albuminuria: time to focus on accuracy

4
Editorial Albuminuria: Time to Focus on Accuracy Related Articles, p. 405 and p. 415 A lbuminuria is common in chronic kidney disease (CKD). It is often the earliest marker of kidney damage and, in many circumstances, precedes any decline in glomerular ltration rate (GFR). The recent Kidney Disease: Improving Global Outcomes (KDI- GO) guideline for the Evaluation and Management of Chronic Kidney Disease revised the classication system for CKD to include both albuminuria level and GFR for staging the severity of CKD. 1 This revi- sion was based in part on data from large scale meta-analyses of studies of people with and without CKD that demonstrated the prognostic importance of albuminuria for kidney disease outcomes as well as for cardiovascular disease and mortality. 2-5 The level of albuminuria is also relevant for management decisions. Patients with elevated levels of albuminuria may benet from lower blood pressure targets and medications that block the renin angiotensin system, and in patients with nephrotic syndrome, physicians initiate therapy and monitor its response by the level of albuminuria or proteinuria. 6 Successful implementation of these guidelines into clinical practice requires a simple and accurate way to measure albuminuria in routine care. The gold stan- dard measure of albuminuria is albumin excretion rate (AER) in 24-hour urine specimens, but 24-hour urine collections samples are cumbersome and prone to errors. Thus, spot urine samples are recommended to quantify albuminuria; these adjust the measured urine albumin level for the urine creatinine level to account for differences in urine concentration or dilution. 7-9 The use of this simple equation (urine albumin- creatinine ratio [ACR] 5 urine albumin in mg/dL divided by urine creatinine in g/dL) is based on the assumption that creatinine excretion rate (CER) is 1 gram per day. Accordingly, an ACR of 1,000 mg/g would translate to an AER of 1,000 mg/d. However, CER varies substantially among individuals, so ACR is only a rough approximation of AER. Strategies to improve the accuracy of assessing albuminuria from measurements in spot samples could enhance its utility in both clinical practice and in research studies. The same issues apply to assessing protein- uria from measurements in spot samples. The studies by Fotheringham et al and Abdelmalek et al in this issue of AJKD, both of which propose methods to estimate AER from ACR, are timely and open the eld to this important discussion. 10,11 Both studies address the question in similar ways. They both focus on that fact that variation in CER, which is largely a function of creatinine generation by muscle and diet, is the major source of error when using ACR to assess albuminuria. Accordingly, both studies calculate estimated CER (eCER) from pre- diction equations and then multiply that value by the ACR to compute an estimated AER (eAER). Across the 2 studies, 3 eCER equations (Box 1) are used: one developed in the Modication of Diet in Renal Dis- ease (MDRD) Study population (its development is described by Fotheringham et al; it is referred to by Abdelmalek et al as eCER Ellam and referred to here as eCER MDRD ), one developed in a pooled dataset of several CKD populations (referred to by Abdelmalek et al as eCER Ix and referred to here as eCER CKD-EPI ), and one previously developed and used in clinical practice (eCER Walser ). 10-12 The equations differ slightly but are all based on demographic factors related to creatinine generation, such as sex, race, age, and weight. In particular, eCER MDRD does not use weight, whereas the other 2 equations do. Figure 1A shows the difference in eAER by level of measured ACR for white women and men of varying weights. Regardless of the equation used, for women of average weight, differences between ACR and eAER are relatively small. In contrast, for women at extremes of weight, eCER CKD-EPI and eCER Walser provide a substantially different eAER for any given ACR than eCER MDRD , and these differences are magnied at higher ACR levels. For men, eAER and ACR are similar for eCER Walser and eCER CKD-EPI at low weight and lower levels of ACR, but all equations give a higher eAER than ACR at both average and higher body weight, particularly at higher levels of ACR. Similar patterns were seen for black men and women. Thus using ACR and eCER from different equations will provide different eAERs, particularly at higher levels of ACR and for certain subgroups. Fotheringham and colleagues evaluated 2 of the equations (eCER MDRD and eCER CKD-EPI ) in 2 sepa- rate populations: CRIC (Chronic Renal Insufciency Cohort), which includes individuals with CKD, and DCCT (Diabetes Control and Complications Trial), which includes individuals with diabetes who pri- marily do not have CKD. Abdelmalek and colleagues applied all 3 equations to a general population cohort, the PREVEND (Prevention of Renal and Vascular End-Stage Disease) Study. All 3 cohorts had fairly Address correspondence to Lesley A. Inker, MD, Tufts Medical Center, 800 Washington St, Box 391, Boston, MA 02111. E-mail: [email protected] Ó 2014 by the National Kidney Foundation, Inc. 0272-6386/$36.00 http://dx.doi.org/10.1053/j.ajkd.2014.01.002 378 Am J Kidney Dis. 2014;63(3):378-381

Upload: lesley-a

Post on 30-Dec-2016

212 views

Category:

Documents


0 download

TRANSCRIPT

Editorial

Albuminuria: Time to Focus on Accuracy

Related Articles, p. 405 and p. 415

Albuminuria is common in chronic kidney disease(CKD). It is often the earliest marker of kidney

damage and, in many circumstances, precedes anydecline in glomerular filtration rate (GFR). The recentKidney Disease: Improving Global Outcomes (KDI-GO) guideline for the Evaluation and Managementof Chronic Kidney Disease revised the classificationsystem for CKD to include both albuminuria leveland GFR for staging the severity of CKD.1 This revi-sion was based in part on data from large scalemeta-analyses of studies of people with and withoutCKD that demonstrated the prognostic importanceof albuminuria for kidney disease outcomes as wellas for cardiovascular disease and mortality.2-5 Thelevel of albuminuria is also relevant for managementdecisions. Patients with elevated levels of albuminuriamay benefit from lower blood pressure targets andmedications that block the renin angiotensin system,and in patients with nephrotic syndrome, physiciansinitiate therapy and monitor its response by the levelof albuminuria or proteinuria.6

Successful implementation of these guidelines intoclinical practice requires a simple and accurate way tomeasure albuminuria in routine care. The gold stan-dard measure of albuminuria is albumin excretion rate(AER) in 24-hour urine specimens, but 24-hour urinecollections samples are cumbersome and prone toerrors. Thus, spot urine samples are recommended toquantify albuminuria; these adjust the measured urinealbumin level for the urine creatinine level to accountfor differences in urine concentration or dilution.7-9

The use of this simple equation (urine albumin-creatinine ratio [ACR] 5 urine albumin in mg/dLdivided by urine creatinine in g/dL) is based on theassumption that creatinine excretion rate (CER) is1 gram per day. Accordingly, an ACR of 1,000 mg/gwould translate to an AER of 1,000 mg/d. However,CER varies substantially among individuals, so ACRis only a rough approximation of AER. Strategiesto improve the accuracy of assessing albuminuriafrom measurements in spot samples could enhanceits utility in both clinical practice and in researchstudies. The same issues apply to assessing protein-uria from measurements in spot samples. The studies

Address correspondence to Lesley A. Inker, MD, Tufts MedicalCenter, 800 Washington St, Box 391, Boston, MA 02111. E-mail:[email protected]� 2014 by the National Kidney Foundation, Inc.0272-6386/$36.00http://dx.doi.org/10.1053/j.ajkd.2014.01.002

378

by Fotheringham et al and Abdelmalek et al in thisissue of AJKD, both of which propose methods toestimate AER from ACR, are timely and open thefield to this important discussion.10,11

Both studies address the question in similar ways.They both focus on that fact that variation in CER,which is largely a function of creatinine generation bymuscle and diet, is the major source of error whenusing ACR to assess albuminuria. Accordingly, bothstudies calculate estimated CER (eCER) from pre-diction equations and then multiply that value by theACR to compute an estimated AER (eAER). Acrossthe 2 studies, 3 eCER equations (Box 1) are used: onedeveloped in the Modification of Diet in Renal Dis-ease (MDRD) Study population (its development isdescribed by Fotheringham et al; it is referred to byAbdelmalek et al as eCEREllam and referred to here aseCERMDRD), one developed in a pooled dataset ofseveral CKD populations (referred to by Abdelmaleket al as eCERIx and referred to here as eCERCKD-EPI),and one previously developed and used in clinicalpractice (eCERWalser).

10-12 The equations differ slightlybut are all based on demographic factors related tocreatinine generation, such as sex, race, age, andweight. In particular, eCERMDRD does not use weight,whereas the other 2 equations do. Figure 1A shows thedifference in eAER by level of measured ACR forwhite women and men of varying weights. Regardlessof the equation used, for women of average weight,differences between ACR and eAER are relativelysmall. In contrast, for women at extremes of weight,eCERCKD-EPI and eCERWalser provide a substantiallydifferent eAER for any given ACR than eCERMDRD,and these differences are magnified at higher ACRlevels. For men, eAER and ACR are similar foreCERWalser and eCERCKD-EPI at low weight and lowerlevels of ACR, but all equations give a higher eAERthan ACR at both average and higher body weight,particularly at higher levels of ACR. Similar patternswere seen for black men and women. Thus using ACRand eCER from different equations will providedifferent eAERs, particularly at higher levels of ACRand for certain subgroups.Fotheringham and colleagues evaluated 2 of the

equations (eCERMDRD and eCERCKD-EPI) in 2 sepa-rate populations: CRIC (Chronic Renal InsufficiencyCohort), which includes individuals with CKD, andDCCT (Diabetes Control and Complications Trial),which includes individuals with diabetes who pri-marily do not have CKD. Abdelmalek and colleaguesapplied all 3 equations to a general population cohort,the PREVEND (Prevention of Renal and VascularEnd-Stage Disease) Study. All 3 cohorts had fairly

Am J Kidney Dis. 2014;63(3):378-381

Box 1. Equations for the Estimation of Creatinine Excretion Rate

Walser Equation for Estimated Creatinine Excretion

Rate

eCERWalser (mg/d) 5Male: (28.2 2 0.1723 age)3 weight (kg)

Female: (21.92 0.1153 age)3 weight (kg)

MDRD Study (Ellam) Equation for Estimated Creatinine

Excretion Rate

eCERMDRD (mg/d) 5Male/black: 1413.91 (23.2 3 age) 2 (0.33 age2)

Female/black: 1148.61 (15.63 age) 2 (0.33 age2)

Male/nonblack: 1307.31 (23.1 3 age) 2 (0.33 age2)

Female/nonblack: 1051.31 (5.33 age) 2 (0.1 3 age2)

CKD-EPI (Ix) Equation for Estimated Creatinine Excre-

tion Rate

eCERCKD-EPI (mg/d) 5879.891 12.513 [weight (kg) 2 6.19]3 age

1 34.51 (if black) 2 379.42 (if female)

Editorial

low levels of albuminuria (range of median AER,7-64 mg/d). Studies evaluating prediction equationsfocus on 3 characteristics: bias, precision, and accu-racy. Bias refers to a systematic difference betweenthe estimated and measured values, precision is thespread or magnitude of the differences, and accuracyrefers to the combination of bias and precision.Across all 3 populations, we can make 2 generalconclusions about the performance of the eAERequation compared to measured AER. First, the eAERprovides more accurate estimates than ACR alone.Second, eAER is unbiased compared to measuredAER. However, accuracy, defined as the percentageof eAER values that differ by less than 30% frommeasured AER, varies across the populations (Fig 1).If the equations are unbiased and accuracy is variable,then this implies variable precision. There are poten-tially important differences in the gold standard andtest sample used across these studies that maypartially explain this variation. ACR was calculated inboth CRIC and DCCT from 24-hour urine samples;accordingly, these studies may have seen better per-formance of eAER compared to measured AER thanwould have been found if spot samples had beenused. In PREVEND, the 24-hour urine collections andthe spot urine sample were not simultaneous, and truebiological variation may partially explain some of thedifference between eAER and measured AER in thispopulation.My overall interpretation is that all 3 eAER equations

are imprecise and therefore will be inaccurate whenapplied to at least some populations. In addition, thereappears to be insufficient data to conclude whether oneequation is substantially better than the others, al-though in the study by Abdelmalek et al, it appears thateCER equations that include weight (eCERCKD-EPI andeCERWalser) may be slightly more accurate than those

Am J Kidney Dis. 2014;63(3):378-381

do not include weight (eCERMDRD) at higher levels ofweight. It would have been helpful to see these data inthe other 2 populations as well.The cause of imprecision could be error in the gold

standard (measured AER), the estimating equation, orthe urine albumin or creatinine assay.13 Errors both inthe gold standard used to develop the CER equation aswell as errors in the gold standard used to evaluate theeAER are relevant. Both gold standards are derivedfrom values determined from 24-hour urine collec-tions, which, as mentioned above, are prone to errordespite investigators’ best efforts. One importantcause of imprecision in the estimating equations isindividual variation in determinants of urine creatinineexcretion that are not fully accounted for by the vari-ables included in the equations; this could lead tobias in subgroups of the population and imprecisionoverall. Indeed, the higher accuracy in the DCCTpopulation may suggest that these equations havereasonable accuracy in people with average weightand low levels of albuminuria. Finally, differences increatinine assays between those used to develop theequation and those used to measure the ACR couldpotentially lead to error. However, this is more likelyto lead to bias or systematic error than imprecision andis probably not a major factor in these studies. Sincealbumin was measured for the AER and ACR in thesame lab and at approximately the same time, differ-ences in albumin assays are not likely to be factors.What should we do now? First, the concept of

estimating AER is sound and it is reasonable for cli-nicians to incorporate this process into clinical prac-tice. In so doing, clinicians should understand thatthere remains much inaccuracy in the eAER, espe-cially in particular populations. If we use the exampleof estimated GFR to guide us, then the next questionthat should be asked is whether one of these equationsshould be reported by clinical laboratories whenevera spot urine albumin is requested. In my opinion, thisseems premature given the variable accuracy andparticularly that there are no data at the higher albu-minuria range; however, I do advocate a well-designed study to develop an equation suitable forgeneral use and laboratory reporting.The ideal study design for the development of an

estimating equation for AER should follow similarrecommendations for the development of GFR esti-mating equations.13,14 First, it is critical to start withan accurate gold standard measurement. Given theerrors in 24-hour urine measurements, the bestapproach would be to collect the urine under directobservation for 24 hours, repeat the collection, eval-uate any large discrepancies between the collections,and consider the average of these 2 measurements tobe the gold standard value. Second, given that the goalis to use these equations in populations that either

379

Figure 1. A) Differences in esti-mated albumin excretion rate (eAER)by level of albumin-creatinine ratio(ACR) for the CERCKD-EPI, CERMDRD,and CERWalser equations for a70-year-old white woman (top) andman (bottom) at 3 different weights(40, 70, or 200 kg). Note that CERMDRD

only has one line as the estimatesdo not vary by weight. The thin blackdiagonal line is the line of identity.Variables in CERCKD-EPI includeage, sex, race, and weight. Variablesin CERMDRD include age, sex, andrace. Variables in CERWalser includeage, sex, and weight. B) Percentageof eAERs within 30% of mea-sured AER. Abbreviations: CKD-EPI,Chronic Kidney Disease Epidemi-ology Collaboration; CRIC, ChronicRenal Insufficiency Cohort; DCCT,Diabetes, Complications and ControlTrial; eCER, estimated creatinineexcretion rate; MDRD, Modificationof Diet in Renal Disease Study; NR,not reported; PREVEND, Preventionof Renal and Vascular End-StageDisease.

Lesley A. Inker

have CKD or are at risk for CKD, representativesfrom diverse populations should be included in thedevelopment population. Third, careful considerationshould be given to the variables used and their formsin developing the equation. Neither the paper byFotheringham et al nor the study by Abdelmaleket al shows results by ACR level or creatinine ex-cretion, but, reflecting that there may be differentrelationships by sex and weight between eAER andACR (Fig 1), it is possible that equations that incor-porate more complex variable forms and interactionsamong variables will better capture the true relation-ship. Fourth, evaluation of equation performanceshould be performed using spot samples and 24-hourcollections conducted on the same day and should beperformed in separate validation populations thanthose from which the equations were derived. Finally,use of an assay in both the gold standard and test

380

samples should be traceable to higher level referencematerials.In conclusion, Fotheringham and colleagues and

Abdelmalek and colleagues should be congratu-lated for bringing this important topic to the forefront.Improved tools to estimate AER from spot urinesamples may help us to prognosticate more efficiently,develop better treatments, and make better treatmentdecisions.

Lesley A. Inker, MDTufts Medical Center

Boston, Massachusetts

ACKNOWLEDGEMENTSSupport: None.Financial Disclosure: Dr. Inker’s institution receives funding

from Pharmalink AB.

Am J Kidney Dis. 2014;63(3):378-381

Editorial

REFERENCES1. Kidney Disease: Improving Global Outcomes (KDIGO)

CKD Work Group. KDIGO 2012 clinical practice guideline forthe evaluation and management of chronic kidney disease. KidneyInt Suppl. 2013;3(1):1-150.

2. Astor BC, Matsushita K, Gansevoort RT, et al. Lowerestimated glomerular filtration rate and higher albuminuria areassociated with mortality and end-stage renal disease. A collabo-rative meta-analysis of kidney disease population cohorts. KidneyInt. 2011;79(12):1331-1340.

3. Matsushita K, van der Velde M, Astor BC, et al. Associationof estimated glomerular filtration rate and albuminuria with all-cause and cardiovascular mortality in general population cohorts:a collaborative meta-analysis. Lancet. 2010;375(9731):2073-2081.

4. van der Velde M, Matsushita K, Coresh J, et al. Lowerestimated glomerular filtration rate and higher albuminuria areassociated with all-cause and cardiovascular mortality. A collab-orative meta-analysis of high-risk population cohorts. Kidney Int.2011;79(12):1341-1352.

5. Gansevoort RT, Matsushita K, van der Velde M, et al.Lower estimated GFR and higher albuminuria are associated withadverse kidney outcomes in both general and high-risk pop-ulations. A collaborative meta-analysis of general and high-riskpopulation cohorts. Kidney Int. 2011;80(1):93-104.

6. Kidney Disease: Improving Global Outcomes (KDIGO)Glomerulonephritis Work Group. KDIGO clinical practice guide-line for glomerulonephritis. Kidney Int Suppl. 2012;2:139-274.

7. Ralston SH, Caine N, Richards I, O’Reilly D, Sturrock RD,Capell HA. Screening for proteinuria in a rheumatology clinic:comparison of dipstick testing, 24 hour urine quantitative protein,

Am J Kidney Dis. 2014;63(3):378-381

and protein/creatinine ratio in random urine samples. Ann RheumDis. 1988;47(9):759-763.

8. Claudi T, Cooper JG. Comparison of urinary albuminexcretion rate in overnight urine and albumin creatinine ratio inspot urine in diabetic patients in general practice. Scand J PrimHealth Care. 2001;19(4):247-248.

9. Ginsberg JM, Chang BS, Matarese RA, Garella S. Use ofsingle voided urine samples to estimate quantitative proteinuria.N Engl J Med. 1983;309(25):1543-1546.

10. Fotheringham J, Campbell MJ, Fogarty DG, El Nahas M,Ellam T. Estimated albumin excretion rate versus urine albumin-creatinine ratio for the estimation of measured albumin excretionrate: derivation and validation of an estimated albumin excretionrate equation. Am J Kidney Dis. 2014;63(3):405-414.

11. Abdelmalek JA, Gansevoort RT, Lambers Heerspink HJ,Ix JH, Rifkin DE. Estimated albumin excretion rate versus urinealbumin-creatinine ratio for the assessment of albuminuria: adiagnostic test study from the Prevention of Renal and VascularEndstage Disease (PREVEND) Study. Am J Kidney Dis. 2014;63(3):415-421.

12. Ix JH, Wassel CL, Stevens LA, et al. Equations to estimatecreatinine excretion rate: the CKD epidemiology collaboration.Clin J Am Soc Nephrol. 2011;6(1):184-191.

13. Stevens LA, Zhang Y, Schmid CH. Evaluating the per-formance of equations for estimating glomerular filtration rate.J Nephrol. 2008;21(6):797-807.

14. Earley A, Miskulin D, Lamb EJ, Levey AS, Uhlig K. Esti-mating equations for glomerular filtration rate in the era of creatininestandardization: a systematic review. Ann Intern Med. 2012;156(11):785-795.

381