Good afternoon. My name is Christian Bimi. I'm down from Tucson, two hours away. It's a pleasure to present this talk. I'm a pulmonary critical care doctor. I'm not a nephrologist, but I do use a lot of nephrology insights in the ICU. That's why I'm passionate about this. And like Dr. South said, I was very instrumental in bringing the issue of pulmonary function test disparities that we published in the last major issue of the Blue Journal last year. I do not have anything to disclose. So estimating or measuring renal function is based on, you know, assessing the rate of glomerular filtration. And the gold standard way to do this is to measure urine over a certain period of time and then see how much clearance of an exogenous marker, such as inulin or Ixorhexol or Iothalamate. And that gives you an exact filtration rate of the kidneys. The problem is it's very cumbersome. It takes a long time. You can't do it at bedside. You can't get multiple measurements over a short period of time. And so about 50 years ago, we had to start coming up with alternatives. And the easiest alternative was to look at an endogenous biomarker, which you could easily measure in blood, such as creatinine. Nowadays, we have Cystatin C or Engol, as you can see. The problem is if you measure the biomarker in blood, that doesn't give you what is happening at the level of the glomerular filtration. You do need to then go back and see if that biomarker gives you an estimate of what is actually happening in real time. And so you have the measured GFR, which is the gold standard, but challenging to do at bedside as I've explained. And then you have what we have now come to call the estimated GFR. That's the clearance of an endogenous filtration marker. The first attempt to get an equation that gave us an estimated GFR was at Veterans Hospital in Canada. It was done in 1973, published in 1976. If you remember, you know, if you read textbooks, the Cockroft and Gold Equation in all physiology textbooks, that's what we all read. And it's important to point out where this study was done. You know, in Canada, 1973, you can imagine the demographic of that population where the initial study was done. And then the next iteration was an upgrade of that equation because the original equation did have some weaknesses. And so the MDRD equation was developed in 1999. And that's the first time when RACE was introduced in that estimation equation. There were some criticisms of that equation, and therefore in 2009, the equation that is most used in most EMRs, the CKD-EPI equation was developed. At the same time, they upgraded some of the variables, but RACE was kept as part of that equation. And I'll talk about the details of how RACE came to be part of these equations. Recently, because of, you know, a lot of the highlighted issues that we've talked about here, there have been some attempts to improve on that. And looking back at some of the data, we are able to identify areas where RACE really shouldn't be part of this equation. And I think that going forward, that's where we're going to be looking at. So just to give you an example of these equations, so the MDRD equation and the CKD-EPI equations, the main variables will be age, sex, and, you know, for those two equations, they introduce RACE. And assumption was that the estimated GFR for black patients required that you include a correction factor for RACE because of an overestimation of the GFR by this measured creatinine level. Well, there's a problem when you introduce RACE like that. First of all, you know, you could create a positive bias. If you over-diagnose CKD by introducing a positive bias, then you could cause all kinds of problems all the way down to the ICU if, you know, I mean, we tend to think of, you know, EGFR in the clinic with chronic kidney disease. But in the ICU, it could be very problematic. For example, when patients come in and we're dosing antibiotics, if you're going to determine if a patient needs to get contrast or not, you're looking at the GFR. And most clinicians just go by what is reported in their EMR. We don't even think about the fact that it's a correction factor. We just go by whatever it is. Whoever inputted that data, they don't know. Somebody's lying in bed. They didn't give us their demographics. You choose what race they are. If they look dark enough to you, call them black. So those are problems that occur in our ICU as we speak. And we don't tend to think about how these equations were developed. So as you can see, unsafe use of nephrotoxic medications could be a negative bias. I'm not going to talk about the chronic issues that you see there, late listing for kidney transplant. That's more on the chronic basis. I want to focus more on, you know, the ICU side. You know, for patients that are on dialysis, if you delay the initiation of dialysis, that could be a problem. If you do not list the patient for transplant at the right time, that could be a problem. How do you initiate renal replacement therapy in the ICU? At what point, if you're using a GFR, are you using the right equation? So there are many issues with, you know, using a race-based equation. So in 2021, a couple of studies came out that tried to solve this problem. The first one went back at a dataset called the Chronic Renal Insufficiency Cohort Creek. And what they did was they introduced two things. So they introduced cystatin C as another biomarker and tested it against creatinine. The other thing they did was introduce ancestry informative markers to quantify the amount of ancestry that correlates with your race to see if that could actually identify this issue of race admixing or misclassification of race. What you're looking at here on the left side is on the y-axis is, you know, the median difference between the GFR that's measured and that's calculated. And on the left side is a creatinine-based equation, and on the right-hand side is a cystatin C-based equation. So what you see here is that when you use an equation that only includes age and sex, there's an overestimation of renal function when you compare the estimated GFR and the measured GFR. If you introduce race, then statistically you solve that problem. If you use ancestry to an extent, you solve the problem. But there's a better way to do it, because if you just use cystatin C as your biomarker, that problem goes away. And if you introduce race, you begin to see that even the cystatin C, which was a good biomarker, begins to show more of a discrepancy, which tells you that there's more to this than just race. Because, you know, and I'll talk about this, the reasons for this overcorrection or overestimation are based on very flawed studies that were pretty much just made up by prior studies that stated something as fact, and it was just propagated for subsequent iterations. So what we are seeing there is if you change the serum biomarker from creatinine to cystatin C, all of a sudden you do not need any race correction at all. This is from the study. It's just a supplement that shows you exactly the problem of, you know, misclassification. So the upper panel is for black patients. That's the percentage of genetic ancestry versus self-report. So these are blacks. As you can see, there's a high proportion of European ancestry by genetic ancestry informative markers in this study. So someone who looks black may have a high proportion, up to 40% of European ancestry. And so who are they? Are they black or are they non-black? And as you can see below is for those that are reported, self-identified as non-black. But you can see there's a high proportion for some people, all the way to 70% of European ancestry. So it gets really confusing if you just go by skin tone or what somebody self-reports. Now if you go back into this study, look at, you know, the percent agreement. So the P30 is the agreement between the measured and the estimated EGFR within 30%. That is pretty good because you're casting a wide range. What that tells you is that there's a lot of variation, but up to 30% you're still fine. When you narrow the percent agreement to 10%, you fall down to the 30, 40% range. So there's a lot of variability in this data, even though we are still able to estimate GFR. But there's a lot of variability between what you're actually measuring and the actual GFR that's estimated. And then fast forward to the most recent equation. So because, you know, there's been a lot of work by the National Kidney Foundation and the Nephrology Societies, they really went back and did a lot of evaluation of how these equations came to be. What I'm showing you here is the 2009 equation data, 2012, and then the subsequent data in 2021. And what you can see is that it's the same group of studies that are being accumulated over time. And this is important because, you know, we see these data and we do not know exactly how they came to be. Well, it's based on cohorts that were developed by researchers. I told you about the initial cohort that was developed in, you know, a hospital in Canada. Well, subsequent cohorts do not have much representation as well. And so you end up with cohorts that are accumulated over time, but they still perpetuate this bias where there's not enough representation of persons of color. And so if you run a regression model with such data, you're not always going to have the kind of, you know, representation in terms of race. But it's a bigger problem because some of those cohorts are cohorts of patients that have diseases. They have preexisting conditions like diabetes, heart disease. And the proportion of representation of different races in these cohorts is not exactly the same. And so even though these equations in 2021 are an improvement over previous equations, it does not solve the problem of representation in some of the studies that were used to derive these equations. And that's very important. With that being said, if you look at the latest CKD EPI study for 2021 using creatinine as a biomarker, here from these graphs from left to the right, this one is estimating on the y-axis the measured GFR and then on the x-axis the estimated GFR. And you can see that if you use all three covariates of age, sex, and race, that's what we currently do, then you have an overestimation of GFR in black patients with a bias of up to three milliliters per minute per liter of surface area. For non-blacks, it's minimal. So that equation is tailored to favor non-blacks. If you then, as most people advocated in the beginning, to just forget about the race-based equation and use the reference data for non-blacks, you make that overestimate, you create a worse overestimation of EGFR among blacks. Of course, the whites remain the same because that's the equation that you use. How about you take race out of the equation? You actually still have some overestimation, but then it goes both ways because now, you know, you have an underestimation for whites and overestimation for black patients. So creatinine is not very ideal. How does an equation that combines creatinine and cystatin C do? A lot better. You see there's an underestimation for blacks and whites, but it's minimal. It's within the margin of error. So it's much better if you combine both biomarkers. If you use the reference for whites only, that disparity is still there. So the solution is not to abandon the current race-based equation and use only the reference for whites. The best solution will be to get rid of race altogether because when you do that, the discrepancy is okay for blacks, a little bit unfair for white patients. So right now, the recommendation as it stands is to use the new equation that does not include race for creatinine or if you have cystatin C, that would be ideal because if you use cystatin C, as mentioned, the bias is minimal, negative 0.1 for blacks and 0.7. That's quite minimal given the range of changes in EGFR. So the nephrology community has done an extensive piece of work to really demonstrate that the foundational use of race in those initial equations, the MDRD and the CKD-EPI in 2009 was very flawed. And so, you know, just summarizing this data on these equations, you can see that the current EGFR with race is problematic because as you can see, it's overestimating blacks. When you do not use race, there's much improvement. And the percent agreement, you know, for blacks and non-blacks is within 30% is actually much better when you eliminate race from the equation. I'm going to put this slide just to give a big summary of the journey that we've come with these equations. So again, about 50 years ago, I started with all white men in a hospital in northern Canada. That's the original equation. And this is the equation right here. And there was no race at that point because all the subjects were white. And race is introduced here and there's a correction factor based on the regression equation that was used. And at that time, the reason, the justification to explain the higher estimated GFR for blacks was because they claimed that blacks had more muscle mass. Well, the reality is that that was not true. Nobody had measured that. To measure muscle mass is a cadaveric study and nobody has done that before. So it was just assumed based on stereotypes, as was mentioned, that blacks had, you know, bigger muscles, just like blacks didn't feel pain. And so, that's how it came to be. But the reality is much more complex than that. It could be because of diet. It could be because of metabolic changes, how you metabolize different types of diet. So it's much more complicated than just saying blacks globally that way. And as you can see, when we evolved to the latest CKD EPI study in 2021, there's much more accurate estimation of GFR without including race in the equation. Where we go from here, I do not know. But I think that, you know, we have to go with the biomarker that is routinely measured in all of our samples, in all of our EHRs, which is creatinine, and hope that in the future, maybe we start using cystatin C as an additional biomarker. That would certainly improve things from where we are right now. And obviously, this is just a conversation that is beginning. We haven't come to the end of this conversation. So in summary, the estimated GFR has its limitations. Use of race coefficients is very problematic both ways. It's not just for blacks. It's also for blacks and for whites. And just eliminating race coefficients as, you know, just using one reference equation for whites for everybody is very problematic as well. The ideal scenario would be to combine the populations, probably get better studies that include more diversity in patients that are included in the derivation equations and the validation samples. And then, you know, finally, I think cystatin C or maybe other NOVA biomarkers that are race agnostic are much better. So thank you.

race-based equations

kidney function

glomerular filtration rate

cystatin C

biomarkers