Content uploaded by Sergio Luis-Lima
Author content
All content in this area was uploaded by Sergio Luis-Lima on May 12, 2021
Content may be subject to copyright.
A reliable determination of renal function
is fundamental in many clinical situations,
including the evaluation of patients with
renal diseases, staging of chronic kidney
disease (CKD), predicting the risk of disease
progression, monitoring renal function
over time, determining the need for dialysis
therapy, screening living kidney donors and
enabling dose adjustment of nephrotoxic
agents (drugs or contrast media) in patients
with impaired renal function. In addition,
the accurate and precise evaluation
of glomerular filtration rate (GFR) is
particularly important in clinical studies
that include renal function as a primary
outcome measure.
Serum creatinine level is the most widely
used marker of renal function in routine
clinical practice1. Creatinine clearance over
a 24 h period (24 h CrCl) is considered a
surrogate marker of GFR. However,
measurement of 24 h CrCl is burdensome
and frequently compromised by incomplete
data on urinary volume. To overcome
these limitations, more than 70 equations
have been developed to estimate GFR
continued to be published, notably the Mayo
Clinic Quadratic (MCQ) equation in 2004
(REF.3), the CKD Epidemiology Collaboration
formula (CKD- EPI) in 2009 (REF.4), and
an eGFR equation based on normalized
values of serum creatinine in 2016 (REF.14).
In 1985, cystatin C was first proposed as a
marker of renal function15. Since then,
more than 15 eGFR formulae based on
cystatin C levels have been published
(Supplementary Table 1).
In addition, several eGFR formulae have
combined creatinine and cystatin C16–20,
used multiple equations21,22, or included
biomarkers (such as β- trace protein or
β2-microglobulin23) to increase the accuracy
and precision of GFR estimation. The most
recent equations available were published
in 2018: the full age spectrum (FAS)
formulae are based on either creatinine or
cystatin C levels, or both21. Finally, some
eGFR equations were developed in specific
populations, such as renal transplant
recipients24, patients with type 2 diabetes
mellitus (T2DM)25,26, and specific ethnic
groups, including African American18,27,
Chinese20,28–33, Japanese34 or Thai35 patients.
Most modern formulae (algorithms)
used to calculate eGFR (such as the MCQ
equation) are based on serum creatinine
and/or cystatin C levels, age, weight and sex.
Typically, these formulae were developed
in large populations covering the whole
spectrum of renal function3,4,14,17,18,36 and,
therefore, are considered more reliable
than early equations such as the MDRD,
which was derived from studies of patients
with CKD. However, the reliability of these
equations remains a matter of debate.
Several studies performed in different
clinical settings have criticized the
performance of eGFR in reflecting real renal
function37–44. Moreover, without detracting
from the valuable work dedicated to
developing these equations, the observation
that new eGFR formulae continue to be
published implies that a definitive equation
has not yet been developed. In fact, the
performance of modern eGFR formulae
(when assessed using appropriate statistical
methods) has not improved substantially on
that of the first equation described in 1957
— even with the use of complex algorithms
and cystatin C45,46, as we discuss in further
detail herein.
(Supplementary Table 1). These equations
have increased in mathematical complexity
over time, from simple ratios between a
constant and creatinine level2, to expressions
involving exponentials or logarithms
combined with conditional parameters
and corrective factors for sex, ethnicity or
renal function3–6.
The evolution of GFR estimation
methods can be considered from an
historical perspective. The first formula
based on serum levels of creatinine used
to calculate estimated GFR (eGFR) was
published in 1957 and was mainly intended
to facilitate dose adjustments of medications
cleared by the kidneys, such as digitalis and
antineoplastic drugs7. Other equations based
on serum creatinine levels followed8–11, and
in 1976 Cockcroft and Gault published an
eGFR formula that became very popular
among physicians12. The eGFR formula
developed in the Modification of Diet in
Renal Disease (MDRD) study, published
in 2000, became the reference equation
for several years13. However, additional
creatinine- based eGFR formulae have
OPINION
Estimated GFR: time for a critical
appraisal
EstebanPorrini, PieroRuggenenti, SergioLuis- Lima, FabiolaCarrara,
AlejandroJiménez, AikodeVries, ArmandoTorres, FlavioGaspari
and GiuseppeRemuzzi
Abstract | Since 1957 , over 70 equations based on creatinine and/or cystatin C
levels have been developed to estimate glomerular filtration rate (GFR). However,
whether these equations accurately reflect renal function is debated. In this
Perspectives article, we discuss >70 studies that compared estimated GFR (eGFR)
with measured GFR (mGFR), involving ~40,000 renal transplant recipients and
patients with chronic kidney disease (CKD), type 2 diabetes mellitus or polycystic
kidney disease. Their results show that eGFR often differed from mGFR by ±30% or
more, that eGFR values incorrectly staged CKD in 30–60% of patients, and that
eGFR and mGFR gave different rates of GFR decline. Errors were unpredictable,
and comparable for equations based on creatinine and/or cystatin C. We argue,
therefore, that the persistence of these errors (despite intensive research) suggests
that the problem lies with using creatinine and/or cystatin C as markers of renal
function, rather than with the mathematical methods used for GFR estimation.
PERSPECTIvES
Nature reviews
|
Nephrology
In this Perspectives article, we critically
appraise the results of over 70 published
studies that compared the performance of
any eGFR formulae against any measured
GFR (mGFR) reference method in renal
transplant recipients and patients with
CKD, T2DM or polycystic kidney disease.
The results of studies that assessed a large
number of patients or eGFR formulae,
used appropriate statistics, or evaluated
the agreement between eGFR decline and
mGFR decline, are included in TABLES1–4.
Those that did not fulfil these criteria are
included in Supplementary Tables 2–4. We
argue that the persistence of substantial
errors in all GFR estimation formulae
developed to date suggests that the problem
lies with serum creatinine and cystatin C,
which are poor markers of real renal
function, rather than with the mathematical
approach used to develop the formulae.
Finally, we discuss the implications of eGFR
errors for clinical practice and research.
Markers for estimating GFR
An ideal marker of GFR must not be bound
to proteins and must be inert, freely filtered
by glomeruli, constantly produced, and
eliminated in the urine but not reabsorbed,
secreted or metabolized by renal cells1. The
rate of plasma clearance from the circulation
must equal that of urinary excretion1, and
the relationship between its concentrations
in serum with mGFR must be reciprocal
(Supplementary Fig. 1). Unfortunately,
neither creatinine nor cystatin C meets
these conditions47,48.
Creatinine
Serum creatinine is not bound to proteins
and is freely filtered by the glomeruli1.
However, synthesis of creatinine from
creatine is not constant, since it is affected
by the daily intake of protein and by muscle
turnover. Meat is the major dietary source
of creatine; meat intake influences the
total creatine pool and, therefore, serum
creatinine levels49. The major endogenous
source of creatinine is muscle, and changes
in muscle mass also influence the production
of creatinine. Dietary protein deficiency
leads to loss of muscle mass, which reduces
levels of creatinine50,51. In other conditions
linked to reduced muscle mass (such as
anorexia, cirrhosis or sarcopenia), reduced
production of creatinine leads to diminished
serum creatinine levels and overestimation
of GFR. By contrast, conditions associated
with increased muscle mass or high dietary
protein intakes, such as vigorous exercise52,53,
chronic glucocorticoid therapy54 and
hyperthyroidism55, can increase the levels of
creatinine, leading to underestimation
of GFR.
Furthermore, secretion and reabsorption
of creatinine by renal tubular cells and
extrarenal clearance also influence serum
creatinine levels. Under normal conditions,
tubular secretion accounts for about 10%
of total urinary creatinine excretion, but
this percentage increases in parallel with
decreases in GFR1,47, reaching 80–100% in
patients with advanced CKD. Thus, during
the course of CKD, increased tubular
secretion of creatinine limits the increase
in serum creatinine level, which masks the
reduction in GFR. Data on the effect of
tubular reabsorption of creatinine on serum
creatinine levels are limited56–58. In addition,
extrarenal clearance of creatinine (which
includes recycling of creatinine to creatine
and degradation of creatine to products
other than creatinine) increases with the
decrease in renal function, which further
contributes to the overestimation of GFR
in patients with advanced renal disease59,60.
Similar limitations also apply to 24 h CrCl.
These factors explain why the
relationship between creatinine and mGFR
is not linear but curvilinear and why a given
value of serum creatinine can be associated
with a wide range of mGFR values (FIG.1).
For example, a serum creatinine level of
1.5 mg/dl can be found in patients with
GFRs between 30 ml/min and 90 ml/
min (FIG.1), illustrating the limitations of
serum creatinine levels for estimating GFR.
Interestingly, this curvilinear relationship
between serum creatinine and GFR
(and the resulting lack of accuracy and
precision of eGFR) has been known
since 1985 (REF.47) (FIG.2).
Errors in the assays used to measure
creatinine levels might also influence the
correlation between creatinine and GFR.
The Jaffé reaction (a colorimetric method
of measuring creatinine concentration in
blood and serum) can also be induced by
organic compounds in plasma other than
creatinine, which can lead to overestimation
of creatinine levels61–63. Modern enzymatic
assays, in which creatinine levels are
standardized by calibration traceable
to isotopic dilution mass spectrometry
(IDMS) reference measurement procedures,
overcome this issue and have been proposed
as a reference method64. In addition, the
low coefficient of variation of creatinine
measurement could influence the estimation
of GFR65. However, in our opinion, the
choice of method for serum creatinine
measurement does not appreciably affect
the relationship between the two variables,
as the same curvilinear correlation between
creatinine level and GFR can be observed
irrespective of whether creatinine is
measured with the IDMS (FIG.1) or Jaffé
(FIG.2) methods.
Cystatin C
Cystatin C is an inhibitor of various cysteine
proteinases that is encoded by CST3, a
regulatory gene expressed in nucleated
cells66. Cystatin C has a constant rate of
production and is freely filtered across
glomeruli66,67. However, it is completely
reabsorbed and metabolized by renal tubular
cells, which prevents its use in calculations
of 24 h urinary clearance66,67. High serum
levels of cystatin C can be observed in
individuals with morbid obesity, T2DM,
hypertension and metabolic syndrome68.
Thus, changes in cystatin C levels might
reflect the patient’s clinical status, in
addition to their renal function. A possible
explanation for the association between high
cystatin C levels and conditions related to
obesity is that cystatin C is the endogenous
inhibitor of cathepsins, which have a role
in adipogenesis; thus, cystatin C might be
secreted by adipocytes in a homeostatic
response to the expansion of adipose tissue68.
As with creatinine and GFR, cystatin C
has a curvilinear relationship with GFR
such that a given value of cystatin C reflects
a range of GFR values (FIG.1). For example,
cystatin C levels of 1.5 mg/l are associated
with GFR values of 30–90 ml/min (FIG.1).
Finally, the low coefficient of variation of
cystatin C could influence the estimation
of GFR65. To sum up, we consider that the
extremely wide margins of error for both
creatinine- based and cystatin C- based
estimation of GFR make it impossible to
prove that these values can be corrected by
mathematical modelling.
Agreement between eGFR and mGFR
Agreement between mGFR and eGFR values
has been evaluated by various statistical
methods, including correlation coefficients,
mean and mean percent errors, Student’s
t- test, linear regression, quadratic mean,
and the percentage of estimations falling
within defined margins of error (TABLES1–4;
Supplementary Tables 2–4). All these
approaches have methodological limitations.
For example, Pearson’s correlation
coefficient does not take into account any
differences in values between the groups
being compared. Thus, correlation can be
good (r > 0.9), even when eGFR values differ
from mGFR values by twofold or more
(Supplementary Fig. 2). The mean and mean
percentage errors represent the mean of the
difference between eGFR and mGFR for
www.nature.com/nrneph
PersPectives
Nature reviews
|
Nephrology
PersPectives
Table 1 | Studies that evaluated estimated GFR in patients with CKD
Study Population Gold standard Formulae Statistics and error Comments Refs
Cross- sectional studies
Selistre
(2016)
9,430 white adults
(including 1,550 renal
transplant recipients)
Inulin Cr: CKD- EPI,
Schwartz
• P30 overall: CKD- EPI 80,
Schwartz 78
• P30 CKD stage 1: CKD- EPI
95, Schwartz 86
• P30 CKD stage 2: CKD- EPI
82, Schwartz 86
• P30 CKD stages 3–5: CKD-
EPI 69, Schwartz 67
• P10 overall: CKD- EPI 38,
Schwartz 33
In 20% of patients, eGFR
error was more than ±30%
of mGFR
81
Fan (2014) 1,119 white patients
(including 594 with
diabetes mellitus)
NS Cr: CKD- EPI P30: Overall, 87; CKD stage
1–2, 91; CKD stage 3, 87; CKD
stage 4–5, 75
• In 10–15% of patients,
eGFR error was more
than ±30% of mGFR
• Similar error in all CKD
stages
82
Cy: CKD- EPI P30: Overall, 79; CKD stage
1–2, 93; CKD stage 3, 82; CKD
stage 4–5, 72
Cr+Cy: CKD- EPI P30: Overall, 91; CKD stage
1–2, 96; CKD stage 3, 91; CKD
stage 4–5, 81
Evans (2013) 2,198 white patients
with CKD stage 4 and
5 (including 729 with
diabetes mellitus)
Iohexol Cr: MDRD, CG,
CKD- EPI, Rule,
Lund- Malmö
• P30: MDRD 65, CG 54, CKD-
EPI 67 , Rule 67 , Lund- Malmö
76
• L A: MDRD −8 to 14, CG −6
to 18, CKD- EPI −9 to 14,
Rule −9 to 16; Lund- Malmö
−7 to 11
In 35–45% of patients,
eGFR error was more than
±30% of mGFR
83
Inker (2016) 3,551 black and white
patients (including 273
with diabetes mellitus)
Iothalamate Cr: CKD- EPI P30: CKD- EPI 84 • In 10–25% of patients,
eGFR error was more
than ±30% of mGFR
• BTP and B2M
comparable to old
formulae
23
Cy: CKD- EPI P30: CKD- EPI 83
Cr+Cy: CKD- EPI,
BTP, B2M, BTP- B2M
P30: CKD- EPI 89, BTP 76,
B2M 82, BTP- B2M 85, CKD-
EPI+BTP- B2M 90
van Deventer
(2011)
100 black South
African patients
51Cr- EDTA Cr: MDRD, CKD-
EPI, van Deventer
P30: MDRD 74, CKD- EPI 72,
van Deventer 70
In 15–30% of patients,
eGFR error was more than
±30% of mGFR
84
Cy: van Deventer P30: van Deventer 84
Stevens
(2007)
1,737 black patients Iothalamate Cr: MDRD (with
corrective factor)
P30: 82 if mGFR <60 ml/min/
1.73 m2; 88 if mGFR
>60 ml/min/1.73 m2
• In 10–20% of patients,
eGFR error was more
than ±30% of mGFR
• Similar P30 in the white
population (not shown)
85
Feng (2013) 788 Chinese patients 99mTc- DTPA Cr: CG, MDRD • P30: CG 55, MDRD 44
• L A: CG −52 to 47 , MDRD
−60 to 52
• In 25–55% of patients,
eGFR error was more
than ±30% of mGFR
• Wide L A for all formulae
• In 30–50% of patients,
eGFR led to an error in
the classification of CKD
20
Cy: CKD- EPI, CCG,
Feng
• P30: CKD- EPI 63, CCG 65,
Feng 72
• L A: CKD- EPI −25 to 36, CCG
−34 to 32, Feng −28 to 31
CG, CCG, Feng • P30: CKD- EPI 63, CCG 59,
Feng 74
• L A: CKD- EPI −33 to 39, CCG
−43 to 38, Feng −29 to 28
Murata
(2011)
2,324 white patients Iothalamate Cr: MDRD,
CKD- EPI
P30: MDRD 75, CKD- EPI 75 • In 25% of patients, eGFR
error was more than
±30% of mGFR
• In 50–60% of patients,
eGFR led to an error in
the classification of CKD
86
Froissart
(2005)
2,095 white patients 51Cr- EDTA Cr: MDRD, CG • P30 overall: MDRD 87 , CG
79
• P30 if mGFR >60 ml/min:
MDRD 92, CG 88
• P30 if mGFR <60 ml/min:
MDRD 83, CG 69
• In 15–20% of patients,
eGFR error was more
than ±30% of mGFR
• In 30% of patients, eGFR
led to an error in the
classification of CKD
87
each patient. For example, in a hypothetical
group of five patients, eGFR and mGFR
values might differ by −20, −15, 20, 15 and
−5 ml/min. The mean error is −1 because
numerically identical values of opposite
signs (such as −20 and 20) cancel each other
out, blunting the deviation of eGFR from
mGFR (Supplementary Fig. 3). In the paired
t- test, only mean values are compared, and
the differences between individual data
points are, therefore, largely underestimated.
To assess the acceptability of a formula
that estimates GFR, it is necessary to define
apriori clinically meaningful margins
of error, to determine the proportion of
estimations that fall within these margins,
and to define how much disagreement
is tolerable. The parameter used most
frequently to assess the performance of GFR
estimation equations is P30, which is defined
as the percentage of GFR estimations falling
within ±30% of mGFR values (TABLES1–4;
Supplementary Tables2–4). Although
other parameters (usually P10 and P20,
representing the percentages of estimations
falling within ±10% and ±20% of mGFR,
respectively) have also been used, P30
deserves special consideration. Notably, P30
was defined in the absence of any clinical
or statistical rationale for considering it an
acceptable margin of error.
However, in our opinion, the limits of
±30% are extremely wide and not acceptable
for evaluating GFR estimation formulae. For
instance, in a patient with mGFR 60 ml/min/
1.73 m2, a P30 margin of error means
that eGFR values of 42–78 ml/min/
1.73 m2 (FIG.3a) are considered acceptable.
Furthermore, for most eGFR equations, the
proportion of measurements falling within
the nominally acceptable P30 limit varies
from 50% to 90% (TABLES1–4).Thus, in
10–50% of measurements, the error in eGFR
values is >30%. A more rigorous evaluation
of eGFR than the P30 is needed.
Two issues must be considered to
establish new limits of acceptability for
eGFR: first, the reproducibility of the
reference method used to measure GFR,
which is influenced by both the variation in
measurements made by the same instrument
under similar conditions (technical
and clinical) within a short period of
time, and the biological variability of GFR;
and second, the clinical relevance of the
proposed limits in the evaluation of GFR
values and GFR decline. The reproducibility
of mGFR is 5.0–8.0% for iohexol69,70, 4.0%
for EDTA71,72 and 8.5% for iothalamate73.
Moreover, the variability of an external
quality assessment scheme for iohexol
(used by clinical laboratories to calibrate
their GFR reference measurements) is
about 8%74. Thus, the variability of all
GFR measurement methods is 4–8%. This
range includes both the intrinsic error of
the method and the biological variability
of GFR. To ensure that eGFR formulas are
reliable enough to replace mGFR, estimated
values must be comparable to measured
values. Of course, error is present in all
measurement techniques, but intrinsic
errors must be minimized. Accordingly, we
suggest that an acceptable limit of agreement
between eGFR and mGFR should not
exceed ±10% of mGFR and, importantly,
that 90% of estimations should fall within
this margin of error (FIG.3b). Acceptance of
www.nature.com/nrneph
PersPectives
Study Population Gold standard Formulae Statistics and error Comments Refs
Hojs (2011) 764 white patients 51Cr- EDTA • Cr: MDRD,
CKD- EPI, CG
• Cy: Hojs
• P30 CKD stage 3: MDRD 78,
CG 59, CKD- EPI 75, Hojs 78
• P30 CKD stage 4: MDRD 53,
CG 44, CKD- EPI 50, Hojs 70
• P30 CKD stage 5: MDRD 50,
CG 51, CKD- EPI 44, Hojs 53
In 30–50% of patients,
eGFR error was more than
±30% of mGFR
88
Longitudinal studies
Wang (2006) 1,094 black patients 125I- iothalamate Cr: AASK Poor concordance (r = 0.60) Poor agreement between
eGFR and mGFR
89
Xie (2008) 542 white patients 125I- iothalamate Cr: MDRD Poor correlation between
eGFR and mGFR decline
(r = 0.60, r2 = 0.42). Between-
patient difference between
mGFR slope and eGFR slope:
58% from −2 to +2 ml/min per
year ; 20% from −2 to −4 ml/
min per year or larger
eGFR decline slower than
mGFR decline
90
Tent (2012) 65 white patients 125I- iothalamate Cr: CKD- EPI,
MDRD
• Reduced r2 between mGFR
decline and eGFR decline:
MDRD: 0.45, CKD- EPI: 0.52
• L A: from +3 to −3
• Median 9 mGFRs per
patient (range 9–16)
• Poor agreement between
eGFR and mGFR
104
Padala (2012) 3,531 white American
and African- American
patients
125I- iothalamate Cr: CKD- EPI
(pooled data from
four studies),
AASK , DCCT, CSG,
MDRD
Mean eGFR decline 15%
faster in the AASK study and
18% slower in the CSG study.
Comparable results for the
DCCT study
• All patients had 3–15
mGFRs
• Poor agreement between
eGFR decline and mGFR
decline
91
Lee (2009) 146 white patients with
stage 3–5 CKD
51Cr- EDTA
99mTc- DTPA
Cr: MDRD, CG,
MDRD6
Bias between formulae and
mGFR ranged from +30 to
−30 ml/min at baseline,
12 and 24 months
Poor agreement between
eGFR decline and mGFR
decline
92
AASK , The African American Study of Kidney Disease and Hypertension; BTP, β- trace protein; B2M, β2-microglobulin; CCC, concordance correlation coefficient;
CCG, Chinese Collaborative Group formula; CG, Cockcroft–Gault; CKD, chronic kidney disease; CKD- EPI, CKD Epidemiology Collaboration; Cr, creatinine; CrCl,
creatinine clearance; Cy , cystatin C; DTPA , pentetic acid; eGFR , estimated GFR; GFR , glomerular filtration rate; L A , limits of agreement (values in ml/min/1.73 m2
for cross- sectional studies and in ml/min/1.73 m2 per year for longitudinal studies); mGFR , measured GFR; MDRD, Modification of Diet in Renal Disease; NS, not
specified; P, percentage of estimations included within an error of either ±30% (P30) or ±10% (P10) of measured GFR; TDI, total deviation index.
Table 1 (cont.) | Studies that evaluated estimated GFR in patients with CKD
Nature reviews
|
Nephrology
PersPectives
Table 2 | Studies that evaluated estimated GFR in patients with type 2 diabetes mellitus
Study Population Gold
standard
Formulae Statistics and error Comments Refs
Cross- sectional studies
Iliadis (2011) 448 white
patients
51Cr- EDTA • Cr: MDRD, CKD- EPI
• Cy: Rule, Perkins,
Tan, Arnal, Stevens,
Grubb, MacIsaac,
Tidman, Flodin
• Cr+Cy: Stevens
P30: MDRD 79, CKD- EPI 81,
Rule 53, Perkins 21, Arnal 45,
Stevens (Cy) 54, MacIsaac 46,
Tan 79, Grubb 69, Tidman 40,
Flodin 43, Stevens (Cr+Cy) 70
• In 20–80% of patients,
eGFR error was more than
±30% of mGFR
• Similar bias between
Cr- based and Cy- based
formulae
46
Maple- Brown (2014) 224 indigenous
Australian
patients
Iohexol Cr: MDRD, CKD- EPI, CG P30: MDRD 80, CKD- EPI 87 ,
CG 71
In 15–20% of patients, eGFR
error was more than ±30%
of mGFR
108
Rigalleau (2006) 200 white
patients
51Cr- EDTA Cr: MDRD, CG, MCG L A: Bias NS In 30–50% of patients,
eGFR led to an error in the
classification of CKD
109
MacIsaac (2015) 199 white
patients
99mTc- DTPA Cr: MDRD, CKD- EPI • P30: MDRD 86, CKD- EPI 90
• L A: MDRD −34 to 31,
CKD- EPI −30 to 27
• In 10–15% of patients,
eGFR error was more than
±30% of mGFR
• Wide L A
• In 10–30% of patients,
eGFR led to an error in the
classification of CKD
37
Inker (2012) 1,726 patientsa
from a
multi- ethnic
cohort
NS • Cr: CKD- EPI
• Cy: CKD- EPI
• Cr- Cy: CKD- EPI
P30: ~90 for all three
equations
In 10% of patients, eGFR
error was more than ±30%
of mGFR
18
Li (2010) 91 Chinese
patients
99mTc- DTPA • Cr: MDRD
• Cy: Stevens, Ma, Rule,
MacIsaac, Perkins
P30: MDRD 69, Stevens 70,
Ma 61, Rule 47 , MacIsaac 55,
Perkins 44
• In 60–80% of patients,
eGFR error was more than
±30% of mGFR
• Similar bias between
Cr- based and Cy- based
formulae
102
Longitudinal studies
Gaspari (2013) 600 white
patientsa
Iohexol Cr: MDRD, CG, CKD-
EPI, Rule, Effersøe, Hull,
Mawer, Gates, Ibrahim,
Bjornsson, Davis-
Chandler, Jelliffe-1,
Jelliffe-2, Walser,
Edwards- Whyte
GFR at baseline:
• TDI: from Rule 32 to
Jelliffe-2 93; CG 51, CKD- EPI
41; MDRD 52; Effersøe 55
• CCC: from Jelliffe-2 0.21
to Rule 0.52; CG 0.43,
CKD- EPI 0.43, MDRD 0.38,
Effersøe 0.31
GFR decline:
• CCC: from Effersøe −0.21
to Hull 0.36; CKD- EPI 0.28,
MDRD 0.32, CG 0.35
• 449 patients had multiple
mGFRs
• Poor agreement between
eGFR and mGFR
• Hyperfiltration rarely
detected
• Effersøe similar to CG and
CKD- EPI
• eGFR decline slower than
mGFR decline
45
Rossing (2006) 383 white
patients
51Cr- EDTA Cr: MDRD, CG GFR at baseline:
• L A µAlb: CG −58 to 31,
MDRD −66 to 20
• L A proteinuria: CG −39 to
33; MDRD −47 to 25
GFR decline
(mean ±SD ml/min per year):
• µAlb mGFR: −4.1 ± 4.2, CG
−3.4 ± 3.2, MDRD −2.9 ± 2.8
• Proteinuria mGFR: −5.2
± 4.1, CG −4.6 ± 4.1,
MDRD −4.2 ± 3.8
• Wide L A for eGFR and
eGFR decline
• eGFR decline slower than
mGFR decline
38
Wood (2016) 152 white
patients
99mTc- DTPA Cr: CKD- EPI GFR decline
(ml/min per year): mGFR −2.6;
CKD- EPI −1.6
eGFR decline slower than
mGFR decline
110
µAlb, microalbuminuria; CCC, concordance correlation coefficient; CG, Cockcroft–Gault; CKD, chronic kidney disease; CKD- EPI, CKD Epidemiology
Collaboration; Cr, creatinine; CrCl, creatinine clearance; Cy , cystatin C; DTPA , pentetic acid; eGFR , estimated GFR; GFR , glomerular filtration rate; L A , limits of
agreement (values in ml/min/1.73 m2 for cross- sectional studies and in ml/min/1.73 m2 per year for longitudinal studies); mGFR , measured GFR; MCG, modified
Cockcroft–Gault; MDRD, Modification of Diet in Renal Disease; NS, not specified; P, percentage of estimations included within an error of either ±30% (P30) or
±10% (P10) of measured GFR; TDI, total deviation index. aOnly the patients with diabetes mellitus were considered.
www.nature.com/nrneph
PersPectives
Table 3 | Studies that evaluated the performance of estimated GFR in patients after renal transplantation
Study Population Gold standard Formulae Statistics and error Comments Refs
Cross- sectional studies
Luis- Lima
(2015)
193 white
patients
Iohexol Cr: 34 formulae • CCC: Rowe 0.04, Rule 0.66, CG
0.70, Effersøe 0.77 , CKD- EPI 0.79,
MDRD 0.82, Matsuo 0.84
• TDI: Matsuo >49%, MDRD 53%,
Effersøe and CKD- EPI 61%, CG
81%, Rule 101%
• For all formulae, eGFR showed
poor agreement with mGFR
• In 30–60% of patients, eGFR led
to an error in the classification of
CKD stage
79
Cy: 14 formulae • CCC: Perkins 0.67; Rule, Tan,
Hoek 0.90
• TDI: Hoek 34%, Perkins 83%
Cr+Cy:
3 formulae
• CCC: Ma 0.85, Stevens 0.90,
CKD- EPI 0.87
• TDI: Ma 49%, Stevens 37%,
CKD- EPI 47%
Masson
(2013)
825 white
patients
• Inulin
• 51Cr- EDTA
Cr: MDRD,
CKD- EPI
• P30: MDRD 80, CKD- EPI 74
• Similar errors within each CKD
stage
• L A: MDRD −23 to 26, CKD- EPI
−21 to 32
• In 20–25% of patients, eGFR error
was more than ±30% of mGFR
• Wide L A for all formulae
• In 30% of patients, eGFR led to
an error in the classification of
CKD
115
Masson
(2013)
670 white
patients
Inulin Cr: CKD- EPI P30: 75 In 15–25% of patients, eGFR error
was ±30% of mGFR
116
Cy: CKD- EPI P30: 81
Cr+Cy: CKD- EPI P30: 86
Mariat
(2004)
294 white
patients
Inulin Cr: MDRD, CG,
Walser, Jelliffe,
Nankivell, 24 h
CrCl
• P10: MDRD 30, CG 24, Walser 28,
Jelliffe 32, Nankivell 23, 24 h CrCl 26
• L A: CG −35 to 27 , Walser −23 to
33, MDRD −29 to 28, Jelliffe −26
to 29, Nankivell −37 to 21, 24 h
CrCl −39 to 26
• In 70% of patients, eGFR error
was more than ±30% of mGFR
• Wide L A for all formulae
40
Mariat
(2005)
284 white
patients
Inulin Cr: CG, aMDRD,
MDRD-7
P30: CG 59, aMDRD 73, MDRD-7 75 • In 25–40% of patients, eGFR
error was more than ±30% of
mGFR
• Wide L A for all formulae
• In 30–40% of patients, eGFR led
to an error in the classification
of CKD
41
Longitudinal studies
Bosma
(2005)
798 white
patients
125I- iothalamate Cr: MDRD,
Jelliffe-1, Jelliffe-2,
Gates, Nankivell,
CG, Hull, Mawer,
Rule, 24 h CrCl
• P30 at baseline: MDRD 88,
Jelliffe-1 87 , Jelliffe-2 88, Gates
87 , Hull 74, Nankivell 76, CG 76;
Mawer 73, Rule 72, 24 h CrCl 66;
P10: <40% for all equations
• mGFR decline: −1.9 ± 15
• eGFR decline: from 0.5 ± 12.0
(Jelliffe-1) to −2.3 ± 12.0 (Hull);
P30: <24% for all equations
• 478 patients had 3 mGFRs in
5 years
• In 10–30% of patients, eGFR
error was more than ±30% of
mGFR
• Poor agreement between eGFR
and mGFR decline
42
Gaspari
(2004)
81 white
patients
Iohexol Cr: CG, Nankivell,
Bjornsson,
Edwards- White,
Davis- Chandler,
Hull, Jelliffe-1,
Jelliffe-2, Mawer,
Walser, MDRD
• P10 at baseline: from
Davis-Chandler 20 to Walser 46,
CG 31, MDRD 44
• mGFR decline: −3.0 ml/min/
1.73 m2 per year
• eGFR decline: from Walser −5.0
to Davis-Chandler −7.4 ml/
min/1.73 m2 per year
• All patients had 3 mGFRs
• In 20–55% of patients, eGFR
error was more than ±10% of
mGFR
• eGFR decline was faster than
mGFR decline
43
Gera (2007) 684 white
patients
125I- iothalamate Cr: MDRD, CG,
Rule
• mGFR decline: −3.9 ± 13
• eGFR decline:. MDRD 2.2 ± 10,
CG: −1.1 ± 11, Rule −1.5 ± 12
(% per year)
• Correlation (r2) between MDRD
and mGFR 0.36; between Rule
and mGFR 0.49
• 360 patients had >3 mGFRs in
3 years
• eGFR decline was slower than
mGFR decline
• Poor correlation between eGFR
and mGFR
117
Fauvel
(2013)
631 white
patients
Inulin Cr: MDRD, CKD-
EPI, CG, 1/Cr
L A (% of variation in GFR decline):
from −40% to 40% for all formulae
• All patients had 2–7 mGFRs
• Extreme variability between
eGFR and mGFR decline
118
wider limits of agreement (>10%) would
both increase the variability between eGFR
and mGFR and reduce the reliability of
GFR estimations.
Another important point to consider is
the large effect of errors in eGFR values
on the calculation of GFR decline. This factor
explains why eGFR declines can be either
faster or slower than mGFR declines in
published studies, a topic that is discussed
further in the following sections. So, eGFR
values must be comparable to mGFR values,
to avoid errors in the evaluation of renal
function and in its change over time.
Furthermore, appropriate statistics must
be used to assess the agreement between
mGFR and eGFR. Such statistics must
simultaneously consider precision and
accuracy, and must provide confidence
intervals75–77. Appropriate parameters
include the limits of agreement78, total
deviation index (TDI), concordance
correlation coefficient (CCC) and coverage
probability75–77. The limit of agreement is a
range that encompasses most differences
between two measurements (the reference
interval is defined as mean ±1.96 × 1 SD).
The narrower these limits are, the better the
agreement between the measurements.
CCC combines elements of accuracy
and precision; a score >0.90 (range 0–1)
reflects optimal concordance between
measurements. TDI is a measure that
captures a large proportion of data within
a boundary representing the allowable
difference between two measurements75.
For example, a TDI of 60% represents poor
agreement (because it means that 90% of
eGFR values fall within ±60% of mGFR,
which is a very large margin of error)75.
The ideal situation would be a TDI of <10%,
meaning that 90% of eGFR values fall within
±10% of mGFR, a much smaller margin
of error (FIG.3b). Coverage probability
values range from 0 to 1; this statistic
estimates whether a given TDI is less than a
prespecified percentage75. All these statistics
provide confidence intervals that enable
generalization of the results. However, these
tests have rarely been used in the study of
eGFR45,79,80 (TABLES1–4).
Evidence of eGFR errors
Chronic kidney disease. Over 30 studies,
involving about 30,000 participants, have
evaluated the performance of eGFR in
patients with CKD17,20,23,28–35,81–103 (TABLE1;
Supplementary Table 2). In cross- sectional
studies, P30 ranged from 60% to 90%, and
P10 averaged 40%. Thus, most eGFR values
showed ±30% discordance with mGFR
values. Importantly, in 10–40% of patients,
the discordance with mGFR exceeded
±30%. The margin of error was comparable
for creatinine- based and cystatin C- based
equations, showing that cystatin C did
not improve the performance of eGFR
calculations17,20,23,82,97. For example, two
different studies observed that CKD- EPI
equations using creatinine, cystatin C, or
both, all exhibited similar P30 values of
83–89%23,97. The limits of agreement between
eGFR and mGFR were extremely wide
(from −40 ml/min/1.73 m2 to 40 ml/min/
1.73 m2)98. The only study that used statistics
of agreement showed poor concordance
(CCC <0.75) between mGFRs and eGFRs
calculated using the MDRD, CKD- EPI and
Cockcroft–Gault equations93.
Agreement between eGFR and mGFR
was poor for every stage of CKD81,82,88,97.
P30 values tended to be higher in CKD
stage 1–2 than in CKD stage 3–5. This error
markedly affected the classification of CKD
stage, resulting in about one in three patients
being incorrectly classified on the basis of
eGFR into a lower or higher CKD stage than
they would be by mGFR20,86,101. The MDRD
equation is considered appropriate for use in
patients with CKD, because it was developed
mostly in individuals with this condition.
However, P30 values of 92% and 88%
(obtained in patients with mGFRs above and
below 60 ml/min, respectively87), and a lower
P30 of 50% in patients with CKD stages 4–5,
have been reported with this formula88,100.
Moreover, the P30 values of formulae
developed in populations that include the
whole spectrum of GFR values (such as
CKD- EPI, variants of which are based on
creatinine, cystatin C or both, MCQ and
Cockcroft–Gault) are similar to or even
worse than those of the MDRD equation
in patients with CKD82,83,86,88 (TABLE1).
The same equation can also yield discrepant
results in similar populations; for
example, CKD- EPI had P30s of 67%83
and 87%82 in two separate studies. Similar
variability was observed for MDRD86,87 and
cystatin C-based formulae17,82,83,97,100.
Interestingly, eGFR has shown wider
margins of error in non- white than in
white populations. However, the use of
ethnicity- specific corrective factors or
population- specific equations did not
improve the accuracy or precision of eGFR
measurements. In African Americans with
CKD, eGFRs calculated using the African-
American Study of Hypertension and
Kidney Disease (AASK) formula exhibited
great variability compared with mGFRs25,27.
In patients with eGFRs <60 ml/min/1.73 m2,
an MDRD variant adapted to African
Americans and the CKD- EPI equation4 both
showed similar P30 values of ~82%85. In
black South Africans, the standard MDRD
exhibited a higher P30 than the adapted
formula84 (TABLE1). Similarly, adapted
equations or population- specific formulae
did not improve the accuracy of eGFRs
in Chinese or Japanese patients20,29–35,90.
Finally, equations based on β- trace protein
or β2-microglobulin did not outperform
previous formulae23.
Few studies have compared the
performance of different eGFR formulae
in the monitoring of renal function over
time, an issue of particular relevance in
patients with progressive diseases. From a
methodological point of view, it is important
to note that GFR decline is frequently
Nature reviews
|
Nephrology
PersPectives
Study Population Gold standard Formulae Statistics and error Comments Refs
Hossain
(2010)
99 white
patients
99mTc- DTPA Cr: MDRD, CG • mGFR decline: −6.0
• eGFR decline: MDRD −6.8, CG
−6.2 (% per year)
eGFR decline was faster than
mGFR decline
119
Buron (2011) 1,249 white
patients
Inulin Cr: MDRD, CKD-
EPI, CG, Nankivell
• P30 at baseline: MDRD 85, CKD-
EPI 81, CG 77 , Nankivell 64
• L A: MDRD −22 to 21, CKD- EPI
−21 to 29, CG −20 to 30, Nankivell
−13 to 35
• In 15–35% of patients, eGFR error
was more than ±30% of mGFR
• Wide L A for all formulae
• No statistical analysis of GFR
decline
120
aMDRD, abbreviated Modification of Diet in Renal Disease; CCC: concordance correlation coefficient; CG: Cockcroft–Gault; CKD, chronic kidney disease;
CKD- EPI, CKD Epidemiology Collaboration; Cr, creatinine; CrCl, creatinine clearance; Cy , cystatin C; DTPA , pentetic acid; eGFR , estimated GFR; GFR , glomerular
filtration rate; L A , limits of agreement (values in ml/min/1.73 m2 for cross- sectional studies and in ml/min/1.73 m2 per year for longitudinal studies, except where
stated); mGFR , measured GFR; MCG, modified Cockcroft–Gault; MDRD, Modification of Diet in Renal Disease; P, percentage of estimations included within an
error of either ±30% (P30) or ±10% (P10) of mGFR; TDI, total deviation index.
Table 3 (cont.) | Studies that evaluated the performance of estimated GFR in patients after renal transplantation
evaluated using linear regression analysis,
because the reliability of this calculation is
highly dependent on having a large number
of GFR determinations. When GFR decline
is evaluated with only 2–3 data points, as it
generally is in published studies, the result is
highly dependent on the last recorded value
of GFR. This factor has to be considered
when interpreting the results of available
studies, which have consistently shown
poor agreement between eGFR- based and
mGFR-based rates of decline (TABLE1).
Mean eGFR decline was 15% or 18%
faster than mGFR decline in two different
studies91. In another study, MDRD- based
eGFR decline was poorly correlated
(r = 0.60, r2 = 0.42) with mGFR decline90.
Accordingly, in 58% of these patients, eGFR
decline differed from mGFR decline by
between −2 ml/min/1.73 m2 per year and
+2 ml/min/1.73 m2 per year, and another
20% of patients showed differences between
−2 ml/min/1.73 m2 per year and −4 ml/min/
1.73 m2 per year or more90. Comparable
results were observed in other studies, which
revealed either poor concordance (r = 0.60)
between eGFR decline and mGFR decline89,
or wide limits of agreement (ranging from
+3 ml/min/1.73 m2 per year to −3 ml/min/
1.73 m2 per year)104. Thus, the rate of
change in eGFR is not a reliable method
for detecting changes in renal function
over time. This factor must be taken into
consideration both in clinical practice and
in the design of clinical trials investigating
the prevention of renal function loss.
Very few studies have analysed the
effect of GFR decline on important clinical
outcomes such as cardiovascular disease or
end- stage renal disease (ESRD). In one
study of patients with GFR <30 ml/min,
mGFR was a better predictor of all- cause
mortality than CKD- EPI-based eGFR:
an improvement in GFR of ~7 ml/min was
associated with reductions in all- cause
mortality of 29% when mGFR was used
versus 9% when eGFR was used105. This
disparity clearly indicates that errors in
eGFR calculations affect predictions of the
consequences of CKD. Moreover, this study
used a simplified mGFR technique based on
a single measurement of plasma clearance
of iohexol, a method shown elsewhere to
be inaccurate in 25% of patients106. The
use of a more accurate GFR measurement
method might further increase the observed
difference between mGFR- based and eGFR-
based predictions of all- cause mortality
in patients with CKD. In a subanalysis of
the AASK study89, the incidence of major
adverse renal events (defined as either a
50% reduction in GFR or ESRD) differed
according to whether eGFR or mGFR was
used. A total of 280 events were detected
using mGFR, of which eGFR only detected
223 (80%). Therefore, 57 of the patients
identified by mGFR as having experienced
a major adverse renal event would remain
undiagnosed by eGFR. In addition, a major
adverse renal event was diagnosed in 23
patients by eGFR but not detected by mGFR.
Therefore, eGFR proved to be inaccurate in
the diagnosis of major adverse renal events.
By contrast, in a study that evaluated
the effect of GFR decline (both eGFR-
based and mGFR- based) on the risk of
cardiovascular events, death and ESRD107,
mGFR declines over a 2-year period did
not outperform eGFR declines for these
outcomes. Methodological limitations might
explain these intriguing and counterintuitive
results, including the determination of GFR
decline (both eGFR and mGFR) using only
two measurements in 24 months, which
as we noted above is an unreliable method
of measuring the rate of decline in renal
function. In addition, eGFR (not mGFR)
values were used to define ESRD and to
prompt the initiation of dialysis, which leads
to bias in the results favouring eGFR- based
measurements. Finally, this study was not
designed to evaluate the non- inferiority
of either method with regard to outcome
prediction. Therefore, the results of this
study must be interpreted cautiously.
Type 2 diabetes mellitus. Several studies,
involving about 3,500 patients with T2DM,
have evaluated the performance of eGFR
versus mGFR over a wide spectrum
of renal function: normoalbuminuria,
microalbuminuria, overt proteinuria,
www.nature.com/nrneph
PersPectives
Table 4 | Studies that evaluated the performance of estimated GFR in autosomal dominant polycystic kidney disease
Study Population Gold standard Formulae Statistics and error Comments Refs
Cross- sectional studies
Orskov
(2009)
101 white
patients
51Cr- EDTA Cr: CG,
MDRD,
CKD- EPI Cy:
MDRD- Cy
P30: CG 90, MDRD 83, CKD- EPI 90;
MDRD- Cy 97
• In 10–25% of patients, eGFR
error was more than ±30%
of mGFR
• Similar error in eGFR
whether mGFR was above
or below 60 ml/min
136
Longitudinal studies
Spithoven
(2013)
121 white
patients
125I- iothalamate Cr: MDRD,
CKD- EPI
GFR at baseline:
• P30: MDRD 97%, CKD- EPI 99%; P10:
MDRD 50%, CKD- EPI 54%
• L A: MDRD −15 to 28; CKD- EPI −15 to 22
GFR decline (ml/min per year):
• mGFR −2.8 ± 3.4; CKD- EPI −3.2 ± 3.3;
MDRD −2.9 ± 3.2
• L A: MDRD −5 to 5; CKD- EPI −5 to 5
• In 1–3% of patients, eGFR
error was more than ±30%
of mGFR
• In 46–50% of patients, eGFR
error was more than ±10%
of mGFR
• Wide L A between mGFR
and eGFR decline
137
Ruggenenti
(2012)
111 white
patients
Iohexol Cr: MDRD,
CKD- EPI
GFR at baseline:
• L A: MDRD −30.6 to 19.5; CKD- EPI −21.3 to 27
GFR decline (ml/min per year):
• mGFR −8 ± 10; CKD- EPI −5 ± 9; MDRD −4.5 ± 10
• L A: MDRD −22 to 29; CKD- EPI −21 to 28
• Poor agreement between
eGFR and mGFR decline
• mGFR decline faster than
eGFR decline
44
CG, Cockcroft–Gault; CKD- EPI, Chronic Kidney Disease–Epidemiology Collaboration; Cr, creatinine; Cy , cystatin C; eGFR , estimated GFR; GFR , glomerular
filtration rate; L A , limits of agreement (values in ml/min/1.73 m2 for cross- sectional studies and in ml/min/1.73 m2 per year for longitudinal studies); mGFR ,
measured GFR; MDRD, Modification of Diet in Renal Disease; P, percentage of estimations included within an error of either ±30% (P30) or ±10% (P10) of mGFR .
normal renal function, hyperfiltration
and CKD18,37,38,45,46,102,108–110 (TABLE2;
Supplementary Table3). In cross- sectional
studies, P30 was 40–90% and P10 was
8–48%, indicating that most eGFRs differed
from mGFRs by ±30% or more. The limits
of agreement between eGFR and mGFR
were wide and sometimes reached extreme
values; for example, from −50 ml/min to
35 ml/min for the MDRD109 (TABLE2).
In 600 patients with T2DM and
normoalbuminuria, microalbuminuria,
normal renal function or hyperfiltration45,
15 creatinine- based eGFR formulae all
showed very poor concordance with mGFR
(CCC <0.52). In addition, the average TDI
was 50%, indicating that 90% of eGFRs
fell within a margin of error of ±50%
compared with mGFR. In addition, 70%
of the patients with mGFR- confirmed
hyperfiltration were not identified by any
eGFR formula. In patients with T2DM and
overt nephropathy or CKD, Cockcroft–
Gault and MDRD eGFR formulae showed
wide limits of agreement with mGFR (from
−47 ml/min to 33 ml/min)38. This error led
to the misclassification of CKD stage in
35% of patients, when staging was based
on eGFRs derived by the Cockcroft–Gault,
MDRD and MCQ equations109. Similar
results were observed in patients with
T2DM and CKD46,110,111. Thus, compared
with mGFR, eGFR formulae have failed to
identify hyperfiltration as well as normal
kidney function or various degrees of
renal insufficiency in patients with T2DM
(reviewed elsewhere114).
Cystatin C- based eGFR formulae have
shown no improvements in precision or
accuracy versus serum- creatinine-based
formulae in patients with T2DM. Indeed,
many cystatin C- based formulae have
lower P30 values than those for MDRD or
CKD-EPI formulae, although a few have P30
values comparable to those of creatinine-
based formulae (TABLE2). Also, the three
versions of CKD- EPI (those based on
creatinine, cystatin C, or both markers) all
have comparable P30 values of about 90%,
Nature reviews
|
Nephrology
PersPectives
a
c
b
d
0.0
010 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180
0.0
0.5
010 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7. 0
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
Serum creatinine (mg/dl)Serum cystatin C (mg/l)
mGFR (ml/min)
mGFR (ml/min)
0.0
010 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180
0.0
0.5
010 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7. 0
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
Serum creatinine (mg/dl)Serum cystatin C (mg/l)
mGFR (ml/min)
mGFR (ml/min)
Fig. 1 | Relationship between serum levels of
creatinine or serum cystatin C and measured
glomerular filtration rate in patients with and
without renal disease. The cause of renal impair-
ment in these two cohorts of patients was type 2
diabetes mellitus, renal transplantation, autoso-
mal dominant polycystic kidney disease or
chronic kidney disease. Measured glomerular fil-
tration rate (mGFR) was assessed by plasma clear-
ance of iohexol. a | The curvilinear relationship
between serum creatinine level and mGFR in
3,146 patients with renal impairment. b | In these
3,146 patients, serum creatinine 1.5 mg/dl is asso-
ciated with mGFRs of 30–90 ml/min. c | The curvi-
linear relationship between serum cystatin C level
and mGFR in 597 patients with and without renal
impairment. d | In these 597 patients, serum
cystatin C 1.5 mg/l is associated with mGFRs of
30–90 ml/min. Data for this figure were provided
by the Laboratory of the Aldo e Cele Daccó Center
for Rare Diseases, Mario Negri Institute, Bergamo,
Italy and the Laboratory of Renal Function,
Hospital Universitario de Canarias, University of La
Laguna, Tenerife, Spain.
and similar results have also been observed
in other studies111 (TABLE2). So, as observed
for patients with CKD, the use of cystatin C
did not improve the accuracy of eGFR
formulae in patients with T2DM.
Decline in renal function was slower
when determined by eGFR than by mGFR.
For example, in one longitudinal study
in patients with T2DM, the mean mGFR
decline was −3.37 ml/min/1.73 m2 per year,
whereas the mean eGFR decline was 50–60%
slower (from −1 ml/min/1.73 m2 per year to
−1.50 ml/min/1.73 m2 per year) with various
creatinine- based formulae45. Similar results
were observed for the CKD- EPI110, MDRD38
and Cockcroft–Gault38 equations, in patients
with normoalbuminuria, microalbuminuria
or overt proteinuria38,45. The fact that eGFR
decline is slower than mGFR decline limits
the use of eGFR formulae in clinical practice.
Moreover, clinical trials that use eGFR to
measure renal function decline will have a
reduced statistical power to detect potential
benefits of treatments designed to prevent
renal function loss in patients with T2DM.
Renal transplantation. The performance of
eGFR formulae in renal transplant recipients
has been evaluated in more than 30 studies,
including ~6,000 patients40–43,79,115–135 (TABLE3;
Supplementary Table 4). In cross- sectional
studies, P30 was 30–90% and P10 was
8–48%. Thus, most eGFR formulae erred
by ±30% compared to mGFR, and errors
were frequently even greater. The limits
of agreement between eGFR and mGFR
were wide, from −30 ml/min to 30 ml/min
on average40,115,122,131. This error was similar
for creatinine- based and cystatin C- based
equations116,127,131–135. In 193 patients evaluated
using 51 eGFR formulae, all equations
showed poor concordance with mGFR
(CCC 0.04–0.90)79. The average TDI was
about 60–70%, indicating that 90% of
estimations fell within a margin of error of
60–70% with regard to mGFR. The error for
24 h CrCl was similar to that for eGFR42,130–132.
Furthermore, CKD staging based on eGFR
values led to incorrect classification in
30–60% of patients79,132–135. The use of eGFR
values for CKD staging purposes has a major
confounding effect in this population.
Studies that compared creatinine- based
and cystatin C- based formulae yielded
contradictory results. One report suggested
that P30 values were comparable between
the three different CKD- EPI formulae116.
By contrast, several cystatin C- based
equations performed better than creatinine-
based formulae, demonstrated by a reduction
in TDI from 70% to 40%79. However, this
numerical change, though substantial from
a mathematical point of view, is not sufficient
to indicate a clinically relevant improvement
in agreement between eGFR and mGFR.
A TDI of 40% still represents a very wide
margin of error that cannot be accepted in
day- to-day clinical practice or research.
Longitudinal studies have shown opposite
results: in some publications, eGFR decline
was faster, and in others it was slower, than
mGFR decline. One study reported an mGFR
decline of ~2 ml/min per year, but most
eGFR equations showed slower declines42.
Another report showed that eGFR decline
was 30–50% slower than mGFR decline117.
Finally, eGFR decline measured using
creatinine- based eGFR formulae was twofold
faster than the mGFR decline43. Thus, eGFR
has poor reliability for monitoring kidney
function over time in this population.
Autosomal dominant polycystic kidney
disease. Few studies have tested the
performance of eGFR formulae in patients
with autosomal dominant polycystic
kidney disease (ADPKD)44,136,137 (TABLE4).
Two reports showed that most eGFRs
(calculated using MDRD, Cockcroft–Gault
and CKD-EPI equations) fell within a
margin of error of ±30%136,137. In longitudinal
studies, eGFR decline showed wide limits of
agreement with mGFR decline137. Moreover,
eGFR decline was 50% slower than mGFR
decline: −8 ml/min per year versus −4 ml/min
per year44. This difference indicates that eGFR
formulae fail to detect patients who progress
rapidly towards CKD in this population.
Flaws of eGFR in clinical trials
In clinical trials of interventions designed
to slow the rate of decline in renal function
over time, the use of a reliable tool to
determine GFR is imperative. The fact
that chronic nephropathies are slowly
progressive must be taken into account.
www.nature.com/nrneph
PersPectives
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170
180
0.5
0.0
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7. 0
7. 5
8.0
8.5
Serum creatinine (mg/dl)
mGFR (ml/min)
0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170
180
0.5
0.0
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
7. 0
7. 5
8.0
8.5
Serum creatinine (mg/dl)
mGFR (ml/min)
a
b
Fig. 2 | Relationship between serum creatinine level and measured glomerular filtration rate.
Measured glomerular filtration rate (mGFR) was assessed by plasma clearance of 99T c- labelled pentetic
acid (DTPA). a | The curvilinear relationship between serum creatinine level and mGFR in 171 patients
with renal disease. b | In these 171 patients, serum creatinine 1.5 mg/dl is associated with mGFRs of
20–70 ml/min. Adapted with permission from REF.47, Elsevier.
The 2012 Kidney Disease: Improving
Global Outcomes (KDIGO) clinical practice
guideline on CKD defined rapid progression
to CKD as a GFR decline faster than
−5 ml/min per year138. Clinical interventions
that achieve a 40–50% reduction in GFR
decline (for example, from −5.0 ml/min
per year to −2.0 or −2.5 ml/min per year)
are considered to have proven efficacy in
preventing renal function loss in high-
risk patients. Although this difference
in the rate of decline is clinically important,
the change is numerically small, so very
precise and accurate tools are needed
to detect it. The margin of error of all
eGFR formulae is wide, and errors are
unpredictable and occur randomly. Thus,
in serial eGFR values, variations of ±30%
from mGFR severely affect the calculation
of GFR decline, especially when only a few
measurements are obtained.
Several studies in renal transplant
recipients and patients with CKD, T2DM or
ADPKD showed important discrepancies in
renal function decline as assessed by eGFR
and mGFR. In fact, the rate of change in
renal function using eGFR formulae proved
to be 50–100% faster or slower than with
mGFR38,42–45,90,91,110,117. This variability can
mask the effects of interventions that slow
GFR decline. For example, data from the
ALADIN study139 showed a 50% reduction
in GFR decline versus placebo in patients
with ADPKD receiving somatostatin
(−2 ml/min per year versus −4 ml/min per
year; P = 0.03). However, recalculation of
GFR decline using the abbreviated MDRD
equation led to slower rates of decline in
both groups, particularly in the patients
receiving somatostatin, which masked the
difference between groups (−3.0 ml/min per
year versus −2.8 ml/min; P not significant)
for somatostatin versus placebo
(P.R., unpublished observations).
In the light of the evidence from
published studies, the European Medicines
Agency recommended that mGFR should be
prioritized over eGFR whenever the precise
determination of GFR is considered essential
(that is, in prospective studies) to avoid the
great variability in eGFR140.
mGFR in clinical practice
A new approach to the evaluation of renal
function in day- to-day clinical practice
is needed. According to the evidence we
discuss here, the development of further
eGFR equations similar to those already
published might not be the correct strategy.
However, the incorporation of biomarkers of
renal damage into existing equations might
improve the precision and accuracy of eGFR.
Another possibility could be that mGFR
could be used in circumstances in which a
reliable determination of renal function is
needed. No international recommendations
in this field are available. We speculate —
based on the available literature — that
mGFR could be implemented to evaluate
renal function, its evolution over time and
CKD staging in patients with established
renal disease, and to prevent or avoid the
toxic effects of drugs or procedures involving
contrast media. These circumstances might
include, for example: the evaluation of living
kidney donors before nephrectomy141,142;
evaluation of the progression of renal
function over time in patients with
established CKD20,79,86,101,109,132–135; dose
adjustment of potentially nephrotoxic drugs
(such as chemotherapeutic agents) according
to renal function; evaluation of patients
with CKD scheduled to undergo procedures
that require the use of contrast media; to
guide therapeutic decisions based on renal
function, such as the initiation of dialysis
in patients with advanced (stage 4–5) CKD;
to investigate the possibility of performing
nephron- sparing surgery in patients with
CKD scheduled to undergo nephrectomy;
and evaluation of patients with CKD in
whom urine collection is not possible.
Several different methods have been used
for measuring GFR: the clearance of inulin,
51Cr- EDTA,99Tc- pentetic acid (DTPA), iohexol
or iothalamate143. All these methods have been
criticized as impractical, difficult, burdensome
and time-consuming, which has limited and
even discouraged their use in clinics and
research. However, these limitations mostly
relate to measuring the clearance of inulin,
which requires continuous marker infusion
and urine collection. In a review of studies
that compared clearance of 51Cr- EDTA,
99Tc- DTPA, iohexol or iothalamate with
inulin clearance, the authors concluded that
the evidence suggests that renal clearance
of iothalamate, renal or plasma clearance of
EDTA, and plasma clearance of iohexol are
reliable alternatives to inulin clearance in the
measurement of GFR144. This conclusion is
especially important because inulin is now
very expensive. Additionally, 51Cr- EDTA
and 99Tc- DTPA have the disadvantage of
requiring administration of a radiolabelled
substance. However, iothalamate and iohexol
are non- radioactive markers and clearance
of these agents is simple and practical to
measure in the clinic69,70. In fact, measuring
the plasma clearance of iohexol is safe145 and
only requires a single intravenous injection
of the marker (5 ml in 2 min) with minimal
blood sampling over the subsequent 4–8 h
depending on the patient’s renal function69,70.
Nature reviews
|
Nephrology
PersPectives
a b
mGFR
ml/min
P30 (%)
30 36 42 78 84 90
mGFR
ml/min 42 48 54 66 72 78
–50 –40 –30 30 40 50
60
90%
60
P10 (%) –30 –20 –10 10 20 30
Fig. 3 | The clinical relevance of margins of error in parameters used to assess agreement
between estimated and measured glomerular filtration rates. a | P30 is defined as the percentage
of estimated glomerular filtration rates (eGFRs) that lie within ±30% of the measured glomerular fil-
tration rate (mGFR). In a patient with mGFR 60 ml/min, eGFRs of 42–78 ml/min are within the P30
limit and are therefore considered to be in good agreement with mGFR . b | P10 is defined as the per-
centage of eGFR values that lie within ±10% of mGFR . In a patient with mGFR 60 ml/min, eGFRs of
54–66 ml/min fall within the P10 limit and are considered to be in good agreement with mGFR . To be
clinically meaningful, 90% of eGFRs should fall within this range.
Glossary
Accuracy
The degree of closeness of the determined value (in this
case, estimated glomerular filtration rate) to the true
value (measured glomerular filtration rate) under
prescribed conditions. Accuracy is also sometimes
termed trueness.
Bias
The difference between an estimated value (such as
estimated glomerular filtration rate) and a true value
(measured glomerular filtration rate), which is also
termed error. A statistic is biased if it is calculated in
such a way that it is systematically different from the
parameter being estimated.
Coefficient of variation
The variation obtained when measurements are
repeated under the same conditions. A low value
indicates that the technique is both accurate and
precise.
Precision
The closeness of agreement (that is, the degree of
scatter) in a series of determinations (that is, estimated
glomerular filtration rate values) obtained from multiple
sampling of the same homogenous sample under the
prescribed conditions.
Reproducibility
The precision of results compared between two
laboratories. Reproducibility also refers to the precision
of a particular method when used under the same
operating conditions over a short period of time.
Importantly, plasma clearance of inulin and
iohexol result in comparable mGFR values
when multiple blood samples are taken146.
Moreover, iohexol-based GFR measurement
is not a particularly expensive procedure; it
costs about €100–200 (REFS69,70).
If methodological complexity is claimed
as an argument against the clinical use of
mGFR, the most appropriate approach must
be to simplify the reference method without
losing its accuracy and precision. One group
simplified the measurement of plasma
clearance of iohexol by replacing venous
blood samples with dried capillary blood
samples deposited on filter paper147. The
dried blood sampling approach is simple and
safe, and increases patient comfort owing to
the use of a painless finger prick to collect
capillary blood. This sampling method also
reduces the number of venipunctures as well
as the cost of iohexol clearance measurement
(since tubes, syringes and sample
centrifugation are not needed). Moreover,
dried blood on filter paper is stable at room
temperature and does not need to be stored
in the freezer. Agreement between the dried
blood sampling approach and the standard
method for measuring plasma clearance of
iohexol was excellent (TDI 9%), suggesting
that the two methods are interchangeable147.
Thus, the dried blood sampling approach
represents a valuable simplification of a
reference method for measuring GFR that
could help to promote the use of mGFR in
clinical practice and research.
Finally, we consider that the suggestion
that GFR measurement is time- consuming
cannot defend the continued use of
suboptimal and unreliable measures of
kidney function. No one would today
challenge the role of morphological and
functional evaluations in the diagnosis
and monitoring of cardiovascular and
gastroenterological diseases that require
time- consuming and costly procedures,
such as angiography, colonoscopy,
magnetic resonance neurography or
histological evaluations.
Conclusions
Our review of comparative studies reveals
that the margin of error for all eGFR
formulae is unexpectedly and unacceptably
wide across the whole spectrum of kidney
function, from hyperfiltration and normal
renal function to moderate and advanced
CKD. The performance of eGFR formulae
(in terms of recognized metrics such as
P30 values) has not improved over the
past 60 years, and no improvements can
be attributed to the use of cystatin C- based
formulae. Thus, eGFR is an unreliable tool to
assess renal function in health and disease, as
well as in clinical practice and research.
Whenever feasible, we recommend the
direct measurement of GFR insituations
where careful monitoring of kidney function
and/or its changes over time are required at
the individual level. In the era of precision
medicine, the ongoing use of unreliable
eGFRs rather than mGFR, a key marker of
kidney function, remains a paradox.
EstebanPorrini1,2*, PieroRuggenenti3,4,
SergioLuis-Lima1,2, FabiolaCarrara4,
AlejandroJiménez1, AikodeVries5, ArmandoTorres1,2,
FlavioGaspari4 and GiuseppeRemuzzi3,4,6
1Hospital Universitario de Canarias, Tenerife, Spain.
2University of La Laguna, ITB: Instituto de Tecnologías
Biomédicas, Tenerife, Spain.
3Nephrology and Dialysis Unit, Azienda Socio Sanitaria
Territoriale Papa Giovanni XXIII, Bergamo, Italy.
4Istituto di Ricerche Farmacologiche Mario Negri
IRCCS, Centro di Ricerche Cliniche per le Malattie Rare
“Aldo e Cele Daccò”, Ranica, Bergamo, Italy.
5Leiden University Medical Centre, Department of
Medicine, Division of Nephrology, Leiden, Netherlands.
6Department of Biochemical and Clinical Sciences ‘
L. Sacco’, University of Milan, Milan, Italy.
*e- mail: esteban.l.porrini@gmail.com
https://doi.org/10.1038/s41581-018-0080-9
Published online xx xx xxxx
1. Perrone, R. D., Madias, N. E. & Levey, A. S. Serum
creatinine as an index of renal function: new insights
into old concepts. Clin. Chem. 38, 1933–1953
(1992).
2. Kaji, D., Strauss, I. & Kahn, T. Serum creatinine in
patients with spinal cord injury. Mt. Sinai J. Med. 57,
160–164 (1990).
3. Rule, A. D. etal. Using serum creatinine to estimate
glomerular filtration rate: accuracy in good health and
in chronic kidney disease. Ann. Intern. Med. 141,
929–937 (2004).
4. Levey, A. S. etal. A new equation to estimate
glomerular filtration rate. Ann. Intern. Med. 150,
604–612 (2009).
5. Björk, J. etal. Prediction of relative glomerular
filtration rate in adults: new improved equations based
on Swedish Caucasians and standardized plasma-
creatinine assays. Scand. J. Clin. Lab. Invest. 67,
678–695 (2007).
6. Björk, J., Grubb, A., Sterner, G. & Nyman, U. Revised
equations for estimating glomerular filtration rate
based on the Lund–Malmö Study cohort. Scand. J.
Clin. Lab. Invest. 71, 232–239 (2011).
7. Effersøe, P. Relationship between endogenous 24-hour
creatinine clearance and serum creatinine
concentration in patients with chronic renal disease.
Acta Med. Scand. 156, 429–434 (1957).
8. Edwards, K. D. & Whyte, H. M. Plasma creatinine level
and creatinine clearance as tests of renal function.
Australas. Ann. Med. 8, 218–224 (1959).
9. Jelliffe, R. W. Estimation of creatinine clearance when
urine cannot be collected. Lancet 1, 975–976
(1971).
10. Mawer, G. E., Lucas, S. B., Knowles, B. R.
& Stirland, R. M. Computer- assisted prescribing of
kanamycin for patients with renal insufficiency.
Lancet 1, 12–15 (1972).
11. Jelliffe, R. W. Letter: creatinine clearance: bedside
estimate. Ann. Intern. Med. 79, 604–605 (1973).
12. Cockcroft, D. W. & Gault, M. H. Prediction of
creatinine clearance from serum creatinine. Nephron
16, 31–41 (1976).
13. Levey, A. S., Greene, T., Kusek, J. W. & Beck, G. J.
A simplified equation to predict glomerular filtration rate
from serum creatinine [abstract]. J. Am. Soc. Nephrol.
11, 115A (2000).
14. Pottel, H. etal. An estimated glomerular filtration
rate equation for the full age spectrum. Nephrol. Dial.
Transplant. 31, 798–806 (2016).
15. Grubb, A., Simonsen, O., Sturfelt, G., Truedsson, L.
& Thysell, H. Serum concentration of cystatin C, factor
D and β2-microglobulin as a measure of glomerular
filtration rate. Acta Med. Scand. 218, 499–503
(1985).
16. Ma, Y. C. etal. Improved GFR estimation by combined
creatinine and cystatin C measurements. Kidney Int.
72, 1535–1542 (2007).
17. Stevens, L. A. etal. Estimating GFR using serum cystatin
C alone and in combination with serum creatinine:
a pooled analysis of 3,418 individuals with CKD. Am. J.
Kidney Dis. 51, 395–406 (2008).
18. Inker, L. A. etal. Estimating glomerular filtration rate
from serum creatinine and cystatin C. N. Engl. J. Med.
367, 20–29 (2012).
19. Schaeffner, E. S. etal. Two novel equations to estimate
kidney function in persons aged 70 years or older.
Ann. Intern. Med. 157, 471–481 (2012).
20. Feng, J. F. etal. Multicenter study of creatinine- and/or
cystatin C- based equations for estimation of
glomerular filtration rates in Chinese patients with
chronic kidney disease. PLOS ONE 8, e57240 (2013).
21. Pottel, H. etal. Estimating glomerular filtration rate
for the full age spectrum from serum creatinine and
cystatin C. Nephrol. Dial. Transplant. 32, 497–507
(2017).
22. Björk, J. etal. Comparison of glomerular filtration rate
estimating equations derived from creatinine and
cystatin C: validation in the age, gene/environment
susceptibility- Reykjavik elderly cohort. Nephrol. Dial.
Transplant. 33, 1380–1388 (2018).
23. Inker, L. A. etal. GFR estimation using β- trace protein
and β2-microglobulin in CKD. Am. J. Kidney Dis. 67,
40–48 (2016).
24. Nankivell, B. J., Gruenewald, S. M., Allen, R. D.
& Chapman, J. R. Predicting glomerular filtration rate
after kidney transplantation. Transplantation 59,
1683–1689 (1995).
25. Ibrahim, H. etal. An alternative formula to the
Cockcroft–Gault and the modification of diet in renal
diseases formulas in predicting GFR in individuals with
type 1 diabetes. J. Am. Soc. Nephrol. 16, 1051–1060
(2005).
26. MacIsaac, R. J. etal. Estimating glomerular filtration
rate in diabetes: a comparison of cystatin-
C− and
creatinine- based methods. Diabetologia 49,
1686–1689 (2006).
27. Lewis, J. etal. Comparison of cross- sectional renal
function measurements in African Americans with
hypertensive nephrosclerosis and of primary formulas
to estimate glomerular filtration rate. Am. J. Kidney
Dis. 38, 744–753 (2001).
28. Xie, P., Huang, J. M., Li, Y., Liu, H. J. & Qu, Y.
The modified CKD- EPI equation may be not more
accurate than CKD- EPI equation in determining
glomerular filtration rate in Chinese patients with chronic
kidney disease. J. Nephrol. 30, 397–402 (2017).
29. Yang, M. etal. Performance of the creatinine and
cystatin C- based equations for estimation of GFR in
Chinese patients with chronic kidney disease.
Clin. Exp. Nephrol. 21, 236–246 (2017).
30. Changjie, G. etal. Evaluation of glomerular filtration
rate by different equations in Chinese elderly with
chronic kidney disease. Int. Urol. Nephrol. 49,
133–141 (2017).
31. Guo, X. etal. Improved glomerular filtration rate
estimation using new equations combined with
standardized cystatin C and creatinine in Chinese adult
chronic kidney disease patients. Clin. Biochem. 47,
1220–1226 (2014).
32. Li, J. T. etal. Relative performance of two equations
for estimation of glomerular filtration rate in a Chinese
population having chronic kidney disease. Chin. Med.
J. 125, 599–603 (2012).
33. Liu, X. etal. Comparison of prediction equations to
estimate glomerular filtration rate in Chinese patients
with chronic kidney disease. Intern. Med. J. 42,
e59–e67 (2012).
34. Matsuo, S. etal. Revised equations for estimated GFR
from serum creatinine in Japan. Am. J. Kidney Dis. 53,
982–992 (2009).
35. Praditpornsilpa, K. etal. The need for robust
validation for MDRD- based glomerular filtration rate
estimation in various CKD populations. Nephrol. Dial.
Transplant. 26, 2780–2785 (2011).
36. Grubb, A. etal. Generation of a new cystatin C- based
estimating equation for glomerular filtration rate by
use of 7 assays standardized to the international
calibrator. Clin. Chem. 60, 974–986 (2014).
37. MacIsaac, R. J. etal. The Chronic Kidney Disease-
Epidemiology Collaboration (CKD- EPI) equation does
not improve the underestimation of glomerular
www.nature.com/nrneph
PersPectives
filtration rate (GFR) in people with diabetes and
preserved renal function. BMC Nephrol. 16, 198
(2015).
38. Rossing, P., Rossing, K., Gaede, P., Pedersen, O.
& Parving, H. H. Monitoring kidney function in type 2
diabetic patients with incipient and overt diabetic
nephropathy. Diabetes Care 29, 1024–1030 (2006).
39. Fontsere, N. etal. Are prediction equations for
glomerular filtration rate useful for the long- term
monitoring of type 2 diabetic patients? Nephrol. Dial.
Transplant. 21, 2152–2158 (2006).
40. Mariat, C. etal. Assessing renal graft function in
clinical trials: can tests predicting glomerular filtration
rate substitute for a reference method? Kidney Int.
65, 289–297 (2004).
41. Mariat, C. etal. Predicting glomerular filtration rate in
kidney transplantation: are the K/DOQI guidelines
applicable? Am. J. Transplant. 5, 2698–2703 (2005).
42. Bosma, R. J., Doorenbos, C. R., Stegeman, C. A.,
vander Heide, J. J. & Navis, G. Predictive performance
of renal function equations in renal transplant
recipients: an analysis of patient factors in bias. Am. J.
Transplant. 5, 2193–2203 (2005).
43. Gaspari, F. etal. Performance of different prediction
equations for estimating renal function in kidney
transplantation. Am. J. Transplant. 4, 1826–1835
(2004).
44. Ruggenenti, P. etal. Measuring and estimating GFR
and treatment effect in ADPKD patients: results and
implications of a longitudinal cohort study. PLOS ONE
7, e32533 (2012).
45. Gaspari, F. etal. The GFR and GFR decline cannot be
accurately estimated in type 2 diabetics. Kidney Int.
84, 164–173 (2013).
46. Iliadis, F. etal. Glomerular filtration rate estimation in
patients with type 2 diabetes: creatinine- or cystatin
C- based equations? Diabetologia 54, 2987–2994
(2011).
47. Shemesh, O., Golbetz, H., Kriss, J. P. & Myers, B. D.
Limitations of creatinine as a filtration marker in
glomerulopathic patients. Kidney Int. 28, 830–838
(1985).
48. Laterza, O. F., Price, C. P. & Scott, M. G. Cystatin C:
an improved estimator of glomerular filtration rate?
Clin. Chem. 48, 699–707 (2002).
49. Crim, M. C., Calloway, D. H. & Margen, S. Creatine
metabolism in men: urinary creatine and creatinine
excretions with creatine feeding. J. Nutr. 105,
428–438 (1975).
50. Heymsfield, S. B., Arteaga, C., McManus, C., Smith, J.
& Moffitt, S. Measurement of muscle mass in humans:
validity of the 24-hour urinary creatinine method.
Am. J. Clin. Nutr. 37, 478–494 (1983).
51. Bleiler, R. E. & Schedl, H. P. Creatinine excretion:
variability and relationships to diet and body size.
J. Lab. Clin. Med. 59, 945–955 (1962).
52. Irving, R. A., Noakes, T. D., Irving, G. A.
& Van Zyl-Smit, R. The immediate and delayed
effects of marathon running on renal function. J. Urol.
136, 1176–1180 (1986).
53. Rennie, M. J. etal. Effect of exercise on protein
turnover in man. Clin. Sci. 61, 627–639 (1981).
54. Horber, F. F., Scheidegger, J. & Frey, F. J.
Overestimation of renal function in glucocorticosteroid
treated patients. Eur. J. Clin. Pharmacol. 28,
537–541 (1985).
55. Friedman, R. B., Anderson, R. E., Entine, S. M.
& Hirshberg, S. B. Effects of diseases on clinical
laboratory tests. Clin. Chem. 26, 1D–476D
(1980).
56. Miller, B. F., Leaf, A., Mamby, A. R. & Miller, Z.
Validity of the endogenous creatinine clearance as a
measure of glomerular filtration rate in the diseased
human kidney. J. Clin. Invest. 31, 309–313 (1952).
57. Baldwin, D. S., Sirota, J. H. & Villarreal, H. Diurnal
variations of renal function in congestive heart failure.
Proc. Soc. Exp. Biol. Med. 74, 578–581 (1950).
58. Chesley, L. C. Renal excretion at low urine volumes
and the mechanism of oliguria. J. Clin. Invest. 17,
591–597 (1938).
59. Jones, J. D. & Burnett, P. C. Creatinine metabolism in
humans with decreased renal function: creatinine
deficit. Clin. Chem. 20, 1204–1212 (1974).
60. Mitch, W. E., Collier, V. U. & Walser, M. Creatinine
metabolism in chronic renal failure. Clin. Sci. 58,
327–335 (1980).
61. Miller, W. G. etal. Creatinine measurement: state of
the art in accuracy and interlaboratory harmonization.
Arch. Pathol. Lab. Med. 129, 297–304 (2005).
62. Panteghini, M. Enzymatic assays for creatinine: time
for action. Clin. Chem. Lab. Med. 46, 567–572
(2008).
63. Peake, M. & Whiting, M. Measurement of serum
creatinine — current status and future goals.
Clin. Biochem. Rev. 27, 173–184 (2006).
64. Myers, G. L. etal. Recommendations for improving
serum creatinine measurement: a report from the
Laboratory Working Group of the National Kidney
Disease Education Program. Clin. Chem. 52, 5–18
(2006).
65. Delanaye, P., Cavalier, E., Depas, G., Chapelle, J. P.
& Krzesinski, J. M. New data on the intraindividual
variation of cystatin C. Nephron. Clin. Pract. 108,
c246–c248 (2008).
66. Abrahamson, M. etal. Structure and expression of the
human cystatin C gene. Biochem. J. 268, 287–294
(1990).
67. Grubb, A. Diagnostic value of analysis of cystatin C
and protein HC in biological fluids. Clin. Nephrol. 38
(Suppl. 1), S20–S27 (1992).
68. de Vries, A. P. & Rabelink, T. J. A possible role of
cystatin C in adipose tissue homeostasis may impact
kidney function estimation in metabolic syndrome.
Nephrol. Dial. Transplant. 28, 1628–1630 (2013).
69. Delanaye, P. etal. Iohexol plasma clearance for
measuring glomerular filtration rate in clinical practice
and research: a review. Part 1: how to measure
glomerular filtration rate with iohexol? Clin. Kidney J.
9, 682–699 (2016).
70. Delanaye, P. etal. Iohexol plasma clearance for
measuring glomerular filtration rate in clinical practice
and research: a review. Part 2: why to measure
glomerular filtration rate with iohexol? Clin. Kidney J.
9, 700–704 (2016).
71. Chantler, C. & Barratt, T. M. Estimation of glomerular
filtration rate from plasma clearance of 51-chromium
edetic acid. Arch. Dis. Child. 47, 613–617 (1972).
72. Brochner- Mortensen, J. & Rodbro, P. Selection of
routine method for determination of glomerular
filtration rate in adult patients. Scand. J. Clin. Lab.
Invest. 36, 35–43 (1976).
73. Hall, P. M. & Rolin, H. Iothalamate clearance and its
use in large- scale clinical trials. Curr. Opin. Nephrol.
Hypertens. 4, 510–513 (1995).
74. Equalis. External quality assessment (EQA) schemes:
Iohexol (024). equalis https://www.equalis.se/en/
products- and-services/external- quality-assessment-
eqa/eqa- schemes/g- l/iohexol-024/ (2014).
75. Lin, L., Hedayat, A. & Wu, W. Statistical Tools for
Measuring Agreement (Springer Science+Business
Media, 2012).
76. Lin, L. I. A concordance correlation coefficient to
evaluate reproducibility. Biometrics 45, 255–268
(1989).
77. Lin, L., Hedayat, A., Sinha, B. & Yang, M. Statistical
methods in assessing agreement: models, issues, and
tools. J. Am. Stat. Assoc. 97, 257–270 (2002).
78. Bland, J. M. & Altman, D. G. Statistical methods for
assessing agreement between two methods of clinical
measurement. Lancet 1, 307–310 (1986).
79. Luis- Lima, S. etal. Estimated glomerular filtration rate
in renal transplantation: the nephrologist in the mist.
Transplantation 99, 2625–2633 (2015).
80. Ahlstrom, M. G., Kjaer, A., Gerstoft, J. & Obel, N.
Agreement between estimated and measured renal
function in an everyday clinical outpatient setting of
human immunodeficiency virus- infected individuals.
Nephron 136, 318–327 (2017).
81. Selistre, L. etal. Comparison of the Schwartz and
CKD- EPI equations for estimating glomerular filtration
rate in children, adolescents, and adults: a
retrospective cross- sectional study. PLOS Med. 13,
e1001979 (2016).
82. Fan, L. etal. Glomerular filtration rate estimation
using cystatin C alone or combined with creatinine as
a confirmatory test. Nephrol. Dial. Transplant. 29,
1195–1203 (2014).
83. Evans, M. etal. Glomerular filtration rate- estimating
equations for patients with advanced chronic kidney
disease. Nephrol. Dial. Transplant. 28, 2518–2526
(2013).
84. van Deventer, H. E., Paiker, J. E., Katz, I. J.
& George, J. A. A comparison of cystatin C− and
creatinine-based prediction equations for the estimation
of glomerular filtration rate in black South Africans.
Nephrol. Dial. Transplant. 26, 1553–1558 (2011).
85. Stevens, L. A. etal. Evaluation of the modification of
diet in renal disease study equation in a large diverse
population. J. Am. Soc. Nephrol. 18, 2749–2757
(2007).
86. Murata, K. etal. Relative performance of the MDRD and
CKD- EPI equations for estimating glomerular filtration
rate among patients with varied clinical presentations.
Clin. J. Am. Soc. Nephrol. 6, 1963–1972 (2011).
87. Froissart, M., Rossert, J., Jacquot, C., Paillard, M. &
Houillier, P. Predictive performance of the modification
of diet in renal disease and Cockcroft–Gault equations
for estimating renal function. J. Am. Soc. Nephrol. 16,
763–773 (2005).
88. Hojs, R., Bevc, S., Ekart, R., Gorenjak, M. & Puklavec, L.
Kidney function estimating equations in patients with
chronic kidney disease. Int. J. Clin. Pract. 65, 458–464
(2011).
89. Wang, X. etal. Validation of creatinine- based
estimates of GFR when evaluating risk factors
in longitudinal studies of kidney disease. J. Am.
Soc. Nephrol. 17, 2900–2909 (2006).
90. Xie, D. etal. A comparison of change in measured and
estimated glomerular filtration rate in patients with
nondiabetic kidney disease. Clin. J. Am. Soc. Nephrol.
3, 1332–1338 (2008).
91. Padala, S. etal. Accuracy of a GFR estimating
equation over time in people with a wide range of
kidney function. Am. J. Kidney Dis. 60, 217–224
(2012).
92. Lee, D., Levin, A., Roger, S. D. & McMahon, L. P.
Longitudinal analysis of performance of estimated
glomerular filtration rate as renal function declines in
chronic kidney disease. Nephrol. Dial. Transplant. 24,
109–116 (2009).
93. Hossain, F., Kendrick- Jones, J., Ma, T. M.
& Marshall, M. R. The estimation of glomerular
filtration rate in an Australian and New Zealand
cohort. Nephrology (Carlton) 17, 285–293 (2012).
94. Ebert, N. etal. Cystatin C standardization decreases
assay variation and improves assessment of
glomerular filtration rate. Clin. Chim. Acta 456,
115–121 (2016).
95. Poggio, E. D., Wang, X., Greene, T., Van Lente, F.
& Hall, P. M. Performance of the modification of diet
in renal disease and Cockcroft–Gault equations in the
estimation of GFR in health and in chronic kidney
disease. J. Am. Soc. Nephrol. 16, 459–466
(2005).
96. Bevc, S., Hojs, R., Ekart, R., Gorenjak, M. & Puklavec, L.
Simple cystatin C formula compared to serum
creatinine- based formulas for estimation of glomerular
filtration rate in patients with mildly to moderately
impaired kidney function. Kidney Blood Press. Res. 35,
649–654 (2012).
97. Bevc, S., Hojs, R., Ekart, R., Gorenjak, M. & Puklavec, L.
Simple cystatin C formula compared to sophisticated
CKD- EPI formulas for estimation of glomerular filtration
rate in the elderly. Ther. Apher. Dial. 15, 261–268
(2011).
98. Hojs, R., Bevc, S., Ekart, R., Gorenjak, M. & Puklavec, L.
Serum cystatin C as an endogenous marker of renal
function in patients with mild to moderate impairment
of kidney function. Nephrol. Dial. Transplant. 21,
1855–1862 (2006).
99. Barroso, S., Martinez, J. M., Martin, M. V., Rayo, I.
& Caravaca, F. Accuracy of indirect estimates of renal
function in advanced chronic renal failure patients.
Nefrologia 26, 344–350 (2006).
100. Lopes, M. B. etal. Estimation of glomerular filtration
rate from serum creatinine and cystatin C in
octogenarians and nonagenarians. BMC Nephrol. 14,
265 (2013).
101. Brown, M. A. etal. Inaccuracies in estimated
glomerular filtration rate in one Australian renal
centre. Nephrology (Carlton) 16, 486–494 (2011).
102. Li, H. X., Xu, G. B., Wang, X. J., Zhang, X. C. & Yang, J. M.
Diagnostic accuracy of various glomerular filtration rates
estimating equations in patients with chronic kidney
disease and diabetes. Chin. Med. J. 123, 745–751
(2010).
103. El- Minshawy, O. M. & El- Bassuoni, E. Validation of
El- Minia equation for estimation of glomerular
filtration rate in different stages of chronic kidney
disease. Iran. J. Kidney Dis. 6, 262–268 (2012).
104. Tent, H. etal. Performance of MDRD study and
CKD-EPI equations for long- term follow- up of
nondiabetic patients with chronic kidney disease.
Nephrol. Dial. Transplant. 27 (Suppl. 3), 89–95 (2012).
105. Methven, S., Gasparini, A., Carrero, J. J., Caskey, F. J.
& Evans, M. Routinely measured iohexol glomerular
filtration rate versus creatinine- based estimated
glomerular filtration rate as predictors of mortality in
patients with advanced chronic kidney disease: a
Swedish Chronic Kidney Disease Registry cohort study.
Nephrol. Dial. Transplant. 32 (Suppl. 2), 170–179
(2017).
106. Gaspari, F. etal. Glomerular filtration rate determined
from a single plasma sample after intravenous iohexol
injection: is it reliable? J. Am. Soc. Nephrol. 7,
2689–2693 (1996).
Nature reviews
|
Nephrology
PersPectives
107. Ku, E. etal. Change in measured GFR versus eGFR and
CKD outcomes. J. Am. Soc. Nephrol. 27, 2196–2204
(2016).
108. Maple- Brown, L. J. etal. Performance of formulas for
estimating glomerular filtration rate in indigenous
Australians with and without type 2 diabetes: the
eGFR Study. Diabet. Med. 31, 829–838 (2014).
109. Rigalleau, V. etal. A simplified Cockcroft–Gault
formula to improve the prediction of the glomerular
filtration rate in diabetic patients. Diabetes Metab.
32, 56–62 (2006).
110 . Wood, A. J. etal. Estimating glomerular filtration rate:
performance of the CKD- EPI equation over time in
patients with type 2 diabetes. J. Diabetes
Complications 30, 49–54 (2016).
111. Beauvieux, M. C. etal. New predictive equations
improve monitoring of kidney function in patients with
diabetes. Diabetes Care 30, 1988–1994 (2007).
112 . Silveiro, S. P. etal. Chronic Kidney Disease
Epidemiology Collaboration (CKD- EPI) equation
pronouncedly underestimates glomerular filtration
rate in type 2 diabetes. Diabetes Care 34,
2353–2355 (2011).
113 . Perkins, B. A. etal. Detection of renal function decline
in patients with diabetes and normal or elevated GFR
by serial measurements of serum cystatin C
concentration: results of a 4-year follow- up study.
J. Am. Soc. Nephrol. 16, 1404–1412 (2005).
114 . Bjornstad, P., Cherney, D. Z. & Maahs, D. M. Update
on estimation of kidney function in diabetic kidney
disease. Curr. Diab. Rep. 15, 57 (2015).
115 . Masson, I. etal. MDRD versus CKD- EPI equation to
estimate glomerular filtration rate in kidney transplant
recipients. Transplantation 95, 1211–1217 (2013).
116 . Masson, I. etal. GFR estimation using standardized
cystatin C in kidney transplant recipients. Am. J.
Kidney Dis. 61, 279–284 (2013).
117 . Gera, M. etal. Assessment of changes in kidney
allograft function using creatinine- based estimates
of glomerular filtration rate. Am. J. Transplant. 7,
880–887 (2007).
118 . Fauvel, J. P., Hadj- Aissa, A., Buron, F., Morelon, E.
& Ducher, M. Performance of estimated glomerular
filtration rates to monitor change in renal function in
kidney transplant recipients. Nephrol. Dial. Transplant.
28, 3096–3100 (2013).
119 . Hossain, M. A., Attia, A. & Shoker, A. Measurement
error in estimated GFR slopes across transplant
chronic kidney disease stages. Am. J. Nephrol. 31,
151–159 (2010).
120. Buron, F. etal. Estimating glomerular filtration rate in
kidney transplant recipients: performance over time of
four creatinine- based formulas. Transplantation 92,
1005–1011 (2011).
121. Attia, A., Zahran, A. & Shoker, A. Comparison of
equations to estimate the glomerular filtration rate in
post- renal transplant chronic kidney disease patients.
Saudi J. Kidney Dis. Transpl. 23, 453–460 (2012).
122. Goerdt, P. J., Heim- Duthoy, K. L., Macres, M. &
Swan, S. K. Predictive performance of renal function
estimate equations in renal allografts. Br. J.
Clin. Pharmacol. 44, 261–265 (1997).
123. Harzallah, K. etal. Creatinine clearance estimation
after kidney transplantation: an analysis of three
methods. Transplant. Proc. 39, 2571–2573
(2007).
124. Kukla, A. etal. GFR- estimating models in kidney
transplant recipients on a steroid- free regimen.
Nephrol. Dial. Transplant. 25, 1653–1661 (2010).
125. Maillard, N. etal. Cystatin C- based equations in renal
transplantation: moving toward a better glomerular
filtration rate prediction? Transplantation 85,
1855–1858 (2008).
126. Poge, U. etal. MDRD equations for estimation of GFR
in renal transplant recipients. Am. J. Transplant. 5,
1306–1311 (2005).
127. Poge, U. etal. Cystatin C- based calculation of
glomerular filtration rate in kidney transplant
recipients. Kidney Int. 70, 204–210 (2006).
128. Poge, U., Gerhardt, T., Stoffel- Wagner, B., Sauerbruch, T.
& Woitas, R. P. Validation of the CKD- EPI formula in
patients after renal transplantation. Nephrol. Dial.
Transplant. 26, 4104–4108 (2011).
129. Poggio, E. D. etal. Assessing glomerular filtration
rate by estimation equations in kidney transplant
recipients. Am. J. Transplant. 6, 100–108
(2006).
130. Raju, D. L., Grover, V. K. & Shoker, A. Limitations of
glomerular filtration rate equations in the renal
transplant patient. Clin. Transplant. 19, 259–268
(2005).
131. Risch, L. & Huber, A. R. Assessing glomerular filtration
rate in renal transplant recipients by estimates derived
from serum measurements of creatinine and cystatin
C. Clin. Chim. Acta 356, 204–211 (2005).
132. White, C. etal. Estimating glomerular filtration rate
in kidney transplantation: a comparison between
serum creatinine and cystatin C- based methods.
J. Am. Soc. Nephrol. 16, 3763–3770 (2005).
133. White, C. etal. Chronic kidney disease stage in renal
transplantation classification using cystatin C and
creatinine- based equations. Nephrol. Dial. Transplant.
22, 3013–3020 (2007).
134. Yeo, Y. etal. Suitability of the IDMS- traceable MDRD
equation method to estimate GFR in early
postoperative renal transplant recipients. Nephron
Clin. Pract. 114 , c108–c117 (2010).
135. Zahran, A., Qureshi, M. & Shoker, A. Comparison
between creatinine and cystatin C- based GFR
equations in renal transplantation. Nephrol. Dial.
Transplant. 22, 2659–2668 (2007).
136. Orskov, B. etal. Estimating glomerular filtration rate
using the new CKD- EPI equation and other equations
in patients with autosomal dominant polycystic kidney
disease. Am. J. Nephrol. 31, 53–57 (2010).
137. Spithoven, E. M. etal. Tubular secretion of creatinine
in autosomal dominant polycystic kidney disease:
consequences for cross- sectional and longitudinal
performance of kidney function estimating equations.
Am. J. Kidney Dis. 62, 531–540 (2013).
138. [No authors listed.] Chapter 2: definition,
identification, and prediction of CKD progression.
Kidney Int. Suppl. 3, 63–72 (2013).
139. Caroli, A. etal. Effect of longacting somatostatin
analogue on kidney and cyst growth in autosomal
dominant polycystic kidney disease (ALADIN):
a randomised, placebo- controlled, multicentre trial.
Lancet 382, 1485–1495 (2013).
140. European Medicines Agency. Guideline on the clinical
investigation of medicinal products to prevent
development/slow progression of chronic renal
insufficiency. EMA http://www.ema.europa.eu/docs/en_
GB/document_library/Scientific_guideline/2016/10/
WC500214980.pdf (2016).
141. British Transplantation Society. Guidelines for living
donor kidney transplantation. BTS https://bts.org.uk/
wp- content/uploads/2018/07/FINAL_LDKT- guidelines_
June-2018.pdf (2018).
142. The Renal Association. Clinical practice guideline post-
operative care in the kidney transplant recipient. renal
https://renal.org/wp- content/uploads/2017/06/FINAL-
Post-Operative- Care-Guideline.pdf (2017).
143. Schaeffner, E. Determining the glomerular filtration
rate — an overview. J. Ren. Nutr. 27, 375–380
(2017).
144. Soveri, I. etal. Measuring GFR: a systematic review.
Am. J. Kidney Dis. 64, 411–424 (2014).
145. Gaspari, F. etal. Safety of iohexol administration to
measure glomerular filtration rate in different patient
populations: a 25-year experience. Nephron 140,
1–8 (2018).
146. Sterner, G. etal. Determining ‘true’ glomerular
filtration rate in healthy adults using infusion of inulin
and comparing it with values obtained using other
clearance techniques or prediction equations.
Scand. J. Urol. Nephrol. 42, 278–285 (2008).
147. Luis- Lima, S. etal. Iohexol plasma clearance simplified
by dried blood spot testing. Nephrol. Dial. Transplant.
33, 1597–1603 (2018).
Acknowledgements
The authors acknowledge research support from the
DIABESITY working group of the ERA- EDTA, the IMBRAIN
(CIBICAN) project (FP7-RE6-POT-2012-CT2012-31637-
IMBRAIN) funded under the 7th Framework Programme
(capacities); Instituto de Salud Carlos III (ISCIII) grants
PI13/00342 and PI16/01814 to E.P. and A.T., REDINREN
RD16/0009 and PI10/02428 grants to E.P. and A.T.; and
funding from the IRSIN (Instituto Reina Sofia de Investigacion)
and FEDER (both to A.T.). S.L.L. is a research fellow
supported by ISCIII grant CM15/00214 for Río Hortega spe-
cialized health- care post- training contracts. E.P. is a
researcher supported by the ISCIII Ramón y Cajal Programme
and Fundación Caja Canarias grant DIAB05. The authors
thank F. G. Rinne for preparation of the figures, N. N. Mena
for performing the iohexol method in the Laboratory of Renal
Function, and M. L. McLean for technical assistance.
Author contributions
E.P., F.C. and F.G. researched data for the article, made sub-
stantial contributions to discussions of its content, wrote the
manuscript and reviewed or edited the manuscript before
submission. P.R., A.d.V. and G.R. made substantial contri-
butions to discussions of the article content, wrote the
manuscript and reviewed or edited the manuscript before sub-
mission. S.L.-L. researched data for the article, contributed
substantially to discussions of its content, and reviewed or
edited the manuscript before submission. A.J. researched data
for the article and contributed substantially to discussions of
its content. A.T. substantially contributed to discussions of the
article content.
Competing interests
The authors declare no competing interests.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
Reviewer information
Nature Reviews Nephrology thanks E. Cavalier, L. Dubourg
and the other anonymous reviewer(s) for their contribution to
the peer review of this work.
Supplementary information
Supplementary information is available for this paper at
https://doi.org/10.1038/s41581-018-0080-9.
www.nature.com/nrneph
PersPectives
