Answers are grounded in this paper and cite their sources.
In 2002, the Kidney Disease Outcomes Quality Initiative (KDOQI) published a new chronic kidney disease (CKD) classification based on five categories of estimated glomerular filtration rate (eGFR) [1]. For near-normal kidney function, additional signs of kidney damage were required (Table 1). Previously, there was no commonly accepted definition of CKD, and serum creatinine values from 120 to 350 µmol/L had been used as cut-off [2,3]. The new classification was, therefore, nothing else than a paradigm shift in nephrology: the main attention was moved from end-stage CKD to less advanced stages of the disease with the aim to improve early CKD detection. The latter is fundamental to enable preventive measures. In 2004, the Kidney Disease Improving Global Outcome (KDIGO) foundation conducted a survey to nephrologists worldwide (12% of 10 000 responded) on their opinion on the new classification and later hosted two conferences (2004 and 2006) to reanalyse the KDOQI CKD classification. Global use of the KDOQI 2002 classification system was endorsed after minor modifications (i.e. specifying the presence of dialysis or transplantation) [4], and the classification has over the subsequent years had a very strong and positive impact on CKD awareness, research and public health policy. Definition of CKD Stages 1–5 according to the KDOQI 2002 classification system Findings should be present for more than 3 months. Definition of CKD Stages 1–5 according to the KDOQI 2002 classification system Findings should be present for more than 3 months. The prevalence of CKD has been found to be high worldwide [5–7]. Furthermore, CKD has emerged as a strong independent risk factor for cardiovascular and total mortality [8–10]. CKD is, therefore, increasingly recognized as a global public health problem and not only a sub-speciality end-organ damage problem needing renal replacement therapy (RRT) [11]. CKD screening has been suggested to improve the clinical outcome in CKD patients and to potentially reduce the need of RRT, which has increased strongly during the last decades and imposes a major burden on health budgets [12]. However, there are still some important aspects of CKD screening that need to be improved before a general recommendation can be made. The major questions are who and how we should test for CKD and are those cases identified really at risk of progression to kidney failure? These aspects are closely related to the CKD definition, and the KDOQI 2002 classification is, therefore, currently under vivid debate. A substantial number of letters and editorials have been published lately challenging the current CKD definition [13–17]. No common solution has been reached so far, but all seem to agree that a revision of the KDOQI 2002 classification system is needed, including the leadership of KDOQI and KDIGO [18]. The appropriateness of the KDOQI definition was questioned already in 2004. Catherine Clase warned against relying on eGFR only and suggested the inclusion of proteinuria [19]. This excellent and clear-sighted article did not, however, generate much debate on the topic. It was not until Glassock and Winearls revitalized the debate in a series of articles during 2008 and 2009 [16,17,20–22] that serious discussions about the KDOQI 2002 classification started. Some of the major concerns put forward in this debate are highlighted in Table 2. Major problems with the current KDOQI 2002 classification system of CKD CKD, chronic kidney disease; RR, relative risk; eGFR, estimated glomerular filtration rate; MDRD and CKD-EPI, equations for estimating GFR. Major problems with the current KDOQI 2002 classification system of CKD CKD, chronic kidney disease; RR, relative risk; eGFR, estimated glomerular filtration rate; MDRD and CKD-EPI, equations for estimating GFR. eGFR is the backbone, but also the soft spot, of the current CKD classification. The initial Modification of Diet in Renal Disease (MDRD) study equation led to a substantial underestimation of GFR and overinflated CKD prevalence estimates. This has been improved with more recent equations, but the low accuracy of eGFR is still a problem leading to misclassification of individual patients. The prevalence of CKD, especially in the elderly, has, therefore, been claimed to be implausibly high, and age- and sex-specific cut-off values for defining abnormal GFR have been suggested [17,23,24]. However, the history of medicine has often shown that the 2.5th and 97.5th percentiles do not necessarily reflect good health and low risk. For instance, guidelines on hypertension and hypercholesterolaemia have repeatedly lowered their levels for intervention below and beyond these percentiles over the past decades. Furthermore, subjects of very good health, e.g. potential kidney donors, have a significantly lower loss of kidney function per year compared to the general population [25]. Based on data from the population-based HUNT 2 study (n = 65 123) and the new Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation, which has a minimal negative bias even at near-normal GFR levels, we find that the donor-based age-specific fifth percentile equals 60 mL/min/1.73 m2 at age 87 (Figure 1). Therefore, an eGFR < 60 mL/min/1.73 m2 should clearly be considered as abnormal even at high age, supporting at least parts of the current CKD definition. Distribution of baseline eGFR values by age and risk of progression to kidney failure over 10 years in HUNT 2 participants with eGFR below a fixed cut-off at eGFR 60 mL/min/1.73m2, below the age-specific fifth percentile in healthy kidney donors, and below the age-specific fifth percentile in the general population. However, a very low risk of progression to kidney failure among CKD Stage 3 patients has been used as an argument against the current CKD classification. Only 1–2% of such patients will need RRT over a 10-year period [26]. Likewise, the relative risk associated with lower eGFR is substantially reduced in the elderly while the prevalence of CKD increases with age [23]. These findings illustrate the shortcomings of the current CKD classification system and point towards the need for improvements. Focus should be on patient prognosis and outcomes should include cardiovascular disease and mortality in addition to RRT. Albuminuria has long been known as a very strong risk factor for the progression of kidney disease, and especially, the Prevention of Renal and Vascular End-stage Disease (PREVEND) Study Group in the Netherlands has published a large number of excellent articles on this topic. They have, shown that albuminuria is a stronger predictor of a rapid decline in kidney function than baseline eGFR level [27] and that even increased levels in the normoalbuminuric range are associated with an increased risk for an adverse cardiovascular and renal outcome [28], coming to the conclusion that pre-screening for albuminuria could be a useful strategy for CKD detection [29]. Figure 1 illustrates some of the aspects of eGFR and albuminuria. Subjects with eGFR below the fifth percentile are displayed as small dots and those progressing to kidney failure are indicated as large dots (n = 124). If we use eGFR = 60 mL/min/1.73 m2, the age-specific fifth percentile based on healthy subjects eligible for kidney donation and the age-specific fifth percentile based on the general population for defining abnormality, we would diagnose 2677, 7900 and 4007 subjects as CKD patients, respectively. Regardless of the cut-off level used, it is obvious that only a small percentage progress to kidney failure. Likewise, a substantial number of future kidney failure patients had eGFR above these cut-offs at baseline in the HUNT 2 study—but, something that is of major importance—most of them had microalbuminuria (Figure 1, marked as yellow circles) or macroalbuminuria (Figure 1, marked as red circles). Reduced eGFR and increased albuminuria are related variables, but the association between the combined eGFR–albuminuria variable and a hard kidney endpoint had not been described until the last 5 years. Data from three reputable studies consistently showed that lower eGFR and higher urinary albumin excretion were independently associated with kidney failure [30–32]. The relative risk associated with the category of lowest eGFR and highest albuminuria ranged from 30-fold in studies investigating high-risk populations (African American Study of Kidney Disease, Multiple Risk Factor Intervention Trial) [31,32] to 1000-fold in a low-risk population-based study (Okinawa Study) [30]. In April 2008, Ron Gansevoort and Paul de Jong suggested that albuminuria should be used to adequately define CKD Stage 3 [15]. In October 2008, new British guidelines suggested a new CKD classification where CKD Stage 3 was subdivided into Stages 3a and 3b and the presence or absence of macroalbuminuria was indicated at all levels of kidney function [33]. These suggestions were mainly expert opinion-based as (i) there were no data available on how they would perform compared to the KDOQI 2002 system and (ii) at that time-point only one population-based study had been published on the topic (eGFR ± macroalbuminuria in the Okinawa study) [30]. In May 2009, Hallan et al. documented that all levels of eGFR should be complemented by urinary albumin–creatinine ratio (ACR) measurements to get optimal kidney failure risk prediction [34]. This finding was based on the 10-year follow-up of 65 123 subjects from the Norwegian population-based HUNT 2 study using various receiver operating characteristics and efficacy analyses in addition to the traditional multi-adjusted relative risk analysis. A new CKD classification system using four categories of eGFR (≥60, 45–59, 30–44 and 15–30 mL/min/1.73 m2) complemented by three categories of albuminuria (normoalbuminuria, microalbuminuria and macroalbuminuria) was suggested for the description of low, moderate and high risk of progression to kidney failure (Table 3). It has to be pointed out that the data shown in Table 3 are characterized by wide 95% confidence intervals, which to a certain degree weakens the validity of risk categorization performed on the basis of these data. Suggestion for new CKD classification system All eGFR levels are complemented by information on urinary albumin excretion using a matrix of 4 eGFR categories × 3 ACR categories. Hazard ratios (95% CI) for progression to kidney failure (n = 124/65 123) are adjusted for age, sex, blood pressure, antihypertensive medication, diabetes, HDL cholesterol and physical activity. Data are adapted with permission from Hallan et al. [34]. Low, medium and high renal risk categories are based on arbitrary relative risk cut-offs (1–99, 100–999 and 1000+, respectively) as well as on clinical experience for how to handle CKD patients (controlled and treated in general practice; cooperation between general practitioner and nephrologist; or nephrology outpatient clinic). Data on prevalence of the three risk categories and the observed kidney failure incidence (per 100 000 population per year) for the three risk categories are also shown. Suggestion for new CKD classification system All eGFR levels are complemented by information on urinary albumin excretion using a matrix of 4 eGFR categories × 3 ACR categories. Hazard ratios (95% CI) for progression to kidney failure (n = 124/65 123) are adjusted for age, sex, blood pressure, antihypertensive medication, diabetes, HDL cholesterol and physical activity. Data are adapted with permission from Hallan et al. [34]. Low, medium and high renal risk categories are based on arbitrary relative risk cut-offs (1–99, 100–999 and 1000+, respectively) as well as on clinical experience for how to handle CKD patients (controlled and treated in general practice; cooperation between general practitioner and nephrologist; or nephrology outpatient clinic). Data on prevalence of the three risk categories and the observed kidney failure incidence (per 100 000 population per year) for the three risk categories are also shown. In February 2010, the relative risk estimates underlying the new classification system suggested above were confirmed by Hemmelgarn et al. in the much larger Alberta study [35]. Using the same 4 × 3 matrix, they found a very similar pattern of relative risks for kidney failure as in the HUNT 2 study, but it has to be pointed out that it would be statistically incorrect to directly compare the relative risks found in these two studies. In the Alberta study, the potential of improved risk classification based on eGFR–albuminuria even extended to other important outcomes, since robust data for mortality and myocardial infarction risks were also given. Results from a meta-analysis of 50 CKD cohorts (including the HUNT 2 and the Alberta studies) using the same analytical approach as described above are expected to be published soon giving more precise risk estimates [36]. This KDIGO initiative will also provide data on the risk for progressive CKD and acute kidney injury, which are other important clinical outcomes that need to be taken into account. A new CKD classification system will then likely be agreed on. Since (i) there are no generally accepted criteria for the categorization of renal risk and (ii) the HUNT 2 study results had rather wide confidence intervals, the new CKD classification system may slightly differ from the one proposed in Table 3. However, the main principles are clearly shown in our suggestion, and it would anyway be a significant improvement over the current KDOQI 2002 system regarding CKD prevalence distribution, risk stratification and efficacy. In most classification systems, there is a gradual decline in prevalence with increasing severity of stages. However, based on the KDOQI system and the MDRD study equation for estimating GFR, we find the following non-standardized distribution of CKD Stages 1–4 in the HUNT 2 population: 3.4, 3.9, 4.5 and 0.2%. Some improvement is obtained using the new CKD-EPI equation (5.3, 4.1, 3.9 and 0.2%), but a new classification system incorporating albuminuria as suggested in Table 3 gives a much more plausible frequency distribution (12.6, 1.3 and 0.2%). Furthermore, it gives a much better resolution of associated renal risk. The multi-adjusted relative risk of future kidney failure range between 10.4, 36.2, 81.5 and 982 for KDOQI Stages 1–4, while the new system assigns relative risks of 26.8, 239.8 and 2186.4 to low-, medium- and high-risk CKD, respectively. Estimates from the HUNT 2 study indicate that the number of CKD patients assigned to intensive follow-up could be reduced without losing predictive power for incident kidney failure [34]. These findings have been highlighted as potentially important for a much more efficient handling of the large number of CKD patients [36,37]. Whether the new classification system also leads to better risk prediction of other outcomes of interest in CKD (e.g. infections [38], cognitive impairment [39] and fractures [40]) remains to be studied. The KDOQI 2002 classification system of CKD has several weaknesses and will very likely be revised in the near future. In our opinion, the new CKD classification system must be evidence-based, focus on the risk of progression to renal as well as cardiovascular end-points, combine all levels of eGFR with ACR measurements and condense the information into a simple nomenclature to facilitate cooperation with general practitioners and other non-nephrologists. Information on the background, key studies that will form the scientific basis for the revised CKD system and an example of a revised classification system for the definition of low-, medium- and high-risk CKD patients have been presented here. We are convinced that physicians should immediately start to evaluate renal and cardiovascular risk in their patients by adopting the combined eGFR–ACR approach, particularly in the large group of patients with eGFR 30–59 mL/min/1.73 m2. Conflict of interest statement. None declared.
No takes yet. Share an insight, caveat, or question.
Hallan et al. (2010) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: