Open Access Research

Chronic kidney disease: a large-scale

population-based study of the effects of
introducing the CKD-EPI formula for
eGFR reporting
Christopher A OCallaghan,1 Brian Shine,2 Daniel S Lasserson3

To cite: OCallaghan CA, ABSTRACT

Shine B, Lasserson DS.
ARTICLE SUMMARY
Objective: To evaluate the effects of introducing the
Chronic kidney disease: Chronic Kidney Disease-Epidemiology Collaboration
a large-scale
Article focus
(CKD-EPI) formula for estimated glomerular filtration - Estimated glomerular filtration rates form the
population-based study of
rate (eGFR) reporting in the adult population in routine basis for clinical and health policy decisions in
the effects of introducing the
CKD-EPI formula for eGFR clinical practice with clinician-directed testing. chronic kidney disease.
reporting. BMJ Open 2011;1: Design: Retrospective study of all creatinine - The new CKD-EPI formula for estimated glomer-
e000308. doi:10.1136/ measurements and calculation of eGFRs using ular filtration rates estimates renal function better
bmjopen-2011-000308 Modification of Diet in Renal Disease (MDRD) and than the Modification of Diet in Renal Disease
CKD-EPI formulae. formula in current use.
< Prepublication history and
Setting: General population, Oxfordshire, UK. - We have studied the effects of using the CKD-EPI
additional tables and figures
for this paper are available Participants: An unselected population of around formula in a UK population of over half a million.
online. To view these files 660 000.
Key messages
please visit the journal online Interventions: Reporting of eGFRs using MDRD or
(http://bmjopen.bmj.com).
- Overall, the CKD-EPI formula produces higher
CKD-EPI formulae.
better estimated glomerular filtration rates, which
Primary and secondary outcome reduces the diagnosis of chronic kidney disease.
Received 11 August 2011
Accepted 4 November 2011
measures: Evaluation of the effects of the CKD-EPI However, in men older than 70 years, it produces
formula on the prevalence of different stages of lower worse estimated glomerular filtration rates
This final article is available chronic kidney disease (CKD). and increases the number with chronic kidney
for use under the terms of Results: The CKD-EPI formula reduced the prevalence disease stages 3e5.
the Creative Commons of CKD (stages 2e5) by 16.4% in patients tested in - Our results predict a net reduction of around
Attribution Non-Commercial primary care. At the important stage 2estage 3 cut-off, 200 000 in the numbers with chronic kidney
2.0 Licence; see there was a relative reduction of 7.5% in the
http://bmjopen.bmj.com
disease stages 3e5 in the UK. This would reduce
prevalence of CKD stages 3e5 from 15.7% to 14.5%. the primary care chronic kidney disease regis-
The CKD-EPI formula reduced the prevalence of CKD ters, inappropriate disease labelling and patient
stages 3e5 in those aged <70 but increased it at ages monitoring.
>70. Above 70 years, the prevalence of stages 3e5 - The chronic kidney disease classification of up to
was similar with both equations for women (around 3 million patients could be altered by the use of
41.2%) but rose in men from 33.3% to 35.5%. CKD the CKD-EPI formula in the UK.
stages 4e5 rose by 15% due exclusively to increases
in the over 70s, which could increase specialist referral Strengths and limitations of this study
rates. The CKD classification of 18.3% of all individuals - The study is large and unbiased. All primary care
who had a creatinine measurement was altered by samples taken during the study period were
1
a change from the MDRD to the CKD-EPI formula. In analysed, so the results represent current clinical
Nuffield Department of testing practice.
the UK population, the classification of up to 3 million
Clinical Medicine, University
patients could be altered, the prevalence of CKD could - Estimated glomerular filtration rates are suffi-
of Oxford, John Radcliffe
be reduced by up to 1.9 million and the prevalence of cient to diagnose chronic kidney disease stages
Hospital, Oxford, UK
2
Department of Clinical CKD stages 3e5 could fall by around 200 000. 3e5, but stages 1e2 also require proteinuria or
Biochemistry, John Radcliffe Conclusions: Introduction of the CKD-EPI formula for a structural abnormality, which cannot be
Hospital, Oxford, UK eGFR reporting will reduce the prevalence of CKD in assessed in this study. However, a change in
3
Department of Primary Care a primary care setting with current testing practice but estimated glomerular filtration rate can still alter
Health Sciences, University will raise the prevalence in the over 70s age group. the classification of stage 1 or 2.
of Oxford, Oxford, UK This has implications for clinical practice, healthcare
policy and current prevalence-based funding
Correspondence to
Professor Christopher A arrangements. INTRODUCTION
OCallaghan; chris. Chronic kidney disease (CKD) is common
[email protected] and important.1 People with CKD have an

OCallaghan CA, Shine B, Lasserson DS. BMJ Open 2011;1:e000308. doi:10.1136/bmjopen-2011-000308 1

Healthcare implications of introducing the CKD-EPI formula for eGFR reporting

increased overall mortality and an increased risk of formula for routine eGFR reporting by laboratories in
cardiovascular events, even after known risk factors are the USA.11 We have tested whether the use of the new
controlled for.2 Renal impairment also affects the safe CKD-EPI equation would have a significant effect on the
prescribing of many common drugs. Some patients will CKD stage classification in a large representative popu-
progress to end-stage renal disease with its substantial lation of patients tested in primary care. Such an effect
associated morbidity, mortality and cost. Milder degrees could have major implications for patients and for the
of renal impairment are more common but can also healthcare system.
cause morbidity, especially from fluid retention, hyper-
tension and a range of metabolic disturbances such as METHODS
hyperuricaemia.3 It is unfortunate that there is no We analysed all requests for creatinine measurement in
routinely available method for the reliable measurement the Oxford University Hospitals Trust Clinical
of glomerular filtration rate (GFR), which is the key Biochemistry laboratories from the Oxfordshire region
index of renal function. Serum creatinine is easy to during the period October 2009 to January 2011
quantify but only rises substantially when there is a major and identified all requests from primary care on patients
reduction in GFR. Furthermore, creatinine production aged $18 years. Assays were conducted using a
is influenced by age, sex, muscle mass and ethnicity. For kinetic Jaffe method (Siemens, Camberley, UK)
these reasons, formulae have been developed that calibrated to give isotope dilutionemass spectrometry
attempt to incorporate these factors and produce an (IDMS)-compatible results. We recorded the gender,
estimated glomerular filtration rate (eGFR). In the UK date of birth and the National Health Service (NHS)
and many other countries, versions of the Modification number of the patient and the date of the specimen. For
of Diet in Renal Disease (MDRD) formula have been each patient, we identified the first specimen and
adopted for routine reporting of eGFRs.4 While there is calculated the eGFRs using the MDRD and the CKD-EPI
a consensus that eGFR reporting has been clinically equations and compared the CKD stage classifications
helpful, there are problems associated with these based on these. We excluded creatinine measurements
eGFRs.5 In particular, they are inaccurate in early CKD of <10 mmol/l. The formulae are provided as online
such that the true GFR can vary substantially from the supplementary information (http://bmjopen.bmj.com).
eGFR.6 Estimates for the UK population were scaled on the size
CKD has been classified into stages based principally of the catchment area we studied and the most recently
upon eGFR and the assessment of proteinuria.7 Esti- available UK population census information.12 The
mates of the prevalence of CKD in different populations similarity of the Oxfordshire population to that of the
vary from around 5% to 10% of the total population.8 In UK is shown as online supplementary information
many countries, guidelines have been developed, such as (http://bmjopen.bmj.com).
the guidance from NICE in the UK or the National The calculation of eGFR is affected by ethnicity, but
Kidney Foundation K/DOQI guidelines in the USA, for given the imperfect recording of ethnicity in primary
the management of patients with CKD.5 7 Such guide- care and that a small percentage of Oxfordshire renal
lines emphasise the value of regular monitoring of function requests require adjustment for ethnicity,
patients with estimations of renal function, typically bootstrapping methods were used to test the effects of
annually for CKD stage 2 and 6 monthly for CKD stage ethnicity.13 To do this, we first needed to establish the
3.5 However, there are problems associated with this, variability in creatinine measures by analysing the results
including labelling millions of asymptomatic people with from patients who had more than one specimen
the term disease, the cost of repeated monitoring and collected within 3 months of each other. The distribu-
the inconvenience to the affected individual. In the UK, tion of test frequency is shown in online supplementary
primary care physicians have a financial incentive to figure 1 (http://bmjopen.bmj.com). The mean and SD
diagnose and follow-up patients with CKD under the were calculated for each patient, and from this, we
Quality and Outcomes Framework scheme (QOF), calculated the median SD within increasing 10 mmol/l
whereby funding allocations reflect the practices bins from 10 to 200 mmol/l and pooled measurements of
prevalence of chronic conditions including CKD.9 >200 mmol. The relevant median SD was used to add
Given the recognised inadequacies of the MDRD a random normal deviate to each creatinine value in
equation, it is important to use the most accurate Monte Carlo simulations designed to investigate the
possible estimates of GFR to avoid potential misclassifi- effects of ethnicity.
cation of patients, with the concomitant costs to the The Monte Carlo bootstrap simulations were
healthcare system and the individuals concerned. performed by random selection of samples of 100 000
Recently, a new formula, the Chronic Kidney Disease- patients with replacement from the pool of first or only
Epidemiology Collaboration (CKD-EPI), has been devel- specimens of 175 671 individual patients. All simulations
oped that shows a much improved concordance between were based on the first creatinine value in this study
eGFR and true measured GFR, especially in the earlier period for those with >1 measure of creatinine or on the
stages of CKD.10 The US National Kidney Foundation only measure for those with just one. For each patient
has recently recommended the adoption of the CKD-EPI sampled, we added a random normal deviate using the

2 OCallaghan CA, Shine B, Lasserson DS. BMJ Open 2011;1:e000308. doi:10.1136/bmjopen-2011-000308

 Healthcare implications of introducing the CKD-EPI formula for eGFR reporting

relevant median SD. We randomly allocated black strates that use of the CKD-EPI formula generally results
African ethnicity to 2.8% (on the basis of the ethnic in a higher eGFR than that obtained with the MDRD
distribution in Oxfordshire as shown in table 2) of the formula. The discrepancy increases as the eGFR rises but
patients in each sample of the total data set that was used even within the range where CKD classification depends
in each bootstrap simulation and used these data to on eGFR, the disparity between the two formulae can be
generate MDRD and CKD-EPI eGFR values. We large. Overall, 27.3% (47 882) of all patients had a higher
performed 10 000 simulations to derive approximate and better eGFR using the CKD-EPI formula compared
95% confidence limits for the proportion of patients with the MDRD formula. On the basis of these eGFR
who would be allocated to each CKD group. We used this results, we stratified the patients by the CKD stage that
method to generate pairs of data for each patient to their eGFR corresponds to in the conventional K/DOQI
define the mean proportion of patients who would be classification system.7 Overall, 18.3% (32 167of 175 671)
allocated to different CKD classes by successive of all individuals tested were reclassified on the basis of
measurements using each method. eGFR by a change from the use of the MDRD to the CKD-
EPI formula. Eight-four per cent of reclassifications were
RESULTS to a milder CKD stage with a higher eGFR and only 16%
Overview to a more severe CKD stage with a lower eGFR.
We analysed all creatinine results arising from requests
in our Oxfordshire catchment area during the time Effect on overall prevalence of CKD in primary care using
period from 1 October 2009 to 4 February 2011. This eGFR criteria
area covers a population of around 660 000, and 738 348 When the MDRD formula was used, 70.6% (124 187) of
requests were received during this time period. Of these all patients tested were shown to have eGFRs that
requests, 321 964 requests on 175 671 patients aged at correspond to CKD stage 2 or worse (table 1). However,
least 18 years were from primary care. The median SD of this number fell to only 59.1% (103 738) of patients
creatinine measurements in repeat requests is shown in when the CKD-EPI formula was used. This reduction
online supplementary table 1 (http://bmjopen.bmj. resulted in 16.4% of patients who would be labelled as
com). As our laboratory uses an isotope dilutionemass having CKD stage 2 or worse on the basis of eGFR
spectrometry-standardised serum creatinine assay, we criteria having that stage altered to stage 1 (normal
calculated the eGFR for each result using the revised renal function, eGFR>90 ml/min/1.73m2) or having
four-variable MDRD formula and the new CKD-EPI the disease label removed completely when the CKD-EPI
formula.6 10 formula was used.
As figure 1 illustrates, the use of the CKD-EPI formula The boundary between CKD stages 2 and 3 at the
compared with the MDRD formula results in a general eGFR cut-off of 60 ml/min/1.73m2 is important because
shift to the right, corresponding to higher, and so better, this is the level at which current policies and guidelines
eGFR values. The eGFR values derived from the MDRD recommend that patients are logged on a CKD register
and CKD-EPI formulae were highly correlated, the MDRD in primary care and monitored more frequently.5
values being more skewed. Figure 2 shows Patients with an eGFR of <60 ml/min/1.73m2 will
a BlandeAltman plot of this relationship. This demon- currently be listed on the primary care kidney disease

Figure 1 Prevalence of different

estimated glomerular filtration
rates (eGFRs) with the
Modification of Diet in Renal
Disease (MDRD) and Chronic
Kidney Disease-Epidemiology
Collaboration (CKD-EPI)
formulae. The prevalence within
the samples studied is indicated
on the y-axis with eGFR on the
x-axis. eGFRs were calculated
using the MDRD formula (dashed
line) or the CKD-EPI formula
(continuous line). Overall, eGFRs
calculated using the CKD-EPI
formula are higher and better than
those calculated using the MDRD
formula.

OCallaghan CA, Shine B, Lasserson DS. BMJ Open 2011;1:e000308. doi:10.1136/bmjopen-2011-000308 3

Healthcare implications of introducing the CKD-EPI formula for eGFR reporting

been classified as CKD stage 3 rather than stage 2. The

use of the CKD-EPI formula resulted in the removal of
1.2% of the entire population tested in primary care from
eligibility for CKD registers. This would equate to a fall in
the size of CKD registries in primary care by 7.5%,
a reduction in the associated monitoring requirements
and the proportionate loss of the associated QOF income
to general practices. Recent guidance has emphasised the
prognostic value of dividing CKD stage 3 into 3A
(45e59 ml/min/1.73m2) and 3B (30e44 ml/min/
1.73m2).5 Larger numbers of reclassifications are seen in
stage 3A than 3B with the use of the CKD-EPI formula,
and this mainly reflects reclassification to a lower better
CKD stage (online supplementary table 2, http://
Figure 2 BlandeAltman plot of the relationship between the bmjopen.bmj.com). Current guidance also emphasises
Modification of Diet in Renal Disease (MDRD) and Chronic that patients with CKD stages 4 and 5 should generally be
Kidney Disease-Epidemiology Collaboration (CKD-EPI) referred to a nephrologist for a specialist opinion. As
estimates of glomerular filtration rate The x-axis indicates the
table 1 shows, the number of patients with these stages of
average of the estimated glomerular filtration rates calculated
using the MDRD and CKD-EPI formulae and the y-axis
CKD rose by 338 (15%).
indicates the difference between these two estimates.
Influence of age and gender on CKD classification using
MDRD or CKD-EPI
registers which form part of the QOF funding arrange- Differences in the results obtained for eGFR with the two
ment in the UK.9 Overall, 15.7% (27 579) of all patients formulae were explored further by analysing specific age
tested were classified as CKD stage 3 or worse on eGFR groups for both men and women. Figure 3 shows the
criteria using the MDRD formula, but this number prevalence of CKD stages 2e5 classified by gender and
fell to 14.5% (25 504), a relative fall of 7.5%, when the age group, and the underlying data are provided in
CKD-EPI formula was used. online supplementary table 3 (http://bmjopen.bmj.
Three thousand and forty-eight patients (1.7% of all com). For every group except those over 70 years of age,
patients, 11.1% of those classified as CKD stage 3 or the prevalence of each stage of CKD is lower with the
worse) were reclassified as CKD stage 2 using the CKD-EPI CKD-EPI formula. This shift arises predominantly from
formula, while 973 patients (0.6% of all patients and a reduction in the numbers with CKD stage 2 in all
1.9% of patients classified as CKD stage 2) would have groups except the over 70s where there is a slight

Table 1 CKD classification by MDRD and CKD-EPI

eGFR MDRD <15 15e29 30e59 60e89 >90 Totals
CKD-EPI CKD stage 5 4 3 2 1/0
A. Numbers
<15 5 328 53 0 0 0 381
15e29 4 4 1812 368 0 0 2184
30e59 3 0 30 21 936 973 0 22 939
60e89 2 0 0 3048 71 565 3621 78 234
>90 1/0 0 0 0 24 070 47 863 71 933
Totals 332 1895 25 352 96 608 51 484 175 671
B. Percentages
<15 5 0.187 0.03 0 0 0 0.217
15e29 4 0.002 1.031 0.209 0 0 1.242
30e59 3 0 0.017 12.488 0.554 0 13.058
60e89 2 0 0 1.735 40.738 2.061 44.534
>90 1/0 0 0 0 13.702 27.246 40.948
Totals 0.189 1.078 14.431 54.994 29.307 100
Columns indicate results obtained using the MDRD formula and show the redistribution of patients in these CKD groups on the basis of eGFR
(ml/min/1.73m2) when assessed using the CKD-EPI formula, shown in rows. A shows the numbers of patients in each group and B shows
these numbers as a percentage of the total number of those tested. Figures in bold indicate those who do not change CKD category on the basis
of eGFR with a change from the MDRD to the CKD-EPI formula. Below these, underlined figures indicate the numbers who move into a better
CKD category and above and italicised figures indicate the numbers of patients who move into a worse CKD stage. Both CKD stages 1 and 2
require a structural or other abnormality in addition to the eGFR criteria.
CKD, chronic kidney disease; CKD-EPI, Chronic Kidney Disease-Epidemiology Collaboration; eGFR, estimated glomerular filtration rate;
MDRD, Modification of Diet in Renal Disease.

4 OCallaghan CA, Shine B, Lasserson DS. BMJ Open 2011;1:e000308. doi:10.1136/bmjopen-2011-000308

 Healthcare implications of introducing the CKD-EPI formula for eGFR reporting

Figure 3 Graph of the

prevalence of each stage of
chronic kidney disease (CKD)
within the samples tested grouped
by age and gender. The y-axis
indicates the percentage of the
patients tested in each age group
with estimated glomerular filtration
rates in the ranges corresponding
to the indicated CKD stages with
each formula (M indicates MDRD,
Modification of Diet in Renal
Disease formula and C indicates
CKD-EPI, Chronic Kidney
Disease-Epidemiology
Collaboration formula). Numbers
on the x-axis indicate ages in
years.

increase. Similarly, there is a smaller reduction in the Effects of ethnicity on eGFR prediction and CKD
numbers with stage 3 in all groups except the over 70s classification
men where there is a slight increase. At all ages <70, use An issue of relevance is the ethnic distribution of the
of the CKD-EPI formula reduced the number of people population we studied. From the perspective of eGFR
with CKD stages 3e5 (figure 3, online supplementary estimation, the key ethnicity of importance is that of
figure 2 and supplementary tables 3 and 4, http:// black African ethnicity where adjustments to the eGFR
bmjopen.bmj.com). In contrast, in men over 70, there calculations are made to correct for recognised ethnic
were increases in the percentage of patients with each differences in the relationship between serum creatinine
stage of CKD from stages 2 to 5 with the CKD-EPI levels and true measured GFR.14 15 Omitting ethnicity
formula and the number with CKD stages 3e5 rose from from the calculation of eGFR in an ethnically mixed
33.3% to 35.5%. In women over 70, the percentage of population would lead to a net underestimation of eGFR
patients with stages 3e5 CKD was similar with both and overestimation of CKD. Table 2 demonstrates that
equations (around 41.2%), although percentages of the ethnic structure of the Oxfordshire population
patients with stages 2, 4 and 5 were increased in women
with the CKD-EPI equation. Apart from this, the results
for men and women are essentially analogous for each
age group. However, although eGFRs are generally lower
in both men and women aged over 70 with the CKD-EPI
formula, the increase in CKD stages 3e5 in this age
group is due to an increase in the number of men rather
than women.
At the important eGFR cut-off of 60 ml/min/1.73m2,
the percentage of reclassifications is greatest in the
younger age group and greater in women than in men
(figure 4). For men and women of all ages <75, there is
a net shift to a better higher eGFR and so to the better
lower CKD stage 2. However, it is important to note that
in older patients (>80 years of age), the opposite is true
as there is a net shift to a worse lower eGFR and so to
a worse higher CKD stage. In older patients, there is
greater reclassification of men than of women into the Figure 4 Changes at the estimated glomerular filtration rates
CKD 3 or higher stages. Use of the CKD-EPI formula (eGFR) cut-off boundary of 60 ml/min/1.73m2 The x-axis
represents patient groups divided according to age in 5-year
rather than the MDRD formula reduced the proportion
groupings. The y-axis represents the percentage change in the
of younger patients with CKD stage 3 or worse but
number of people with an eGFR of <60 ml/min/1.73m2
increased the proportion of older patients with CKD occurring with a change from the use of the MDRD
stage 3 or worse. In younger patients, the reduction in (Modification of Diet in Renal Disease) formula to the CKD-EPI
the severity of CKD will be greatest in women; in older (Chronic Kidney Disease-Epidemiology Collaboration formula.
patients, the increase in the severity of CKD will be The percentage change is negative if there is a reduction in the
greatest in men. number of people with an eGFR <60 ml/min/1.73m2.

OCallaghan CA, Shine B, Lasserson DS. BMJ Open 2011;1:e000308. doi:10.1136/bmjopen-2011-000308 5

Healthcare implications of introducing the CKD-EPI formula for eGFR reporting

Predictions for the UK

Table 2 Ethnic distribution in Oxfordshire and the UK
Based on the latest UK census information indicating
Oxfordshire % UK % a UK population of 61 792 000 in mid-2009,12 our results
White/Asian 96.9 96.1 indicate that a change to the use of the CKD-EPI formula
Black African 0.83 2 would alter the classification of over 3 million
Other, mixed, not recorded 1.93 1.6 (3 011 611) individuals (table 5). Specifically, our
Total 99.63 99.7 modelling of the UK population indicates that on the
basis of eGFR criteria, a change in formula would result
in a net reduction of 1.9 million (1 914 522) in the
studied is similar to that of the UK and likely to be number of people identified as having reduced renal
representative of the overall UK population. To include function and so potentially classified as having CKD
the effect of ethnic distribution, we undertook a boot- stage 2 or worse. At the important eGFR cut-off of
strap analysis by performing a set of 10 000 Monte Carlo 60 ml/min/1.73m2 which separates CKD stages 2 and 3,
simulations to estimate the 95% confidence limits for we predict that there will be a net reduction in the
the proportion of people with each stage of CKD with number of people with CKD stage 3 or worse of around
each formula and the proportion of people who would 200 000 (194 270) individuals. This is the number of
be classified differently with each formula on two occa- people in the UK who are predicted to be removed from
sions within 3 months. Table 3 confirms that when the current CKD registers that form part of the UK QOF
ethnicity is included, using the CKD-EPI formula would funding structure. Our results also indicate that the
still result in a lower proportion of people being classi- number of people with CKD stages 4e5 will rise by
fied as having CKD overall, with a substantial reduction around 32 000, which could result in an increase in
in the proportion being classified as having CKD stage 2 referrals to specialists. The number of people with CKD
(43.8% vs 53.3%) and stage 3 (13.1% vs 14.5%) but stage 2 is predicted to fall by up to 1.7 million
increases in both stages 4 and 5. The number of patients (1 720 252) and the number with CKD stage 3 by
with CKD stages 4e5 rose by 16.1% in relative terms 225 915. These estimates are based on eGFR, and the
from 1.25% to 1.46% of those tested. When considered distinction between the different CKD stages is based on
as a percentage of the total population of 660 000 in the eGFR, but clearly a diagnosis of CKD stage 1 or 2 also
geographical region studied, these figures equate to requires the presence of proteinuria or a structural
a fall in the overall prevalence of CKD stages 3e5 abnormality.
detected using current testing policies from 4.2% with
the MDRD formula to 3.9% with the CKD-EPI formula.
DISCUSSION
Reclassification on repeat testing Our study demonstrates that introduction of the CKD-
Analysis of repeated measurements shows that using EPI formula would reduce the number of patients in the
either formula, there would, as expected, be some UK who are classified as having CKD on the basis of
reclassification of CKD stages on repeat testing (table 4). a reduced eGFR. Importantly, it would reduce the
For instance, with the MDRD formula, 12.1% of patients number of people with an eGFR of <60 ml/min/1.73m2
would be classified as having CKD stage 3 on their first and so reduce the numbers who are classified as having
and second tests but 2.2% would be reclassified as CKD CKD stage 3 or worse.
stage 2 on retesting and 0.22% as CKD stage 4. Overall, A key strength of our study is that it is based on
there is more consistency with the CKD-EPI formula such a complete, unbiased and very large data set of patients
that the classification would be unchanged for about 82% seen in primary care who have creatinine requests. The
of patients using MDRD and about 85% of patients using population from which the studies were drawn is rela-
CKD-EPI. tively typical of the total UK population. Blood samples

Table 3 Confidence limits for CKD stage distribution when ethnicity is taken into account
CKD stage eGFR MDRD % CKD-EPI %
3e5 <15e59 15.8 (15.5e16.00) 14.6 (14.4e14.8)
5 <15 0.193 (0.167e0.220) 0.225 (0.197e0.255)
4 15e29 1.06 (1.00e1.13) 1.23 (1.16e1.3)
3 30e59 14.5 (14.3e14.7) 13.1 (12.9e13.3)
2 60e89 53.3 (53.0e53.6) 43.8 (43.5e44.1)
1/0 $90 41.6 (41.3e41.9) 36.2 (35.9e36.5)
Values indicate the median percentages of those tested, who are diagnosed with each CKD stage when ethnicity is taken into account using
either the MDRD or the CKD-EPI formula. Values are show with 2.5% and 97.5% confidence limits. Medians and confidence limits are derived
from a bootstrapping analysis. The first row shows the values for CKD stages 3e5 inclusive. Both CKD stages 1 and 2 require a structural or
other abnormality in addition to the eGFR criteria.
CKD, chronic kidney disease; CKD-EPI, Chronic Kidney Disease-Epidemiology Collaboration; eGFR, estimated glomerular filtration rate;
MDRD, Modification of Diet in Renal Disease.

6 OCallaghan CA, Shine B, Lasserson DS. BMJ Open 2011;1:e000308. doi:10.1136/bmjopen-2011-000308

 Healthcare implications of introducing the CKD-EPI formula for eGFR reporting

Table 4 Reclassification on repeat testing

First
Second <15, stage 5 15e29, stage 4 30e59, stage 3 60e89, stage 2 >90, stage 1/0
A. MDRD testing
<15, stage 5 0.16 0.03
15e29, stage 4 0.03 0.81 0.22
30e59, stage 3 0.22 12.1 2.2
60e89, stage 2 2.2 46.1 5.0
$90, stage 1/0 5.0 25.8
B. CKD-EPI testing
<15, stage 5 0.19 0.04
15e29, stage 4 0.04 0.94 0.25
30e59, stage 3 0.25 11.1 1.8
60e89, stage 2 1.8 37.8 4.2
$90, stage 1/0 4.2 37.4
Values indicate the percentage of all patients tested. Rows and columns refer to eGFR (ml/min/1.73m2) and CKD stage. Each column shows
the distribution of people who were originally in that eGFR category group and the rows in the column indicate their subsequent reclassification
on repeat measurement of creatinine using the same formula (MDRD in A or CKD-EPI in B). Figures in bold indicate the percentage who do not
change classification with repeat testing. Both CKD stages 1 and 2 require a structural or other abnormality in addition to the eGFR criteria.
CKD, chronic kidney disease; CKD-EPI, Chronic Kidney Disease-Epidemiology Collaboration; GFR, estimated glomerular filtration rate; MDRD,
Modification of Diet in Renal Disease.

were taken for a wide range of clinical reasons and were Our study is the largest study of the effect of changing
not performed as a population screen for the purposes to the use of the CKD-EPI formula in the UK population
of establishing the population prevalence of CKD. and is the only such study to factor in the effects of
Therefore, this is not a population prevalence analysis. ethnicity. Furthermore, it is based on a well-defined
As no CKD screening policy is in place, the strength of population in a single region and on accurate
our results is that they are the product of current clinical measurements from a single laboratory service. The
practice and are a complete picture of primary care Oxfordshire population has a relatively typical ethnic
testing in the population served by our laboratory. It ageegender distribution compared with the rest of the
remains the case that CKD will only be detected if UK (table 2, online supplementary figure 4, http://
a clinician decides to test an individual, so studies based bmjopen.bmj.com) and only a slightly lower level of
on current practice for renal function testing could estimated deprivation (online supplementary informa-
underestimate the true prevalence. There may well be tion, http://bmjopen.bmj.com). We have not attempted
unidentified people with CKD who are not coming to to assess the accuracy of the estimation of true GFR by
medical attention; this would also apply in other the different equations as this has been extensively
geographical areas. We did not have estimates of albu- studied already.4 10 Rather, our aim has been to take
minuria, but this was not necessary for our aim, which a large representative sample of the UK population
was to explicitly test the consequences for CKD classifi- being tested in the primary care setting and study the
cation based on eGFR of a change from the MDRD to practical effects for clinical practice and for patients of
the CKD-EPI formula. All other considerations in CKD a change from the use of the MDRD formula to the CKD-
classification are unchanged by the choice of formula. EPI formula.

Table 5 CKD classification estimates for the UK population by MDRD and CKD-EPI
eGFR MDRD <15 15e29 30e59 60e89 >90
CKD-EPI CKD stage 5 4 3 2 1/0 Totals
<15 5 30 709 4962 0 0 0 35 671
15e29 4 374 169 647 34 454 0 0 204 475
30e59 3 0 2808 2 053 741 91 096 0 2 147 645
60e89 2 0 0 285 367 6 700 219 339 013 7 324 599
>90 1/0 0 0 0 2 253 536 4 481 137 6 734 673
Totals 31 083 177 417 2 373 562 9 044 851 4 820 150 16 447 063
The table design and coding is based on that of table 1. Columns indicate the MDRD results and show the redistribution of patients in these CKD
groups on the basis of eGFR when assessed using the CKD-EPI formula, shown in rows. Numbers in bold indicate those who do not change
CKD category with a change from the MDRD to the CKD-EPI formula. Below these, underlined figures indicate the numbers who move into
a better CKD category and above and italicised figures indicate the numbers of patients who move into a worse CKD stage. Both CKD stages 1
and 2 require a structural or other abnormality in addition to the eGFR criteria.
CKD, chronic kidney disease; CKD-EPI, chronic kidney disease-epidemiology collaboration; eGFR, estimated glomerular filtration rate; MDRD,
Modification of Diet in Renal Disease.

OCallaghan CA, Shine B, Lasserson DS. BMJ Open 2011;1:e000308. doi:10.1136/bmjopen-2011-000308 7

Healthcare implications of introducing the CKD-EPI formula for eGFR reporting

A smaller study from Kent did not take into account min/1.73m2.10 A further study by the same group
the effect of ethnicity in the analysis of the laboratory examined the relative performance of the MDRD and
results.16 This may have led to an overestimate of the CKD-EPI using a validation data set of 3896 patients and
numbers with CKD as 19.1% of patients tested had CKD found that in the eGFR range 30e59 ml/min/1.73m2,
stages 3e5 with the MDRD formula and 17.2% with the the bias was decreased from 4.9 to 2.1 ml/min/1.73m2.21
CKD-EPI formula on the basis of eGFR. This equated to In a study of 1992 individuals recruited for a population
a prevalence of CKD stages 3e5 in the population screening exercise, a BlandeAltman analysis indicated
studied of 4.9% with the MDRD formula and 4.4% with a mean difference between the MDRD and CKD-EPI
the CKD-EPI formula, which are higher than our esti- formulae of 2.667 ml/min/1.73m2.22 Consistent with
mates of 4.2% and 3.9%, respectively, when ethnicity is our data, there was a reduction in the number of
included. When scaled up, these differences would patients with CKD stage 3 or worse from 11.04% to
amount to substantial increases in the estimates of the 7.98%. As these results are based on small numbers of
UK prevalence of around 430 000 for the MDRD formula individuals recruited to a screening study, they do not
and 310 000 for the CKD-EPI formula compared with the allow clear estimates of the effects of a change in formula
prevalence predicted from our study after factoring in on clinical practice. Application of the CKD-EPI formula
ethnicity. A smaller study from Scotland examined rather than the MDRD formula to data derived from the
changes in the numbers of patients with CKD detected NHANES study resulted in a reduction in the percentage
by the MDRD and CKD-EPI formulae in 2004 and 2009.17 of study participants with CKD stage 3 or worse from
Ethnicity was not included in this study, although the 8.2% to 6.7%.10 The major change was in CKD stage 3
population is relatively homogeneous and as such is not that fell from 7.8% to 6.3% of the population studied.
typical of that of the rest of the UK. In both years, the However, the NHANES data are also derived from
numbers of patients with CKD stages 3e5 were less with a specific study group rather than from routine clinical
the use of the CKD-EPI equation. Between the 2 years activity. Although the NHANES data are detailed, the
studied, the population from which the samples were sample number used was relatively small (16 032 partic-
taken increased by 0.5%, but the number of creatinine ipants) compared with the current study; serum creati-
measurements increased by 20%. Despite this increased nine values were re-calibrated from measurements made
sampling, the prevalence of CKD stages 3e5 in the with a kinetic rate Jaffe method. In addition, there are
population from which samples were taken rose by only well-recognised differences in a range of relevant
0.19% with the MDRD formula and was static with the parameters between the USA and UK populations
CKD-EPI formula. including body mass index and ethnic distribution. An
In participants in the Kidney Early Evaluation Program analysis of the US patients in the Kidney Early Evaluation
(KEEP) study in the USA, a change from the MDRD to Program study identified a reduction of 2.1% in the
CKD-EPI formula was associated with generally higher prevalence of CKD stage 2 or worse when the CKD-EPI
eGFRs.18 This study was not based on routine clinical formula was used.23 However, this was a targeted popu-
practice but on a screening programme targeting people lation screening study and 31.8% of the study group
at high risk of kidney disease, who were identified as were AfricaneAmericans, which is very different to the
those with hypertension, diabetes or a first-order relative ethnic distribution of the UK population.
with hypertension, diabetes or kidney disease. In this Of 11 247 individuals who were recruited to the
selected population, CKD staging based on the CKD-EPI Australian Diabetes, Obesity and Lifestyle study of adults
formula provided better prediction of mortality than over 25 years of age, 13.4% were classified as having CKD
staging based on the MDRD formula. Compared with stage 2 or worse using the MDRD formula compared
patients whose CKD stage was not altered, reclassifica- with 11.5% using the CKD-EPI formula; the prevalence of
tion to a better lower CKD stage with the CKD-EPI CKD stage 3 or worse fell from 7.8% to 5.8%.20 The CKD-
formula was associated with reduced mortality and EPI formula produces a good estimate of GFR in Japa-
reclassification to a worse higher CKD stage was associ- nese patients, and the distribution of patients with
ated with increased mortality. Similar correlations of different stages of CKD was calculated using data from
improved outcome for people who were reclassified to a Japanese annual health check programme.24 These
a better CKD stage using the CKD-EPI formula were also data are not directly comparable to our own as they are
seen for mortality, renal progression and cardiovascular based on people who are attending regularly for routine
disease in the Atherosclerosis Risk in Communities health checks, regardless of their clinical state, rather
(ARIC) study and for mortality in the Australian Dia- than clinician-directed testing. Nevertheless, the number
betes, Obesity and Lifestyle study.19 20 These studies of people classed as CKD stage 3 or worse fell with the
suggest that the more accurate estimation of eGFR use of the CKD-EPI formula from 7.7% to 5.4%. Specif-
allows more accurate prediction of prognosis. ically, CKD stage 3 fell from 7.5% to 5.2%.
A number of other small studies have examined the Our study is important because it contains an analysis
use of CKD-EPI in different contexts. The CKD-EPI of real working clinical data and thus offers a robust
formula certainly reduces bias between measured GFR analysis of the impact on the UK National Health Service
and eGFR especially in those with an eGFR of <60 ml/ of a change from the MDRD to the CKD-EPI formula.

8 OCallaghan CA, Shine B, Lasserson DS. BMJ Open 2011;1:e000308. doi:10.1136/bmjopen-2011-000308

 Healthcare implications of introducing the CKD-EPI formula for eGFR reporting

Such a change would be justifiable given the superior patients with CKD stages 4e5 will include specialist
performance of the formula in estimating GFR.10 Using referral to a nephrologist. The rise in the number of
the largest data set examined to date in this context, we patients with these stages when the CKD-EPI formula is
demonstrate that the use of the CKD-EPI formula will used could clearly lead to an increase in the workload in
bring about a substantial reduction in the number of secondary care with training and resource implications.
people who will be classified as having CKD. Scaled up to As indicated in figure 3 and online supplementary table
the UK population, this represents a reduction in the 3 (http://bmjopen.bmj.com), this rise is due exclusively
number of people labelled as having CKD using eGFR to increases in the numbers of both men and women in
criteria by 1.9 million. Of particular relevance to primary the over 70s age group, who are identified as having
care, we observed a fall in the prevalence of CKD stage 3 CKD stages 4 and 5 using the CKD-EPI formula. It will be
or worse, which would represent a reduction by around important to establish whether CKD-EPI performs better
200 000 individuals in the UK. It is likely that major as a monitoring tool for CKD progression than MDRD,
reductions in the numbers of patients with CKD will be especially in the elderly population. Up to 3 million
found in other countries with the use of the CKD-EPI people in the UK are predicted to have their CKD clas-
formula. These estimates are based on eGFR and sification altered by the change in formula, and the cost
a diagnosis of CKD stage 1 or 2 also requires the pres- of altering their healthcare records or informing them
ence of proteinuria or a structural or other abnormality. about this change would be substantial. The clinical
Current UK guidelines are that all patients with CKD implications of reclassification will vary and caution is
stage 3 or above should be on a primary care CKD certainly necessary in the interpretation of eGFRs >60
register and should be regularly monitored and in some where the inaccuracies of the formulae are greatest and
cases investigated. The current QOF provides primary other features such as proteinuria are required for CKD
care physicians in the UK with a financial incentive to do classification. In many cases, reclassification will also
this as the practice prevalence of major chronic condi- trigger a need for a review of medication choice and
tions (including CKD) contributes to the funding drug doses, many of which change with CKD. Overall,
received.9 A reduction in the incidence of CKD stage 3 the introduction of the CKD-EPI formula would generate
and above would therefore mean that there would be substantial relabelling of individuals but could ultimately
a cut in funding to primary care arising from the use of reduce the workload attributable to CKD, principally
the CKD-EPI formula, other things being equal, of around stage 3, and allow more effective targeting of
around 400 for an average practice.25 However, in evidence-based therapies to patients who will benefit
parallel with this, there may be a reduction in the work from them.
required to care for these patients who would then no
longer fall within the current NICE guidance for CKD Funding This work is underpinned by support from the National Institute for
stage 3 or above, depending on co-morbid conditions Health Research (NIHR) Oxford Comprehensive Biomedical Research Centre
and other monitoring and surveillance programmes (BRC) and the NIHR School for Primary Care Research. The NIHR BRC and
NSPCR had no role in the design and conduct of the study, data collection,
incentivised by QOF, for example, diabetes or ischaemic
management, analysis and interpretation of the data or preparation, review or
heart disease. At present, NICE guidelines recommend approval of the manuscript.
typically monitoring people with CKD stage 2 annually,
Competing interests All authors have completed the Unified Competing
stage 3 every 6 months and stage 4 every 3 months.5 Interest form at www.icmje.org/coi_disclosure.pdf (available on request from
Thus, shifts in CKD classification will have significant the corresponding author) and declare that (1) they have no support from
implications for both patients and doctors, especially in companies for the submitted work, (2) they have no relationships with
primary care where the care of most patients with CKD companies that might have an interest in the submitted work in the previous
takes place. 3 years, (3) their spouses, partners or children have no financial relationships
that may be relevant to the submitted work and (4) they have no non-financial
In principle, the CKD-EPI formula should lead to interests that may be relevant to the submitted work.
better estimates of eGFR and so greater precision in the
Ethical approval Ethical approval was not required for this study. Clinical
placement of patients into the different CKD stages,
databases were only accessed by those with permission to do so. No
which should ultimately make the classification more individual data points were examined manually, and the data were anonymised
useful by facilitating better prediction of outcome from at all stages of the analysis.
CKD staging. Although the use of the CKD-EPI formula Contributors CAOC, BS and DSL contributed to the study design and concept,
will lead to a reduction in the overall numbers of were involved in analysis and interpretation of data and preparation of the
patients with CKD stages 3e5, the number of elderly manuscript. CAOC is guarantor.
men in this category will rise leading to an increase in Provenance and peer review Not commissioned; externally peer reviewed.
the numbers of older patients on CKD registers.
Data sharing statement No additional data available.
Whether this generates additional work in primary care
will depend on whether these patients already require
REFERENCES
monitoring and input for other co-morbidities and so 1. Coresh J, Selvin E, Stevens LA, et al. Prevalence of chronic kidney
the extent to which monitoring and management of disease in the United States. JAMA 2007;298:2038e47.
2. Go AS, Chertow GM, Fan D, et al. Chronic kidney disease and the
their CKD increases overall workloads. Current UK risks of death, cardiovascular events, and hospitalization. N Engl J
guidelines emphasise that the default management of Med 2004;351:1296e305.

OCallaghan CA, Shine B, Lasserson DS. BMJ Open 2011;1:e000308. doi:10.1136/bmjopen-2011-000308 9

 Healthcare implications of introducing the CKD-EPI formula for eGFR reporting

3. Shemin D, Dworkin LD. Sodium balance in renal failure. Curr Opin 17. Gifford FJ, Methven S, Boag DE, et al. Chronic kidney disease
Nephrol Hypertens 1997;6:128e32. prevalence and secular trends in a UK population: the impact of
4. Levey AS, Greene T, Kusek J, et al. A simplified equation to predict MDRD and CKD-EPI formulae. QJM 2011;104:1045e53.
glomerular filtration rate from serum creatinine [Abstract]. J Am Soc 18. Stevens LA, Li S, Kurella Tamura M, et al. Comparison of the CKD
Nephrol 2000;11:155A. Epidemiology Collaboration (CKD-EPI) and Modification of Diet in
5. NICE. Chronic Kidney Disease. Early identification and management Renal Disease (MDRD) study equations: risk factors for and
of chronic kidney disease in adults in primary and secondary care. complications of CKD and mortality in the Kidney Early Evaluation
2008. http://www.nice.org.uk Program (KEEP). Am J Kidney Dis 2011;57(3 Suppl 2):S9e16.
6. Levey AS, Coresh J, Greene T, et al. Using standardized serum 19. Matsushita K, Selvin E, Bash LD, et al. Risk implications of the new
creatinine values in the modification of diet in renal disease study CKD Epidemiology Collaboration (CKD-EPI) equation compared
equation for estimating glomerular filtration rate. Ann Intern Med with the MDRD Study equation for estimated GFR: the
2006;145:247e54. Atherosclerosis Risk in Communities (ARIC) Study. Am J Kidney Dis
7. National Kidney Foundation. K/DOQI clinical practice guidelines for 2010;55:648e59.
chronic kidney disease: evaluation, classification, and stratification. 20. White SL, Polkinghorne KR, Atkins RC, et al. Comparison of the
Am J Kidney Dis 2002;39(2 Suppl 1):S1e266. prevalence and mortality risk of CKD in Australia using the CKD
8. Coresh J, Stevens LA, Levey AS. Chronic kidney disease is common: Epidemiology Collaboration (CKD-EPI) and Modification of Diet in
what do we do next? Nephrol Dial Transplant 2008;23:1122e5. Renal Disease (MDRD) Study GFR estimating equations: the
9. Quality and Outcomes Framework (QOF) 2007/8. http://www.ic.nhs. AusDiab (Australian Diabetes, Obesity and Lifestyle) Study. Am J
uk (accessed 2 Oct 2011). Kidney Dis 2010;55:660e70.
10. Levey AS, Stevens LA, Schmid CH, et al. A new equation to estimate 21. Stevens LA, Schmid CH, Greene T, et al. Comparative performance
glomerular filtration rate. Ann Intern Med 2009;150:604e12. of the CKD Epidemiology Collaboration (CKD-EPI) and the
11. Becker BN, Vassalotti JA. A software upgrade: CKD testing in 2010. Modification of Diet in Renal Disease (MDRD) Study equations for
Am J Kidney Dis 2010;55:8e10. estimating GFR levels above 60 mL/min/1.73 m2. Am J Kidney Dis
12. Office for National Statistics. United Kingdom Population Estimates - 2010;56:486e95.
mid-2009. London: HMSO, 2010. 22. Delanaye P, Cavalier E, Mariat C, et al. MDRD or CKD-EPI study
13. The NHS Information Centre for Health and Social Care. http://www. equations for estimating prevalence of stage 3 CKD in
ic.nhs.uk/statistics-and-data-collections/audits-and-performance/the- epidemiological studies: which difference? Is this difference relevant?
quality-and-outcomes-framework (accessed 17 May 2011). BMC Nephrol 2010;11:8.
14. Stevens LA, Coresh J, Feldman HI, et al. Evaluation of the 23. McFarlane SI, McCullough PA, Sowers JR, et al. Comparison of the
modification of diet in renal disease study equation in a large diverse CKD Epidemiology Collaboration (CKD-EPI) and Modification of Diet
population. J Am Soc Nephrol 2007;18:2749e57. in Renal Disease (MDRD) study equations: prevalence of and risk
15. Stevens LA, Claybon MA, Schmid CH, et al. Evaluation of the Chronic factors for diabetes mellitus in CKD in the Kidney Early Evaluation
Kidney Disease Epidemiology Collaboration equation for estimating Program (KEEP). Am J Kidney Dis 2011;57(3 Suppl 2):S24e31.
the glomerular filtration rate in multiple ethnicities. Kidney Int 24. Horio M, Imai E, Yasuda Y, et al. Modification of the CKD
2011;79:555e62. epidemiology collaboration (CKD-EPI) equation for Japanese:
16. Carter JL, Stevens PE, Irving JE, et al. Estimating glomerular filtration accuracy and use for population estimates. Am J Kidney Dis
rate: comparison of the CKD-EPI and MDRD equations in a large UK 2010;56:32e8.
cohort with particular emphasis on the effect of age. QJM 25. NHS Employers. Quality and Outcomes Framework guidance for
2011;104:839e47. GMS contract 2011/12. London: NHS Employers, 2011.

10 OCallaghan CA, Shine B, Lasserson DS. BMJ Open 2011;1:e000308. doi:10.1136/bmjopen-2011-000308