am. j. epidemiol. 2008 lambers heerspink 897 905

7/27/2019 Am. J. Epidemiol. 2008 Lambers Heerspink 897 905

1/9

American Journal of Epidemiology

The Author 2008. Published by the Johns Hopkins Bloomberg School of Public Health.

All rights reserved. For permissions, please e-mail: [email protected].

Vol. 168, No. 8

DOI: 10.1093/aje/kwn209

Advance Access publication September 5, 2008

Practice of Epidemiology

Albuminuria Assessed From First-Morning-Void Urine Samples Versus 24-Hour

Urine Collections as a Predictor of Cardiovascular Morbidity and Mortality

Hiddo J. Lambers Heerspink, Auke H. Brantsma, Dick de Zeeuw, Stephan J. L. Bakker, Paul E. deJong, and Ron T. Gansevoort for the PREVEND Study Group

Received for publication April 8, 2008; accepted for publication June 13, 2008.

Screening for albuminuria has been advocated because it is associated with cardiovascular morbidity and all-

cause mortality. The gold standard to assess albuminuria is 24-hour urinary albumin excretion (UAE). Because24-hour urine collection is cumbersome, guidelines suggest measuring albuminuria in a first morning void, either as

urinary albumin concentration (UAC) or adjusted for creatinine concentration, the albumin:creatinine ratio (ACR).

To decide which albuminuria measure to use in clinical practice, it is essential to know which best predicts clinical

outcome. In a sample representative of the Groningen (the Netherlands) population (n 3,414), the authors

compared UAC, ACR, and UAE as predictors of cardiovascular events and all-cause mortality. During a median

follow-up of 7.5 years, which ended December 31, 2005, they observed 278 events (a major adverse cardiovas-

cular event or mortality). The area under the receiver operating characteristic curve predicting events was 0.65 for

UAE, 0.62 for UAC (P 0.06 vs. UAE), and 0.66 for ACR (P 0.80 vs. UAE; P 0.01 vs. UAC). When sex-

specific subgroups were considered, UAE was superior to UAC in predicting outcome (P 0.04) for females,

whereas, for males as well as females, no difference was found between ACR and UAE. To predict cardiovascular

morbidity and all-cause mortality, measuring ACR in a first-morning-void urine sample is a good alternative to

measuring 24-hour UAE.

age groups; albuminuria; cardiovascular diseases; creatinine

Abbreviations: ACR, albumin:creatinine ratio; ICD-9, International Classification of Diseases, Ninth Revision; PREVEND,

Prevention of REnal and Vascular End-stage Disease; ROC, receiver operating characteristic.

Various large cohort studies have shown that microalbu-minuria is a strong risk predictor for cardiovascular morbid-ity and all-cause mortality. Recent epidemiologic studieshave demonstrated that even a small increase in urinaryalbumin excretion is associated with increased risk of car-

diovascular morbidity and mortality (13). This associationholds true for patients with diabetes and hypertension, andeven in the general population (15). Screening for highurinary albumin levels has therefore been advocated.

Because urinary albumin excretion follows a circadianrhythm, the preferred method to collect urine for albuminassessment is to do so for 24 hours. However, 24-hour urinecollection is inconvenient for patients. Easier and morepractical alternatives have been proposed (6). These alter-

natives include measurement of urinary albumin concentra-tion or albumin:creatinine ratio (ACR) in a first morningvoid (Table 1). Guidelines for estimating urinary albuminlevels indeed propose using the ACR (7).

The rationale for using these alternative albuminuria

measures is based on cross-sectional comparison studiesshowing good correlations of urinary albumin concentrationand ACR in urine portions and urinary albumin excretion in24-hour urine collection (6, 8). However, it is unknownwhich of these 3 albuminuria measures best identifies sub-

jects at increased risk of cardiovascular morbidity or mor-tality (9). Therefore, we investigated prospectively whichalbuminuria measure serves best to predict cardiovascularmorbidity and all-cause mortality. An important issue to

Correspondence to Dr. H. J. Lambers Heerspink, Department of Clinical Pharmacology, University Medical Center Groningen, P.O. Box 30.001,

9700 RB Groningen, The Netherlands (e-mail: [email protected]).

897 Am J Epidemiol 2008;168:897905


2/9

consider is that creatinine is a waste product of musclecatabolism. Since the ACR depends not only on urinaryalbumin but also on urinary creatinine excretion, it will beaffected by gender and age because muscle mass is lower infemales than in males and decreases with age. Therefore, in

a secondary analysis, we also studied the impact of genderand age on the performance of urinary albumin excretion,urinary albumin concentration, and ACR in predicting out-come, as well as the relation between gender and age and the3 albuminuria measures.

MATERIALS AND METHODS

Population

This study was conducted among subjects who participatein the Prevention of REnal and Vascular End-stage Disease(PREVEND) study, which began in 1997. This prospective

cohort study in the city of Groningen (the Netherlands) in-vestigates the natural course of urinary albumin excretionand its relation to renal and cardiovascular disease. Detailsof the study protocol have been published elsewhere (10, 11).In summary, all inhabitants of the city of Groningen aged2875 years were sent a questionnaire and a vial to collecta first-morning-void urine sample (prescreening). Of thesesubjects, 40,856 responded (47.8%) and returned a vial toa central laboratory for urinary albumin and urinary creati-nine assessment. From these 40,856 subjects, the PREVENDcohort was selected with the aim to create a cohort enrichedfor the presence of high albuminuria. After exclusion ofpatients with type 1 diabetes mellitus (defined as requiringthe use of insulin) and pregnant females (defined by self-report), all subjects with a urinary albumin concentration of!10 mg/L (n 7,768) were invited, and 6,000 participated.Furthermore, a randomly selected control group with a uri-nary albumin concentration of300

Table 2. Characteristics of the Prescreening Cohort and the

Current Study Population, PREVEND, Groningen, the Netherlands,

19972005a

OriginalPopulation

(n5 40,856)

CurrentPopulation(n5 3,414)

Age, years 49.5 (12.8) 48.5 (12.3)

Male 45.6 45.1

Cardiovascular disease history 3.6 3.2

Hi story of myocardial infarction 3.0 2.7

History of cerebrovascularaccident

0.8 0.6

Family cardiovascular diseasehistory

30.6 32.4

Smoked during the last 5 years 42.2 41.3

Use of blood-pressure-loweringagents

11.2 8.7

Use of lipid-lowering agents 4.7 4.1

Urinary albumin concentration,mg/L

6.0 (3.79.8) 5.7 (3.49.8)

Abbreviation: PREVEND, Prevention of REnal and Vascular End-

stage Disease (a study that began in 1997).a Values are given as mean (standard deviation), percentage, or

mean (interquartile range).

898 Lambers Heerspink et al.

Am J Epidemiol 2008;168:897905


3/9

sleep) in a polyethylene tube and to store this sample at 4Cprior to sending it to the central laboratory. For the baselinesurvey, participants filled out an extended questionnaire re-garding demographics, race, cardiovascular and renal his-tory, and smoking status. Furthermore, anthropomorphicmeasurements (body weight and height) were performed,blood pressure was recorded, and blood samples were taken.

After thorough written and oral instructions on how to per-form a 24-hour urine collection, participants collected urineduring 24 hours. They were instructed to postpone urinecollection in case of a urinary tract infection or menstruationand to refrain from heavy exercise during the collectionperiod. Participants were instructed to store urine in poly-ethylene containers at 4C for a maximum of 4 days prior tothe visit. For this study, urinary albumin concentration andACR in a first-morning-void sample obtained during pre-screening were used, as well as the 24-hour urinary albuminexcretion obtained at the baseline screening. The PREVENDstudy has been approved by the local ethics committee andis performed in accordance with Declaration of Helsinkiguidelines.

Analytical methods

Urinary albumin and creatinine concentrations in freshurine samples were determined within 24 hours after deliv-ery. Urinary albumin concentration was measured with theBehring BNII analyzer (Dade Behring, Marburg, Germany)in a central laboratory by using standard laboratory meth-ods. The intraassay and interassay coefficients of variationfor urinary albumin measured by the BNII analyzer were2.2% and 2.6%, respectively, with a threshold of 2.3 mg/L.Urinary creatinine was determined by Kodak Ektachem drychemistry (Eastman Kodak, Rochester, New York), withintraassay and interassay coefficients of variation of 0.9%and 2.9%, respectively.

Outcome

We determined the combined incidence of cardiovascularmorbidity and all-cause mortality during follow-up after thebaseline screening. Mortality data are obtained from theDutch Central Bureau of Statistics. Information on hospital-ization for cardiovascular morbidity was received fromPRISMANT, the Dutch national registry of hospital dis-charge diagnoses. The validity of this database has beenshown to be good, with 84% of primary diagnoses and87% of secondary diagnoses matching the diagnoses re-corded in patients charts (13, 14). All data were codedaccording to the International Classification of Diseases,Ninth Revision (ICD-9) and the classification of interven-tions. For this study, cardiovascular outcomes were definedas the combined incidence of acute myocardial infarction(ICD-9 code 410), acute and subacute ischemic heart dis-ease (ICD-9 code 411), coronary artery bypass grafting(ICD-9 code 414) or percutaneous transluminal coronaryangioplasty, subarachnoid hemorrhage (ICD-9 code 430),intracerebral hemorrhage (ICD-9 code 431), other intracra-nial hemorrhage (ICD-9 code 432), occlusion or stenosis of

the precerebral (ICD-9 code 433) or cerebral (ICD-9 code434) arteries, other vascular interventions such as percuta-neous transluminal angioplasty or bypass grafting of aortaand peripheral vessels, and mortality. Survival time for par-ticipants was defined as the period from the date of urinecollection by the participant to the date of the first event orDecember 31, 2005 (end of follow-up). Subjects were cen-

sored if they moved to an unknown destination.

Statistical analyses

Receiver operating characteristic (ROC) curves wereused to compare the discriminative power of the 3 differentalbuminuria measures to predict cardiovascular morbidityand mortality. The ACR depends on urinary creatinine ex-cretion and thus on muscle mass. Because muscle massdiffers between the sexes and declines with age, not onlygender-specific but also age-specific cutoff values for theACR indicating abnormal values (e.g., microalbuminuria)have been proposed (1517). To investigate the effect ofgender and age on the predictive performance of the differ-ent albuminuria measures, we analyzed males and femalesseparately, as well as subjects above and below the medianvalue of age. In addition, we repeated the ROC analysis byadding age to the model according to the method proposedby Pepe (18). ROC curves do not take into account censor-ing for subjects lost to follow-up. To test whether censoringaffected our findings, we calculated time-dependent ROCcurves (19).

Subsequently, sensitivity and specificity at the clinicallyused cutoff values indicating microalbuminuria were calcu-lated for the 3 different albuminuria measures (20). Of note,sensitivity and specificity of the ACR were calculated at theguideline-recommended sex-independent cutoff value indi-

cating microalbuminuria (30 mg/g) and at the sex-specificcutoff values for microalbuminuria (17 mg/g for males and25 mg/g for females) that have been proposed (21). Sensi-tivity was defined as the proportion of subjects experiencingan event during follow-up who had urinary albumin levelsindicating microalbuminuria at the prescreening or baselinesurvey. Specificity was defined as the proportion of subjectsnot experiencing an event during follow-up who had urinaryalbumin levels below the cutoff value for microalbuminuriaat the prescreening or baseline survey. A McNemar test wasused to test statistical differences between sensitivity andspecificity among the 3 different albuminuria measures.Multivariate Cox regression analysis was applied to determinethe hazard ratioadjusted for gender, age, blood pressure,cholesterol, and smokingfor subjects with urinary albuminlevels in the microalbuminuric range compared with subjectswhose urinary albumin levels were in the normoalbuminuricrange.

To examine the relation between the 3 different albumin-uria measures and age, we divided the study population intosex-specific quintiles of age. A percentage difference at eachquintile of age, with the lowest quintile as the reference, wasdetermined for each albuminuria measure, urinary creati-nine concentration, as well as 24-hour urinary creatinineexcretion. The Wilcoxon rank test with Bonferroni correc-tion for multiple testing was applied to assess statistical

Which Albuminuria Measure Best Predicts Outcome? 899



4/9


5/9

3 albuminuria measures, Table 5). Of note, when cardiovas-cular outcomes were analyzed separately, essentially similarresultswere obtained. Finally, time-dependent ROC analysesrevealed no change in the area under the ROC curves for eachalbuminuria measure during follow-up, indicating that cen-soring did not affect our findings.

Sensitivity and specificity of the 3 albuminuria

measures at clinically used cutoff values

The sensitivity of urinary albumin excretion did not differsignificantly from that of urinary albumin concentration andACR when sex-specific cutoff values were taken into ac-count (this finding held true in the overall population and

in gender- and age-specific subgroups (Table 6)). However,sensitivity of the ACR, calculated at the sex-independentcutoff value of 30 mg/g, was significantly lower comparedwith urinary albumin excretion as well as urinary albuminconcentration. Again, this finding was true in the overallpopulation and in gender- and age-specific subgroups.

The specificity of urinary albumin excretion was signifi-

cantly higher when compared with urinary albumin concen-tration in the overall population, as well as in gender- andage-specific subgroups. In contrast, specificity of urinaryalbumin excretion was similar to that of the ACR whensex-specific cutoff values were taken into account but waslower when adopting the sex-independent cutoff value forthe ACR.

Association between age and albuminuria measures

The association between age and the 3 albuminuria mea-sures, urinary creatinine concentration, and 24-hour urinarycreatinine excretion are presented in Figure 1. As shown,24-hour urinary creatinine excretion decreased slightly with

age in both males and females. The pattern for urinary cre-atinine concentration in a 24-hour collection was similar tothat for 24-hour urinary creatinine excretion (data notshown). In contrast, urinary creatinine concentration in a firstmorning void decreased more steeply. Urinary albumin con-centration in a first morning void tended to decrease withaging, showing a relative increase at only the highest agequintiles. A different pattern was observed for 24-hour uri-nary albumin excretion, as well as for the ACR derived froma first morning void. Both albuminuria measures increasedsteadily with aging, with a more pronounced increase for theACR.

DISCUSSION

Albuminuria has been established as a valuable riskmarker for cardiovascular morbidity as well as all-cause

Table 4. Incidence of the Composite Endpoint and Individual

Componentsa

No. ofEvents

Composite endpoint 278

Individual components

Cardiac eventsb

130Cerebrovascular eventsc 36

Peripheral vascular diseased 7

Mortality 105

a Only firstcardiovascular events thatoccurred during follow-up are

presented.b Myocardial infarction (International Classification of Diseases,

Ninth Revision (ICD-9) code 410), acute and subacute ischemic heart

disease (ICD-9 code 411), coronary artery bypass grafting (ICD-9

code 414), or percutaneous transluminal coronary angioplasty.c Subarachnoid hemorrhage (ICD-9 code 430), intracerebral hem-

orrhage (ICD-9 code 431), other intracranial hemorrhage (ICD-9 code

432), or occlusion or stenosis of the precerebral (ICD-9 code 433) or

cerebral (ICD-9 code 434) arteries.d Bypass grafting of the aorta and peripheral vessels.

Table 5. Analysis of the Predictive Performance of Various Albuminuria Measures by Comparison of Area Under

the ROC Curve

No. ofSubjects

No. ofEvents

24-Hour Urine First Morning Void

UAE, mg/24 hours UAC, mg/L ACR, mg/g

Mean 95% CI Mean 95% CI Mean 95% CI

Overall 3,414 278 0.65 0.62, 0.69 0.62 0.59, 0.66 0.66 0.62, 0.70a

Subgroups

Male 1,540 185 0.64 0.59, 0.69 0.62 0.57, 0.67 0.68 0.63, 0.73a

Female 1,874 93 0.66 0.60, 0.71 0.59 0.52, 0.65b 0.66 0.60, 0.73a

Aged 47.0 years 1,710 39 0.58 0.48, 0.67 0.52 0.43, 0.61 0.52 0.41, 0.62

Aged >47.0 years 1,704 239 0.65 0.61, 0.69 0.64 0.60, 0.68 0.64 0.60, 0.69

Overall age-adjusted ROC curve 3,414 278 0.80 0.77, 0.82 0.80 0.78, 0.83 0.80 0.77, 0.83

Abbreviations: ACR, albumin:creatinine ratio; CI, confidence interval; ROC, receiveroperating characteristic; UAC,

first morning void urinary albumin concentration; UAE, 24-hour urinary albumin excretion.a P< 0.05 vs. UAE.b P< 0.05 vs. UAC.




6/9

Table 6. Sensitivity and Specificity of Various Albuminuria Measures at Clinically Used Cutoff Values Indicating Microalbuminuria in Predicting

Cardiovascular Outcome and Mortalitya

24-Hour Urine:UAE > 30

mg/24 Hours

First Morning Void

UAC> 20 mg/L

ACR> 30 mg/g

ACR> Sex-specific

Cutoff

Overall

No. 207 268 128 187

Event rate/1,000 person-years1 38.1 30.3 39.4 42.6

Event rate for nonalbuminurics/1,000 person-years1

9.8 10.0 7.9 12.1

Hazard ratio (95% CI) 1.57 (1.14, 2.17) 1.53 (1.12, 2.09) 1.59 (1.08, 2.35) 1.86 (1.33, 2.59)

Sensitivity, % 18.9 19.8 14.4b,c 22.7

Specificity, % 95.0 93.2b 96.3b,c 94.8c

Subgroups

Male

No. 124 145 72 112


Event rate for the albumin-negative group/1,000 person-years1

14.9 15.0 19.3 17.9

Hazard ratio (95% CI) 1.51 (1.03, 2.21) 1.37 (0.94, 2.00) 1.50 (0.94, 2.38) 1.78 (1.20, 2.64)

Sensitivity, % 21.7 21.6 15.6b,c 26.0


Female

No. 83 123 56 75



5.6 6.1 7.8 7.7

Hazard ratio (95% CI) 1,96 (1.08, 3.57) 2.05 (1.17, 3.62) 1.97 (0.97, 4.01) 2.27 (1.21, 4.24)

Sensitivity, % 13.2 16.1 12.0 16.0

Specificity, % 96.0 93.9b 96.8c 95.7c

Aged 47.0 years

No. 59 89 28 42



3.1 2.9 3.8 3.8

Hazard ratio (95% CI) 1.20 (0.36, 4.01) 1.50 (0.52, 4.33) 1.05 (0.14, 7.91) 1.43 (0.33, 6.11)

Sensitivity, % 10.3 12.8 6.3b,c 9.4


Aged >47.0 years

No. 148 179 100 145



17.6 17.6 19.2 20.7

Hazard ratio (95% CI) 1.63 (1.17, 2.28) 1.56 (1.12, 2.17) 1.63 (1.10, 2.43) 1.90 (1.35, 2.68)

Sensitivity, % 20.3 20.9 15.7b,c 24.9

Specificity, % 93.1 91.2b 94.5b,d 92.3c

Abbreviations: ACR, albumin:creatinine ratio; CI, confidence interval; UAC, first morning void urinary albumin concentration; UAE, 24-hour

urinary albumin excretion.a Sensitivity and specificity for the ACR are given at the sex-independent cutoff value of 30 mg/g (as advocated by guidelines) and at recently

advocated sex-specific cutoff values (17 mg/g for males and 25 mg/g for females).b P< 0.05 vs. UAE > 30 mg/24 hours.c P< 0.05 vs. UAC > 20 mg/L.d P< 0.05 vs. ACR sex-specific cutoff.




7/9

mortality. Measurement of 24-hour albumin excretion isconsidered the gold standard to assess albuminuria. Forpractical reasons, the ACR for the first morning void hasbeen advocated as the standard method to estimate albumin-uria. To our knowledge, the present study demonstrates forthe first time that performance of the ACR in a first morningvoid to predict outcome is similar to that of 24-hour urinaryalbumin excretion. In contrast, the predictive performanceof 24-hour urinary albumin excretion is superior to that ofurinary albumin concentration in a first morning void, withborderline statistical significance. Furthermore, we foundthat the predictive performance of the ACR was statisticallysignificantly higher than that for urinary albumin concentra-tion. Essentially similar results were obtained in sex-specificsubgroups and in the overall population. The differences inarea under the ROC curves were eliminated when age-adjusted ROC curves were taken into account. Lastly, weshowed that, at the clinically used cutoff value indicatingmicroalbuminuria, sensitivity and specificity of 24-hour uri-nary albumin excretion did not differ significantly from theACR in a first morning void when gender-specific cutoffvalues for the ACR were applied.

Urinary albumin concentration in urine samples is notonly dependent on the amount of albumin lost but is alsoaffected by variations in hydration status. If a subject is wellhydrated and consequently his or her urine volume is high,urinary albumin concentration will be low, and vice versa.These hydration-status-dependent variations in urinary al-bumin concentration are likely to influence the predictiveperformance. Of note, creatinine per unit of time is assumedto be fairly stable over 24 hours (22). If urinary albuminconcentration is divided by urinary creatinine concentration,it will mathematically result in a correction for intraindi-vidual variations in hydration status. Thus, the obtainedACR is expected to be better in predicting cardiovascularmorbidity and mortality. Indeed, we found that the predic-tive performance of the ACR is significantly higher than thatof albumin concentration.

In addition, our data show that the ACR-predicted out-come was even slightly better than 24-hour urinary albuminexcretion, although not statistically significant. This findingis unexpected for 2 reasons. First, because of the circadianrhythm of albuminuria, which furthermore differs betweensubjects, we anticipated that albuminuria measures in

First Morning Void:

Males

60

40

20

0

20

40

60

80

35 45 55 65 75

Age, years

UAC

UCrC

ACR

24-Hour Urine Collection:

Males

60

40

20

0

20

40

60

80

35 45 55 65 75

Age, years

Difference,

%

UAE

UCrE

ACR

First Morning Void:

Females

60

40

20

0

20

40

60

80

35 45 55 65 75

Age, years

UAC

UCrC

ACR

24-Hour Urine Collection:

Females

60

40

20

0

20

40

60

80

35 45 55 65 75

Age, years

Difference,

%

Difference,

%

Difference,

%

UAE

UCrE

ACR

Figure 1. Association between age and urinary albumin concentration (UAC), albumin:creatinine ratio (ACR), and urinary creatinine concentra-tion (UCrC) in a first morning void (left panel), and the association between age and 24-hour urinary albumin excretion (UAE), ACR, and urinarycreatinine excretion (UCrE) (right panel) among subjects participating in the Prevention of REnal and Vascular End-stage Disease (PREVEND,started in 1997) Cohort, Groningen, the Netherlands. Age quintiles for males:


8/9

first-morning-void urine samples would predict outcomeless well than 24-hour urinary albumin excretion. Second,although the ACR adjusts for hydration status, creatinineexcretion depends on muscle mass. Differences in musclemass between individuals will also affect the predictive per-formance of the ACR.

Which factors may explain why the ACR predicted out-

come even slightly better than 24-hour urinary albuminexcretion did? First, we cannot exclude the possibility thatsubjects made errors during 24-hour urine collection, whichwill negatively affect the predictive performance of 24-hoururinary albumin excretion. Similarly, the predictive perfor-mance of the ACR in the 24-hour urine samples was foundto be exactly similar to that of the ACR in the first morningvoid. Second, the ACR depends on not only urinary albuminconcentration but also urinary creatinine concentration. Wedemonstrated in a secondary analysis that, in both males andfemales, urinary creatinine concentration in a first morningvoid decreases at higher age. This finding is explained notonly by a decrease in muscle mass with aging, since 24-hoururinary creatinine excretion, a surrogate for total musclemass, decreases less with aging than urinary creatinine con-centration in a first morning void. These data indicate thatat a higher age, nighttime urinary albumin concentrationdecreases because of dilution. Several factors have beendescribed in the literature that explain the nighttime dilutionthat occurs with aging. These factors include a change indiurnal rhythm of vasopressin and atrial natriuretic peptideand an inability of the kidney to retain sodium at higher ages(2329). As mentioned above, it has been advocated to usethe ACR to correct for intraindividual variations in urinevolume. However, our data show that, because of the lossof muscle mass, the ACR in first-morning-void urine sam-ples increases with aging more steeply than 24-hour urinary

albumin excretion. These considerations may seem theoret-ical but are of clinical importance because they may explainthat the ACR performs the best in predicting cardiovascularevents and mortality. Since the ACR in a first morning voidincreases more steeply than 24-hour urinary albumin excre-tion with aging, it incorporates to a certain extent the pre-dictive value of higher age for cardiovascular morbidity andmortality. The opposite will be true for albumin concentra-tion in a first morning void, since, because of dilution, itincreases less with aging when compared with 24-hour uri-nary albumin excretion. These hypotheses are corroboratedby our observation that, after adding age to the predictivemodel, the area under the ROC curves was exactly similarfor all albuminuria measures.

Since the predictive performance of the ACR in a firstmorning void is at least similar to that of 24-hour urinaryalbumin excretion, the ACR provides a feasible alternativeto the more cumbersome collection of a 24-hour urine sam-ple. The next question to be addressed is what cutoff valuesfor the ACR indicating microalbuminuria should be used.According to treatment guidelines from diabetic and ne-phrology associations, the cutoff value for ACR indicatingmicroalbuminuria is 30 mg/g (20, 30). However, because ofdifferences between the sexes in muscle mass and hence inurinary creatinine excretion, some authors advocate the useof gender-specific cutoff values: 17 mg/g for males and

25 mg/g for females (21). In our study, application of theguideline-recommended albumin:creatinine cutoff value of30 mg/g resulted in a low sensitivity to identify subjects atrisk of cardiovascular morbidity and all-cause mortalitycompared with the sensitivity at a gender-specific cutoffvalue. As a consequence, the number of false-negative testresults is high, which indicates that a considerable propor-

tion of subjects at risk of adverse outcomes will not beidentified if a cutoff value of 30 mg/g is used. Therefore,we advocate applying gender-specific cutoff values for theACR indicating microalbuminuria: 17 mg/g for males and25 mg/g for females.

A few issues need to be considered when interpreting ourfindings. First, this study was designed to assess which al-buminuria measure can be used for initial population screen-ing for albuminuria from the perspective of early detectionof cardiovascular disease. In this sense, it is not a prerequi-site that microalbuminuria be an independent risk factor.Determination of whether the ACR kept its prognosticpower after adjustment for established cardiovascular riskmarkers was therefore not the aim of this study. Second, theabsolute values of the ROC curves may seem low in thisstudy. However, when we performed ROC curve analysesfor systolic blood pressure and cholesterol to predict cardio-vascular outcomes and all-cause mortality, the areas underthe ROC curves were 0.69 and 0.64, respectively. Theseexamples illustrate that accepted cardiovascular risk factors,such as systolic blood pressure and cholesterol, by them-selves have a modest impact on the area under the ROCcurve (31). Third, one might argue that spot urine collec-tions would have been more appropriate, since they can becollected during consultation and are more practical thanfirst-morning-void collections. The rationale to use firstmorning voids in this study was based on the KDOQI guide-

lines, advocating the use of first-morning-void collectionsover spot urine collections for testing of microalbuminuria(20). Finally, keep in mind that these results were obtainedin a predominantly nondiabetic, Caucasian cohort. Furtherresearch in other populations, including diabetic and hyper-tensive populations, should corroborate our findings.

In conclusion, measuring the ACR in a first morning voidis at least as reliable as measuring the gold standard, 24-hoururinary albumin excretion for prediction of cardiovascularmorbidity and mortality. These results are clinically impor-tant because they imply that albuminuria measured in first-morning-void collection, which is clinically more feasiblethan collecting a 24-hour urine sample, is valid for estimatingpatients risk.

ACKNOWLEDGMENTS

Author affiliations: Department of Clinical Pharmacol-ogy, University Medical Center Groningen, University ofGroningen, Groningen, The Netherlands (Hiddo J. LambersHeerspink, Dick de Zeeuw); and Division of Nephrology,University Medical Center Groningen, University ofGroningen, Groningen, The Netherlands (Auke H. Brantsma,Stephan J. L. Bakker, Paul E. de Jong, Ron T. Gansevoort).




9/9

The authors thank Dade Behring (Marlburg, Germany)for supplying equipment (Behring Nephelometer II) andreagents for nephelometric measurement of urinary albuminconcentration.

Conflict of interest: none declared.

REFERENCES

1. Hillege HL, Fidler V, Diercks GF, et al. Urinary albuminexcretion predicts cardiovascular and noncardiovascularmortality in general population. Circulation. 2002;106(14):17771782.

2. Arnlov J, Evans JC, Meigs JB, et al. Low-grade albuminuriaand incidence of cardiovascular disease events in nonhyper-tensive and nondiabetic individuals: the Framingham HeartStudy. Circulation. 2005;112(7):969975.

3. Romundstad S, Holmen J, Kvenild K, et al. Microalbuminuriaand all-cause mortality in 2,089 apparently healthy individuals:a 4.4-year follow-up study. The Nord-Trondelag Health Study(HUNT), Norway. Am J Kidney Dis. 2003;42(3):466473.

4. Klausen K, Borch-Johnsen K, Feldt-Rasmussen B, et al. Very lowlevels of microalbuminuria are associated with increased risk ofcoronary heart disease and death independently of renal function,hypertension, and diabetes. Circulation. 2004;110(1):3235.

5. Gerstein HC, Mann JF, Yi Q, et al. Albuminuria and risk ofcardiovascular events, death, and heart failure in diabetic andnondiabetic individuals. JAMA. 2001;286(4):421426.

6. Zelmanovitz T, Gross JL, Oliveira JR, et al. The receiveroperating characteristics curve in the evaluation of a randomurine specimen as a screening test for diabetic nephropathy.Diabetes Care. 1997;20(4):516519.

7. Levey AS, Eckardt KU, Tsukamoto Y, et al. Definition andclassification of chronic kidney disease: a position statementfrom Kidney Disease: Improving Global Outcomes (KDIGO).Kidney Int. 2005;67(6):20892100.

8. Jermendy G, Farkas K, Nadas J, et al. Practical aspects ofmeasuring microalbuminuria in diabetic patients. DiabetesNutr Metab. 2001;14(4):195200.

9. Gansevoort RT, Brinkman J, Bakker SJ, et al. Evaluation ofmeasures of urinary albumin excretion. Am J Epidemiol. 2006;164(8):725727.

10. Hillege HL, Janssen WM, Bak AA, et al. Microalbuminuria iscommon, also in a nondiabetic, nonhypertensive population,and an independent indicator of cardiovascular risk factors andcardiovascular morbidity. J Intern Med. 2001;249(6):519526.

11. Pinto-Sietsma SJ, Janssen WM, Hillege HL, et al. Urinaryalbumin excretion is associated with renal functional abnor-malities in a nondiabetic population. J Am Soc Nephrol. 2000;11(10):18821888.

12. Gansevoort RT, Verhave JC, Hillege HL, et al. The validity of

screening based on spot morning urine samples to detectsubjects with microalbuminuria in the general population.Kidney Int Suppl. 2005;(94):S28S35.

13. PRISMANT. Investigating the reliability of the nationalmedical registration [in Dutch]. Utrecht, the Netherlands:PRISMANT; 2002.

14. Stricker BH, Herings RM. Plea for the retention of the DutchNational Medical Registration (LMR) to provide reliable in-formation regarding public health and healthcare. Ned TijdschrGeneeskd. 2006;150(35):19161917.

15. Jacobs DR Jr, Murtaugh MA, Steffes M, et al. Gender- andrace-specific determination of albumin excretion rate usingalbumin-to-creatinine ratio in single, untimed urine speci-mens: the Coronary Artery Risk Development in Young Adults

Study. Am J Epidemiol. 2002;155(12):11141119.16. Mattix HJ, Hsu CY, Shaykevich S, et al. Use of the albumin/

creatinine ratio to detect microalbuminuria: implications ofsex and race. J Am Soc Nephrol. 2002;13(4):10341039.

17. Verhave JC, Hillege HL, De ZD, et al. How to measure theprevalence of microalbuminuria in relation to age and gender?Am J Kidney Dis. 2002;40(2):436437.

18. Pepe MS. Regression modelling framework for receiver op-erating characteristic curves in medical diagnostic testing.Biometrika. 1997;84:595608.

19. Heagerty PJ, Zheng Y. Survival model predictive accuracy andROC curves. Biometrics. 2005;61(1):92105.

20. National Kidney Foundation. KDOQI Clinical PracticeGuidelines and Clinical Practice Recommendations for Dia-betes and Chronic Kidney Disease. Am J Kidney Dis. 2007;49(2 suppl 2):S12S154.

21. Warram JH, Gearin G, Laffel L, et al. Effect of duration of typeI diabetes on the prevalence of stages of diabetic nephropathydefined by urinary albumin/creatinine ratio. J Am Soc Nephrol.1996;7(6):930937.

22. Borst JG, de Vries LA. The three types of natural diuresis.Lancet. 1950;2(6618):16.

23. Kirkland JL, Lye M, Levy DW, et al. Patterns of urine flowand electrolyte excretion in healthy elderly people. Br Med J(Clin Res Ed). 1983;287(6406):16651667.

24. Weiss JP, Blaivas JG, Stember DS, et al. Nocturia in adults:etiology and classification. Neurourol Urodyn. 1998;17(5):467472.

25. Natsume O. A clinical investigation of nocturnal polyuria inpatients with nocturia: a diurnal variation in arginine vaso-pressin secretion and its relevance to mean blood pressure.J Urol. 2006;176(2):660664.

26. Asplund R. Diuresis pattern, plasma vasopressin and bloodpressure in healthy elderly persons with nocturia and nocturnalpolyuria. Neth J Med. 2002;60(7):276280.

27. Matthiesen TB, Rittig S, Norgaard JP, et al. Nocturnal polyuriaand natriuresis in male patients with nocturia and lower uri-nary tract symptoms. J Urol. 1996;156(4):12921299.

28. Cugini P, Lucia P, Di Palma L, et al. Effect of aging oncircadian rhythm of atrial natriuretic peptide, plasma reninactivity, and plasma aldosterone. J Gerontol. 1992;47(6):B214B219.

29. Ohashi M, Fujio N, Nawata H, et al. High plasma concentra-tions of human atrial natriuretic polypeptide in aged men.

J Clin Endocrinol Metab. 1987;64(1):8185.30. American Diabetes Association. Standards of medical care indiabetes2007. Diabetes Care. 2007;30(suppl 1):S4S41.

31. Cook NR. Use and misuse of the receiver operating charac-teristic curve in risk prediction. Circulation. 2007;115(7):928935.



am. j. epidemiol. 2008 lambers heerspink 897 905

Documents