Validation of the DCA® 2000 Microalbumin:Creatinine Ratio Urinanalyzer for Its Use in Pregnancy and Preeclampsia

Objective. To determine the accuracy of the DCA® 2000 albumin/creatinine ratio urinanalyzer (Bayer Corp., Elkhart, IN) in uncomplicated pregnancy and preeclampsia. Methods. This was a prospective observational study in a large teaching maternity hospital. Ninety one uncomplicated pregnant women and 100 women referred for assessment of de novo hypertension in pregnancy had albumin concentrations, creatinine concentrations, and albumin/creatinine ratios (ACR) compared between the DCA 2000 and the laboratory gold standard assays (Dade Dimension® clinical chemistry autoanalyzer), for both early morning urines (EMU) and 24‐hr urine collections. Results. The interassay and intra‐assay variability for the DCA 2000 were less than 5.1%. In uncomplicated pregnancy the mean difference in ACR between the DCA 2000 and the laboratory assay was 0.08 mg/mmol (SD 0.28; 95% limits of agreement, −0.47, 0.63) for EMU and 0.06 (SD 0.23; 95% limits of agreement, −0.39, 0.51) for 24‐hr samples. In the hypertensive cohort the ACR mean differences were −0.82 (SD 7.13; 95% limits of agreement, −14.79, 13.15) for EMU and −0.76 (SD 4.14; 95% limits of agreement, −8.87, 7.35) for 24‐hr samples. The mean differences between assays in the hypertensive group had broader 95% limits of agreement due to greater variability in the samples with high albumin concentrations (>40 mg/L). Conclusions. The DCA 2000 is accurate for the measurement of albumin creatinine ratios in the uncomplicated pregnant population. In the hypertensive population the DCA 2000 remains accurate though when the albumin concentration is greater than 40 mg/L the 95% limits of agreement are broader. We would recommend that all other automated urinalysis devices be validated by similar protocols to allow meaningful comparisons of accuracy.

Keywords: Microalbuminuria; Pregnancy; Albumin/creatinine ratio

The hypertensive disorders of pregnancy affect more than 10% of the antenatal population [1]. The development of proteinuria is accepted as a poor prognostic sign and is associated with increasing maternal and perinatal mortality and morbidity [2], [3].

At present blood pressure measurement and urinalysis are the commonest screening tests performed in pregnancy. Most definitions of preeclampsia require that proteinuria be present [4], and it has been suggested that a 24‐hr total protein excretion be measured to quantify proteinuria [5], [6]. It remains common however for some authors to rely on dipstick technology to define preeclampsia [7]. The use of visually read dipsticks has been shown to be associated with significant observer errors [8], [9], [10], and the introduction of automated dipstick reading devices has been shown to improve accuracy [11]. Other methods of quantifying proteinuria such as the protein/creatinine ratio or the albumin/creatinine ratio (ACR) have not been included in any classification system currently in use.

It has been recommended that automated blood pressure measurement devices should undergo a process of validation according to pre‐determined protocols before they can be recommended for clinical use [12], [13]. Despite the errors associated with point of care, (outpatient) urinalysis, and the variety of different dipstick manufacturers available, no such protocols or recommendations exist for validating either manual or automated urinalysis equipment.

During uncomplicated pregnancy proteinuria increases with increasing gestation and the largest single component of physiological proteinuria is albumin [14], [15]. In preeclampsia a different electrophoretic pattern, with a broader range of molecular weight proteins has been described [15]. Because laboratory assays have varying abilities to detect different molecular weight proteins and hence different reference ranges, any method will always be an estimate of total protein excretion [16]. These changes in protein excretion, either physiological with gestation or pathophysiological with pre‐eclampsia, suggest that any proteinuria screening tool must be accurate in a variety of different clinical circumstances.

Microalbuminuria has been variably defined depending on the type of urine sample collected. For "spot" urine samples, it is accepted that expression of albumin excretion as a ratio to creatinine concentration (the ACR) correlates well with 24‐hr albumin excretion rates [17]. In preeclampsia some authors have suggested a phase of microalbuminuria might precede overt proteinuria in preeclampsia [18]. In studies where this has not been observed it is possible that the sampling procedure was too infrequent to detect this early and possibly short‐lived sign [19]. A point of care urinanalyzer might provide the means to test spot urine samples, thus allowing frequent testing of the antenatal population for microalbumin.

We have therefore designed a protocol for the assessment of automated urinalysis devices in pregnancy and preeclampsia. This protocol has been applied to the DCA® 2000 (Bayer Corp., Elkhart, IN) for the measurement of microalbumin/creatinine ratio.

For the purpose of this validation we recruited 1) 33 women (of varying parity) who had uncomplicated pregnancies and 2) 135 women who were referred to the day assessment unit for assessment of de novo hypertension in pregnancy.

Subjects were asked to collect a 24‐hr urine specimen. On the day the collection started the first void was discarded, and the collection was started with the second voided sample. The final sample was the first void of the following day. Before this sample was added to the 24‐hr collection a 10‐mL aliquot was removed and represented the early morning urine (EMU) sample for comparative purposes.

The subjects were asked to collect both a paired early morning urine specimen and 24‐hr urine specimen four times during pregnancy. The gestation bands were 10–14, 18–24, 26–34, and 36–40 weeks.

The subjects were asked to collect a single paired EMU and 24‐hr urine sample.

We have assessed the performance and accuracy of the DCA 2000 in three stages: 1) inter‐ and intra‐assay variability, 2) accuracy compared to a laboratory "gold standard" in uncomplicated pregnancy, and 3) accuracy compared to a laboratory gold standard in hypertensive pregnancy.

The DCA 2000 has a restricted reference range for microalbumin measurement (5–300 mg/L). Thirty samples with ACR measurements from across this reference range (5–300 mg/L) but measured by the laboratory assay were selected. Two consecutive estimations of ACR on the same DCA 2000 were used to calculate intra‐assay variability, and two simultaneous estimations of ACR on different DCA 2000 machines were used to calculate inter assay variability.

The samples (EMU and 24‐hr) were tested at point of care on the DCA 2000 (by JW and CB) and then sent to the chemical pathology department of the Leicester Royal Infirmary for laboratory assessment of the microalbumin/creatinine ratio. All point of care analysis was completed within 2 hr of the samples being collected and all laboratory assays within 24 hr. Aliquots of each sample were stored at 4°C until testing and then stored at −80°C.

This new system for microalbumin estimation utilizes a cartridge system and a 40 μL sample of urine. It uses an immunoturbidometric assay for albumin (calibrated against the Reference Preparation for Proteins in Human Serum CRM470) in the presence of polyethylene glycol, with the resultant complexes increasing the turbidity. The colorimetric assay for creatinine is based on the colored complex produced when creatinine complexes with 3,5‐dinitrobenzoic acid. Both assays are monitored at 531 nm, and the test takes 7 min to complete. The intraassay and interassay coefficients of variation (CV) are reported as being within the range 0.5–5% for microalbumin and 2–8% for ACR (in nonpregnant populations) [20].

This was determined by automated immunoprecipitin analysis using a SPQ™ test system (Diasorin, Stillwater, MN) and the Dade Dimension® clinical chemistry autoanalyzer. Standards, controls, (SPQ™ Calibration/Control set: Cat no.: 86106) and patient samples are pipetted into sample cups. Incubation is for 5 min, and absorbance is measured at a wavelength of 340 nm. An assay of five standards produces a standard curve from which sample values are derived.

Compared to other methods of detection the regression coefficients are as follows: 0.99 compared to a radioimmunoassay for albumin; 0.999 compared to a turbidometric adaptation of a manual spectroscopic assay; and 0.999 compared to a radial immunodiffusion method. The intra‐assay CV is 3% at albumin concentrations of 16.3 mg/L and 1.7% at albumin concentrations of 125 mg/L. The interassay CV is 2.4% at concentrations of 200–1500 mg/L and 4.3% at concentrations of 20–50 mg/L [21].

The creatinine method uses a modification of the kinetic Jaffe reaction. This method is reported to be less susceptible than conventional methods to interference from noncreatinine Jaffe‐positive compounds [22].

In the presence of a strong base (NaOH), lithium picrate reacts with creatinine to form a red chromophore. The rate of increasing absorbance at 510 nm wavelength due to the formation of the chromophore is directly proportional to the creatinine concentration in the sample. It is measured using a bi‐chromatic (510, 600‐nm wavelength) rate technique [23].

Compared to other automated techniques (Beckman CX® 7; Beckman Instruments Inc., Fullerton, CA 92834‐3100) the regression coefficient is 1. At creatinine concentrations of 6.4 mmol/L the intraassay CV is 5.2% and the interassay CV is 8.2%. At creatinine concentrations of 13.5 mmol/L the intra‐assay CV is 3.2% and the interassay CV is 6.1%.

The mean and standard deviations for albumin concentration, creatinine concentration, and the ACR for both the EMU and 24‐hr samples were calculation. Agreement between the methods of measurement was assessed by calculation of the mean and standard deviations of the differences. Limits of agreement were used to help quantify the differences [24].

For the uncomplicated pregnancy group there was the potential problem that there are three sample pairs per woman, which could induce a correlation structure. This was investigated using multilevel models [25], but made virtually no difference to the results, because the correlation between the differences measured on the same woman was negligible. Therefore, only the more simple analyses described above are presented.

Table 1 shows the demographic data on the women recruited to the study. The DCA 2000 has a restricted reference range for microalbumin measurement (5–300 mg/L). Samples with albumin concentrations that are below 5 mg/L or above 300 mg/L are compared separately from those that were within this range. As such, this meant 91 samples from uncomplicated pregnancies and 100 samples from hypertensive pregnancies were available for direct comparison. The hypertensive cohort was further stratified into low (<40 mg/L) and high (>40 mg/L) albumin concentrations. In the high‐albumin cohort the median value was 106.5 mg/L, with a range of 40.4 to 271 mg/L. (An albumin concentration of 40 mg/L is twice the 95th centile for an uncomplicated pregnant population; Ref. [26]).

Table 1. Demographics

217 All ACRs are from laboratory analysis.

At micro‐albumin concentrations of less than 40 mg/L the intra‐assay CV for 2.1% and the interassay CV is 3.7%. At microalbumin concentrations greater then 40 mg/L the intra‐assay CV is 4.6% and the interassay CV is 5.1%.

The intra‐assay variability is 4.2% and the interassay variability is 6% across the range 3.0–23.5 mmol/L.

At microalbumin concentrations of <40 mg/L, the intra‐assay CV is 2.9% and the interassay CV is 3.9%. At microalbumin concentrations of >40 mg/L the intra‐assay CV is 4.8% and the interassay CV is 5.5%.

In the uncomplicated group 29 women had a DCA 2000 estimate of microalbumin of <5 mg/L. When compared with the laboratory measurements in this group, 26 had a laboratory microalbumin of <5 mg/L, and the remaining three were between 5 and 6 mg/L. The ACR measurements on these samples were all comparable to within 0.05 mg albumin/mmol creatinine. No women had an albumin concentration of >40 mg/L.

In the hypertensive group 14 women had a DCA estimation of microalbumin of <5 mg/L. There was 100% agreement with the laboratory measurement of albumin, and ACR measurements were all comparable to within 0.05 mg albumin/mmol creatinine. Nine women had a measurement of >300 mg albumin/L on the DCA 2000. These women all had laboratory measurement of albumin of >300 mg/L.

Table 2 shows a comparison between the mean albumin concentration, creatinine concentration and ACR for EMU, and 24‐hr samples from the laboratory analysis. In the uncomplicated cohort the difference between albumin concentrations is not significant, (mean difference 0.88 mg/L; 95% confidence limits, −0.46, 2.22). The creatinine concentration is significantly higher in the EMU samples (mean difference 2.26 mmol/L; 95% confidence limits, 1.31, 3.21). When the ACR is calculated and compared, the differences between the two samples are also significant (mean difference −0.15 mg/mmol; 95% confidence limits, −0.27, −0.02).

Table 2. Laboratory values for EMU and 24‐hr samples

218 Values for albumin concentration, creatinine concentration, albumin/creatinine ratio and the difference between the EMU and 24‐hr values for the three parameters for samples with albumin concentrations between 5 and 300 mg/L. aValues are mean±SD.

In the hypertensive group both the albumin concentration differences and the creatinine concentration are statistically different (albumin mean difference, −11.93 mg/L; 95% confidence limits, −16.12, −7.74; creatinine mean difference, 1.67 mmol/L; 95% confidence limits, 0.48, 2.86). When the ACR is calculated for this group, the differences are not statistically different (mean difference, −3.27 mg/mmol; 95% confidence limits, −6.63, 0.09).

Table 3 shows the mean values for albumin, creatinine, and ACR as well as the mean differences with 95% limits of agreement between the DCA 2000 and the laboratory assay for the uncomplicated and the hypertensive group. These are expressed graphically as Bland Altman Plots (for the ACR) in Figures 1 and 2.

Table 3. Values for DCA 2000 and laboratory gold standard

219 Values are albumin concentration (mg/L), creatinine concentration (mmol/L), and albumin/creatinine ratio (mg/mmol) for DCA 2000 and the laboratory gold standard and the difference between the two assays for EMU and 24‐hr samples for samples with albumin concentrations between 5 and 300 mg/L. aValues are mean±SD.

Graph: Figure 1. Bland Altman plots for uncomplicated pregnancy (—, 95% limits of agreement).

Graph: Figure 2. Bland Altman plots for hypertensive pregnancy (total cohort) (—, 95% limits of agreement).

The mean differences and 95% limits of agreement between the DCA 2000 and the laboratory assay are greater in the hypertensive pregnancy cohort than the uncomplicated group. To further explore this difference, we stratified the hypertensive data by albumin concentration (above or below 40 mg/L). Table 4 shows that for the group with lower albumin concentrations, the mean differences are comparable to the uncomplicated group. It can be seen from the Bland Altman Plots of the stratified data (Figures 3 and 4) that the scatter of values and mean differences are greater in the high albumin concentration group.

Table 4. Albumin/creatinine ratios (mg/mmol) for the total hypertensive cohort and stratified by albumin concentration of <40 mg/L for EMU and 24‐hr samples

220 aValues are mean±SD.

Graph: Figure 3. Bland Altman Plots for hypertensive cohort with urine albumin concentrations of <40 mg/L (—, 95% limits of agreement).

Graph: Figure 4. Bland Altman Plots for hypertensive cohort with urine albumin concentrations of >40 mg/L.

From these data presented, the DCA 2000 is accurate for the measurement of albumin creatinine ratios in the uncomplicated pregnant population. In the hypertensive population the DCA 2000 remains accurate though when the albumin concentration is greater than 40 mg/L the 95% limits of agreement are broader. We, and others, have observed this albumin concentration to be above the 95th centile for the uncomplicated pregnant population.

When assessing an automated device for accuracy, it is imperative that the device is tested on an appropriate population, that a reliable gold standard is used for comparison, and that the methodology of the comparison is robust. At the same time as differences between assays are measured, these discrepancies must be placed within the context of the clinical application of the new test. Previous studies on microalbumin excretion in pregnancy have used radioimmunoassays to quantify albumin but have been limited by sampling frequency. We have validated the DCA 2000 in two populations. If this device is to be used for screening a low‐risk population, it must be accurate, reliable, and easy to use. If, as is more likely, it is to be used as an additional test for the screening of pregnancies that are at increased risk of hypertensive disorders in pregnancy, then it must also be accurate in this very different population. Given the altered electrophoretic pattern of proteinuria in preeclampsia compared to physiological proteinuria [16], it is not surprising that the device has larger errors associated with the wider scatter of results in this group. It is however encouraging that when we look at hypertensive pregnancies with albumin concentrations of <40 mg/L, the 95% limits of agreement are similar to the uncomplicated group. Albumin concentrations of >40 mg/L are above the 95th centile for uncomplicated pregnant women.

Urinalysis is performed on a variety of different urine samples (early morning, random voids, and timed collections). It is therefore necessary to assess accuracy in different types of urine samples as urine protein composition varies in a circadian pattern [17]. The protein/creatinine ratio on samples collected throughout the day has been shown to vary with time. The strongest correlation to total protein excretion is found for samples collected between 8 AM to 12 AM in nonpregnant populations [27] and after periods of recumbency rather than exercise in pregnant populations [28]. We report as have others that EMU and 24‐hr samples when compared in the laboratory exhibit different albumin concentrations, and hence they may have different protein excretion patterns. As such, we have sought to validate the DCA 2000 with EMU samples as well as 24‐hr samples. This is essential if the device is to be used at point of care, because 24‐hr samples are notoriously poorly collected and delay any diagnosis by 24–48 hrs. It was again reassuring to note that the device is accurate with both samples. We would also recommend that future studies to determine reference ranges and thresholds for urinary proteins must be specimen specific.

The main methodological problem with the validation of automated urinalysis devices is the acceptance of a gold standard. For the DCA® 2000 microalbumin assessment we chose the laboratory assay for microalbuminuira in use at our hospital. The laboratory already had an established automated assay for microalbumin and ran the assay daily for other departments. This choice is both a pragmatic approach and in view of the very low intra‐assay variability in the laboratory, a reliable one.

The accepted protocols for the validation of blood pressure measurement devices have either a grading system based on the number of readings that are within set thresholds of accuracy (BHS) [12], or an acceptable mean difference (AAMI) [13] that determines machine accuracy. No such standards exist for urinalysis. It is accepted that the CV for microalbumin measurement is always higher with turbidometric methods and that this is usually worse at low albumin concentrations [14]. We could find only one recommendation for the measurement of urinary albumin [29]. The authors suggest that any method for monitoring microalbuminuria should have an interassay precision of <12% over the concentration range 5–200 mg/L and be sensitive enough to reliably detect changes of 10 mg/L in the concentration range 5–35 mg/L. The DCA 2000 meets these criteria when tested in both uncomplicated and hypertensive pregnancy, but like any method of estimation, larger differences between methods are seen at higher protein concentrations, which are significantly above clinical thresholds for defining normal populations. The Bland Altman Plots with 95% limits of agreement demonstrate how device accuracy is dependent on protein concentration.

The DCA 2000 is accurate for the measurement of albumin creatinine ratios in the uncomplicated pregnant population. In the hypertensive population the DCA 2000 remains accurate, though when the albumin concentration is greater than 40 mg/L, the 95% limits of agreement are broader. Now that accuracy has been established further studies are required to determine the reference ranges and thresholds for the prediction of clinical endpoints with this new technology.

The authors thank the Chemical Pathology laboratories at Leicester Royal Infirmary. Dr. J. Waugh was part funded by a grant in aid from Bayer Corporation, Elkhart, IN, who also provided the urinalysis equipment.