Glomerular filtration rate (GFR) is considered the best overall measurement of kidney function and is an important clinical variable used in determining the optimal management of a patient with chronic kidney disease, determining the risks of radiologic procedures that require the administration of contrast, and determining doses of drugs with significant renal excretion.1 Current practice is to estimate GFR using an estimation equation that takes serum creatinine, age, and sex as variables, and to present that result whenever serum creatinine is reported.2 It has become standard practice to report a higher level of GFR for Black individuals than White individuals for a given level of serum creatinine using the same multiplicative factor irrespective of age, sex, or level of GFR. This practice has come under scrutiny in recent years as part of a general reevaluation of the use of race in medical decision making.3 Accurate measurement of GFR is cumbersome and impractical in day-to-day clinical medicine as it requires infusion of an exogenous substance that is excreted exclusively by glomerular filtration. Creatinine, an endogenous substance that is produced from metabolism of creatine, an important energy source in muscle, is excreted in large part via filtration through the glomerulus, so is measured routinely in clinical practice as a reflection of GFR.4 While percentage changes in serum creatinine in a given patient may be a good approximation (in the steady state) of percentage changes in GFR, a given level of serum creatinine can represent a wide range of GFR values.5 In 1 data set, a serum creatinine of 1.5 mg/dL was associated with measured GFRs from 40 to 90 mL/min.6 The serum creatinine in the steady-state condition is that level that will allow daily creatinine excretion to equal creatinine production. A given serum creatinine may reflect a low GFR if creatinine production is low, or a high GFR if creatinine production is high. Daily creatinine excretion has been shown to range from ≈500 to 2000 mg per day.7 Creatinine excretion is thought to be a reflection of muscle mass to a great extent, though there is a dietary component related to meat intake.8 For many years, a better approximation of GFR than could be obtained from serum creatinine alone was generated through a collection of urine for 24 hours in order to measure creatinine excretion over that period of time. The total 24-hour creatinine excretion divided by the serum creatinine gives what is referred to as creatinine clearance. It is the volume of blood that would have to be hypothetically completely cleared of creatinine in that time period (generally expressed as mL/min) to give the observed excretion rate. If a substance is freely filtered by the glomerulus and excreted solely by the kidney through filtration and without tubular reabsorption, its urinary clearance rate will equal GFR. Creatinine is also secreted by the tubules, and the fraction of excretion by this route increases as GFR decreases. Therefore, creatinine clearance will overestimate GFR, with the degree of overestimation increasing as GFR decreases.4 Twenty-four-hour urine collection is also impractical in most clinical situations where an estimate of GFR is needed, and that collection is often improperly obtained, so a number of equations for estimating GFR based on creatinine, and other clinical variables, including age, sex, and in some cases weight, have been developed.9 These equations have been clearly shown to predict measured GFR better than simple serum creatinine measurement, and may be in general more accurate than measuring creatinine clearance via 24-hour urine as a GFR estimate.9 One of the first equations to undergo widespread use was the Cockcroft-Gault (CG) equation first reported in 1974.10 This equation was developed by performing regressions of serum creatinine and other variables against measured creatinine clearance, so one limitation is the problem of overestimation of GFR by creatinine clearance. The CG equation uses serum creatinine, age, and sex as variables, as well as weight as a proxy for muscle mass. There is concern about the accuracy of this given the increased incidence of obesity since the equation was developed. Attempts to use ideal body weight to account for weight differences related to fat were actually less accurate than the original.11 The CG equation was replaced for most uses by the Modification of Diet in Renal Disease (MDRD) equation developed in 1999,12, 13 and this was largely replaced by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation developed in 2009.1 These equations were generated using regressions of serum creatinine and other variables against measured GFR. Currently, the CKD-EPI equation is used to calculate GFR using the serum creatinine, age, and sex of a patient, and the result is reported with lab results when serum creatinine is ordered. Results of the CG equation are presented as milliliters per minute. It has been observed that kidney size and GFR vary directly with body size. Presumably, this reflects the need for greater filtration to process metabolic products of a larger body, so that normal GFR in a larger individual should be higher than in a smaller individual.14 It has been customary for many uses to present GFR as mL/min/1.73 m2, which was thought to be average body surface area when the body surface formula was developed in 1928.15 This normalization is done in theory to allow better comparison of measured or calculated GFR with normal values. This is how GFR is customarily expressed in the MDRD and CKD-EPI equations. However, the use of body surface area to normalize GFR has never been rigorously tested. A poor correlation between body surface area and GFR has been found, and other physiologic variables have been suggested as alternatives.16 Calculation of body surface area has limited accuracy, and may be particularly inaccurate in heavier patients.17 The standard for acceptable accuracy of these equations has been that at least 90% of the calculated values are within 30% of the true GFR.18 Thus, if true GFR were 60 mL/min, an equation would be considered acceptable if most of the calculated GFRs were between 42 and 78. Because of this relative lack of accuracy, some have questioned the use of these equations at all.6 A group instrumental in developing these equations recognizes the limitations but argues that they are still useful.19 They have been shown to predict GFR better than serum creatinine alone. The daily creatinine production in women is on the order of 30% lower than in men on average. Similarly, the daily creatinine production in people in their 80s is about 30% lower than people in their 20s.7 Thus, the calculated GFR that factors in sex and age accounts for a great deal of the variability in creatinine production that makes serum creatinine alone such a poor predictor of GFR. Data used in the development of the MDRD and CKD-EPI equations showed that measured GFR for a given level of serum creatinine was higher in Blacks than non-Blacks. Both of these equations include a multiplicative factor that increases the calculated GFR in Black individuals compared to non-Black individuals for a given level of serum creatinine and identical age and sex.1, 12 It has been assumed that at least a large part of the explanation is the increased muscle mass that had been found in Blacks in some studies.20 This factor is 1.21 in the MDRD equation, which was developed using a population of patients with renal disease with a mean serum creatinine of 2.13 mg/dL, in which only 12% of the patients were Black.12 This factor is 1.16 in the CKD-EPI equation, which was developed using a population with a broader range of kidney function, in which 35% of the patients were Black1 (though a large percentage of the Black patients had been enrolled in a study of chronic renal disease,21 so average serum creatinine in this group was likely higher than in the rest of the study population22). There are several clinical situations in which an estimate of GFR is needed to deliver optimal care. One is in the care of patients with chronic renal disease, in particular in deciding when referral to a nephrologist is warranted. One of the criteria for referral is a GFR <30 mL/min.2 There are a number of studies documenting better outcomes with earlier referral to nephrology.23 GFR has also been used to determine the safety of certain radiologic procedures using intravenous contrast. Recent data though has shown that the cause of renal failure following the administration of intravenous contrast for computerized tomography scans and angiograms is mostly related to factors other than the contrast material itself. There is still thought to be some increased risk in patients with GFR <30 mL/min/1.73 m2, but the magnitude of that risk is uncertain.24 Renal insufficiency has been thought to be a risk factor for the development of (nephrogenic) systemic fibrosis following the administration of gadolinium for magnetic resonance imaging. The risk appears to have dropped dramatically with the development of newer agents.25 Probably, the most common use of estimated GFR in day-to-day clinical medicine is to adjust dosages for drugs that are excreted to a large extent through the kidneys. As GFR declines with age even in the absence of specific renal disease,25 choosing appropriate doses in the face of a reduced GFR is a problem faced routinely in clinical practice. Dose adjustments based on GFR are presented in the manufacturer's insert based on data obtained during initial trials.26 Data may also be obtained in postmarketing studies. Results are also presented in standard texts such as Up to Date, and in publications devoted specifically to this issue, though frequently these present conflicting data.27 The recommendations are often based on limited data.26 Dose adjustments for reductions in GFR are important for drugs in which a substantial percentage of the drug and/or active or potentially toxic metabolites is excreted by the kidney. The effect of reduction of GFR on drug pharmacodynamics is complex. Some drugs are excreted via tubular secretion, and alterations in this pathway may not parallel the decline in GFR. GFR reductions may also affect volume of distribution and nonrenal elimination pathways, though it is not clear how much GFR needs to be reduced to affect these parameters. Thus, calculating drug dosing using theoretical calculations based on the expected change in half-life or clearance from the GFR reduction is only approximate.28 Older studies of pharmacokinetics used measured creatinine clearance as an estimate of GFR,29 whereas more recent studies have generally used the CG equation to estimate GFR.30 This is appropriate in that the absolute rather than normalized GFR is the relevant variable that determines the drug excretion rate.31 However, the CKD-EPI equation has become the standard for estimation of GFR in clinical reporting. These 2 equations may give different GFR estimates, even after converting to absolute GFR from the normalized CKD-EPI equation result.9, 30 The majority of currently available online calculators use the CKD-EPI method, and provide normalized estimates of GFR. Nonnormalized estimates will also be provided, but require entry of height and weight in addition to serum creatinine, age, sex, and race. Though the Food and Drug Administration recommends using nonnormalized values of GFR to decide on drug dosing,31 the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) suggests using nonnormalized GFR specifically in patients at extremes of body size, though the range of those extremes is not specified.32 First, at extremes of body weight, the formulas for estimated body surface area may not be accurate.17 Second, calculated body surface area varies considerably, even among people who are not at extremes of size. For example, a person 67 inches in height who weighs 139 pounds would have a calculated body surface area of 1.73 m2 (using the Dubois and Dubois formula), so nonnormalized and normalized GFR would be identical. But a person 70 inches in height weighing 180 pounds would have a calculated body surface area of 2.2 m2, so that the nonnormalized GFR would be 27% higher than the normalized GFR reported. More fundamentally, the inaccuracy of the CG (and other GFR estimating equations) limits the validity of the recommendations of drug dosing that are arrived at. Recommendations group patients into broad categories of GFR, with a greater number of categories (with narrower GFR ranges) given for drugs with a narrower therapeutic window. Discordance between recommended dose based on measured GFR versus estimated GFR have been shown to result in misclassification of a significant number of patients, particularly for those drugs with narrower GFR ranges that have the narrower therapeutic window.30 Drug dosing guidelines based on estimated GFR must be viewed as an approximation at best. For drugs with a narrow therapeutic window such as chemotherapeutic agents, more accurate estimates of GFR may be advisable.18 The practice of using a multiplier to calculate GFR for Black individuals has come under scrutiny in recent years not only in part because of concerns about the effects of the use of race in medical decision making3 but also because of questions about the scientific basis of this.33 A lower correction factor was found in European Blacks.34 In a study of Black South Africans, the MDRD equation was found to be a better predictor without the race coefficient.35 This is further complicated particularly in the United States by increasing numbers of people of mixed race.36 Though some of the variability in GFR for a given level of serum creatinine may be related to diet, it has been presumed that a large part of this is related to muscle mass. In one study using dual-energy X-ray absorptiometry to estimate muscle mass, mean skeletal muscle mass at age 27 (average age of peak muscle mass) in Black males was 6.6% higher than that of White males, and 5.9% higher in Black women than in white women.20 Skeletal muscle mass at age 27 in Black men was 65% higher than in Black women, and 64% higher in White men than in White women. At the same time, the rate of decline in skeletal muscle after age 27 was found to be more rapid in Black men and women then in their White counterparts. The rate of loss of skeletal muscle mass after age 27 averaged 0.181 kg per year in Black men, and 0.126 kg per year in White men, so that mean skeletal muscle was 28.26 kg (a 20% loss) in a 67-year-old Black man, and 28.26 kg (a 15% loss) in a 67-year-old White man. The decline after age 27 was also more rapid in Black women compared to White women, 0.111 kg per year and 0.065 kg per year respectively, so that mean muscle mass at age 67 was 17.06 kg (a 20.7% loss) in Black women and 17.7 kg (a 12.8% loss) in White women. Studies of the decline in creatinine excretion show comparable age-related losses, though results were not stratified by race.7 The differences in muscle mass between men and women and between young and old are much larger than the differences between Black and White individuals. The latter difference in this study decreased with age, so that mean muscle mass was identical in Blacks and Whites at age 67. The corresponding GFRs calculated by the CKD-EPI equations for males and females at age 27 and 67 years, respectively, with a serum creatinine of 1.5 mg/dL are shown in Table 1. The calculated GFR for men is about 30% higher than women of the same age and race, and about 30% higher for at 27-year-old than a 67-year-old of the same sex and race. As discussed above, the calculated GFR for Black individuals is 16% higher than for non-Black individuals of the same sex and age. Some writers objecting to the use of race in medical algorithms base their objection on the concept that race is a social construct rather than a scientifically verifiable biological category.37 The concept of race in science has a complex history. Linnaeus, who classified much of the biological world during the 18th century, classified humans by geographical region and skin color, along with commentary on behaviors of these groups, without explicit ranking.38 In the middle of the 19th century, the American School of Anthropology claimed to have scientific evidence of the superiority of the White race.38 The eugenics movement, founded by Francis Galton after reading his half cousin Charles Darwin's book The Origin of Species, aimed to give “the more suitable races or strains of blood a better chance of prevailing speedily over the less suitable than they otherwise would have.”39 Rather than relying on natural selection, the program involved selective breeding and forced sterilization, the culmination of which were the Nazi atrocities performed in the name of racial purity. Writers such as DuBois and Boas countered the eugenics concept with the argument that the concept of race was largely what is now termed a “social construct.”40 These arguments gained traction when, in reaction to the Nazi atrocities, writers such as anthropologist Ashley Montagu further challenged the biological basis for race.41 Montagu was one of the authors of the United Nations Educational, Scientific, and Cultural Organization statement in 1950 that race “is not so much a biological phenomenon as a social myth.” Because of backlash to this statement, a revised statement was released the next year that decried statements “regarding purity of the races and the hierarchy of inferior and superior races to which this leads,” but also stated that studying differences among races was not inherently racist.40 Studies abounded over the next several decades that claimed to show a biological link between race and intelligence,42 In 1962, anthropologist Frank Livingstone proclaimed, “There are no races, there are only clines”—clines being gradations of traits rather than clear demarcation among groups—to which evolutionary biologist Theodosius Dobazhansky commented: “Since human populations … differ in the frequencies of one or more, usually several to many, genetic variables, they are by this test racially distinct.”43 Richard Lewontin performed studies of proteins in people from different racial groups and found that 85% of the variation could be found within these groups, and only 15% of the variation was attributed to differences across groups. He concluded, “Human racial classification is now seen to be of no social value and is positively destructive of social and human relations. Since such racial classification is now seen to be of virtually no genetic or taxonomic significance either, no justification can be offered for its continuance.”44 Edwards issued a rebuttal claiming that by looking at differences in individual proteins rather than the “correlation structure in the data,” the true extent of the differences in these groups was hidden.45 The advances in our understanding of genetics on a molecular level have added a great deal of data to help resolve or add fuel to this issue, depending on one's point of view. Initial studies mapping the human genome showed that humans are identical in 99.9% of their genes. In the genes that do display variation, the variation across individuals within any racial group is much greater than the variation between groups.46 The leaders of both projects that sequenced the human genome, Francis Collins and Craig Venter, interpreted the results as showing that there is no genetic basis for the race concept.40 Several years later, Francis Collins would say, “Increasing scientific evidence however, indicates that genetic variation can be used to make a reasonably accurate prediction of geographic origins of an individual. … As those ancestral origins in many cases have a correlation, albeit often imprecise, with self-identified race or ethnicity, it is not strictly true that race or ethnicity has no biological connection.” He did go on to say the connection is “generally blurry” in part because of “the lack of defined boundaries between populations.”47 A prominent geneticist, David Reich, writing in the New York Times in 2018, stated, “I have deep sympathy for the concern that genetic discoveries could be misused to justify racism. But as a geneticist I also know that it is simply no longer possible to ignore average genetic difference among ‘races.’ ”48 A rebuttal was penned by 67 scientists and researchers. They stated, “For centuries race has been used as potent category to determine how differences between human beings should and should not matter. But science and the categories it constructs do not operate in a political vacuum. Population grouping becomes meaningful to scientists in large part because of their social and political salience, including, importantly, their power to produce and enforce hierarchies of race, sex and class.” They go on to say that Reich's statement “misrepresents the many scientists and scholars who have demonstrated the scientific flaws of considering ‘race’ a biological category. Their robust body of scholarship recognizes the existence of geographically based genetic variation in our species, but shows that such variation is not consistent with biological definitions of race. Nor does that variation map precisely onto ever changing socially defined racial groups.”49 Despite some pleas to discontinue this type of research out of concern that this will help perpetuate racism,50 research of this kind seems unlikely to go away. There are clearly poorer medical outcomes related to race that are attributable to differences in the social determinants of health, including access to care.51 Aside from the egregious examples such as the Tuskegee experiment, there are also a number of documented inequities in health care delivery related to race. For example, Black patients have been found to be less likely to receive appropriate cardiac procedures, less likely to receive a kidney transplant, and less likely to given adequate pain medication than similar White patients.52 One interpretation is that these findings reflect implicit bias rather than overt racism.53 Some of the research into genetic and physiological differences among races has the goal of identifying genetic or biochemical markers that might confer increased risk for a given disease, or affect responses to drugs, with a goal of improving health outcomes. A number of clinical algorithms have used race as a factor in assigning risk or planning treatment for a number of diseases, with or without specific knowledge of the genetic basis of this. A review of some of these practices has found that they often had an unclear scientific basis or were based on biased data.3 Perhaps the classic algorithm that uses race as a branching point is the treatment of hypertension. One of the early approaches to understanding hypertension involved categorizing hypertensives based on plasma renin levels. It was noted that hypertensive Blacks tended to have lower plasma renin levels than hypertensive Whites, a fact still commonly cited. In actuality, while the rate of low-renin hypertension in Black individuals has been reported to be approximately 50% to 70%, the rate in White individuals was found to be 30% to 50%.54, 55 This is a large difference, but the rate in White individuals is substantial, so differences in renin levels cannot be used to justify different treatment approaches in different racial groups. The finding of low plasma renin levels in hypertensive patients led to speculation that the hypertension in this group was related to increased blood volume, so that diuretics would be the most effective treatment. The 2017 American College of Cardiology/American Heart Association hypertension guidelines do recommend thiazides or calcium channel blockers as the initial choice in treating hypertension in Blacks, calling the evidence for this moderate.56 Up to Date makes the same recommendation, but gives the recommendation a grade of 2c, which they describe as “a very weak recommendation; other alternatives may be equally reasonable.”57 This recommendation is based largely on the ALLHAT study, which randomized patients to initial treatment with chlorthalidone (a thiazide-like diuretic), amlodipine (a dihydropyridine calcium-channel blocker), and lisinopril (an angiotensin-converting enzyme inhibitor).58 There was no difference among treatments in the primary outcome of combined fatal coronary heart disease or nonfatal myocardial infarction. There were some differences in secondary outcomes, with a higher incidence of stroke in Black patients in the lisinopril group vs the chlorthalidone group. This difference is likely explained, at least in part, by better blood pressure control in the chlorthalidone group. Conclusions of outcome differences among antihypertensive agents is critically dependent on the achievement of similar degrees of blood pressure control with each agent. As there is some evidence that complications of hypertension are not linearly related to blood pressure, similar mean blood pressures may not by themselves indicate equivalent degrees of control.59 The mean blood pressure in the lisinopril group was higher than that in the chlorthalidone group, and the standard deviation of blood pressures in the lisinopril group was higher than in the other 2 groups. Some studies have shown that Black individuals do respond to angiotensin-converting enzyme inhibitors, but may need higher doses.60 The evidence justifying a different approach to hypertension based on race is weak. With the genomic revolution, the idea of personalized medicine has been developed: medical care tailored for the individual based on genetic differences. The field of pharmacogenomics has a goal of tailoring drug therapy to the individual based on relevant genetic variants. The ultimate goal would be to sequence a patient's entire genome, but as that is not yet practical, race has been asked to serve as a proxy for genetic differences. So far, that approach has not been fruitful,61 and is felt to be problematic for a number or reasons.62 This does not eliminate the possibility that factoring race into medical decision making will have net clinical benefit in some cases. But given the concern that factoring race into medical decision making may foster ongoing implicit bias, Eneanya has suggested that to justify this the potential clinical benefits need to be clear-cut and substantial, and not achievable by other means.37 The clinical costs and benefits of using the multiplier must be weighed against the costs and benefits of not using it. Not using the multiplier will potentially lead to underestimation of GFR, which might lead to underdosing and therapeutic failure, whereas using it might cause overestimation of GFR which might lead to drug toxicity. Both can potentially be mitigated to some degree through careful monitoring of response (eg, temperature and white blood cell count in the case of antibiotics) or toxicity (eg, drug concentration monitoring or electrocardiographic QT interval assessment in the case of certain antiarrhythmics), but there is no way to assess the overall potential for benefit vs harm of using the multiplier on drug-related outcomes. Other approaches to estimating GFR such, as using equations based on another endogenous substance cystatin C, or even direct measurement of GFR, might be needed in some cases (eg, when planning longer-term treatment with drugs that would have a high cost of underdosage or overdosage, such as chemotherapy). As discussed above, the use of estimated GFR to guide drug dosing can only be viewed as an approximation. The use of the multiplier will add little in reducing this error. Judgment needs to be used as to whether underdosing or overdosing might lead to the more serious error for the given drug and clinical situation. Overestimation of GFR may result in exposure to potentially harmful contrast agents for radiologic studies, whereas underestimation may lead to withholding of necessary diagnostic tests. As discussed above, concerns for toxicity related to intravenous contrast dye have lessened with more careful analysis of postcontrast acute renal insufficiency and with the development of safer forms of gadolinium. One of the biggest concerns is the potential impact of using the multiplier on the care of patients with chronic kidney disease. Assigning a higher GFR for a given level of serum creatinine based on race may delay appropriate referral to nephrology, thereby resulting in poorer outcomes, and delay access to transplant waiting lists.63, 64 Additionally, the prevalence of chronic kidney disease in Black individuals may be underestimated, leading to inadequate attention to this as a public health issue.65 On the other hand, there is concern that underestimating GFR may cause undue anxiety.37 Here, there would seem to be a clearer asymmetry between the cost of using the race multiplier and not using it. Blacks are known to have a higher incidence of renal disease, with poorer outcomes.66 Certainly as GFR falls, other approaches to estimating GFR would be helpful to ensure optimal management. But using the race multiplier may delay consideration of referral and more careful appraisal of GFR. Decisions need to be made based on kidney function in clinical medicine on a regular basis. Serum creatinine alone gives a poor estimate of GFR, given the wide range of GFR values observed for a given serum creatinine, reflecting a wide range of creatinine excretion rates. Given the large difference in creatinine excretion between men and women, and between the young and the elderly, equations that estimate GFR based on sex and age will account for a good deal of the variation in observed creatinine production. Given the lack of a clear scientific basis for using a constant race multiplier across all ages and over the full spectrum of GFR, and the small correction involved relative to corrections based on sex and age, there is no reason to feel that using this factor will add precision to the calculation. The increase in the number of multiracial people in the United States and elsewhere raises further questions about the use of this correction factor.36 Given the concerns that the continued use of race in medical decision making will help perpetuate inequities in health care, it is hard to justify the continued use of this factor given the lack of clear clinical benefit. Several major teaching hospitals have stopped using this factor in clinical reports.67 Since much of the variation in GFR for a given level of creatinine is thought to be related to muscle mass, new approaches need to be developed for estimating muscle mass. Many, perhaps most, clinicians are likely not aware of the imprecision of these GFR estimates. Until better formal methods become available to estimate muscle mass that can be incorporated in these equations, clinicians will need to use the GFR estimating equations along with a clinical assessment of muscle mass for all patients to judge whether they are likely to be overestimates or underestimates. This approach is being used at Zuckerberg San Francisco General Hospital,67 and could be formally tested. The author has no conflicts of interest to disclose.
Search from 250 M+ research papersSearch from 250 M+ research papers
Open Access
Apr 3, 2024
A. S. Nozdricheva + 6