AKI is common in children undergoing cardiac surgery (CS) and is associated with prolonged length of stay (LOS) and mechanical ventilation (1,2). In noncardiac pediatric intensive care unit (ICU) patients, AKI is independently associated with mortality (3–5). Previous AKI therapeutic trials have been negative, in part, due to a delayed rise in serum creatinine (SCr), missing the therapeutic window for mitigating AKI (6). In search of early AKI biomarkers, we found that postoperative urine neutrophil gelatinase-associated lipocalin (NGAL), IL-18, and serum cystatin C (CysC), obtained within hours after CS, predict AKI before SCr rise (7,8).
Postoperative urine albumin is a promising early AKI biomarker, reflecting tubular or glomerular damage (9–12). Preoperative albuminuria predicts AKI in adults (13,14). In children, evaluating albuminuria as an AKI risk factor is complicated by variable protein excretion, which is physiologically higher in younger children (15–17).
Using a multicenter pediatric CS cohort, we determined if the preoperative urine albumin/creatinine ratio (ACR) and/or early postoperative ACR predicts CS-AKI development, and if preoperative and postoperative ACRs provide additional benefit for predicting CS-AKI above what is feasible using other early AKI biomarkers.
Materials and Methods
Study Design, Setting, and Patient Selection
This was a multicenter prospective cohort study of the Translational Research Investigating Biomarker Endpoints in Acute Kidney Injury consortium study (TRIBE-AKI) conducted at Yale New Haven Hospital (New Haven, CT), Cincinnati Children's Hospital Medical Center (Cincinnati, OH), and Montreal Children's Hospital (Montreal, Canada). Children aged ≤18 years undergoing CS between 2007 and 2009 were eligible. Children with a renal transplant or receiving chronic dialysis were excluded. Infants aged <4 weeks were excluded from analysis due to unique characteristics of neonates with regards to proteinuria (16). We enriched our cohort with higher risk patients by favoring enrollment of children undergoing risk adjustment for congenital heart surgery-1 (RACHS-1) category ≥2 surgeries. Children were recruited preoperatively and followed postoperatively. The original study goal was to validate novel AKI biomarkers such as urine NGAL, IL-18 (7), and serum CysC (8). This study was a secondary analysis of urinary ACR. Each institution's research ethics boards approved the study. Informed consent was obtained from patients or parents/guardians before participation, with assent where appropriate.
Brief Protocol Description
The study protocol has been previously reported (1,7). We recorded demographic data and medical history and collected blood and urine preoperatively for SCr, CysC, and urine ACR measurement. We used the Schwartz equation (18) to estimate preoperative GFR (eGFR) expressed as percentiles for age (8). We recorded surgical procedure details (e.g., anatomy, procedure, bypass time). Within 6 hours of arrival to the ICU (or “first-postoperative”) and then daily for 3 days, we collected urine and blood to measure urine AKI biomarkers (NGAL and IL-18), urine ACR, SCr, and CysC. Urine samples were collected fresh from the Foley catheter urimeter, using cotton balls in patients wearing diapers or by clean catch in older children without a bladder catheter. We centrifuged biospecimens (3000×g for 10 minutes), aliquoted urine supernatant and plasma, and stored them at −80°C until analyte measurement.
Urine and Serum Biomarker Measurements
SCr was measured in local laboratories by modified Jaffe or enzymatic assays. Urine biomarkers and CysC were measured at the Cincinnati Children's Hospital Biomarker Laboratory. We measured urine albumin by immunoturbidimetry (Siemens Dimension Plus with HM clinical analyzer) per the manufacturer’s instructions and urine creatinine by modified Jaffe reaction. Urine ACR was calculated as milligrams of albumin per grams of creatinine (divided by 8.84 for SI units, mg/mmol). We measured urine NGAL and IL-18 with the ARCHITECT assay (Abbott Diagnostics, Abbott Park, IL), with coefficients of variation (CV) of 5% and 8%, respectively. We measured CysC using a nephelometer (Siemens BN-II, Siemens, AG; www.Siemens.com) (CV=1.1%). Individuals performing measurements were blind to clinical data.
ACR Threshold Definitions
We categorized participants into low-, medium-, or high-risk preoperative ACR threshold groups. Because urine albumin is physiologically higher in children aged <2 years (16), preoperative ACR thresholds differed by age group. Children aged <2 years were classified as low risk (preoperative ACR <30 mg/g [3.4 mg/mmol]), medium risk (ACR ≥30 [3.4 mg/mmol] to <300 mg/g [34 mg/mmol]), and high risk (ACR ≥300 mg/g [34 mg/mmol]) (16). In children aged ≥2 years, we based thresholds on standard microalbuminuria definitions (19,20), with low risk (ACR <10 mg/g [1.1 mg/mmol]), medium risk (ACR ≥10 [1.1 mg/mmol] to <30 mg/g [3.4 mg/mmol]), and high risk (ACR ≥30 mg/g [34 mg/mmol]).
To evaluate first postoperative ACR for early CS-AKI diagnosis, we classified patients into lower, middle, and upper tertile ACR concentrations (for each age group), because there is no standard postoperative albuminuria threshold.
We based our AKI outcomes on the Acute Kidney Injury Network (AKIN) definition (21); stage 1 AKI was a ≥50% or 0.3 mg/dl (27 µmol/L) SCr rise from preoperative value, stage 2 was doubling of SCr, and stage 3 was a tripling of SCr or receipt of dialysis during hospitalization. This study was not designed to collect hourly urine output required to define AKIN urine output criteria. We did not include the requirement for SCr to rise within 48 hours (stated in the AKIN definition); however, we have shown that most AKI occurred in the first 48 hours of ICU admission in these patients (1).
We performed analyses separately for children aged <2 and ≥2 years. We calculated unadjusted and adjusted relative risks (RRs) of preoperative ACR to predict CS-AKI with Poisson regression (22). In adjusted analyses, we controlled for our previously reported clinical AKI predictors as follows: age (continuous in the group aged <2 years; 2–<6 versus ≥6 years in group aged ≥2 years), sex, white versus nonwhite race, elective versus urgent surgery, RACHS-1 (23) score ≥3 versus <3, preoperative eGFR percentile, and study site. We calculated the area under the receiver operating characteristic curve (AUC; with SEM) to determine the ability of preoperative ACR to discriminate between patients with and without CS-AKI and determined if adding preoperative ACR to the predictive clinical model led to a statistically significant AUC rise (24). We calculated the continuous net reclassification improvement (NRI; with SEM) of adding preoperative ACR to the clinical prediction model for CS-AKI, to determine the added value of preoperative ACR for predicting AKI (25,26). A higher NRI results from individuals with AKI being assigned to having a higher AKI risk and from individuals without AKI being assigned a lower AKI risk, when the new biomarker is added to the model.
We used similar analyses to determine if tertiles of first postoperative ACR predicted CS-AKI. For these analyses, we included cardiopulmonary bypass time (CPB; <120 versus ≥120 minutes) in the clinical prediction model. We calculated diagnostic characteristics for different first postoperative ACR threshold values for AKI prediction. Because of our previous findings that first postoperative urine NGAL (7), urine IL-18 (7), and serum CysC (8) were predictive of CS-AKI, we additionally determined the extent to which adding first postoperative ACR to CS-AKI prediction models led to significant increase in the AUC for AKI prediction and to statistically significant NRI. The previously reported biomarkers were collected at the same time points as urine ACR. To evaluate if preoperative ACR affected the postoperative ACR-AKI relation, we repeated the above AUC and NRI analyses expressing postoperative ACR as the raw difference and as the percentage change in postoperative ACR relative to preoperative ACR and by including preoperative ACR in clinical prediction models.
We evaluated if first postoperative ACR (natural logarithm-transformed, not in tertiles) predicted longer hospital and ICU LOS independent of other clinical variables described above, using negative binomial multiple regression. All analyses were performed using SAS (version 9.1; SAS Institute, Cary, NC) and R software (version 2.12.1; R Foundation for Statistical Computing, Vienna, Austria).
Figure 1 displays the study flow. In the group aged <2 years, younger age was associated with higher ACR and ACR differed by study site (Table 1); in the group aged ≥2 years, participants in the high ACR category had higher RACHS-1 scores and longer ICU and hospital stays (Table 1). AKI (stage 1 or worse) first occurred on mean postoperative day 1.6 (SD 0.5) and day 1.5 (SD 1.3) in the groups aged <2 years and ≥2 years, respectively.
Association of Preoperative ACR and CS-AKI
There was little difference in CS-AKI incidence across the preoperative ACR categories in the group aged <2 years; in the group aged ≥2 years, CS-AKI was most common in the highest ACR category, but the difference was not statistically significant (Table 1). There was no statistically significant association between preoperative ACR and postoperative development of mild (stage 1 or worse) or moderate (stage 2 or worse) AKI (Table 2).
A preoperative clinical model predicted CS-AKI with AUCs of 0.68 for the group aged <2 years (SEM 0.04) and 0.77 for the group aged ≥2 years (SEM 0.04). Adding preoperative ACR to the clinical model had no statistically significant effect on AUC, with AUCs of 0.72 for the group aged <2 years (SEM 0.05) and 0.80 for the group aged ≥2 years (SEM 0.04) (P>0.05 compared with clinical model alone). The NRI of adding preoperative ACR to the clinical model was not statistically significant, being 0.21 (SEM 0.19; P=0.26) in the group aged <2 years and 0.03 (SEM 0.18; P=0.86) in the group aged ≥2 years.
Association of First Postoperative ACR and CS-AKI
In patients with and without CS-AKI, first postoperative ACR was statistically significantly higher than preoperative ACR (P<0.001, both age groups). Figure 2 displays the dramatic rise in first postoperative ACR in both age groups and in AKI and non-AKI patients, with a rapid drop in ACR by the second and third postoperative days.
In the group aged ≥2 years, the highest first postoperative ACR tertile (≥169 mg/g, 19.1 mg/mmol) predicted postoperative stage 1 or worse AKI after adjusting for clinical variables (unadjusted RR, 2.29; 95% confidence interval [95% CI], 1.17–4.49) (adjusted RR, 2.10; 95% CI, 1.08–4.08) (Table 3). For both age groups, the highest postoperative ACR tertile predicted development of stage 2 AKI or worse in unadjusted analyses (Table 3). In adjusted analyses, for the group aged <2 years, first postoperative ACR in the third tertile predicted stage 2 AKI development (adjusted RR, 3.41; 95% CI, 1.24–9.37) (Table 3). It was not possible to calculate adjusted risk for postoperative ACR to predict stage 2 AKI in the group aged ≥2 years due to low event numbers. Table 4 displays the diagnostic characteristics of first postoperative ACR thresholds for early AKI diagnosis. The ACR threshold value with maximal sensitivity and specificity in the group aged ≥2 years was almost half of the similar value in the group aged <2 years. In general, negative predictive values were high with low positive predictive values (Table 4).
Added Benefit of Postoperative ACR to Predict CS-AKI above Known Clinical and Biomarker Predictors
Table 5 displays postoperative ACR performance for predicting AKI over and above clinical data, urine (NGAL, IL-18), and serum (CysC) biomarkers. Results on stage 2 AKI or worse for the group aged ≥2 years were omitted due to low event rates. In both age groups, AUCs from the clinical predictive model and first postoperative ACR were extremely similar to the models including clinical risk factors and other biomarkers (shown in Table 5). Adding first postoperative ACR to other predictive models (clinical, clinical plus biomarkers) did not improve AUC for predicting stage 1 AKI or worse in either age group. When urine biomarkers were normalized to creatinine, adding ACR to the model including clinical and urine biomarker data were associated with a significant NRI in the group aged <2 years; however, this effect disappeared once CysC was in the model (all shown in Table 5). In the group aged ≥2 years, adding first postoperative ACR to the predictive models only led to significant NRIs for predicting stage 1 AKI or worse, when CysC was also in the model and urine biomarkers were not normalized to urine creatinine (shown in Table 5). First postoperative ACR performance for predicting AKI was not improved by incorporating preoperative ACR in any of the models (data not shown).
Association of Postoperative Urine ACR with LOS
In the group aged <2 years, first postoperative ACR predicted longer hospital LOS (adjusted P=0.02), independent of other clinical variables, but not longer ICU LOS. In children aged ≥2 years, first postoperative ACR was independently associated with longer hospital LOS (adjusted P=0.02) and ICU LOS (adjusted P=0.01).
Early postoperative, but not preoperative, ACR predicted CS-AKI in children. The utility of postoperative ACR to predict AKI and other outcomes was greater in children aged ≥2 years.
Animal and human studies have demonstrated that urine albumin rises dramatically with AKI (9–11). Albumin is normally filtered by the glomerulus, transported, and degraded in the proximal tubule. With AKI, albumin excretion may rise due to glomerular and/or renal tubular damage or albumin gene upregulation in the renal cortex or tubules (12). Albuminuria may also occur due to CPB-associated inflammation causing increased glomerular permeability (27). Although we could not evaluate inflammatory markers, when we controlled for CPB, postoperative ACR remained predictive of AKI.
The utility of postoperative ACR for AKI prediction was discrepant between the two age groups. In children aged <2 years, postoperative ACR only predicted more severe (stage 2) AKI and was less strongly associated with the LOS outcomes compared with older children. A possible explanation is that albumin excretion changes dramatically in the first 2 years of life, with renal tubular maturation (16,20). This leads to variable urine albumin excretion, making it difficult to distinguish non-AKI from mild AKI patients. These results highlight the need to consider age when studying novel urine AKI biomarkers in children.
We previously showed that early postoperative serum CysC was a good predictor of postoperative AKI development in this cohort (8); this is evidenced in Table 5, in which of all three biomarker methods evaluated (serum CysC, urine AKI biomarkers, and ACR), adding CysC to the clinical model led to the highest point-estimate rise in AUC. In children aged ≥2 years, we found a consistent tendency for urine ACR to provide added usefulness for predicting AKI (based on NRI and not AUC rise, Table 5) only when CysC was also in the predictive model. This may represent interaction between CysC and ACR, which we were unable to evaluate due to sample size limitations. Moreover, although discrepancy between results from the use of the NRI metric versus the AUC rise is known (25,26), with the NRI being more sensitive to capturing improvement in risk classification than AUC does, it must be acknowledged that at present the clinical significance of a high NRI is still somewhat unclear. Ultimately, only clinical trials that incorporate novel biomarkers in clinical care and evaluate the effects of such measures on clinical (e.g., mortality) and morbidity or cost (e.g., LOS, hospital costs) outcomes, will the added benefit of AKI biomarker measurement be known. When urine biomarkers (NGAL, IL-18) were in AKI predictive models, adding ACR provided no additional benefit. It is possible that because ACR rise occurs due to tubular injury, similar to urine biomarkers, it may not provide substantial additional predictive information. The finding that AUC for AKI prediction using the model of clinical factors and ACR was extremely similar to the model including clinical factors and urine biomarkers supports this hypothesis. Overall, our results suggest that in future trials with stage 2 AKI or worse as a study outcome, early postoperative ACR may be used to predict AKI risk in all children, but that postoperative ACR will likely only be beneficial for predicting all AKI (mild to severe) in older children. Our results reported in Table 4 provide reasonable ACR threshold values for screening patients eligible for future AKI trials or for patients in whom AKI preventive measures (e.g., fluid overload or nephrotoxin avoidance) may help. Because urine ACR is a routinely available test and ACR values already have clinical meaning to clinicians, early AKI diagnosis may be a clinically feasible goal sooner than would be with other novel biomarkers, which may require more time for acceptance by clinicians and institutions.
Preoperative ACR did not predict postoperative AKI in our cohort, which contradicts previous findings in adults (13,14). In adults, microalbuminuria is common and considered a manifestation of CKD (28). In children, CKD and thus clinically significant microalbuminuria, is uncommon, as shown in our cohort. This may explain why preoperative ACR may not distinguish patients who are at higher risk for AKI. It is possible that in children with underlying known renal disease, urine albumin may provide similar utility for AKI prediction as found in adults.
Our study had limitations. Our sample size was not large enough to effectively evaluate prediction of more severe AKI in older children. Our findings cannot be generalized to noncardiac children admitted to the ICU. It must also be acknowledged that the analysis may not have adequately accounted for residual confounding such as effects of CPB-associated inflammation on both ACR elevations and AKI. Our study also had several unique strengths. Because this was a multicenter conducted in the United States and Canada, generalizability to other populations was enhanced. We evaluated two clinically reproducible time points for urine ACR measurement: during preoperative assessment and upon arrival to the ICU from surgery. Unlike most previously published AKI biomarker studies, we provide data on the extent to which adding ACR to other known AKI predictors (clinical and biomarker risk factors) improves current AKI prediction. Finally, because urine ACR is a routine test, timely application is potentially feasible.
Early AKI diagnosis after CS will allow for future trials of AKI to be performed in children and may help to reduce complications of AKI by early avoidance of fluid overload or nephrotoxic medication in high-risk individuals. Although benefit for postoperative AKI prediction may be limited to combination with serum CysC, urine ACR provides an easy postoperative diagnostic test for AKI, particularly in older children.
P.D. is a coinventor on patents regarding NGAL as a biomarker of AKI and is a consultant to Abbott Diagnostics and Biosite Inc. These associations had no effect on the research performed in this manuscript.
The research was supported by a American Heart Association Clinical Development award, as well as grants from the National Heart, Lung, and Blood Institute (R01HL-085757) and the National Center for Research Resources (CTSA Grant Number UL1 RR024139). M.Z. was supported by salary support grants from the Kidney Research Scientist Core Education and National Training Program, the Fonds de Recherche en Santé du Québec, and the Montreal Children’s Hospital Research Institute.
Published online ahead of print. Publication date available at www.cjasn.org.
See related editorial, “We Can Diagnose AKI “Early”,” on pages 1741–1742.
1. Li S, Krawczeski CD, Zappitelli M, Devarajan P, Thiessen-Philbrook H, Coca SG, Kim RW, Parikh CRTRIBE-AKI Consortium: Incidence, risk factors, and outcomes of acute kidney injury after pediatric cardiac surgery: A prospective multicenter study. Crit Care Med 39: 1493–1499, 2011
2. Zappitelli M, Bernier PL, Saczkowski RS, Tchervenkov CI, Gottesman R, Dancea A, Hyder A, Alkandari O: A small post-operative rise in serum creatinine predicts acute kidney injury in children undergoing cardiac surgery. Kidney Int 76: 885–892, 2009
3. Akcan-Arikan A, Zappitelli M, Loftis LL, Washburn KK, Jefferson LS, Goldstein SL: Modified RIFLE criteria in critically ill children with acute kidney injury. Kidney Int 71: 1028–1035, 2007
4. Alkandari O, Eddington KA, Hyder A, Gauvin F, Ducruet T, Gottesman R, Phan V, Zappitelli M: Acute kidney injury is an independent risk factor for pediatric intensive care unit mortality, longer length of stay and prolonged mechanical ventilation in critically ill children: A two-center retrospective cohort study. Crit Care 15: R146, 2011
5. Schneider J, Khemani R, Grushkin C, Bart R: Serum creatinine as stratified in the RIFLE score for acute kidney injury is associated with mortality and length of stay for children in the pediatric intensive care unit. Crit Care Med 38: 933–939, 2010
6. American Society of Nephrology: American Society of Nephrology Renal Research Report. J Am Soc Nephrol 16: 1886–1903, 2005
7. Parikh CR, Devarajan P, Zappitelli M, Sint K, Thiessen-Philbrook H, Li S, Kim RW, Koyner JL, Coca SG, Edelstein CL, Shlipak MG, Garg AX, Krawczeski CDTRIBE-AKI Consortium: Postoperative biomarkers predict acute kidney injury and poor outcomes after pediatric cardiac surgery. J Am Soc Nephrol 22: 1737–1747, 2011
8. Zappitelli M, Krawczeski CD, Devarajan P, Wang Z, Sint K, Thiessen-Philbrook H, Li S, Bennett MR, Ma Q, Shlipak MG, Garg AX, Parikh CRTRIBE-AKI consortium: Early postoperative serum cystatin C predicts severe acute kidney injury following pediatric cardiac surgery. Kidney Int 80: 655–662, 2011
9. Devarajan P, Krawczeski CD, Nguyen MT, Kathman T, Wang Z, Parikh CR: Proteomic identification of early biomarkers of acute kidney injury after cardiac surgery in children. Am J Kidney Dis 56: 632–642, 2010
10. Metzger J, Kirsch T, Schiffer E, Ulger P, Mentes E, Brand K, Weissinger EM, Haubitz M, Mischak H, Herget-Rosenthal S: Urinary excretion of twenty peptides forms an early and accurate diagnostic pattern of acute kidney injury. Kidney Int 78: 1252–1262, 2010
11. Rouse RL, Zhang J, Stewart SR, Rosenzweig BA, Espandiari P, Sadrieh NK: Comparative profile of commercially available urinary biomarkers in preclinical drug-induced kidney injury and recovery in rats. Kidney Int 79: 1186–1197, 2011
12. Ware LB, Johnson AC, Zager RA: Renal cortical albumin gene induction and urinary albumin excretion in response to acute kidney injury. Am J Physiol Renal Physiol 300: F628–F638, 2011
13. Hsu CY, Ordoñez JD, Chertow GM, Fan D, McCulloch CE, Go AS: The risk of acute renal failure in patients with chronic kidney disease. Kidney Int 74: 101–107, 2008
14. Huang TM, Wu VC, Young GH, Lin YF, Shiao CC, Wu PC, Li WY, Yu HY, Hu FC, Lin JW, Chen YS, Lin YH, Wang SS, Hsu RB, Chang FC, Chou NK, Chu TS, Yeh YC, Tsai PR, Huang JW, Lin SL, Chen YM, Ko WJ, Wu KDNational Taiwan University Hospital Study Group of Acute Renal Failure: Preoperative proteinuria predicts adverse renal outcomes after coronary artery bypass grafting. J Am Soc Nephrol 22: 156–163, 2011
15. Bangstad HJ, Dahl-Jørgensen K, Kjaersgaard P, Mevold K, Hanssen KF: Urinary albumin excretion rate and puberty in non-diabetic children and adolescents. Acta Paediatr 82: 857–862, 1993
16. Kwak BO, Lee ST, Chung S, Kim KS: Microalbuminuria in normal Korean children. Yonsei Med J 52: 476–481, 2011
17. Sanchez-Bayle M, Rodriguez-Cimadevilla C, Asensio C, Ruiz-Jarabo C, Baena J, Arnaiz P, Villa S, Cocho PNiño Jesus Group: Urinary albumin excretion in Spanish children. Pediatr Nephrol 9: 428–430, 1995
18. Schwartz GJ, Muñoz A, Schneider MF, Mak RH, Kaskel F, Warady BA, Furth SL: New equations to estimate GFR in children with CKD. J Am Soc Nephrol 20: 629–637, 2009
19. Jones CA, Francis ME, Eberhardt MS, Chavers B, Coresh J, Engelgau M, Kusek JW, Byrd-Holt D, Narayan KM, Herman WH, Jones CP, Salive M, Agodoa LY: Microalbuminuria in the US population: Third National Health and Nutrition Examination Survey. Am J Kidney Dis 39: 445–459, 2002
20. Tsioufis C, Mazaraki A, Dimitriadis K, Stefanidis CJ, Stefanadis C: Microalbuminuria in the paediatric age: Current knowledge and emerging questions. Acta Paediatr 100: 1180–1184, 2011
21. Mehta RL, Kellum JA, Shah SV, Molitoris BA, Ronco C, Warnock DG, Levin AAcute Kidney Injury Network: Acute Kidney Injury Network: Report of an initiative to improve outcomes in acute kidney injury. Crit Care 11: R31, 2007
22. Zou G: A modified poisson regression approach to prospective studies with binary data. Am J Epidemiol 159: 702–706, 2004
23. Jenkins KJ, Gauvreau K: Center-specific differences in mortality: Preliminary analyses using the Risk Adjustment in Congenital Heart Surgery (RACHS-1) method. J Thorac Cardiovasc Surg 124: 97–104, 2002
24. DeLong ER, DeLong DM, Clarke-Pearson DL: Comparing the areas under two or more correlated receiver operating characteristic curves: A nonparametric approach. Biometrics 44: 837–845, 1988
25. Pencina MJ, D’Agostino RB Sr, D’Agostino RB Jr, Vasan RS: Evaluating the added predictive ability of a new marker: From area under the ROC curve to reclassification and beyond. Stat Med 27: 157–172, discussion 207–212, 2008
26. Pencina MJ, D’Agostino RB Sr, Steyerberg EW: Extensions of net reclassification improvement calculations to measure usefulness of new biomarkers. Stat Med 30: 11–21, 2011
27. Ascione R, Lloyd CT, Underwood MJ, Lotto AA, Pitsis AA, Angelini GD: Inflammatory response after coronary revascularization with or without cardiopulmonary bypass. Ann Thorac Surg 69: 1198–1204, 2000
28. Tonelli M, Muntner P, Lloyd A, Manns BJ, James MT, Klarenbach S, Quinn RR, Wiebe N, Hemmelgarn BRAlberta Kidney Disease Network: Using proteinuria and estimated glomerular filtration rate to classify risk in patients with chronic kidney disease: A cohort study. Ann Intern Med 154: 12–21, 2011