Raymond T. Krediet; Friedo W. Dekker
Authors and Disclosures
Posted: 11/12/2010; Nat Rev Nephrol. 2010;6(10):563-564. © 2010 Nature Publishing Group
The results of a large observational study seem to confirm an association between early initiation of dialysis and increased mortality that is not fully explained by the presence of comorbidities and other confounding variables. However, this paradoxical association may be an artifact resulting from the inaccuracy of using plasma creatinine levels to estimate glomerular filtration rates in patients with severely impaired renal function.
During the past 8 years, a number of studies have explored whether a patient's estimated creatinine clearance or estimated glomerular filtration rate (eGFR) at the time of initiation of chronic dialysis treatment is associated with survival.[1–3]
In all of these studies, an increased eGFR at the start of dialysis was associated with increased mortality. The researchers tried to explain this unexpected finding by assuming that chronic dialysis was usually initiated at a higher glomerular filtration rate (GFR) in patients who had marked comorbidities as well as renal impairment than in patients who had chronic kidney failure alone, and that patients with chronic kidney failure alone were, furthermore, generally younger and healthier than those with comorbidities.
However, although adjustments for age and comorbidities attenuated the relationship between increased eGFR at dialysis initiation and increased mortality, they could not make it disappear.[1–3] An analysis of the French REIN registry data conducted in 2010 by Lassalle and colleagues shows similar results with regard to both the presence of this association and its persistence despite adjustment for age and comorbidities. Lassalle et al. estimated the GFR of study participants at the start of dialysis using the Modification of Diet in Renal Disease (MDRD) equation. These researchers found that each 5 ml/min/1.73 m2 increment in eGFR at dialysis initiation was associated with a 40% increase in the unadjusted mortality risk over a median follow-up of 21.9 months. Although adjustments for a number of confounding variables, particularly age and comorbidities, substantially decreased the percentage increase in mortality risk associated with increased eGFR (to 8–20%, depending on the parameters used for the adjustment), the association remained statistically significant.
Undoubtedly, starting chronic dialysis treatment in patients who have a GFR of only a few milliliters per minute is associated with uremic complications and a high risk of mortality. Guidelines on when to start dialysis are not evidence-based, however, and in clinical practice other parameters as well as GFR are used to make this determination. These parameters include the severity of hypertension, control of fluid balance, nutritional status, and the presence of disorders of mineral metabolism.
In 2001, Korevaar et al. reported the results of a prospective cohort study in which participants' GFRs at the start of dialysis were not estimated but were measured as the mean of creatinine and urea clearance obtained from 24 h urine collections. After adjustment for age, sex, comorbidities and primary kidney disease, the researchers observed a mortality hazard ratio of 1.22 (P = 0.01) associated with each 1 ml/min decrement in GFR at dialysis initiation. This association resulted in a mean difference of 2.5 months in the estimated duration of survival after 3 years on dialysis in favor of prompt dialysis initiation compared with delayed initiation. However, the researchers stated that the improved survival found to be associated with prompt dialysis initiation might have been a result of lead-time bias rather than an improvement in the course of the disease; that is, that prompt starters had a shorter history of renal impairment than delayed starters.
Collecting urine over a 24 h period is difficult for many patients, especially women. Many formulas have, therefore, been developed to estimate creatinine clearance or GFR. These equations are based on plasma creatinine level and demographic parameters. The equations developed by Cockroft and Gault to estimate creatinine clearance and by Levey et al. to estimate GFR in the MDRD study are the most widely used.
Although both equations have been validated over a wide range of GFRs, they have never been validated in patients with very low (5–20 ml/min) GFRs. However, very little attention has been paid by the nephrology research community to the lack of validation of formulas used to estimate GFR.
Both the Cockcroft–Gault equation and the MDRD equation are vulgar fractions that have plasma creatinine level in their denominator. Thus, a low plasma creatinine level will lead to a high eGFR. The plasma concentration of creatinine is determined by the rate of release of creatinine from muscles and the rate of its removal by glomerular filtration and proximal tubular excretion.
In patients with severe renal failure, in whom the rate at which creatinine is removed is substantially decreased, the effect of muscle mass on plasma creatinine level is likely to be of increased importance. Patients with a low muscle mass, such as elderly patients with severe comorbidities and malnutrition, will release little creatinine from their muscles, which means that their plasma creatinine level will be lower (and consequently their eGFR will be higher) than that of well-nourished patients who have the same magnitude of renal impairment.
This phenomenon was first recognized by Beddhu et al. in 2003, but their findings were largely ignored by the nephrology research community. Beddhu et al. found that the logarithm of plasma creatinine levels correlated positively with creatinine production, and that this association was independent of age, sex, ethnicity and plasma urea levels.
The final proof of the unreliability of eGFR estimated using serum creatinine level in individuals with very low (or indeed, very high) muscle mass would require a study investigating the relationship between eGFR and muscle mass.
The evidence so far indicates that neither plasma creatinine nor its reciprocal should be used to estimate GFR in patients at the start of dialysis, or to determine the level of kidney function at which dialysis should be started.
Instead, GFR should be determined by the measurement of creatinine and urea clearance rates in patients with severe renal failure, as was done by Korevaar and colleagues. Alternatively, serum levels of another analyte that is removed from the body by glomerular filtration but is not dependent on muscle mass could be investigated. Serum cystatin C level is one such parameter.
Our research group showed that use of the serum concentration of this low-molecular-weight protein in patients on dialysis resulted in estimations of residual GFR that were more accurate and more precise than those calculated using the MDRD formula. No investigations have been published, however, on whether a patient's serum cystatin C level at the start of dialysis is related to their duration of survival.
We conclude that estimations of GFR based on plasma creatinine level and demographic parameters should not be used either to determine the start of dialysis treatment or to predict outcomes in patients on dialysis.