Utility of untimed single urine protein/creatinine ratio as a substitute for 24-h proteinuria for assessment of proteinuria in systemic lupus erythematosus
Patient characteristics
A total of 1730 laboratory urine samples from 421 patients were identified. Of these,
497 samples were excluded because of undercollection in 24-h urine samples, and 1233
urine samples from 322 patients included in the final analysis. The ExC values stratified
by ethnicity were as follows: white 17.3?±?2.8 mmol/day, black 17.5?±?2.2 mmol/day,
Asian 17.6?±?2.6 mmol/day, and other 17.5?±?2.6 mmol/day. The patients’ demographics
are presented in Table 1. The majority of the patients were women (83.5 %) with lupus disease duration of
11.76?±?9.71 years and age at first 24hP of 40.26?±?14.76 years. The number of patients
in groups 1, 2, 3, and 4 were 208, 42, 29, and 43, respectively. Eight patients (2.5Â %)
were treated with cyclosporine, three (1.4Â %) in group 1, one (2.4Â %) in group 2,
one (3.4Â %) in group 3, and three (7Â %) in the group 4. No patients were treated with
tacrolimus.
Table 1. Demographic characteristics of the patients included in the study
Validity of PCR in screening for proteinuria
Of the 1233 urine samples, 552 samples had normal PCR and 24hP. The Pearson correlation
coefficient was 0.29 (p??0.0001). PCR Sn and Sp against 24hP were 91Â % and 83Â %, respectively. PPV was 82.5Â %,
NPV was 91.4, and LR+ was 5.4.
Correlation between PCR and 24hP
For all samples (n?=?1233), the correlation was high (r?=?0.79) (Fig. 1). However, the correlation was low for group 1 (r?=?0.4), negligible for group 2 (r?=?0.2), low for group 3 (r?=?0.3), and moderate for group 4 was (r?=?0.6) (all p??0.05) (Table 2). The correlations of both tests stratified by ethnicity were high for all ethnic
groups (white 0.76, black 0.86, Asian 0.78, other ethnicities 0.88; all p??0.0001).
Fig. 1. Scatterplot of correlation between protein content in a 24-h urine collection sample;
(24hP) and untimed sample of urine protein/creatinine ratio for all urine samples.
PCR spot urine protein/creatinine ratio
Table 2. Results of correlations between 24hP and PCR and agreement between 24hP/24hC and PCR
Agreement between PCR and 24hP/24hC
Intraclass correlation coefficients (2, k)
For all urine samples, ICC was 0.87; for group 1, ICC was 0.52; for group 2, ICC was
0.57; for group 3, ICC was 0.73; and for group 4, ICC was 0.77. The agreement was
appropriate for all the urine samples; however, it was poor (0.85) for groups 1,
2, 3, and 4 (Table 2), indicating less than appropriate agreement of paired urine samples in the same
group and poor reproducibility of the measures for each group.
Concordance correlation coefficients
For all urine samples, CCC was 0.85; for groups 1, 2, 3 and 4, CCCs were 0.62, 0.34,
0.55, and 0.44, respectively. The agreement was poor (0.9) for all urine samples
and for groups 1, 2, 3, and 4 (Table 2), indicating that the PCR levels were not equivalent to the corresponding 24hP levels
in the same patient. The CCC results stratified by ethnicity were as follows: white
0.69 (95 % CI 0.65–0.74), black 0.94 (95 % CI 0.92–0.96), Asian 0.73 (95 % CI 0.63–0.82),
and other ethnicities 0.69 (95 % CI 0.60–0.77).
Bland-Altman plot
The Bland-Altman plot showed that all the paired urine samples from group 1 were between
the limits of agreement; however, with increases in 24hP values, a large difference
between the two methods was found that was more obvious in groups 3 and 4 than in
the other groups. In group 4, PCR tended to overestimate the 24hP results. On the
Bland-Altman plot, the dots corresponding to groups 3 and 4 are outside 2 SD, signifying
poor agreement (Fig. 2).
Fig. 2. Bland-Altman plot for ratios of protein content and creatinine content in 24-h urine
collection samples (24H-P/24-H Creatinine) and untimed samples of urine protein/creatinine
ratio (PCR)
Cutoff of PCR predicting a 24hP of 0.5, 1 and 2Â g/day
The contingency table, using 24hP of 0.5Â g/day as the gold standard and PCR of 0.05Â g/mmol
as the index test, showed Sn and Sp values of 91Â % and 83Â %, respectively. The results
of the ROC curve derived from the logistic regression showed that a PCR of 0.08Â g/mmol
(800 mg/g) reflected 24hP of 0.5 g/day, with Sn of 91 % and Sp of 80 % (Table 3). The overall area under the curve was 0.92 (Fig. 3). For the 24hP cutoffs of 1.0 and 2.0 g/day, the analyses showed PCR values of 0.16 g/mmol
(1600Â mg/g; Sn 90Â % and Sp 83Â %) and 0.35Â g/mmol (3500Â mg/g; Sn 91Â % and Sp 85Â %),
respectively.
Table 3. ROC classification for best PCR cutoffs reflecting 24hP 0.5Â g/day using binary 24hP
as the gold standard
Fig. 3. Receiver operating characteristic (ROC) curve for best cutoff of untimed samples of
urine protein/creatinine ratio using binary protein content in a 24-h urine collection
sample (24hP) as the gold standard (24hP cutoff 0.5Â g/day)
Sensitivity analyses for ACE inhibitors and/or ARB
The results for patients taking ACE inhibitors and/or ARB showed that, correlation
among the 164 patients taking ACE inhibitors and/or ARB, 624 paired samples were identified.
Although the correlation between 24hP and PCR for all samples was 0.77, in group 1
(n?=?288) it was 0.4; in group 2 (n?=?129), it was 0.2; in group 3 (n?=?114), it was
0.3; and in group 4 (n?=?95), it was 0.6.
For all urine samples, ICC was 0.91; for groups 1, 2, 3, and 4, ICCs were 0.48, 0.60,
0.64, and 0.84, respectively. For all samples, CCC was 0.93; for groups 1, 2, 3, and
4, CCCs were 0.59, 0.25, 0.56, and 0.85, respectively.
For the 24hP cutoffs of 0.5, 1.0, and 2.0Â g/day, the analyses showed PCR values of
0.07Â g/mmol (700Â mg/g; Sn 91Â % and Sp 82Â %), 0.11Â g/mmol (1100Â mg/g; Sn 90Â % and Sp
73Â %), and 0.15Â g/mmol (1500Â mg/g; Sn 90Â % and Sp 79Â %), respectively.
The analysis for patients not taking ACE inhibitors and/or ARBs showed results similar
to those for the patients taking ACE inhibitors and/or ARBs. CCC was 0.54 for all
urine samples overall; for groups 1, 2, 3, and 4, CCCs were 0.54, 0.36, 0.39, and
0.36, respectively.