The Conundrum of Chronic Kidney Disease Classification and End-Stage Renal Risk Prediction in the Elderly - What Is the Right Approach?

The worldwide high prevalence of chronic kidney disease (CKD) and the increasing number of patients reaching endstage renal disease (ESRD) are a matter of major concern. The most widely accepted classification system of CKD is that proposed by the Kidney Disease Outcomes Quality Initiative (KDOQI) in 2002. When applying this system, it has become apparent that the prevalence of CKD is particularly high in elderly subjects. The fact that this system is mainly based on estimated glomerular filtration rate (eGFR), subdividing the severity of CKD into five stages, is a matter of debate. A main issue is that although a reduced eGFR is often encountered in elderly subjects, most of these subjects do not have a renal disease leading to an increased risk of ESRD, i.e. the predictive power of ESRD is unsatisfactory. Recent advances have been put forward to improve (1) estimation of GFR and (2) prediction of ESRD. In this review, we discuss the currently available data with a focus on the elderly and propose an improved classification system of CKD which is characterized by a substantially better diagnostic accuracy for progression Published online: July 24, 2010 Stein Hallan Department of Medicine, Division of Nephrology, St. Olav’s Hospital Olav Kyrres gt 17 NO–7000 Trondheim (Norway) Tel. +47 7386 7273, Fax +47 7386 9390, E-Mail stein.hallan @ ntnu.no © 2010 S. Karger AG, Basel 1660–2110/10/1164–0307$26.00/0 Accessible online at: www.karger.com/nec D ow nl oa de d by : 54 .7 0. 40 .1 1 10 /6 /2 01 7 4: 59 :1 1 A M Hallan /Orth Nephron Clin Pract 2010;116:c307–c316 c308 that a large proportion of CKD cases, especially among the elderly, may be misclassified and that their low estimated glomerular filtration rate (eGFR) is just a ‘normal’ reduction of kidney function with increasing age. Current CKD Classification The 2002 KDOQI classification system for CKD is mainly based on degrees of reduction in GFR [10] . Stages 1–5 are defined as eGFR 6 90, 60–89, 30–59, 15–29 and ! 15 ml/min/1.73 m 2 , respectively, and the findings should be consistent for more than 3 months. For stages 1–2 signs of kidney damage (albuminuria, hematuria or sonographic abnormalities) are also required. GFR should be estimated with formulas based on serum creatinine taking age, sex, and other demographics into account. The first publications using the KDOQI system were hampered by estimation formula problems [11, 12] . The first Modification of Diet in Renal Disease (MDRD) study formula was based on an unusually low calibrated serum creatinine method [13] , and the use of the CockcroftGault formula results in a too steep decline of kidney function with age [14] . Both problems lead to a substantial underestimation of GFR thereby overestimating the CKD prevalence. A recalibration of the MDRD formula based on isotope dilution mass spectroscopy traceable serum creatinine values improved performance substantially, and recently the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation was published with even better performance [15] . By using the CKD-EPI equation, the mean systematic underestimation at GFR 60 ml/ min/1.73 m 2 and upwards has been reduced from –10.6 to –3.5 ml/min/1.73 m 2 and precision has increased, resulting in 88.3% of the results being within 8 30% of the true GFR (as measured by radiolabeled tracers) compared to 84.7% with the recalibrated MDRD formula [15] . However, for the individual patients there is still a clinically significant inaccuracy (e.g. errors of 8 20% are not unusual), and this can – especially in eGFR-dominant systems – lead to misclassification. When using this new equation, the median eGFR in the general population shifts from 85 to 95 ml/min/1.73 m 2 , and a substantial proportion of subjects is reclassified from the 30–59 ml/ min/1.73 m 2 category to the 60–89 ml/min/1.73 m 2 category, i.e. from stage 3 to stage 2 of the current KDOQI CKD classification [15] . The CKD-EPI equation should be well suited for use in the elderly, because a total number of 1,592 subjects above age 65 was enrolled during development, internal and external validation of this formula [15] . In contrast to middle-aged subjects, there is no shift of the median eGFR in elderly subjects using the CKD-EPI equation. Table 1 shows the prevalence of CKD stages 1 through 4 in elderly subjects from two large population-based studies using the MDRD and the CKD-EPI equations. Independent of the formula used, the CKD prevalence is consistently found to be very high. Controversies over the Current CKD Classification There has been a brisk debate over the last couple of years whether the CKD epidemic is fact or fiction. In a series of articles, Glassock and Winearls [7–9, 16, 17] put forward strong critiques about the current eGFR-focused CKD classification, which in turn, has been strongly defended by the KDOQI chairpersons [5, 6] . The main critique has been related to ‘the improbable high prevalence estimates’, a fixed cutoff for abnormality for all ages, and omission of the possibility that a decline in kidney function may be a normal part of the aging process. The KDOQI 2002 classification system definitely has shortcomings as a renal risk score, but the above mentioned critiques, which are clearly interrelated, also have deficits. First, we have repeatedly learnt over the past 40 years that normality and percentiles are not always the optimal approach for diagnostics and risk stratification. Cutoffs for blood pressure and cholesterol have repeatedly been lowTable 1. Prevalence of CKD among the elderly in the USA and in a European population NHANES (1999–2006) H UNT 2 (1995–1997) MDRD CKD-EPI MD RD CKD-EPI Stage 1 1.2 0.6 4.0 1.5 Stage 2 7.6 8.5 11.5 14.0 Stage 3 35.5 35.3 18.6 18.7 Stage 4 1.9 2.4 0.7 0.9 Total 46.3 46.8 34.8 34.9 Dat a are percentages of the general population. Prevalence estimates were based on eGFR in subjects aged >70 years. GFR was estimated with the MDRD study equation for use with calibrated serum creatinine values and the CKD-EPI equation. The prevalence estimates for the Norwegian HUNT 2 study were based on the same study sample as previously published [2]. D ow nl oa de d by : 54 .7 0. 40 .1 1 10 /6 /2 01 7 4: 59 :1 1 A M CKD in the Elderly Nephron Clin Pract 2010;116:c307–c316 c309 ered beyond the classical 97.5th percentile, and very large proportions of the general population are now defined as hypertensive or as having hypercholesterolemia [18] . A very high prevalence of CKD in the elderly may therefore indeed be plausible. Second, a number of pathological findings are associated with reduced kidney function in elderly subjects. The prevalence of increased albuminuria is much higher in the elderly compared to the young [19] , a phenomenon that is clearly related to pathological processes in the kidney. Histological changes like glomerulosclerosis, tubular atrophy, interstitial fibrosis, and arteriolar sclerosis, which are typical findings of the common pathway of progressive CKD, are increasingly found with higher age even in the absence of hypertension and diabetes [20] . Figure 1 illustrates such histological alterations in a renal biopsy specimen. Likewise, the cortical thickness is reduced and the overall size of the kidney decreases [20] . All of these changes are considered highly pathological when observed in the younger, and it is difficult to argue that this should not also be the case in the elderly. Reduced Kidney Function in the Elderly The reduction of GFR with increasing age deserves special attention. Studies have shown that there is a gradual decline in measured GFR with age [21–23] . The classical work by Davies and Shock [21] showed that mean GFR (measured using the gold standard of inulin clearance) decreased from 123 (SD 16) ml/min/1.73 m 2 at age 20–29 years to 65 (SD 20) ml/min/1.73 m 2 at age 80–89 years. This indicates that there is a true reduction in kidney function among the elderly and that this may not only be a phenomenon created by inaccurate or biased equations for estimating GFR. However, the majority of participants in the previous study were in-house patients with acute or chronic infections, atherosclerosis or other diseases and treatments that potentially could influence kidney function although patients with obvious cardiorenal diseases were excluded. On the contrary, there are good indications that the reduction of GFR at higher age is only modest, as long as the subjects investigated are healthy. For instance, Fliser et al. [24] found that this was the case when comparing 10 healthy subjects with a median age of 70 years (range 61–82) with a group of 15 subjects with a median age of 26 years (range 23–32). They also documented that at least until the age of 80 years renal functional reserve, i.e. the increase in GFR after an amino acid load, was preserved in these subjects independent of gender. Thus, age-related changes of renal function are less severe than previously thought. However, the arterial wall is substantially remodeled with increasing age even in the healthiest subjects. These normal changes caused by the joint effects of biomolecular, cell and matrix modifications are significantly aggravated by classical cardiovascular risk factors such as hypertension, diabetes and smoking, i.e. conditions often encountered in elderly subjects [20, 25] . One of the consequences is increased arterial stiffness, and significant associations have been found even to only small reductions of eGFR and to low-grade albuminuria. The kidneys are, as a low resistance organ, thought to be especially prone to the resulting high pulse pressures, and resulting microcirculation changes lead to kidney dysfunction through ischemia. Decreasing kidney function could then lead to calcium-phosphate depositions and increased oxidative stress setting up a vicious circle as indicated in figure 2 [26, 27] . To separate the effect of toxins and diseases that may injure the kidneys over time and the ‘natural aging’ caused by replicative senescence and oxidative stress, we must study a population that is very unlikely to have any kind of kidney disease and no cardiovascular risk factors. Kidney donors are probably the most valid study group to analyze this problem. These subjects undergo all available noninvasive tests to ensure that donation will have D