphOjem17JtynVO7BHWOlno%3D&st=2024-10-15T23%3A58%3A29Z&se=2024-10-16T04%3A03%3A29Z&sp=r
Duration:
T00H23M33S
Embed URL:
https://stream.cadmore.media/player/11b95c16-281a-4050-9029-26fed7666c07
Content URL:
https://cadmoreoriginalmedia.blob.core.windows.net/11b95c16-281a-4050-9029-26fed7666c07/17379789.mp3?sv=2019-02-02&sr=c&sig=vpk748mNkVeP9hLbXIkYZ6ta2sSd%2BGyLGre5RgrEji8%3D&st=2024-10-15T23%3A58%3A29Z&se=2024-10-16T02%3A03%3A29Z&sp=r
Upload Date:
2022-02-28T00:00:00.0000000
Transcript:
Language: EN.
Segment:0 .
[ Music ] >> Gordon, why don't we start by having you give us your name and title? >> Gordon Guyatt. I'm a Distinguished Professor of Medicine at McMaster University. >> And also the principal editor of the "User's Guide to Medical Literature" correct? >> That is correct. >> All right. We're going to talk about one of the chapters in the book on composite endpoints. To start, can you describe composite endpoints for us?
>> Sure and I'll give a little bit of history. So, the history was that in cardiovascular disease, for instance, patients having myocardial infarction, if you go back 50 years, there were a lot of people dying and a lot of people having recurrent MI and that was bad for the patients. But it was good for the clinical investigators because they could run trials with feasible sample sizes, and they have lots of events. And so, they might demonstrate a reduction in say, death or myocardial infarction.
And then we stopped leaving people in bed for weeks after their myocardial infarction, started walking them around early, started having initial treatments like aspirin that were effective, and the event rates of death and myocardial infarction dropped substantially. Well, that was very good for patients, but it was not so good for the clinical investigators because to run clinical trials, you need lots of events. And now with event rates much lower, they have to get more and more and more patients for their trials, and things started to become unfeasible.
And so, an idea came up. What if instead of looking separately, for instance, at death and myocardial infarction and the need for revascularization, we put them all together and call them a composite endpoint? So, that means a patient has an event if they die, or if they have a myocardial infarction, or if they have a coronary revascularization. If they have any one of those, they count as having an event.
If they have two or three, they still count as only having one event, and that's what we call a composite endpoint. A composite endpoint is when you have a number of possible outcome events, and if the patient has any one of those, they get counted as having an event. And what that does is it increases the number of events, right? If you have -- can have any one of three counted as an event, there's more events, and so you need less patients in your trials and that made the clinical trials feasible.
And it's become what you'll see in the course of this that I have reservations about composite endpoints. And so, I refer to it in cardiovascular, that there is now an epidemic of trials using composite endpoints in cardiovascular disease in particular, probably more than 50% of current trials are using composite endpoints to increase the number of events and make the clinical trials feasible. >> So, what are the pitfalls with using these composite endpoints?
What would be a good one -- a good set of things to put into a composite? What should you avoid aggregating into a composite? >> Well, so the big problem is, let's use the one I started. So, we would have death, myocardial infarction, and revascularization. Well, death is unequivocally very bad. Myocardial infarction nowadays, when all you might have is a little rise in troponin and nothing much else going on. Some myocardial infarctions are serious, but some are not so serious at all, and revascularization is much less of a big deal.
So, the problem is that you can have an intervention that does nothing for deaths, and does nothing for myocardial infarctions, but reduces revascularization. And the report of such a trial may be We reduced the incidence of deaths, myocardial infarctions, and revascularization by 30%. But in fact, the intervention did nothing for deaths and did nothing for myocardial infarctions, and the only effect was on the least of the important endpoints, the revascularization.
And that, as it turns out, is what actually tends to happen. The interventions tend to be least effective in the most important components of the composite and most effective on the least important. So, that means we could get quite a distorted sense of the magnitude of impact of the intervention by presenting it as death, MI, and revascularizations, when nothing at all has happened in terms of death and MI and only for revascularizations.
So, you asked then, what, you know, what might make a good composite? Well, a good composite, the importance to the patients of the components will be very similar. So, for instance, if your composite was death or disabling stroke, I'd have very little problem with that because death is clearly awful, disabling stroke is almost equally awful, some might actually feel that it's even more awful than dying.
So, if you had -- if your components are very much of the same importance to patients, it makes a composite more reasonable. If one of the components, as I've said, is death, and another is revascularization, where I would think for most folks, death would be much more of a concern than revascularization. So, in other words, there's a big gradient of importance to patients between the most important and the least important, that makes us much more nervous.
>> So, what do you do when you're faced with trying to interpret a trial where there is this sort of mixed outcome that you describe? So, the study is powered for the composite and not its components. So, you may look at two of the three components and say, well, they're not a whole lot different. And then the one that's least important is very different, and that drives the entire result. How does a clinician interpret that sort of finding? >> Well, what we would suggest to the clinician is ignore the composite, be on red alert, as soon as you see a composite and say, wait a minute here, I need to look at the components.
So, for instance, in the chapter in the book, we use the trial called syntax, which looked at coronary interventions, PCI, percutaneous interventions versus coronary artery bypass surgery. And they said the composite endpoint, which was death, stroke, MI, or subsequent revascularization was reduced by almost 30% in relative terms with the intervention of coronary artery bypass grafting.
Well, as it turns out, deaths were a little lower in coronary artery bypass grafting, but not significant. Stroke was actually higher in the coronary artery bypass grafting and that was actually significant difference. Myocardial infarctions were more or less the same, but there was a big benefit on the coronary artery bypass grafting in terms of subsequent revascularization. So, what the clinician would do looking at those results, we think appropriately, instead of saying, ah, coronary artery bypass grafting is better.
It reduced the incidence of deaths, strokes, MI, and subsequent revascularization. The clinician would say, well, there's no clear effect on deaths, stroke is going in the wrong direction, and that's not nice to have a stroke. There's more in the coronary artery bypass grafting, no significant effect on MI, big reduction in subsequent revascularization. So, we seem to be trading off less revascularization with coronary bypass grafting, but more strokes.
So, the bottom line of all that is, you got to look at the components. You cannot just look at the composite and put the components aside. You have to look at what's happening in the components. >> So, I think this statistical purist would say that you shouldn't because the study is only powered for the composite itself and not its components and it's very difficult to interpret them. Because if you're saying there's no difference between the components and they're not powered for those endpoints, that the study is inconclusive.
So, how would you respond to the statistical purist who would be critical of that type of interpretation? >> Well, I have what might be considered a extreme response, which is all this sample size business is a hoax, and it's a hoax because let me tell you my view of what really happens. You're going to do a trial and you say, okay, we need to show that it's adequately powered.
How are we going to do that? Well, we will pick a -- what determines the power we have our standard alpha error, beta error, the real thing that drives it is the magnitude of the effect we want to detect, and we say what's really important is deaths. But, you know, we're not going to be powered for deaths, and stroke is very important. There's no way we can get enough people to be powered in strokes. So, we create our composite endpoint, and then beyond that, we say okay, what effect might we expect in our composite endpoint?
Well, an important effect, if it really did reduce deaths and strokes, might be a particular amount but we're in trouble there. So, let's pick a magnitude of effect that is bigger than is actually-- you know, a much smaller effect might be important, but we'll choose this one. And so, we do a little statistical jiggerty pokerty, and we come up with something that will pass through a peer review panel.
I personally would not take that -- I don't take that seriously at all. The -- what the clinician should do is not worry about what's written in the sample size calculation and the power, what the clinician should do is look at the results and sensibly -- and look at what the sensible conclusions from looking at the results would tell you. >> Yeah, I agree with that interpretation, and it's timely because I have been doing battle with a statistician who is very finicky about some certain statistical points.
And as a clinician, you kind of look at these arguments and think, well, you know it's nice to quibble about statistical niceties. But when you have sick patients in front of you, you really can't just discount data because of some statistical artifact. The data is what you have, and you have to make decisions based on it. >> Well, I would -- we seem to see things eye to eye with respect to that. >> Yeah, so let me -- in the article, you discuss a trial that really bothers me, and it's UKPDS.
And I sort of stumbled -- you know, I'm a surgeon, so I'm not that familiar with this literature. But I was doing -- I was editing a paper on diabetes complications and stumbled on the literature about the time the ACP came out with the revised guidelines for how aggressively to control hemoglobin A1C. And the guidelines are based on I think it was six -- five or six major trials. And the one that really drives the decision making for management of type two diabetes is UKPDS. And as you pointed out in the article, it's -- if you look at the primary endpoint of that trial, it's negative and the results, the 12% reduction in diabetes complications that are often quoted and support a multibillion-dollar industry in diabetes management, come down to one of 21 composite outcome variables being statistically significant, twenty were not.
And I thought this is just -- this sort of highlights the problem of these composites lumping a bunch of things together. finding something that you want to highlight and calling it significant or not significant. And I was wondering if you -- because it's in the article, and I saw it. And I thought, you know this is a problem because we base so many decisions on that one trial. >> Well, you're right, it is very it is extremely problematic. And the basic problem with all due respect to my endocrinologist colleagues, they have all wanted to believe that lowering blood glucose does good things in terms of patient important outcomes of macrovascular and microvascular disease.
And my interpretation of the situation is that that is what they wanted to prove. And it was tough to make the case for UKPDS, where you had these 21 components of the composite of which the overall effect was driven by what many would say was the least or second least important of the components. So, unfortunately, people understandably get invested.
The endocrinologists want to believe that targeting low blood glucose is going to be a good thing in type two diabetes, and in my mind that remains at best questionable. So, people, for better or worse use data to make points that reinforce their particular belief systems. >> Yeah, and it's a particular problem because the second largest amount of spending for all medical care in the United States is diabetes control, and it's most -- mostly type two diabetes.
And a lot of money is spent trying to reduce hemoglobin A1C for a very uncertain outcome because it all comes down to that one trial. And the other two trials ACCORD, and I forget the other one, that looked at similar issue of aggressive control of diabetes proved to be harmful. So, you really wonder what are we doing here? >> Well, yeah, and in the best systematic reviews, looking at across all the trials of targeting higher or lower hemoglobin A1C suggests that the only outcome that might be affected is a reduction in myocardial infarction, which is a small relative effect and a very small absolute effect.
And as I understand it, the pay for performance often targets or one pay for performance criteria is targeting a particular hemoglobin A1C threshold, very questionable. And we could talk about that a little more if you wanted, but it actually can be a perverse incentive. So, quite problematic. >> Yeah let's, even though that's not the topic I'm curious, because I've been asked by the medical residents at UCLA to speak about this.
And I'm preparing a talk, I'm not sure if you know offhand which systematic review has that data in it; I'd like to know. >> Well, there's a number of systematic reviews that have summarized all the trials and have a consistent conclusion. >> Yeah. >> Which, I could not cite immediately while on this but could make available. In terms of this pay for performance the -- as I understand, the way it's structured, they might set a threshold of hemoglobin A1C and say, if your patient is above that hemoglobin A1C, you may get penalized, or you get rewarded if they're below the hemoglobin A1C.
And there's two problems with this, one is picture the patient, say your threshold is seven, and the patient is 7.2 and you think I'm going -- if I can get my patient less than seven, then my rewards will improve; however, the truth is that if you say to the patient, here take an extra drug to get less than seven, and the patient says to you, what benefit will I get?
The answer is probably very little or none, and then the patient would say, well, are there any harms? And the answer would be, well, there's actually a greater risk of symptomatic hypoglycemia and hospitalization. And the patient presented with that would say, you know, doctor, I'm -- what you've told me, I'm quite happy keeping my hemoglobin A1C at 7.2. But the poor physician doesn't -- gets either penalized for leaving the patient, or they're not rewarded for getting them lower.
So, here it's actually a conflict of interest between what's in the patient's best interest and what's in the clinician's best interest. So, that's one patient where there's a perverse incentive. Now picture the patient whose hemoglobin A1C is 10. That individual, although there's no randomized trials showing this because it would be very problematic to randomize people to saying the hemoglobin A1C of 10, that person -- getting that person from 10 to eight or 7.5, may actually be of considerable benefit.
Except that the clinician will not be rewarded for doing that, right? They get them from 10 to still 7.5, which is above the threshold, no rewards for that. So, there's no incentive for the clinician to push the patient down under those situations. So, here is where there's a, what you might call a perfect storm between the lack of evidence supporting the particular pay for performance threshold and the actual perverse incentives that are built in.
>> One of the things in the article is the use of composites for competing risks. So, could you explain competing risks and how composites factor into that? >> Yeah, so one of the situations, let's take -- one where it might conceivably apply is consider the fact of you have a patient who's at risk for stroke. And you're thinking of giving the medication or doing some sort of vascular surgery such as carotid endarterectomy to reduce their risk of stroke.
Well, one way of reducing the number of strokes is to kill the patient before they have the stroke, and that sounds sort of amusing and ironic. But the truth is that some of the people who undergo surgery will die at the time of surgery, and if you die at the time of surgery, you no longer are a candidate for having a stroke, and who might die at surgery?
The people who might die at the surgery are the people with the highest degree of vascular disease who might be at the highest risk of having a stroke. So, as it turns out what you're worried about in the competing risks is that death and stroke is a competing risk. And one way to prevent strokes is to have people die. And it turns out in the case of vascular surgery for stroke, that that is actually what happens, that some of the people who would be destined on medical therapy to have a stroke, do not survive their surgery, and so never get around to having their stroke.
And so, that makes the impact of the intervention on the stroke look better, except it's a bad way of achieving the goal of reducing strokes by killing your patients. So, under those situations people have suggested that the composite endpoint of death and/or stroke would be appropriate because it gets around that problem. And people don't get rewarded for killing their patients, as it turns out -- and that's a legitimate argument.
As it turns out, what's required for that to really be a problem is that the competing risk of death has to be very frequent relative to the risk of the less serious but still serious endpoint, in this case, stroke. And that is very unusual situation, it occurs with some things in cerebrovascular diseases. One where it may occur, at least in some extreme situations, neonatal intensive care is one where it may, where you have death and things like blindness may be -- may be a competing risk, but it's very, very seldom.
So, although it's a theoretical consideration, it plays out as a truly practical issue very seldom. >> Is there anything else we need to cover in the chapter that we haven't talked about? >> No, I wouldn't say that. But I just like to reinforce for clinicians, when you see a composite endpoint, beware. Look at what is in the -- what the components of the composite and if there is a big gradient between the most important and the least important, like death and revascularization, that becomes a red alert, and you must go back and look at the component endpoints.
And you may well find, in fact, you often will find that the effect is restricted to the least important of the components and talking about the composite as reducing death and MI and stroke and revascularization is a very misleading way to put it. And you really need to consider the importance by looking at the components. >> That's really terrific. These are very helpful explanations. >> Well, thank you. >> I've been talking to Dr. Gordon Guyatt from the departments of clinical epidemiology, biostatistics, and medicine at McMasters University.
We've been discussing composite endpoints. To learn more about this topic and others like it, please go to jamaevidence.com where you can find more information in our User's Guide to the Medical Literature. This is Ed Livingston, Deputy Editor of Clinical Reviews and Education. Thank you for listening.
Segment:0 .
[ Music ] >> Gordon, why don't we start by having you give us your name and title? >> Gordon Guyatt. I'm a Distinguished Professor of Medicine at McMaster University. >> And also the principal editor of the "User's Guide to Medical Literature" correct? >> That is correct. >> All right. We're going to talk about one of the chapters in the book on composite endpoints. To start, can you describe composite endpoints for us?
>> Sure and I'll give a little bit of history. So, the history was that in cardiovascular disease, for instance, patients having myocardial infarction, if you go back 50 years, there were a lot of people dying and a lot of people having recurrent MI and that was bad for the patients. But it was good for the clinical investigators because they could run trials with feasible sample sizes, and they have lots of events. And so, they might demonstrate a reduction in say, death or myocardial infarction.
And then we stopped leaving people in bed for weeks after their myocardial infarction, started walking them around early, started having initial treatments like aspirin that were effective, and the event rates of death and myocardial infarction dropped substantially. Well, that was very good for patients, but it was not so good for the clinical investigators because to run clinical trials, you need lots of events. And now with event rates much lower, they have to get more and more and more patients for their trials, and things started to become unfeasible.
And so, an idea came up. What if instead of looking separately, for instance, at death and myocardial infarction and the need for revascularization, we put them all together and call them a composite endpoint? So, that means a patient has an event if they die, or if they have a myocardial infarction, or if they have a coronary revascularization. If they have any one of those, they count as having an event.
If they have two or three, they still count as only having one event, and that's what we call a composite endpoint. A composite endpoint is when you have a number of possible outcome events, and if the patient has any one of those, they get counted as having an event. And what that does is it increases the number of events, right? If you have -- can have any one of three counted as an event, there's more events, and so you need less patients in your trials and that made the clinical trials feasible.
And it's become what you'll see in the course of this that I have reservations about composite endpoints. And so, I refer to it in cardiovascular, that there is now an epidemic of trials using composite endpoints in cardiovascular disease in particular, probably more than 50% of current trials are using composite endpoints to increase the number of events and make the clinical trials feasible. >> So, what are the pitfalls with using these composite endpoints?
What would be a good one -- a good set of things to put into a composite? What should you avoid aggregating into a composite? >> Well, so the big problem is, let's use the one I started. So, we would have death, myocardial infarction, and revascularization. Well, death is unequivocally very bad. Myocardial infarction nowadays, when all you might have is a little rise in troponin and nothing much else going on. Some myocardial infarctions are serious, but some are not so serious at all, and revascularization is much less of a big deal.
So, the problem is that you can have an intervention that does nothing for deaths, and does nothing for myocardial infarctions, but reduces revascularization. And the report of such a trial may be We reduced the incidence of deaths, myocardial infarctions, and revascularization by 30%. But in fact, the intervention did nothing for deaths and did nothing for myocardial infarctions, and the only effect was on the least of the important endpoints, the revascularization.
And that, as it turns out, is what actually tends to happen. The interventions tend to be least effective in the most important components of the composite and most effective on the least important. So, that means we could get quite a distorted sense of the magnitude of impact of the intervention by presenting it as death, MI, and revascularizations, when nothing at all has happened in terms of death and MI and only for revascularizations.
So, you asked then, what, you know, what might make a good composite? Well, a good composite, the importance to the patients of the components will be very similar. So, for instance, if your composite was death or disabling stroke, I'd have very little problem with that because death is clearly awful, disabling stroke is almost equally awful, some might actually feel that it's even more awful than dying.
So, if you had -- if your components are very much of the same importance to patients, it makes a composite more reasonable. If one of the components, as I've said, is death, and another is revascularization, where I would think for most folks, death would be much more of a concern than revascularization. So, in other words, there's a big gradient of importance to patients between the most important and the least important, that makes us much more nervous.
>> So, what do you do when you're faced with trying to interpret a trial where there is this sort of mixed outcome that you describe? So, the study is powered for the composite and not its components. So, you may look at two of the three components and say, well, they're not a whole lot different. And then the one that's least important is very different, and that drives the entire result. How does a clinician interpret that sort of finding? >> Well, what we would suggest to the clinician is ignore the composite, be on red alert, as soon as you see a composite and say, wait a minute here, I need to look at the components.
So, for instance, in the chapter in the book, we use the trial called syntax, which looked at coronary interventions, PCI, percutaneous interventions versus coronary artery bypass surgery. And they said the composite endpoint, which was death, stroke, MI, or subsequent revascularization was reduced by almost 30% in relative terms with the intervention of coronary artery bypass grafting.
Well, as it turns out, deaths were a little lower in coronary artery bypass grafting, but not significant. Stroke was actually higher in the coronary artery bypass grafting and that was actually significant difference. Myocardial infarctions were more or less the same, but there was a big benefit on the coronary artery bypass grafting in terms of subsequent revascularization. So, what the clinician would do looking at those results, we think appropriately, instead of saying, ah, coronary artery bypass grafting is better.
It reduced the incidence of deaths, strokes, MI, and subsequent revascularization. The clinician would say, well, there's no clear effect on deaths, stroke is going in the wrong direction, and that's not nice to have a stroke. There's more in the coronary artery bypass grafting, no significant effect on MI, big reduction in subsequent revascularization. So, we seem to be trading off less revascularization with coronary bypass grafting, but more strokes.
So, the bottom line of all that is, you got to look at the components. You cannot just look at the composite and put the components aside. You have to look at what's happening in the components. >> So, I think this statistical purist would say that you shouldn't because the study is only powered for the composite itself and not its components and it's very difficult to interpret them. Because if you're saying there's no difference between the components and they're not powered for those endpoints, that the study is inconclusive.
So, how would you respond to the statistical purist who would be critical of that type of interpretation? >> Well, I have what might be considered a extreme response, which is all this sample size business is a hoax, and it's a hoax because let me tell you my view of what really happens. You're going to do a trial and you say, okay, we need to show that it's adequately powered.
How are we going to do that? Well, we will pick a -- what determines the power we have our standard alpha error, beta error, the real thing that drives it is the magnitude of the effect we want to detect, and we say what's really important is deaths. But, you know, we're not going to be powered for deaths, and stroke is very important. There's no way we can get enough people to be powered in strokes. So, we create our composite endpoint, and then beyond that, we say okay, what effect might we expect in our composite endpoint?
Well, an important effect, if it really did reduce deaths and strokes, might be a particular amount but we're in trouble there. So, let's pick a magnitude of effect that is bigger than is actually-- you know, a much smaller effect might be important, but we'll choose this one. And so, we do a little statistical jiggerty pokerty, and we come up with something that will pass through a peer review panel.
I personally would not take that -- I don't take that seriously at all. The -- what the clinician should do is not worry about what's written in the sample size calculation and the power, what the clinician should do is look at the results and sensibly -- and look at what the sensible conclusions from looking at the results would tell you. >> Yeah, I agree with that interpretation, and it's timely because I have been doing battle with a statistician who is very finicky about some certain statistical points.
And as a clinician, you kind of look at these arguments and think, well, you know it's nice to quibble about statistical niceties. But when you have sick patients in front of you, you really can't just discount data because of some statistical artifact. The data is what you have, and you have to make decisions based on it. >> Well, I would -- we seem to see things eye to eye with respect to that. >> Yeah, so let me -- in the article, you discuss a trial that really bothers me, and it's UKPDS.
And I sort of stumbled -- you know, I'm a surgeon, so I'm not that familiar with this literature. But I was doing -- I was editing a paper on diabetes complications and stumbled on the literature about the time the ACP came out with the revised guidelines for how aggressively to control hemoglobin A1C. And the guidelines are based on I think it was six -- five or six major trials. And the one that really drives the decision making for management of type two diabetes is UKPDS. And as you pointed out in the article, it's -- if you look at the primary endpoint of that trial, it's negative and the results, the 12% reduction in diabetes complications that are often quoted and support a multibillion-dollar industry in diabetes management, come down to one of 21 composite outcome variables being statistically significant, twenty were not.
And I thought this is just -- this sort of highlights the problem of these composites lumping a bunch of things together. finding something that you want to highlight and calling it significant or not significant. And I was wondering if you -- because it's in the article, and I saw it. And I thought, you know this is a problem because we base so many decisions on that one trial. >> Well, you're right, it is very it is extremely problematic. And the basic problem with all due respect to my endocrinologist colleagues, they have all wanted to believe that lowering blood glucose does good things in terms of patient important outcomes of macrovascular and microvascular disease.
And my interpretation of the situation is that that is what they wanted to prove. And it was tough to make the case for UKPDS, where you had these 21 components of the composite of which the overall effect was driven by what many would say was the least or second least important of the components. So, unfortunately, people understandably get invested.
The endocrinologists want to believe that targeting low blood glucose is going to be a good thing in type two diabetes, and in my mind that remains at best questionable. So, people, for better or worse use data to make points that reinforce their particular belief systems. >> Yeah, and it's a particular problem because the second largest amount of spending for all medical care in the United States is diabetes control, and it's most -- mostly type two diabetes.
And a lot of money is spent trying to reduce hemoglobin A1C for a very uncertain outcome because it all comes down to that one trial. And the other two trials ACCORD, and I forget the other one, that looked at similar issue of aggressive control of diabetes proved to be harmful. So, you really wonder what are we doing here? >> Well, yeah, and in the best systematic reviews, looking at across all the trials of targeting higher or lower hemoglobin A1C suggests that the only outcome that might be affected is a reduction in myocardial infarction, which is a small relative effect and a very small absolute effect.
And as I understand it, the pay for performance often targets or one pay for performance criteria is targeting a particular hemoglobin A1C threshold, very questionable. And we could talk about that a little more if you wanted, but it actually can be a perverse incentive. So, quite problematic. >> Yeah let's, even though that's not the topic I'm curious, because I've been asked by the medical residents at UCLA to speak about this.
And I'm preparing a talk, I'm not sure if you know offhand which systematic review has that data in it; I'd like to know. >> Well, there's a number of systematic reviews that have summarized all the trials and have a consistent conclusion. >> Yeah. >> Which, I could not cite immediately while on this but could make available. In terms of this pay for performance the -- as I understand, the way it's structured, they might set a threshold of hemoglobin A1C and say, if your patient is above that hemoglobin A1C, you may get penalized, or you get rewarded if they're below the hemoglobin A1C.
And there's two problems with this, one is picture the patient, say your threshold is seven, and the patient is 7.2 and you think I'm going -- if I can get my patient less than seven, then my rewards will improve; however, the truth is that if you say to the patient, here take an extra drug to get less than seven, and the patient says to you, what benefit will I get?
The answer is probably very little or none, and then the patient would say, well, are there any harms? And the answer would be, well, there's actually a greater risk of symptomatic hypoglycemia and hospitalization. And the patient presented with that would say, you know, doctor, I'm -- what you've told me, I'm quite happy keeping my hemoglobin A1C at 7.2. But the poor physician doesn't -- gets either penalized for leaving the patient, or they're not rewarded for getting them lower.
So, here it's actually a conflict of interest between what's in the patient's best interest and what's in the clinician's best interest. So, that's one patient where there's a perverse incentive. Now picture the patient whose hemoglobin A1C is 10. That individual, although there's no randomized trials showing this because it would be very problematic to randomize people to saying the hemoglobin A1C of 10, that person -- getting that person from 10 to eight or 7.5, may actually be of considerable benefit.
Except that the clinician will not be rewarded for doing that, right? They get them from 10 to still 7.5, which is above the threshold, no rewards for that. So, there's no incentive for the clinician to push the patient down under those situations. So, here is where there's a, what you might call a perfect storm between the lack of evidence supporting the particular pay for performance threshold and the actual perverse incentives that are built in.
>> One of the things in the article is the use of composites for competing risks. So, could you explain competing risks and how composites factor into that? >> Yeah, so one of the situations, let's take -- one where it might conceivably apply is consider the fact of you have a patient who's at risk for stroke. And you're thinking of giving the medication or doing some sort of vascular surgery such as carotid endarterectomy to reduce their risk of stroke.
Well, one way of reducing the number of strokes is to kill the patient before they have the stroke, and that sounds sort of amusing and ironic. But the truth is that some of the people who undergo surgery will die at the time of surgery, and if you die at the time of surgery, you no longer are a candidate for having a stroke, and who might die at surgery?
The people who might die at the surgery are the people with the highest degree of vascular disease who might be at the highest risk of having a stroke. So, as it turns out what you're worried about in the competing risks is that death and stroke is a competing risk. And one way to prevent strokes is to have people die. And it turns out in the case of vascular surgery for stroke, that that is actually what happens, that some of the people who would be destined on medical therapy to have a stroke, do not survive their surgery, and so never get around to having their stroke.
And so, that makes the impact of the intervention on the stroke look better, except it's a bad way of achieving the goal of reducing strokes by killing your patients. So, under those situations people have suggested that the composite endpoint of death and/or stroke would be appropriate because it gets around that problem. And people don't get rewarded for killing their patients, as it turns out -- and that's a legitimate argument.
As it turns out, what's required for that to really be a problem is that the competing risk of death has to be very frequent relative to the risk of the less serious but still serious endpoint, in this case, stroke. And that is very unusual situation, it occurs with some things in cerebrovascular diseases. One where it may occur, at least in some extreme situations, neonatal intensive care is one where it may, where you have death and things like blindness may be -- may be a competing risk, but it's very, very seldom.
So, although it's a theoretical consideration, it plays out as a truly practical issue very seldom. >> Is there anything else we need to cover in the chapter that we haven't talked about? >> No, I wouldn't say that. But I just like to reinforce for clinicians, when you see a composite endpoint, beware. Look at what is in the -- what the components of the composite and if there is a big gradient between the most important and the least important, like death and revascularization, that becomes a red alert, and you must go back and look at the component endpoints.
And you may well find, in fact, you often will find that the effect is restricted to the least important of the components and talking about the composite as reducing death and MI and stroke and revascularization is a very misleading way to put it. And you really need to consider the importance by looking at the components. >> That's really terrific. These are very helpful explanations. >> Well, thank you. >> I've been talking to Dr. Gordon Guyatt from the departments of clinical epidemiology, biostatistics, and medicine at McMasters University.
We've been discussing composite endpoints. To learn more about this topic and others like it, please go to jamaevidence.com where you can find more information in our User's Guide to the Medical Literature. This is Ed Livingston, Deputy Editor of Clinical Reviews and Education. Thank you for listening.
Form Title Download
Media
Transcript
Form Title Bookmark
Entire Media
Segment(s)
Custom Clip
Start At:
End At:
Form Title Share
Entire Media
Segment(s)
Custom Clip
Start At:
End At:
Form Title Accessibility
The cadmore media player is a professional video and audio player experience designed to deliver media as well as relevant transcript, segmentation, and metadata. On the screen is a traditional html audio player with buttons such as play, pause, forward, captioning, language and play speed selection and more. Adjacent to the audio player are generally tabs, one of which shows a scrolling transcript of the audio and the other shows any segmentation of the media, sometimes referred to as chapters. At the top of the audio player is a menu bar containing an Explore button and Tools button. The Explore button allows you to dig deeper into the audio by opening up an explore dialog which contains an about tab which contains rich metadata about the media, a segments tab which contains further information about the audio, and a related tab which contains links to other media related to this media. The tools button opens a menu with options for sharing the media on social media, creating bookmarking links, creating embed codes for your website, generating citations, and more.
Keyboard Shortcuts
Keyboard shortcuts only work while you are in forms mode. By entering forms mode, you will be able to quickly access portions of the media player with single key commands.
Spacebar: Play/Pause Toggle
E: Explore Menu,
O: Tools Menu
T: Transcript Tab,
S: Segments Tab
F: Forward 15,
R: Rewind 15
L: Languages,
P: Playspeed
M: Mute Toggle,
V: Volume Bar
N: Video Only Toggle,
B: Full Screen Toggle
C: Captions Toggle
Escape: Closes dialogs and menus
Z: Next Transcript Chapter
X: Previous Transcript Chapter
Q: Next Transcript Paragraph
W: Previous Transcript Paragraph
Form Title Embed Media
Entire Media
Segment(s)
Custom Clip
Start At:
End At:
Embed Size
-
You Size - Responsive
-
We Size - Responsive
-
Picture Overlay Popup
-
400 x 225
-
512 x 288
-
400 x 700
-
768 x 1024
-
1024 x 512
-
1280 x 608
Form Title Cite
No citation provided.
Info
Segments
Related
Thumbnails
Accessibility
Using forms mode, you can utilize keyboard shortcuts to make usability more efficient.
Space bar will play and pause the media.
You can press A to open the Accessibility Dialog.
This describes many other keyboard shortcuts for media controls.
Exit Clip Mode
The video is currently set-up to show just a clip from 0:00 to video end. You have requested to see a portion of the media outside of the clip.
Click OK to switch to the full media or click if you would like to stay inside the clip. You can always go back to the clip with Tools and Clip Mode