Accuracy of the HumaSens^plus point-of-care uric acid meter using capillary blood obtained by fingertip puncture

This was a monocentric cross-sectional study performed at a diabetology day-care center of a French hospital in Paris. The study was approved by the local ethics committee.

The patients included in the study were all consecutively seen in the diabetology day-care center between January 2016 and March 2016. All were diabetic, 18 years or older, and were able to understand the requirements and purpose of the study. Patients with gout were not excluded. Patients received a written information sheet with details on the study. All gave informed consent before any study procedure. A subset of patients agreed to sign a specific informed consent for LC-MS measurement. A venous puncture for standard blood biochemistry, which included plasma UA, and a fingertip puncture for capillary glycemia measurement were routinely scheduled in the diabetology day-care center. No supplementary puncture was necessary for this study.

Capillary UA was measured using the HumaSens^plus UA meter (European Conformity marked and approved for EU market use only) and compared to a standard UA measurement in the hospital biochemistry laboratory that used a venous sample taken and prepared as Li-Heparin plasma and an Abbott uricase automated colorimetric assay. LC-MS was performed on frozen Li-Heparin samples in AstraZeneca’s Protein Biomarkers Laboratory (Precision Medicine and Genomics, Gothenburg, Sweden). Medical information (demographic characteristics, medication, and co-morbidities) were retrieved from medical files.

The HumaSens^plus meter has a dynamic measurement range of 180–1190 μmol/l (3.0 to 20.0 mg/dl). When the meter read LO or HI, indicating that the individual’s UA value was outside the test range, these values were individually compared to corresponding plasma measurements.

The statistical analysis involved all patients with available capillary and plasma UA measurements. Results are expressed as mean ± SD. Capillary and plasma UA measurements were compared by paired Student t test. Differences between methods were calculated for each participant, and the distribution was graphically displayed with kernel density estimation (a nonparametric estimation of density). Agreement between methods was assessed by the intraclass correlation coefficient (ICC) and the Lin’s concordance criterion [12]. When measuring the agreement between pairs of observations, the ICC represents the between-pair variance expressed as a proportion of the total variance of the observations (i.e., the proportion of the total variability in the observations that is due to the differences between pairs). ICC was calculated from a general linear model and was expressed with a two-sided 95% confidence interval (CI). An understanding of Lin’s concordance correlation coefficient is obtained if the line of best fit to the data comparing two methods is shown in a scatterplot plotting the results of one method against the other. The Pearson correlation coefficient provides a measure that describes the extent to which the points in the scatter diagram conform to the best fitting line. Lin’s coefficient modifies the Pearson correlation coefficient by assessing how close the data are about the line of best fit and also how far that line is from the 45-degree line through the origin, this 45-degree line representing perfect agreement. Lin’s coefficient is 1 when all the points lie exactly on the 45-degree line drawn through the origin and diminishes as the points depart from this line and as the line of best fit departs from the 45-degree line [13]. In addition, agreement between methods was assessed by Bland-Altman graphic representation. Confounding factors were searched among age, gender, medications, and biological parameters including fasting glycemia, glycated hemoglobin, creatinine clearance, total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, triglycerides, and hematocrit in a multiple regression model adjusted on plasma UA with stepwise selection of variables. The best capillary UA thresholds predictive of hyperuricemia (target plasma UA of 360 and 300 μmol/l) were determined with receiver operating characteristic (ROC) curves. To estimate the confounding effect of metformin intake, which was frequent in this population of diabetic patients, the same analyses were repeated in patients with or without metformin intake. Calculations showed that 206 paired measurements were required for an ICC of 80% with precision 0.10 and alpha risk 5%.

Additional analyses were performed in the subset of patients with measurements of HumaSens^plus capillary UA, plasma UA determined by uricase automated colorimetric assay, and plasma UA determined by LC-MS. Differences between methods were calculated. ICC and Lin’s concordance coefficient were calculated between each pair of methods. Results for capillary UA and plasma uricase UA were plotted against plasma LC-MS UA results. Concordance between capillary UA and LC-MS UA data and between plasma uricase UA and LC-MS UA data were graphically assessed by the Bland-Altman method. The analysis involved the use of SAS 9.4 (SAS Institute, Cary, NC, USA).

Nineteen capillary samples were marked LO by the meter (capillary UA < 180 μmol/l). A second measurement was performed with the meter, and values were obtained for four patients: 180, 180, 270, and 400 μmol/l. Those values were included in the statistical analysis; the other 15 patients were excluded and analyzed separately. Among these 15 patients, 12 had UA values below 180 μmol/l according to plasma measurement, and three had values above: 189, 206, and 428 μmol/l. Two samples were marked HI by the meter (capillary UA > 1190 μmol/l). These were excluded from the statistical analysis and were measured at 303 and 213 μmol/l by the laboratory uricase assay.

Mean capillary blood UA was 300 ± 83 μmol/l (5.04 ± 1.40 mg/dl) and mean plasma UA was 309 ± 90 μmol/l (5.19 ± 1.51 mg/dl). The mean difference between capillary blood and plasma UA was −8.6 ± 36.9 μmol/l (-1.4 ± 6.2 mg/dl; p = 0.0010). The linear regression formula was capillary UA = 39.21 + 0.85 × plasma UA (Fig. 1).

Agreement between capillary blood and plasma UA was high: ICC was 0.90 (0.87–0.92) and Lin’s concordance coefficient was 0.91 (0.88–0.93) (Fig. 2). The Bland-Altman plot (Fig. 3) showed good agreement over all tested values. The estimated bias was 8.6 μmol/l (0.144 mg/dl) with the following 95% limits of agreement: −80.9 μmol/l to 63.7 μmol/l, corresponding to −1.36 to 1.07 mg/dl.

Accuracy of the capillary UA measure with the HumaSens^plus meter was 93.2% at the 20% margin.

To detect hyperuricemia with the HumaSens^plus meter, defined by a plasma UA level > 360 μmol/l (6.05 mg/dl), the best capillary UA threshold according to the ROC curve was 340 μmol/l, with sensitivity and specificity 89% (Fig. 4a). With the definition of plasma UA level > 300 μmol/l, the therapeutic target UA level for patients with severe gout, the best capillary UA threshold was 299 μmol/l, with sensitivity 85% and specificity 92% (Fig. 4b). Applying the regression formula, a plasma UA threshold of 360 μmol/l corresponded to capillary UA level of 343 μmol/l and a plasma UA threshold of 300 μmol/l corresponded to capillary UA level of 293 μmol/l.

Among potential confounding factors, only hematocrit significantly influenced capillary UA measurements (p < 0.0001); however, improvement of the determination coefficient when correcting for hematocrit was negligible, from 0.83 to 0.86. The main types of medication patients received are shown in Table 2. No medication appeared to significantly affect the test results. However, although not statistically significant, the difference between capillary and plasma UA measurement seemed higher with than without metformin (−13.6 versus −6.9 μmol/l; p = 0.27).

To better understand some discordances between the capillary and routine plasma UA results, these values were compared with plasma UA measurements with the LC-MS reference method in a subgroup of 67 patients (including 24 of 26 with the most discordant results) (Table 3). Agreement between capillary UA and routine plasma UA measurements was acceptable (ICC 0.83 (0.74–0.89), Lin’s concordance coefficient 0.83 (0.76–0.91)). Similar results were found when comparing capillary UA and LC-MS UA measurements (ICC 0.84 (0.75–0.90) and Lin’s concordance coefficient 0.84 (0.77–0.91)). Agreement between routine plasma UA and LC-MS measurement UA was excellent (ICC 0.96 (0.94–0.98) and Lin’s concordance coefficient 0.96 (0.94–0.98)). Bland-Altman plots (Fig. 5) showed that the difference between plasma LC-MS and plasma uricase UA measurements increased with increasing UA level, whereas the difference between plasma LC-MS and capillary UA measurements was independent of the mean UA level, but with a higher dispersion.