Unexpected random urinary protein:creatinine ratio results–limitations of the pyrocatechol violet-dye method

Accurate proteinuria results are key to the management of adults and children with chronic kidney disease [1], as well as women with a hypertensive disorder of pregnancy [2].

The most commonly used methods of detection of proteinuria in pregnancy are: urinary dipstick by visual or automated testing, random protein to creatinine ratio (PrCr), or 24-hour urine collections. The random PrCr has been recommended as a confirmatory test for pre-eclampsia, defined as a random PrCr ≥30 mg/mmol [2] (http://www.nice.org.uk/cg107). Since 2009 at our institution, we have used this test more and more for outpatients and inpatients as an easy and timely alternative to 24-hour urine collection for proteinuria determination. However, it has been our clinical impression that there are a number of women who have urinary PrCr values above diagnostic threshold, yet, have normal pregnancy outcomes. We are also aware of possible false positive results outside pregnancy in a community screening study [3].

We sought to investigate our impression that an elevated PrCr result may be associated with normal pregnancy outcome, and if so, to consider potential explanations related to physiology and analytical method.

This pragmatic cohort study took place at BC Women’s Hospital & Health Centre in Vancouver, BC from January 27 to March 31, 2011. Consecutive high-risk inpatient or outpatient pregnant women were prospectively evaluated. All women who presented for hypertension in the assessment room or delivery suite or seen at our (primarily morning) ambulatory medicine or high-risk obstetric clinics were included. Women were excluded if they had ruptured membranes or were in labour. As this was a quality improvement study for the hospital laboratory, consent was not required. This study was approved by the University of British Columbia Clinical Research Ethics Board (H10-02691).

Random midstream urine samples obtained as part of normal clinical care were split into two aliquots; women are not routinely asked to provide first-voided urines, but rather, they provide the sample when they arrive at the clinic. The first was used for urinary dipstick testing. The second aliquot was sent to the hospital laboratory where it was centrifuged at a speed of 1500 rpm for 5 minutes, and then tested for urinary PrCr in batches on an automated analyser (Vitros 5,1 FS or Vitros 5600, Ortho-Clinical Diagnostics, Rochester, NY) as is standard laboratory procedure. Testing included urinary creatinine (using an enzymatic method) and protein (using a pyrocatechol violet molybdate dye-binding method), followed by calculation of the urinary PrCr [4, 5]. The manufacturer lists limits of detection of 0.05 g/L for protein and 0.106 mmol/L for creatinine. The coefficients of variation of the protein and creatinine assays are 2.9% at a concentration of 0.3 g/L and 2.0% at a concentration of 5.3 mM, respectively. In addition to the routinely collected, visually interpreted urinary dipstick testing performed in the clinic, clinicians also received the results of the urine PrCr.

Hospital records were used to abstract maternal and perinatal outcomes. Maternal outcomes included demographics, parity, multiple pregnancy, details of the indication for attendance at the high-risk clinic and other medical co-morbidities, antenatal complications such as haemorrhage or preterm labour, and delivery information including mode of delivery. Perinatal outcomes included stillbirth, neonatal death, birthweight, and neonatal intensive care unit (NICU) admission.

In our exploratory analysis, we plotted the creatinine concentration (mM) against the PrCr (mg/mmol) for all random urine samples submitted (as some women provided more than one urine sample). Then we focused on using the last random urine sample submitted before delivery. In response to the results, we performed sensitivity analyses in which we included only urine samples with a specific gravity (SG) >1.010 and samples from women with no dipstick hematuria. Samples from women with normal pregnancy outcome (i.e., no evidence of pre-eclampsia and a term delivery of an appropriately grown infant) were highlighted in all analyses. We then performed a dilution study on a new random urine sample with a protein result in the low-mid area of the Vitros assay analytical range of 0.05 – 2.00 g/L. 1:2, 1:3, 1:4, and 1:5 dilutions of the urine were made using either deionised water or saline as the diluent and tested for urinary protein on the analyser. To verify whether our findings were related to the urinary protein method used, we repeated our statistical analyses in two separate published cohorts of women in which proteinuria was determined using a pyrogallol-based dye-binding urine protein assay [6, 7]. The Pearson’s Chi-squared test and Fisher’s exact test were used to calculate p values where appropriate.

In our pragmatic study of primarily outpatient women attending high-risk pregnancy clinics, we found that the random urinary PrCr was inflated when urinary creatinine concentration was <3 mM, regardless of pregnancy outcome. This relationship was indeed surprising, as taking a ratio of protein and creatinine concentrations should correct for urinary dilution.

In seeking an explanation for increased PrCr results at low urinary creatinine concentrations, we considered both physiological and analytical explanations. First, we saw inflated random urinary PrCr results when analyses were restricted to women with urine samples with SG >1.010 or those with no dipstick hematuria or leukocytes. Second, considering that proteinuria in pregnancy is measured primarily to detect pre-eclampsia, either de novo or superimposed on pre-existing hypertension (present in 15% of our cohort), it must be noted that women with pre-eclampsia tend to be intravascularly volume contracted so if anything, they should have decreased urine volumes (or even oliguria) and elevated (not decreased) urinary creatinine concentrations. Many of these women are placed on bedrest, but this would not account for decreased urinary creatinine concentration because the natural history of pre-eclampsia is too short to have a significant impact on muscle mass (and urinary creatinine excretion); women at term gestations are delivered right away [8] and those remote from term are delivered for maternal/fetal reasons either right away (40%) or on average, within 10–14 days (30%) [9]. Third, in considering the role of analytical methods, in two other patient cohorts from our centre (which used pyrogallol red dye-based protein assays until 2009, and Vitros thereafter) random urinary PrCr results were inflated in samples tested by the Vitros compared to those tested using pyrogallol red. Of particular note, 10% of 24-hour urine samples were dilute [6] and thus, subject to the same problem as the random urine samples in the current study. Finally, dilution studies showed that using our pyrocatechol violet molybdate dye-binding method on the Vitros analyser, there was overestimation of urine protein as dilution increased, and even pure water and saline contained ‘measurable’ protein. This led us to the conclusion that our random urinary PrCr results were inflated in dilute urines because of falsely high urine protein results.

These results helped to explain both our clinical observations that some women with elevated random urinary PrCr results had normal pregnancy outcome, as well as the findings of a previous report of unexplained proteinuria of 0.38 g/d associated with high water intake (and low 24-hour urinary creatinine concentration of 2.9 mM) in 63 people in a community screening study [3]. In that published study, the Vitros 950 autoanalyser was used to measure protein, and when 56 of the subjects agreed to decrease their water intake, urine protein significantly decreased (to 0.16 g/d, associated with an increase in urinary creatinine concentration to 6.9 mM).

The manufacturer of the Vitros urine protein assay (Ortho-Clinical Diagnostics, Rochester, NY) has documented that when urinary SG is ≤1.010, measurement of 24-hour urine protein excretion may be falsely elevated as a limitation of the assay [5]. We found, however, that urine creatinine <3 mM was more effective than urine SG ≤1.010 for identifying dilute samples in which random urinary PrCr was overestimated; the inaccuracy of urine dipstick SG compared to other measures of urine concentration is well documented [10], whereas the use of creatinine to correct for variability in urine concentration is the premise of PrCr assessment [1]. The overestimation is attributed to the fact that the accuracy of pyrocatechol violet molybdate dye-binding relies on the ionic strength of the urine. Lower ionic strength (as seen in dilute urine) may result in more dye-binding and higher measured protein concentrations [11]. This overestimation by Vitros of protein concentration in dilute urine has been previously published [11, 12]. Consistent with the manufacturer’s explanation was our observation that when saline was used as the diluent in our dilution studies (thus, offering some ionic strength from the sodium chloride), inflation of urinary protein measurement seen with use of deionised water was attenuated. Although this study showed that proteinuria results were inflated in dilute urine because of analytical bias, it is to be noted that the use of creatinine may also fail to correct for urinary dilution in very dilute urines for other analytes, for example, as demonstrated for urinary albumin [13].

Our 2008 review of the diagnostic accuracy of the random urinary PrCr concluded that it is a “reasonable” rule-out test for detecting proteinuria of ≥0.3 g/d in hypertensive pregnancy [14]. Our findings in the current study should have no impact on this conclusion. Although that review could not examine the impact of urine protein assay method on the results because of incomplete reporting of methods, none of the included studies used the current Vitros method; five used the dye pyrogallol red.

It should be highlighted that this is not a problem restricted to random urinary samples; 10% of 24-hour urine collections may be dilute (i.e., have urinary creatinine <3 mM) [6] and proteinuria is assayed in 24-hour urine samples by taking a sample, measuring the urine protein concentration, and then multiplying it by the volume of urine submitted to get the value for protein excretion in g/d.

The impact on maternal management of a false positive proteinuria result may include patient anxiety and greater use of health care resources due to enhanced maternal and fetal surveillance (as a diagnosis of pre-eclampsia is associated with both maternal and perinatal risk), hospital admission, and/or earlier delivery [2]. Higher costs related to misclassification of renal function related to creatinine standardization have also been previously published [15].

The first strength of our study is that in our current study cohort, we maximised generalisability by assessing a broad spectrum of (primarily outpatient) high-risk patients with and without significant proteinuria as diagnosed by a random urinary PrCr of ≥30 mg/mmol. We also expanded our analyses to two other cohorts of primarily inpatient women with pre-eclampsia. Second, we performed an additional exploratory dilution study using the Vitros proteinuria method, and validated our result in two separate cohorts of women with proteinuria measurement by a different dye-based method (i.e., pyrogallol red).

Some would point out that the random urinary PrCr result was not compared to 24-hour urinary protein; however, it must be acknowledged that the 24-hour urine collection is frequently incomplete in pregnancy and as such, is no longer considered to be the gold standard for diagnosis of proteinuria in pregnancy [6].

In conclusion, our Vitros random urinary PrCr results were inflated in dilute urines because of overestimation of urine protein as a result of technical limitations of the pyrocatechol violet dye-based assay, as confirmed by dilution studies. We observed this phenomenon in another cohort in which Vitros was used. However, the effect was greatly reduced when examining data from two other independent cohorts from our centre in which pyrogallol red was the dye used in the urine protein assay.