Response

Reply from the Authors: We thank Coresh et al. for their comments. We provided NHANES III population estimates for four common methods of glomerular filtration rate (GFR) assessment: the Modification of Diet in Renal Disease (MDRD), Cockcroft-Gault, Couchoud, and reciprocal of creatinine methods (1). On the basis of the prevalence of low GFR and the consequent potential for laboratory-based reporting of calculated GFR to increase nephrology referrals by an order of magnitude, we recommended against implementation of this strategy before further evaluation. Since we wrote the article, the Kidney Disease Outcomes Quality Initiative (K/DOQI) guidelines (2) have appeared; these guidelines recommend immediate implementation of GFR-based laboratory reports. The first issue concerns the prevalence of low GFR. As Coresh et al. point out, differences between the NHANES III MDRD GFR prevalence estimates that we provide and those recently published as part of the K/DOQI guidelines (2) are largely due to the inclusion in the K/DOQI analyses of a correction factor to account for differences in calibration between the White Sands Research Center laboratory (used in NHANES III) and the Cleveland Clinic laboratory. The Cleveland Clinic laboratory was used in the MDRD study; data from these participants were used to derive and validate the MDRD equations for calculation of GFR and to validate the Cockcroft-Gault formula (3). The calibration difference was recently reported by Coresh et al. (4): creatinine measured at the White Sands laboratory is 0.23 mg/dl (20.3 μmol/L) higher than that measured at the Cleveland Clinic. This information was not available in the public domain at the time of our report. We followed the procedures of the analytic guidelines for NHANES III (5). Previous reports of renal function based on creatinine from the NHANES III data set have not included recalibration to an external standard (6). The use of a correction factor in all the K/DOQI analyses is reported in Appendix 2 (2), but is not, to our knowledge, found in the text and tables of the document. It may not be readily apparent that all K/DOQI analyses are adjusted in this way; for example, in a recent editorial (7), the reason for the difference between our results and the K/DOQI tables was questioned and the question was unanswered. In our article, as noted by Coresh et al., we reported that a calibration difference of the order of 0.2 mg/dl would have a large effect on the prevalence of low GFR, particularly at higher levels of GFR. Using the actual calibration difference, Coresh et al. have recently documented the same issue (4). To highlight this issue and to document the magnitude of the effects that are introduced by inter-laboratory differences in serum creatinine measurements, we present the prevalence of low MDRD GFR in US adults without diabetes (the study population we reported) using the newly published calibration factor. The overall prevalence of MDRD GFR less than 30 ml/min per 1.73 m2 was 0.16%, for GFR 30 to 59 ml/min per 1.73 m2 was 3.2%, and for GFR 60 to 79 ml/min per 1.73 m2 was 14.5% (Figure 1). For reference, we have also duplicated the race-gender stratified analyses using the MDRD formula applied to adjusted creatinine (Figure 2). The largest validation series for the Cockcroft-Gault formula also derived from creatinine measured at the Cleveland Clinic laboratory as part of the MDRD study; therefore, one could argue that Cockcroft-Gault estimates from NHANES III should also be recalibrated to the Cleveland Clinic laboratory. Figure 1. : Nondiabetic adults in NHANES III. Weighted distribution of the level of predicted GFR (MDRD categories) in ml/min per 1.73 m2, by age (in decades), for the total US nondiabetic black and white adult population. Uncalibrated estimates are calculated from serum creatinine as reported in NHANES III database; calibrated estimates are calculated after subtraction of 0.23 mg/dl (20.3 μmol/L) from serum creatinine to account for differences between the White Sands and the Cleveland Clinic laboratories. Methods are otherwise as described in our original publication.Figure 2. : Nondiabetic adults in NHANES III. Weighted distribution of level of predicted GFR (MDRD categories) in ml/min per 1.73 m2, by age (in decades), stratified by race and gender and calibrated to the Cleveland Clinic laboratory by subtraction of 0.23 mg/dl (20.3 μmol/L) from serum creatinine. Methods are otherwise as described in our original publication.In their letter, Coresh et al. argue that the calibration issue explains the unexpected nature of our findings. We do not agree. We discussed the calibration issue in our original paper. Even accounting for differences in inter-laboratory variation, we feel that the adjusted prevalence of low GFR is surprisingly high, particularly in the elderly. For example, our adjusted estimate for the prevalence of GFR between 30 and 60 ml/min per 1.73 m2 (stage 3 disease by the K/DOQI classification [2]) in nondiabetic Americans is 3.2%. The K/DOQI adjusted estimate for GFR 30 to 60 ml/min per 1.73 m2 in the total US population is 4.3%. Among the nondiabetic elderly, 8% of people in their sixties, 16% of people in the seventies, and 34% of people in their eighties (Figure 2), have GFR in this range and, if their values were confirmed on repeat testing, would be classified as having stage 3 chronic kidney disease according to K/DOQI guidelines. We argue that our finding of a high prevalence of GFR between 30 and 60 ml/min per 1.73 m2 (of the same order of magnitude as diabetes mellitus, which has been estimated at 5.1% of the population [8]) should lead the nephrology community to question whether a test result in this range is invariably indicative of disease. On the basis of current evidence, we believe that GFR ranging from 30 to 60 ml/min per 1.73 m2 at a population level might be better conceptualized as a probable risk factor. A thoughtful discussion of the issues that surround the classification of clinical findings as diseases or as risk factors is found in a recent article by Moynihan et al. (9). This issue of terminology applies also to the criteria used to define stage 1 and 2 chronic kidney disease in the K/DOQI classification (a further 3.3% and 3.0% of the population, respectively, according to K/DOQI estimates [2]). The second issue concerns the evidence base surrounding the recommendation for widespread laboratory reporting of GFR. We do not disagree that equation-based estimates are the preferred method of estimating GFR (we wrote, “they represent the best available option for the estimation of clearance in large studies” in the Discussion section of our article [1]) and have nowhere suggested that creatinine clearances based on 24-h urine collection are preferable. GFR is a familiar and helpful concept to nephrologists. However, most healthcare workers who review creatinine data are not nephrologists. Most of the impact of the proposed change will be on professionals who are experienced in the interpretation of serum creatinine but inexperienced in the interpretation of GFR or clearance data. This will result in identification of many people with low GFR who would not previously have been recognized, particularly in the elderly. The rationale behind recommendations for laboratory reporting of GFR is that these patients will benefit from having been recognized, and that ultimately this recognition would lead to reduction in the burden of end-stage renal disease (ESRD). This is an important and testable hypothesis, not a proven strategy. We wish to express our concern that this untried intervention will be implemented before evaluation of its consequences, not all of which will be positive. Implementation of a new strategy on this scale mandates careful consideration of the risk-benefit ratio and of cost-effectiveness. The benefits of specific diagnosis and management in people who are currently not referred to nephrologists have not been studied. Relevant data suggesting that benefits may be more limited in unreferred than in referred populations include the following. Referred populations would be expected to differ substantially from unreferred populations, and there are emerging data to support this hypothesis. Proteinuria is a potent and consistent multivariable predictor of risk of progression of renal insufficiency. In the MDRD study A (GFR 25 to 55 ml/min per 1.73 m2), 27% of patients had greater than 1000 mg/d proteinuria (10). In contrast, in NHANES III, only 3.3% of participants with a GFR between 30 and 60 ml/min per 1.73 m2 demonstrated greater than 288 mg/day of albuminuria (approximated from Couchoud serum creatinine cut-points and random urine albumin-to-creatinine ratio respectively) (11). While the prognostic significance of low GFR in unreferred populations is unknown, on the basis of proteinuria we would expect overall that unreferred low GFR is less of a risk for ESRD than is low GFR in patients seen and studied by nephrologists. The rate of loss of GFR in unreferred populations with low GFR has not, to our knowledge, been reported. Benefits of therapy to reduce the rate of progression of renal insufficiency have been tested only in referred populations. Studies of optimal approaches to the diagnosis of the etiology of renal insufficiency are lacking even for referred populations. Part 9 of the K/DOQI guidelines outlines key clinical considerations and possible diagnostic tests, but cautions that “many of the recommendations in this section have not been adequately studied and therefore represent the opinion of members of the Work Group” (2). The current state of knowledge does not permit the publication of evidence-based diagnostic pathways that could be implemented in a primary care setting. Potential adverse effects of the proposed intervention (including calculated GFR with all creatinine reports) include the following. 5.Labeling effects. Diagnosis and referral for evaluation of hypertension were shown to be associated with subsequent reductions in income and increases in absenteeism up to 5 yr after screening (12). Effects on anxiety levels and social functioning by labeling people with a diagnosis of low GFR might also be anticipated. These factors are particularly important in the light of the K/DOQI recommendation that patient and public education programs be developed encouraging the philosophy “Know your number!” (2). 6.Overlabeling due to inter-laboratory variation of serum creatinine is a potential problem. The College of American Pathologists reported that in laboratories surveyed in 1994, creatinine was overestimated on average by 13 to 17% (0.12 to 0.17 mg/dl; 11 to 15 μmol/L) (13). As we reported in our original article, when creatinine is converted into GFR, a difference of this order of magnitude has a large impact on the prevalence of disease. In NHANES III, the prevalence of GFR 30 to 60 ml/min per 1.73 m2 is increased fourfold (12.5% versus 3.2%) when the calibration factor is not used for creatinine measured at the White Sands laboratory (Figure 1). 7.Increased investigation. Most suggested investigations for the etiology of chronic renal insufficiency are noninvasive, but some, including renal biopsy and angiography, carry a risk of harm. Harm associated with screening strategies that are superficially noninvasive is well documented in other contexts (14,15). Increased investigation will also increase the costs of care. To our knowledge, no estimates of the cost-effectiveness of any investigational strategy for low GFR/CRI are available. 8.Increased referral to nephrologists. In addition to the general issue of increasing healthcare utilization, increased referral poses a specific challenge to a limited societal resource. The continuing and projected growth of ESRD implies that this resource will be severely taxed over the next decade, and it is unclear whether the training of adequate numbers of new nephrologists to account for this expansion will be feasible (16). The Ad Hoc Working Committee estimated that there were about 5000 nephrologists in practice in the United States in 1997 (16). Given the prevalence of MDRD GFR 30 to 60 ml/min in the population (around 4%; 7.6 million Americans by corrected estimates [2]), the potential for increased workload is immense should the proportion of those with GFR in this category who are referred suddenly increase. This would also negatively impact the care of the more selected group of patients who are currently referred and of those with ESRD. We share the concerns about the limitations of serum creatinine expressed by your correspondents and summarized in detail in the K/DOQI guidelines. We also agree that strategies are urgently required to reduce the societal burden of ESRD. Reporting GFR (if this is implemented in conjunction with appropriate calibration to a gold standard) is a potential strategy to this end; our point of disagreement is whether the available data are sufficiently strong that this should be broadly implemented now without further evaluation. We argue that the current state of knowledge of the community epidemiology of renal disease is such that scarce resources would be better devoted to further research into appropriate methods of detection, diagnosis, and treatment of those most at risk for progressive chronic renal insufficiency and ESRD. Yours sincerely,

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.