ORIGINAL ARTICLE JORGE MEDINA-ROSAS, 1 KRISTY S. YAP, 1 MELANIE ANDERSON, 2 JIANDONG SU, 1 AND ZAHI TOUMA 1 INTRODUCTION

Transcription

1 Arthritis Care & Research Vol. 68, No. 9, September 2016, pp DOI /acr VC 2016, American College of Rheumatology ORIGINAL ARTICLE Utility of Urinary Protein-Creatinine Ratio and Protein Content in a 24-Hour Urine Collection in Systemic Lupus Erythematosus: A Systematic Review and Meta-Analysis JORGE MEDINA-ROSAS, 1 KRISTY S. YAP, 1 MELANIE ANDERSON, 2 JIANDONG SU, 1 AND ZAHI TOUMA 1 Objective. To systematically review literature on the utility of spot urinary protein-creatinine ratio (PCR) as a screening test for proteinuria and its ability to accurately measure proteinuria compared with 24-hour urine collection (24H-P) in patients with systemic lupus erythematosus (SLE). Methods. We conducted a literature search ( ) for articles comparing PCR and 24H-P in SLE patients in the databases Medline, Web of Science, and Embase. Included studies and their results were critically appraised and analyzed. Results. Thirteen studies (1,001 patients; 84.01% women) were included. Ten studies reported on Pearson s (range ), and 3 studies reported on Spearman s (range ). The meta-analysis of studies with Pearson s showed a high overall of 0.80 between 24H-P and PCR, yet with high heterogeneity (I %). Correlation analysis is not sufficient to evaluate the utility of a new test against the gold standard test, and analysis on is required. Seven studies reported on : 3 studies analyzed concordance coefficient ( ), 3 analyzed intraclass coefficient ( ), and 1 analyzed kappa coefficient (0.58). These results confirmed that the between 24H-P and PCR was inappropriate. Three studies included Bland-Altman plots, and the results also demonstrated poor between both tests. Conclusion. The PCR has a utility as a screening test for proteinuria in SLE patients. The studies results of 24H-P and PCR showed poor between both tests, signifying that PCR should not be a substitute for the gold standard test (24H-P) to accurately measure proteinuria. INTRODUCTION Lupus nephritis (LN) occurs as a manifestation of systemic lupus erythematosus (SLE) with a cumulative incidence of 54% (1), and may present early (50%) as well as years after the diagnosis of SLE (45%) (2). Proteinuria is the most common manifestation of LN; it has been reported in almost 1 Jorge Medina-Rosas, MD, Kristy S. Yap, MBBS, FRACP, Jiandong Su, MSc, Zahi Touma, MD, FACP, FACR, PhD: University of Toronto Lupus Clinic, Toronto Western Hospital, Centre for Prognosis Studies in the Rheumatic Diseases, Toronto, Ontario, Canada; 2 Melanie Anderson, BA, MLIS: University Health Network, Toronto, Ontario, Canada. Address correspondence to Zahi Touma, MD, FACP, FACR, PhD, Centre for Prognosis Studies in the Rheumatic Diseases, Toronto Western Hospital, EW, 1-412, 399 Bathurst Street, Toronto, Ontario, Canada, M5T 2S8. E- mail: zahi.touma@uhn.ca. Submitted for publication September 16, 2015; accepted in revised form December 15, % of patients, followed by granular casts, cellular casts, hematuria, and reduced renal function (1). Proteinuria is used as a screening test for LN (3), to monitor the response to the therapy (3,4) and to monitor kidney disease progression. The collection of urine for 24 hours is the gold standard method for quantification of proteinuria (24H-P), but this test can be cumbersome for the patients and sometimes is undercollected (5). Since 1983, Ginsberg et al proposed the use of the ratio of the concentrations of the protein and creatinine contents in a single voided urine sample (PCR) instead of 24H-P, reasoning that if the protein excretion remained stable, then PCR would reflect the cumulative protein excretion during a day (6). After the work of Ginsberg et al, many authors have shown moderatehigh between PCR and 24H-P in different diseases, such as diabetes mellitus (7 9), LN (10,11), and chronic kidney disease (CKD) (12,13). The Renal Disease Subcommittee of the American College of Rheumatology recommends the PCR for use in clinical trials of LN (14), and the European League Against Rheumatism/European Dialysis and Transplant Association suggests using PCR for monitoring LN (15). However, the studies on PCR have 1310

2 Assessment of Proteinuria in SLE 1311 Significance & Innovations The majority of studies on the utility of proteincreatinine ratio (PCR) compared to 24-hour urine collection (24H-P) found a high between both tests but with high heterogeneity among studies. It is important to note that analysis is not sufficient to evaluate the utility of a new test against the gold standard test, and analysis on is required. Based on the results from different studies, PCR doesn t provide an accurate measure of proteinuria when compared to 24H-P in systemic lupus erythematosus (SLE) patients. PCR can be used as a screening test for proteinuria in SLE, but all the abnormal results should be confirmed by the 24H-P. some caveats (16 19). Chitalia et al pointed out that the results in many studies are based on the association () between PCR and 24H-P, but do not enable a reliable decision to be made to replace one with the other (18). Bland and Altman highlighted the inappropriate methodology in studies reporting on the of 2 methods of clinical measurement over the entire sample range (16). This may conceal dis between the 2 methods of the same sample at the extreme ranges, and it is possible to have a high with confidence intervals that may be unacceptably wide (16). Clearly, there is not complete among the existing studies on the utility of PCR in screening and monitoring LN. One of the reasons for the variance among the results of the published studies is partly related to the use of inappropriate statistical analyses. We systematically reviewed the literature, critically appraised the methodology and the statistical analyses of included studies, and conducted a meta-analysis to evaluate 1) the accuracy of PCR as a screening test and its ability to detect clinically important proteinuria compared to 24H-P, and 2) to evaluate the accuracy of PCR in quantifying proteinuria when compared to 24H-P in patients with SLE. MATERIALS AND METHODS Literature review. This systematic review was prepared with a protocol, following the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols statement (20). The search of the literature was conducted by a medical librarian (MA) using 3 databases: Medline (1946 to October, week 4, 2014), Embase (1974 to November 3, 2014), and Web of Science (1900 to November 3, 2014). For Medline, we used the MeSH terms Lupus Nephritis, Glomerulonephritis, Lupus Erythematosus, Systemic, Proteinuria and Urinalysis, as well as relevant keywords. For Embase, we used the Emtree terms lupus erythematosus nephritis, proteinuria, urinalysis, and protein, as well as relevant keywords. For Web of Science we used relevant keyword strings in all available databases, including Science Citation Index Expanded (1900 to November 2014), BIOSIS Citation IndexSM (1926 to November 2014), BIOSIS Previews (1980 to November 2014), and the Conference Proceedings Citation Index-Science (1990 to November 2014). Updated searches using the same strategies were performed in these databases on February 19, 2015, as well as searches using all appropriate subject headings and keywords in Medline in Process and PubMed on February 24 and 25, respectively. Complete strategies can be found in Supplementary Table 1, available on the Arthritis Care & Research web site at /acr.22828/abstract. Study selection and assessment. Eligible for inclusion were all the studies of all designs, published in any language and conducted with humans, that compared the results of PCR and 24H-P in patients with LN and included information relevant to the objective of this study. We excluded articles without an English-written abstract and publications written only in abstract form. Letters, commentaries, reviews, and editorials were reviewed but not included, and their references were screened. Data extraction and synthesis. A standardized form was developed and used to extract the necessary data from included publications. In the development of the form we focused on 4 major domains: 1) related to the authors and references (author, year, country, type of study, and number of patients), 2) patients demographics (age, sex, and ethnicity), 3) kidney function tests (creatinine and glomerular filtration rate [GFR]) and information on the tests used to measure proteinuria (documented completeness of the 24H urine collection, type of PCR [first morning, untimed, or a urine spot from a 24H urine collection sample], result of the PCR and 24H-P, or 24H-P/ 24H-creatinine content [24H-P/24H-creatinine content: a ratio of the total amount of protein in the 24H urine sample to the total amount of creatinine in this sample]), and 4) methods and statistical analyses for assessing the diagnostic accuracy of the used tests to measure proteinuria: (method [e.g., Pearson s ] and results) and (method [e.g., intraclass coefficient (ICC)] and results). One author (JM-R) scanned the titles and abstracts for initial selection. Selected articles were retrieved in full, and 2 reviewers (JM-R and KSY) assessed them for eligibility and extracted the data independently. Discrepancies were resolved by consensus and involvement of another author if needed. Two reviewers (JM-R and ZT) assessed the methodological quality of identified studies independently with the Quality Assessment of Diagnostic Accuracy Studies appraisal tool, version 2 (QUADAS-2) (21). QUADAS-2 comprises 4 phases: 1) state the review question, 2) develop review-specific guidance, 3) review the published flow diagram for the primary study or construct a flow diagram if none is reported, and 4) judge on risk of bias and applicability. This tool consists of 4 domains: patient selection, index test, reference standard, and flow and timing of the index test(s). Each domain assesses the

3 1312 Medina-Rosas et al risk of bias (e.g., patient selection, risk of bias related to the conduct or interpretation of the results of the gold standard, and the index tests), and the first 3 domains focus on the applicability in regards to patients, gold standard, and the index tests. To help the reviewers on judgment of risk of bias and applicability, signaling questions are included for each domain. In each domain, the risk of bias and the concerns regarding applicability are scored independently (low, high, and unclear). Descriptive statistics were used to describe the characteristics of the patients and the results of the analyses (mean 6 SD for continuous data) available from the retrieved studies. For the meta-analysis, the s were converted into a standard normal metric using Fisher s r-to-z transformation formula, the standardized meta- was calculated, then the standardized meta- was transformed back to meta- (22). The overall Pearson coefficient (r) was calculated using the Hedges method (fixed-effects model) (22) and the Hunter and Schmidt method (fixed-effects model) (23). The heterogeneity between studies was calculated with the Cochran s Q test (24) and the Higgins s I 2 test statistic (25). The I 2 was interpreted as follows (26): 25 49% for low heterogeneity, 50 74% for moderate heterogeneity, and $75% for high heterogeneity. The s and confidence intervals for every study and the overall from meta-analysis weredisplayedinaforestplot. To explore the results of the included studies, ICC results were interpreted as follows (27): for negligible, for low, for moderate, for high, and for very high; ICC $0.85 reflected good (28). Concordance coefficient (CCC) results were interpreted as,0.9 for poor (29). As Landis and Koch proposed (30), strength of for the kappa coefficient was interpreted as follows: #0 for poor, for slight, for fair, for moderate, for substantial, and for almost perfect. All analyses were done with Microsoft Excel 2010 software and Prism 5 for Windows, version 5.03 (GraphPad Software). RESULTS Citation identification and screening. From the databases initially searched (Medline, Embase, and Web of Science) and the update searches in Medline in Process and PubMed, we identified 3,029 references, including 473 duplicates. Of 2,556 references, 2,408 references were excluded because they were not relevant to our research aims, 120 references were excluded because they did not use the test of interest, and 10 references were excluded because they were abstracts published at meetings. The results of the search strategy and the steps of selection of articles are given in Supplementary Table 1 and Supplementary Figure 1, respectively, and are available on the Arthritis Care & Research web site at wiley.com/doi/ /acr.22828/abstract. From 18 eligible full texts, we excluded 6 for the reasons shown in Supplementary Table 2, available at wiley.com/doi/ /acr.22828/abstract. Patients characteristics in included studies. Atotalof 1,001 patients were identified from 13 studies (10,11,31 41) published between 1983 and 2015, with a majority of female patients (84%; 2 studies did not report on sex) (33,39). Only 6 studies specified the ethnicity of the participants (n 5 532): 279 white, 110 black, 143 Asian, and 65 of other ethnic groups (10,31,34,35,37,40). The weighted average age of the patients was years (10 studies reported on the age of the patients, n 5 893) (10,31,32,34 36,38,40,41). Among the included studies, only 3 studies reported on disease activity: Medina-Rosas et al (114 patients with active lupus nephritis and 208 patients with inactive lupus nephritis) (37), Leung et al (median and interquartile range of Systemic Lupus Erythematosus Disease Activity Index 2000: 5.48 [2 8]) (11), and Fine et al (15 of 41 patients had SLE flare 3 months prior to the study start) (34). None of the included studies reported on phase of the therapy for the patients (i.e., induction versus maintenance). Assessment of kidney function and review of the method of assessment of proteinuria in included studies. Kidney function. Eight studies reported on at least 1 of the kidney function variables (creatinine in 5 studies and calculated GFR in 3 studies). Only one study showed reduction in kidney function (mean 6 SD calculated GFR ml/minute) (10). PCR and 24H-P. Eleven studies calculated the PCR from an untimed urine sample (10,11,31,32,34 37,39 41); 1 study used the first morning urine (38) and 1 study used 24H-P/ 24H-creatinine content (33). The 24H-P/24H-creatinine ratio is obtained by dividing the protein content from a 24H urine sample over the creatinine content in the same sample; this approach was used to adjust for the difference in the magnitude of 24H-P and PCR (24H-P scale is 7.6 times greater than PCR scale) (37). One study reported on ratios of the protein content and the creatinine content in urine samples collected over 6-hour periods (6H-P/6H-creatinine) or 12-hour periods (12H-P/12H-creatinine) (34). Ten studies reported on 24H-P (11,32,33,35 41), and 3 studies reported on 24H- P/24H-creatinine (10,31,34). Accuracy of the 24-hour urine collection. The accuracy of collection of the 24H urine sample was evaluated in 9 of the 13 studies (10,11,31 35,37,41). Five studies excluded under- and overcollected urine samples (10,31 35), while the other 4 studies excluded only undercollected urine samples (10,11,37,41). Quality assessment. The QUADAS-2 tool items are summarized in Table 1. In 5 articles, there was a low risk of bias and applicability concerns (31,34 37). In 2 articles (Choi et al [10]) had the second largest sample size [n 5 275] after Medina-Rosas et al [37] [n 5 1,233]), there was an unclear risk of bias for the index and the reference tests because the authors did not report on the time interval between PCR and 24H-P (10,33). The patient selection item was associated with a higher risk of bias and applicability concerns for the following reasons: Christopher- Stine et al studied patients with LN treated with cyclophosphamide (33), Salesi et al studied only female

4 Assessment of Proteinuria in SLE 1313 Table 1. Quality assessment of the included studies (using QUADAS-2)* Risk of bias Applicability concerns Study Patient selection Index test Reference standard Flow and timing Patient selection Index test Reference standard Birmingham et al, 2007 (31) Low Low Low Low Low Low Low Choi et al, 2013 (10) Low Unclear Unclear Low Low Low Low Chotayaporn et al, 2013 (32) Low Low Low Low Low Low Low Christopher-Stine et al, 2004 (33) High Unclear Unclear Low High Low Low Fine et al, 2009 (34) Low Low Low Low Low Low Low Leung et al, 2007 (11) Unclear Low Low Low Low Low Low Marques et al, 2013 (35) Low Low Low Low Low Low Low Matar et al, 2012 (36) Low Low Low Low Low Low Low Salesi et al, 2009 (38) High Low Low Low High Low Low Sessoms et al, 1983 (39) Unclear Low Low Low Unclear Low Low Solorzano et al, 2012 (40) Unclear Low Low Low Low Low Low Zhang et al, 2015 (41) High Low Low Low High Low Low Medina-Rosas et al, 2015 (37) Low Low Low Low Low Low Low * QUADAS-2 5 Quality Assessment of Diagnostic Accuracy Studies appraisal tool, version 2. patients (38), and Zhang et al studied hospitalized lupus patients (41). Correlation and between PCR and 24H-P. All 13 studies included in this review reported on (Pearson s or Spearman s coefficient) between 24H-P and PCR, but only 7 studies included an method (ICC, CCC, and Bland-Altman plot) (10,11,14,31,32,34,37,41). Correlation (Table 2). All the 13 articles reported on the between PCR and 24H-P for all concurrent urine samples. Two studies reported on the for all urine samples and also for the samples with values between gm/day (10,31). Christopher-Stine et al reported on the for all urine samples and for the samples after excluding under- and overcollected samples (33). Medina-Rosas et al reported on the for all urine samples and for 4 groups based on 24H-P (group I:,0.5 gm/day, group II: gm/day, group III: gm/day, and group IV: $2.0 gm/day) (37). One study reported on the between PCR and 24H-P and between 24H-PCR and the 24H-P (32). Fine et al reported on the between 24H-P and the first morning PCR, 24H-P and 24H-PCR, and 24H-P and 6H-P/6H-creatinine or 24H-P and 12H-P/12H-creatinine (34). Marques et al studied the between 24H- PCR and 24H-P for the entire range of the urine samples and for 3 groups (group I:,0.5 gm/day, group II: gm/day, and group III:.1.0 gm/day) (35). Ten articles reported on Pearson s coefficient (31 35,37 41). For PCR and 24H-P in all the samples, the ranged from moderate to very high (r ), with a mean high positive (r ). For the samples between 0.5 and 3.0 gm/day, the was moderate (r ) (31). For the comparison between 24H-P and the 6H and 12H, the s were very high (r ), with a very high mean (r ) (34). For the study with 4 groups, the s for group I was low (r ), for group II was negligible (r ), for group III was low (r ), and for group IV was moderate (r ) (P, 0.05 for all) (37). Three studies calculated the Spearman coefficient (10,11,36). For PCR and 24H-P in all the urine samples, the ranged from high to very high (r s ), with a high mean (r s ). For the study that used the samples between 0.5 and 3.0 gm/day, the was high (r s ) (10). Agreement (Table 1). Of the 13 studies evaluated, 7 studies performed at least 1 of the following analyses for : ICC, CCC, or Bland-Altman plot. Three studies used CCC (31,34,37). In the first study, the CCC for the entire population was 0.76, and for the group with 24H-PCR gm/day the CCC was 0.48 (poor ) (31). The second study that used 6H-P and 12H-P showed an appropriate (mean CCC 0.94) (34). The third study, by Medina-Rosas et al, showed poor for the entire population (CCC 0.85) and for the subgroups (24H-P, 0.5 gm/day: CCC 0.62; 24H-P gm/day: CCC 0.34; 24H-P gm/day: CCC 0.55; and 24H-P. 2.0 gm/day: CCC 0.44) (37). Three studies used ICC (10,37,41). The first study found good for the entire range of urine samples (ICC 0.95) but poor for the samples between gm/day (ICC 0.66) (10). In the second study, the was poor for all samples (ICC 0.66) (41). In the third study, the was poor for all the samples (ICC 0.86) and for the subgroups: ICC 0.42 for 24H-P, 0.5 gm/day, ICC 0.57 for 24H-P gm/day, ICC 0.65 for 24H-P gm/day, and ICC 0.78 for 24H-P. 2.0 gm/day (37). Chotayaporn et al reported on using kappa coefficient (32) and showed moderate for the entire population (k ), substantial for proteinuria,0.5 gm/day (k ), gm/day (k ), gm/day (k ), and for proteinuria.2.0 gm/day (k ).

5 1314 Medina-Rosas et al Study Birmingham et al, 2007 (31) Choi et al, 2013 (10) Chotayaporn et al, 2013 (32) Christopher-Stine et al, 2004 (33) Fine et al, 2009 (34) Table 2. Summary of studies included in the review: sample size,, and results* Patients, no. Analyzed paired urine samples, no. Kidney function Creatinine Pearson s PCR vs. 24H-P/24H-creatinine All samples: 0.78 Proteinuria between 0.5 and 3.0: cgfr Spearman s PCR vs. 24H-P/24H-creatinine All samples: 0.94 Proteinuria between 0.5 and 3.0: Creatinine Pearson s 24H PCR vs. 24H-P: 1.0 PCR vs. 24H-P: Not reported Not reported Pearson s 24H-P/24H-24H-P/ 24H-creatinine vs. 24H-P All samples: 0.89 Excluding under- and overcollections: Creatinine Pearson s 24H-P vs. PCR First morning urine: 0.97 After completing 24H urine collection: H-P vs. 6H-P/6H-creatinine 0 6 h: h: h: 0.94 CCC PCR vs. 24H-P/24H-creatinine All samples: 0.76 Proteinuria between 0.5 and 3.0: 0.48 Simple linear regression PCR vs. 24H-P/24H-creatinine All samples: 0.88 Proteinuria between 0.5 and 3.0: 0.44 ICC PCR vs. 24H-P/24H-creatinine All samples: 0.95 (95% CI ) Proteinuria between 0.5 and 3.0: 0.66 (95% CI ) Bland-Altman kappa Not calculated All samples Bland-Altman for PCR vs. 24H-P: bias 0.01 (SD 2.24) 24H-P, 2 gm/day Bland-Altman for PCR vs. 24H-P: bias 0.23 (SD 0.96) Kappa PCR vs. 24H-P, 0.5 gm/day: 66.7% PCR vs. 24H-P gm/day: 71.4% PCR vs. 24H-P gm/day: 64.3% PCR vs. 24H-P. 2.0 gm/day: 72.2% PCR vs. 24H-P all samples: 0.58 (95% CI ) CCC 24H-P vs. PCR First morning urine: 0.95 After 24H urine collection: H-P vs. 6H-P/6H-creatinine 0 6 h: h: h: 0.94 (continued)

6 Assessment of Proteinuria in SLE 1315 Study Leung et al, 2007 (11) Marques et al, 2013 (35) Matar et al, 2012 (36) Salesi et al, 2009 (38) Sessoms et al, 1983 (39) Solorzano et al, 2012 (40) Zhang et al, 2015 (41) Table 2. (Cont d) Patients, no. Analyzed paired urine samples, no. Kidney function h: H-P vs. 12H-P/12H-creatinine 0 12 h: h: H-P vs. 18H-P/18H-creatinine 0 18 h: cgfr Spearman s PCR vs. 24H-P: 0.91 Bland- Altman cgfr Pearson s 24H-P/24H-creatinine vs. 24H-P (all samples): 0.99 PCR vs. 24H-P (all samples): 0.85 PCR vs. 24H-P/24H-creatinine All samples: 0.86 Proteinuria,0.5 gm/day: 0.47 Proteinuria gm/day: Proteinuria.1.0 gm/day: Not reported Spearman s PCR vs. 24H-P cross-sectional data (all samples): 0.87 PCR vs. 24H-P longitudinal data: 0.91 (14 patients with paired samples for a period of 2 years); 0.91 (8 patients with paired samples for a period of 3 years) Not reported Pearson s PCR vs. 24H-P First urine sample: 0.83 Second urine sample: 0.82 Not calculated Not calculated Not calculated Creatinine Pearson s PCR vs. 24H-P: 0.81 Not calculated h: H-P vs. 12H-P/12H-creatinine 0 12 h: h: H-P vs. 18H-P/18H-creatinine 0 18 h: H-P, 2.0 gm/day: limits of for PCR and 24H- P: and H-P, 1.5 gm/day: limits of for PCR and 24H- P: and g/day Creatinine Pearson s PCR vs. 24H-P: 0.90 Not calculated Not reported Pearson s PCR vs. 24H-P: 0.67 ICC PCR vs. 24H-P: 0.66, 95% CI ( ) (continued)

7 1316 Medina-Rosas et al Table 2. (Cont d) Study Patients, no. Analyzed paired urine samples, no. Kidney function Medina-Rosas et al, 2015 (37) 322 1,233 Not reported Pearson s PCR vs. 24H-P All samples: H-P, 0.5 gm/day: H-P gm/day: H-P gm/day: H-P. 2 gm/day: 0.6 ICC, CCC Bland-Altman ICC All samples: 0.86 (95% CI ) 24H-P, 0.5 gm/day: 0.42 (95% CI ) 24H-P gm/day: 0.57 (95% CI ) 24H-P gm/day: 0.65 (95% CI ) CCC 24H-P. 2.0 gm/day: 0.78 (95% CI ) All samples: 0.85 (95% CI ) 24H-P, 0.5 gm/day: 0.62 (95% CI ) 24H-P gm/day: 0.34 (95% CI ) 24H-P gm/day: 0.55 (95% CI ) 24H-P. 2.0 gm/day: 0.44 (95% CI ) * PCR: ratio of protein and creatinine contents in a single voided urine sample; 24H-P: protein content in a 24-hour urine collection; 24H-P/24H-creatinine: ratio of protein and creatinine contents in a 24-hour/6-hour/18-hour urine collection; CCC 5 concordance coefficient; cgfr: calculated glomerular filtration rate; ICC 5 intraclass coefficient; 95% CI 5 95% confidence interval. Birmingham et al (31) conducted a subgroup analysis in the following groups: 24H-PCR patients; 24H-PCR 1.0 to, patients, and 24H-PCR 0 to patients. Medina-Rosas et al conducted a subgroup analysis in the following groups: 24H-P, 0.5 g/day in 662 patients; 24H-P g/day in 42 patients; 24H-P g/day in 171 patients; and 24H- P $ 2.0 g/day in 174 patients.

8 Assessment of Proteinuria in SLE 1317 Figure 1. Forest plot of s for the 8 studies included in the meta-analysis for the overall of 24-hour urine collection (24H-P) and protein-creatinine ratio (PCR). 95% CI 5 95% confidence interval. The Bland-Altman plot was used in 3 studies, showing inappropriate between 24H-P and PCR when 24H-P. 1 gm/day (11,32,37). This indicates that for 24H- P. 1 gm/day, the PCR accuracy is suboptimal compared to 24H-P. One study found that PCR tends to overestimate the level of proteinuria compared to 24H-P (39). In 6 studies the was not determined (33,35,36,38 40). Meta-analysis of studies using 24H-P and Pearson s coefficient. Eight articles reported on the Pearson between PCR and 24H-P and were included in the meta-analysis (32,33,35,37 41). The overall obtained was r by the Hedges method and r (95% confidence interval ) by the Hunter and Schmidt method, indicating high overall. The forest plot for s and confidence intervals from individual studies, as well as overall meta-analysis results, are shown in Figure 1. The Cochran s Q test for heterogeneity showed x , 7 df, P, , indicating a significant heterogeneity, and the Higgins s I 2 for heterogeneity result was 97.23%, also indicating a high heterogeneity of the results for the articles included in the meta-analysis. Although the overall was high between 24H-P and PCR in these studies, recommending PCR over 24H-P is not justified if is not studied and found to be appropriate. DISCUSSION The accurate assessment of proteinuria is of major importance for guidance on the treatment of patients with SLE (3). Moreover, the level of proteinuria has a prognostic value for lupus patients; e.g., a proteinuria level of $2 gm/ day predicts late recovery and has an effect on the percentage of patients who will recover (4). Therefore, in LN an accurate quantification of the level of proteinuria is crucial for appropriate management of the patient. This degree of accuracy cannot be achieved with PCR. For the evaluation of proteinuria in SLE, both an appropriate and sensitive screening test (a test done in apparently healthy people to identify those at an increased risk of a disease) and an accurate and specific diagnostic test (a test used in positively screened patients or symptomatic individuals to confirm the presence or absence of the disease) should be used to achieve an optimal management for the patients (42). Given the importance of monitoring of proteinuria in patients with SLE, and the amount of conflicting data about the accuracy of PCR as a surrogate of the 24H-P, we conducted this systematic review to evaluate the usefulness of the PCR to screen for and to quantify proteinuria. Our meta-analysis on (n 5 8 studies) concluded that there is a high overall (r 5 0.8) between PCR and 24H-P, but with high heterogeneity among studies, indicating variation between studies and precluding a conclusion for the overall obtained. Although the majority of the studies determined for all samples, it is more important to study the in samples with different levels of proteinuria. Birmingham et al found a lower between PCR and 24H-P (r 5 0.5, P ) (31). We found that the was high for all the samples, but it dropped for subgroups with different levels of proteinuria (10,31,35,37). Before an index test (e.g., PCR) can replace the gold standard (e.g., 24H-P), the appropriate statistical methods should be used. The coefficient only assesses a linear association between 2 continuous variables, but the appropriate method to be used for the comparison is one that evaluates between the 2 tests (43). In our review, it was not possible statistically to combine the results of the s for all studies (n 5 8) because of the variety of the methods used in these studies; ICC and CCC were reported in 3 studies and the Bland- Altman plot was used in 3 studies. One study showed poor

9 1318 Medina-Rosas et al CCC for all samples (34). Among the studies that assessed ICC, only 1 study showed good for all the samples (10), but in the group between gm/day, as in the other 2 articles (37,41), the was poor. In 3 studies (11,32,37), the Bland-Altman plot showed poor with 24H-P. 1 gm/day. Based on the results from different studies, PCR doesn t provide an accurate measure of proteinuria when compared to 24H-P. The for proteinuria ranges most commonly encountered in clinical practice (24H-P. 1.0 gm/day as defined by Medina-Rosas et al [37] and 24H-P gm/day as defined by Birmingham et al [31]) was especially poor, signifying that PCR is not an accurate test to estimate the level of proteinuria compared to 24H-P. The possible reasons for the lack of between PCR and 24H-P are 1) PCR was initially validated to differentiate between patients with nephrotic proteinuria and patients with nonnephrotic proteinuria and not to accurately measure the level of proteinuria (6); therefore, using PCR for a different intent might result in unreliable results, 2) PCR doesn t account for hour-to-hour and day-by-day variations in urine protein excretion, while 24H-P is a more accurate estimation of the daily exact production and excretion of protein, and 3) the sample size of the studies by Medina- Rosas et al (37), Choi et al (10), and Zhang et al (41) is more appropriate compared to other studies (e.g., Fine et al [34]) and therefore the results of these studies are more valid. The critical appraisal of the methodology of the studies included in our systematic review showed that there were several pitfalls that resulted in erroneous conclusions regarding the utility of PCR versus 24H-P (Table 1). First, some studies had a small sample size, which calls into question the validity of their results. For instance, from the 13 studies included in the systematic review, 6 studies had,100 patients (32 35,39,40). Flahault et al (44) and Chu et al (45) provided useful approaches to calculate the needed sample size based on the normal approximation or the exact method, respectively, for design accuracy in diagnostic tests studies. These methods for sample size calculation have been adopted in a recent study by Medina-Rosas et al (37), which included 322 patients and 1,233 concurrent urine samples of PCR and 24H-P. Second, the statistical analysis was conducted on all urine samples and authors failed to provide analysis for different subgroups of urine samples with different levels of proteinuria. It is important to ensure that studies have a representation of urine samples with different ranges of proteinuria. For example, Medina-Rosas et al studied separately urine samples with proteinuria,0.5 gm/day, 1 2 gm/day, and.2 gm/day (37); Birmingham et al (31) and Choi et al (10) studied all samples and the group with 24H-P between 0.5 and 3.0 gm/day; and Marques et al (35) studied all samples and groups with 24H-P, 0.5 gm/day; gm/day, and $1.0 gm/day. In a recent meta-analysis on the efficacy of combined treatment with angiotensin receptor blockers and angiotensinconverting enzymes inhibitors on diabetic nephropathy (32 studies and 2,596 patients), proteinuria was actually measured by 24H-P and not PCR (46). Price et al conducted a systematic review and reviewed 16 studies (1,781 patients) and concluded that the PCR provides evidence to rule out the presence of significant proteinuria, but when the results of PCR are above the cutoff value, a full 24H collection for proteinuria for accurate quantification is indicated (47). Medina-Rosas et al defined specific cutoffs for PCR to predict 24H-P (37), which differ from the cutoffs determined by Leung et al (11). Yadav et al also highlighted the variability in the cutoff of PCR and recommended that the cutoff should be determined for different patient groups under different laboratory procedures and settings (9). This further confirms the lack of accuracy with PCR. In our systemic review we found important pitfalls in the studies comparing PCR and 24H-P in LN. The small sample size, the lack of information on lupus disease activity, and the phase of therapy of the included patients, as well as the use of inappropriate methods (focus on analyses and not on analyses) were common pitfalls in these studies. Although PCR correlates with 24H-P, these tests did not show adequate to recommend the use of PCR instead of 24H-P in quantifying proteinuria. Therefore, PCR can be used as a screening test, but all the abnormal results should be confirmed by the 24H-P. More importantly, for an accurate measurement of proteinuria, 24H-P should be used. AUTHOR CONTRIBUTIONS All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Touma had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study conception and design. Medina-Rosas, Touma. Acquisition of data. Medina-Rosas, Yap, Anderson, Su, Touma. Analysis and interpretation of data. Medina-Rosas, Su, Touma. REFERENCES 1. Bastian HM, Roseman JM, McGwin G Jr, Alarcon GS, Friedman AW, Fessler BJ, et al. Systemic lupus erythematosus in three ethnic groups: XII. Risk factors for lupus nephritis after diagnosis. Lupus 2002;11: Dye-Torrington D, Urowitz MB, Ibanez D, Gladman DD. Late versus early development of lupus nephritis [abstract]. Arthritis Rheum 2011;63 Suppl: Balow JE. Clinical presentation and monitoring of lupus nephritis. Lupus 2005;14: Touma Z, Urowitz MB, Ibanez D, Gladman DD. Time to recovery from proteinuria in patients with lupus nephritis receiving standard treatment. J Rheumatol 2014;41: Mitchell SC, Sheldon TA, Shaw AB. Quantification of proteinuria: a re-evaluation of the protein/creatinine ratio for elderly subjects. Age Ageing 1993;22: Ginsberg JM, Chang BS, Matarese RA, Garella S. Use of single voided urine samples to estimate quantitative proteinuria. N Eng J Med 1983;309: Rodby RA, Rohde RD, Sharon Z, Pohl MA, Bain RP, Lewis EJ, and the Collaborative Study Group. The urine protein to creatinine ratio as a predictor of 24-hour urine protein excretion in type 1 diabetic patients with nephropathy. Am J Kidney Dis 1995;26: Biradar SB, Kallaganad GS, Rangappa M, Kashinakunti SV, Retnakaran R. Correlation of spot urine protein-creatinine ratio with 24-hour urinary protein in type 2 diabetes mellitus patients: a cross sectional study. J Res Med Sci 2011;16:

10 Assessment of Proteinuria in SLE Yadav BK, Adhikari S, Gyawali P, Shrestha R, Poudel B, Khanal M. Use of protein: creatinine ratio in a random spot urine sample for predicting significant proteinuria in diabetes mellitus. Nepal Med Coll J 2010;12: Choi IA, Park JK, Lee EY, Song YW, Lee EB. Random spot urine protein to creatinine ratio is a reliable measure of proteinuria in lupus nephritis in Koreans. Clin Exp Rheumatol 2013;31: Leung YY, Szeto CC, Tam LS, Lam CW, Li EK, Wong KC, et al. Urine protein-to-creatinine ratio in an untimed urine collection is a reliable measure of proteinuria in lupus nephritis. Rheumatology (Oxford) 2007;46: Wahbeh AM, Ewais MH, Elsharif ME. Comparison of 24- hour urinary protein and protein-to-creatinine ratio in the assessment of proteinuria. Saudi J Kidney Dis Transpl 2009; 20: Hayashi A, Okada T, Matsumoto H, Nagaoka Y, Wada T, Gondo A, et al. Validity of the assessment of urinary protein excretion by spot urine in patients with chronic kidney disease. Nihon Jinzo Gakkai Shi 2013;55: In Japanese. 14. Renal Disease Subcommittee of the American College of Rheumatology Ad Hoc Committee on Systemic Lupus Erythematosus Response Criteria. The American College of Rheumatology response criteria for proliferative and membranous renal disease in systemic lupus erythematosus clinical trials. Arthritis Rheum 2006;54: Bertsias GK, Tektonidou M, Amoura Z, Aringer M, Bajema I, Berden JH, et al. Joint European League Against Rheumatism and European Renal Association-European Dialysis and Transplant Association (EULAR/ERA-EDTA) recommendations for the management of adult and paediatric lupus nephritis. Ann Rheum Dis 2012;71: Bland JM, Altman DG. Comparing two methods of clinical measurement: a personal history. Int J Epidemiol 1995;24 Suppl 1:S Bland JM, Altman DG. Applying the right statistics: analyses of measurement studies. Ultrasound Obst Gynecol 2003;22: Chitalia VC, Kothari J, Wells EJ, Livesey JH, Robson RA, Searle M, et al. Cost-benefit analysis and prediction of 24- hour proteinuria from the spot urine protein-creatinine ratio. Clin Nephrol 2001;55: Restrepo-Escobar M. Correlation versus in two methods for urinary protein quantification in lupus nephritis. Lupus 2012;21: Moher D, Shamseer L, Clarke M, Ghersi D, Liberati A, Petticrew M, et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Syst Rev 2015;4: Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al. QUADAS-2: a revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med 2011;155: Hedges L, Olkin I. Statistical method for meta-analysis. Orlando (FL): Academic Press; Hunter JE, Schmidt FL. Meta-analysis of s corrected individually for artifacts: methods of meta-analysis. Hunter JE, Schmidt FL, eds. Thousand Oaks (CA): Sage Publications; pp Cochran WG. The combination of estimates from different experiments. Biometrics 1954;10: Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ 2003;327: Sedgwick P. Meta-analyses: what is heterogeneity? BMJ 2015;350:h Mukaka MM. Statistics corner: a guide to appropriate use of coefficient in medical research. Malawi Med J 2012;24: Streiner DL, Norman GR. Reliability, generalizability theory and validity. In: Streiner DL, ed. Health measurement scales: a practical guide to their development and use. 4th ed. New York: Oxford University Press; pp McBride GB. Using statistical methods for water quality management. Hoboken (NJ): Wiley-Interscience; Landis JR, Koch GG. The measurement of observer for categorical data. Biometrics 1977;33: Birmingham DJ, Rovin BH, Shidham G, Nagaraja HN, Zou X, Bissell M, et al. Spot urine protein/creatinine ratios are unreliable estimates of 24 h proteinuria in most systemic lupus erythematosus nephritis flares. Kidney Int 2007;72: Chotayaporn T, Kasitanon N, Sukitawut W, Louthrenoo W. Comparison of proteinuria determination by urine dipstick, spot urine protein creatinine index, and urine protein 24 hours in lupus patients. J Clin Rheumatol 2011;17: Christopher-Stine L, Petri M, Astor BC, Fine D. Urine proteinto-creatinine ratio is a reliable measure of proteinuria in lupus nephritis. J Rheumatol 2004;31: Fine DM, Ziegenbein M, Petri M, Han EC, McKinley AM, Chellini JW, et al. A prospective study of protein excretion using short-interval timed urine collections in patients with lupus nephritis. Kidney Int 2009;76: Marques MG, Cotovio P, Ferrer F, Silva C, Botelho C, Lopes K, et al. Random spot urine protein/creatinine ratio: a reliable method for monitoring lupus nephritis? Clin Kidney J 2013;6: Matar HE, Peterson P, Sangle S, D Cruz DP. Correlation of 24- hour urinary protein quantification with spot urine protein: creatinine ratio in lupus nephritis. Lupus 2012;21: Medina-Rosas J, Gladman D, Su J, Sabapathy A, Urowitz M, Touma Z. Utility of untimed single urine protein/creatinine ratio as a substitute for the 24 hour proteinuria for the assessment of proteinuria in systemic lupus erythematosus. Arthritis Res Ther 2015;24: Salesi M, Karimifar M, Farajzadegan Z, Esalatmanesh K, Khosravi S, Fallahi P, et al. The protein-creatinine ratio in spot morning urine samples and 24-h urinary protein excretion in patients with systemic lupus erythematosus. Rheumatol Int 2009;29: Sessoms S, Mehta K, Kovarsky J. Quantitation of proteinuria in systemic lupus erythematosus by use of a random, spot urine collection. Arthritis Rheum 1983;26: Solorzano GT, Silva MV, Moreira SR, Nishida SK, Kirsztajn GM. Urinary protein/creatinine ratio versus 24-hour proteinuria in the evaluation of lupus nephritis. J Bras Nefrol 2012; 34: Zhang Q, Sun L, Jin L. Spot urine protein/creatinine ratio is unreliable estimate of 24 h proteinuria in lupus nephritis when the histological scores of activity index are higher. Lupus 2015;24: Gilbert R, Logan S, Moyer VA, Elliott EJ. Assessing diagnostic and screening tests: part 1. Concepts. West J Med 2001; 174: Bland JM, Altman DG. Statistical methods for assessing between two methods of clinical measurement. Lancet 1986;1: Flahault A, Cadilhac M, Thomas G. Sample size calculation should be performed for design accuracy in diagnostic test studies. J Clin Epidemiol 2005;58: Chu H, Cole SR. Sample size calculation using exact methods in diagnostic test studies [letter]. J Clin Epidemiol 2007; 60: Ren F, Tang L, Cai Y, Yuan X, Huang W, Luo L, et al. Metaanalysis: the efficacy and safety of combined treatment with ARB and ACEI on diabetic nephropathy. Ren Fail 2015: Price CP, Newall RG, Boyd JC. Use of protein:creatinine ratio measurements on random urine samples for prediction of significant proteinuria: a systematic review. Clin Chem 2005;51: