Creatinine measurement from finger stick dried blood spots with a routine chemistry analyzer for estimation of GFR

DBS extraction and creatinine analysis

Previously, accurate measurements of CRP and TSH from DBS on a routine chemistry analyzer were validated [9], 10]. Here, we sought to apply the same protocol to creatinine. DBS extraction was performed as previously reported [9], 10]. 40 µL whole blood was spotted on Whatman 903™ filter paper (Eastern Business Forms LLC, Greenville, USA) and left to dry overnight at room temperature. Three dried EDTA blood spots of 8 mm diameter were transferred to an Eppendorf tube. 450 µL PBS was added and samples were incubated at room temperature for 15 min on an orbital shaker at 600 rpm. After centrifugation at 21,130 g for 3 min at room temperature, the supernatant was transferred to analyzer cups (Tube-Top sample cups, Atellica^® Solution, Siemens Healthcare Diagnostics Inc., Tarrytown, USA) without any modification to the sample volume. The Atellica analyzer processed these eluates as if they were standard plasma samples. Creatinine in DBS extracts was measured enzymatically (ECre3) in triplicate using the Atellica CH Analyzer (Siemens Healthcare Diagnostics Inc., Tarrytown, USA) following the manufacturer’s instructions to obtain a single result. In venous samples, creatinine was measured in lithium heparin (LH) plasma.

The study was performed under the tenets of the Helsinki Declaration and all relevant national regulations and institutional policies. The study protocol was approved by the authors’ institutional review board (23U-0136).

Calibration and quality control

To convert creatinine concentrations from DBS to plasma-equivalent concentrations, two calibration lines were prepared: venous EDTA whole blood vs. venous LH plasma and capillary finger stick whole blood vs. venous LH plasma. The venous EDTA whole blood calibration curve consisted of 20 samples spanning a range of 18–615 μmol/L. The capillary finger stick whole blood calibration curve consisted of 5 samples spanning a range of 56–346 μmol/L. After spotting and drying, DBS were extracted and measured as described above. The venous EDTA whole blood calibration line was used for experiments addressing precision, stability, influence of hematocrite (Hct), and spotting effects. The finger stick calibration line was used for method comparison. DBS creatinine concentrations were corrected for Hct (see below). Subsequently, the corresponding calibration lines were used offline (i.e. post analyzer) and no additional computational adjustments were made. For method comparison the following equation was used: plasma DBS creatinine converted to plasma = (DBS creatinine Hct corrected)/0.07325.

Quality and stability of the analyzer calibration was monitored by measuring quality control (QC) samples both directly and after the spotting and drying process as described above (Liquid Assayed Multiqual Premium Levels (means of 24, 99 and 489 μmol/L), Bio-Rad Laboratories, Clinical Diagnostics Group, CA, USA) before every run. Spotted QC samples were calculated from the off-line calibration curve without Hct correction because of the absence of erythrocytes. For each run, QC of Hct measurement using near-infrared (NIR) spectroscopy was performed with EDTA whole blood DBS samples with known Hct concentrations. QC samples confirmed analyzer calibration accuracy as batches with QC exceeding ± 2 SD were excluded from analysis. This appeared to be the case for one batch.

Precision

Precision was assessed following the CLSI EP15-A3 guideline, utilizing an extended protocol. Venous EDTA whole blood patient pools consisting of three patients with matching blood types were made with three different creatinine concentrations: low (creatinine: 52 μmol/L), medium (creatinine: 103 μmol/L), and high (creatinine: 337 μmol/L). EDTA whole blood samples were spotted and dried as previously described. For seven consecutive days, every pool was spotted and extracted four times and measured in duplicate.

Influence of hematocrit on creatinine concentration from DBS

To investigate the influence of Hct, plasma pools were created and mixed in different ratios with red blood cells to generate a Hct range. Briefly, pooled venous EDTA whole blood samples, consisting of five patient samples per pool, were centrifuged at 2,500 g for 7 min. Plasma was divided into six tubes, and the cell fraction was added to obtain a Hct range of 0.1–0.6 L/L. This protocol was performed for three pools with low, medium, and high plasma creatinine concentration (50, 97, and 292 μmol/L, respectively) to investigate the influence of Hct on different creatinine levels. After mixing samples by inversion, Hct concentrations were measured using a Cell-Dyn Sapphire Hematology Analyzer (Abbott, Abbott Park, IL, USA). Measurement of Hct in DBS was performed by NIR-spectroscopy on a NIRFlex N-500 Fiber Optic Solids (BÜCHI Labortechnik, Hendrik-Ido-Ambacht, the Netherlands) [11]. DBS creatinine levels were plotted against measured Hct. The slope found for the effect of Hct on the measurement of creatinine from DBS was used to correct subsequent creatinine measurements from DBS for this effect with the following equation: Hct corrected DBS creatinine=uncorrected creatinine DBS – (Hct* –23.18+9.27).

Method comparison

In the method comparison, 122 patients presenting to the outpatient clinic of the University Medical Center (UMC) Utrecht were included after giving written informed consent.

Venous blood sampling was performed via venipuncture with the BD Vacutainer^® Eclipse™ Blood Collection needle with Pre-Attached Holder (Becton Dickinson, Franklin Lakes, USA) and collected in VACUTEST^® 3.6 mg K₂EDTA tubes (2 mL, VACUTEST KIMA, Arzergrande, IT) and BD Vacutainer^® Barricor™ LH plasma tubes (3 mL, Becton Dickinson, Franklin Lakes, USA).

In addition to venipuncture, capillary blood samples were obtained by the finger stick technique in accordance with CLSI standard GP42-A6. Briefly, the puncture site was cleaned with a tissue containing isopropyl alcohol 70 % and allowed to dry. The skin was punctured using a BD Microtainer^® contact-activated lancet (Becton Dickinson, Franklin Lakes, USA) to a depth of 2.0 mm and the first drop of blood was wiped off. If necessary, the skin was warmed before the puncture with a gel pack of 42 °C for a maximum of 3 min. Blood drops were collected on Whatman 903™ filter paper. Samples that were multi-spotted, unable to fully penetrate the back side of the filter paper, or <8 mm in diameter were excluded from analysis. Creatinine was measured in LH plasma within 4 h after sampling. Two samples were excluded from analysis; one due to double spotting and one because of the high creatinine level (this data point is reported in the figure legend).

From the creatinine levels, eGFR, P30, and P10 were derived. eGFR was computed with the Chronic Kidney Disease Epidemiology collaboration (CKD-EPI, 2009) formula [12] which corrects for sex, age, and body surface without correcting for race as recommended by the EFLM [13]. P30 and P10 are defined as the percentage of patients for which the creatinine in DBS is within 30 and 10 % range of plasma creatinine levels, respectively [14]. eGFR above 90 mL/min/1.73 m² was depicted as 91 since the formula was validated in patients with chronic kidney disease and values higher than 90 are not reported in clinical practice. Since this will skew regression analysis and P30 and P10 values, we excluded all patients above 90 mL/min/1.73 m² for these analyses. For sensitivity, specificity, positive predictive value, and negative predictive value, 60 mL/min/1.73 m² was used as clinical cut-off.

Time and temperature stability of creatinine from DBS

In clinical practice, DBS have to travel at various temperatures and for different days before reaching the laboratory. To investigate creatinine stability in this process, DBS were spotted and dried at room temperature for 3 h before storing them at temperatures of −20 °C, +4 °C, +20 °C, or +37 °C for 1, 2, 3, 6, 14, and 31 days. During DBS extraction of samples stored at −20 °C, repeated freeze–thaw cycles were analyzed simultaneously. Therefore, day 1 samples were freeze–thawed once, day 2 samples were freeze–thawed twice, etc.

Spot effects on analysis of creatinine from DBS

Outside controlled laboratory environments, spotting variations may occur and hamper accurate diagnostics. Therefore, the effect of double spotting, different blood volumes, and number of spots was investigated (all samples were spotted in triplicate).

Statistical analysis

Graphs were generated using GraphPad Prism 9 (Version 9.3.0; GraphPad Software, LLC, San Diego, USA). EP Evaluator (Data Innovations, VT, USA; version 12.2) was used to evaluate precision, Deming regression analysis, and bias. 95 % confidence interval (CI) for R² was calculated according to R² ± 2 × SER2, where SE R 2 = 4 R 2 1 − R 2 2 n − k − 1 2 n 2 − 1 n + 3 in which ‘‘k’’ is the number of independent variables and ‘‘n’’ is the number of cases.

The Clinical Laboratory Improvement Amendments (CLIA) requirements for analytical quality for creatinine are: target value ± 18 μmol/L or ±10 % (total allowable error (TEa)). TEa was used to analyze spotting effects by comparing the average values (target values) of venous creatinine and DBS measurements. Stability was analyzed by comparison with CLIA acceptance limits.

Precision

Evaluation of precision in patient samples showed that total coefficient of variation was between 4.1 and 17.7 % for creatinine concentrations of 345 and 49 μmol/L, respectively (Table 1).

Dried blood spot creatinine levels are negatively correlated with hematocrit

Since Hct can bias DBS measurements, we examined the potential effect of Hct. A consistent negative correlation between DBS creatinine levels and Hct was observed across all creatinine concentrations studied (Figure 1A and Table 2). To determine the effect of Hct on the measurement of creatinine from DBS, uncorrected DBS creatinine values were plotted against Hct (Figure 1B). Creatinine was corrected for Hct using the formula derived from the intermediate level, as it closely aligns with the clinical cut-off. Using two correction formulas (one for high and one for medium/low levels) did not improve results (data not shown). Thus, creatinine levels in DBS showed a negative correlation with Hct. However, applying the correction effectively minimized the impact of Hct.

Figure 1:

Influence of hematocrit on the concentration of creatinine from DBS. Pools from three patients with three different target concentrations were made. From each pool, a Hct dilution range was created. Then, blood was spotted on DBS cards and dried overnight before analysis in triplicate. (A) Shows plasma equivalent creatinine concentrations from DBS. Uncorrected creatine DBS results plotted against Hct measured with a NIRFlex N-500 are shown in (B).

Method comparison of creatinine levels and eGFR in patients using DBS

Subsequently, creatinine levels in finger stick DBS extracts were compared to venous heparin plasma of 122 patients (including 5 patients for the calibration curve, see Table 3 for patient characteristics). One sample was removed from further analysis because of a spotting failure. To convert creatinine concentrations from DBS to plasma-equivalent concentrations, a linear calibration line with samples across the creatinine range was created (R²>0.997) (Figure 2A). The method comparison showed R² [95 % CI] 0.967 [0.955–0.978], slope 1,014 [0.885–1.143], and intercept −4.32 [−16.52–7.89] (Figure 2B). Moreover, the absolute bias [95 % CI] of 2.878 [−32.44–26.69] was not significant between the compared methods (Figure 2C).

Figure 2:

Method comparison of venous lithium heparin (LH) plasma and capillary finger stick DBS. (A) calibration line for DBS creatinine was calculated with linear regression. Range of plasma creatinine: 56–346 μmol/L; n=5. (B) Method comparison for DBS creatinine with the line of identity (red) and deming linear regression (black). Storage time of DBS cards is displayed in ⚫ (1 day, n=82), ◼ (2 days, n=20), ▲ (3 days, n=13), total n=115. Average plasma creatinine [stdev]=112 [112] μmol/L. The highest point (creatinine level of 947 and 1,061 μmol/L in plasma and DBS respectively) is not shown. (C) Bias plot from method comparison of panel B. (D) Method comparison for DBS eGFR with the line of identity (red) and deming linear regression (black). Average eGFR [stdev]=70 [24] L/min/1.73 m²; n=115. Dotted lines show the reference interval. (E) Bias plot from method comparison of panel D. All 95 % confidence intervals are shown in between brackets. DBS, dried blood spot; eGFR, estimated glomerular filtration rate.

We found high correlation with R² [95 % CI] 0.858 [0.80–0.91] and low absolute bias of 0.31 for eGFR [95 % CI] [−15.05–15.67] (Figure 2D and E). In 97.0 and 49.3 % of the DBS, eGFR was within the 30 % (P30) and 10 % (P10) range of plasma eGFR, respectively. Using the clinical cut-off value of 60 mL/min/1.73 m² for eGFR, sensitivity was 91 %, and specificity was 96 % The data used for this calculation are shown in Supplementary Table S1. Similarly, positive predictive value and negative predictive values were 91 and 96 %, respectively. This indicates good correlation between DBS and plasma creatinine and accurate qualification based on the clinical cut-off.

Without Hct correction, sensitivity and specificity were lower, 85 and 93 % respectively. Similarly, R² for creatinine and eGFR were lower without Hct correction (0.96 and 0.82 respectively, see Supplementary Figure S1). Regression analyses of the venous calibration line (n=20) and method comparison (n=60) of venous heparin plasma vs. venous EDTA whole blood DBS extract yielded similar results (Supplementary Figure S2). Thus, creatinine levels and eGFR were predicted with high accuracy using DBS.

Variations in DBS spotting or storage are compatible with spotting at home

Since telediagnostics deal with different storage conditions and inaccuracies with blood spotting by patients, we investigated how these affect creatinine and eGFR measurement from DBS. Creatinine levels appeared stable up to 30 days when stored at −20 °C and 4 °C. In addition, creatinine showed acceptable stability up to 6 days at 20 °C, and up to one day at 37 °C (Figure 3). Interestingly, creatinine levels tended to increase when DBS-cards were stored at 20 °C and 37 °C (Figure 3C and D). This resulted in a decrease in eGFR over time when DBS were stored at higher temperatures. Overall, eGFR stability was comparable to creatinine stability (Supplementary Figure S3). To study the effect of double spotting, blood from five patients with a range of creatinine concentrations was spotted twice with an interval of 30 and 60 s. After double spotting, samples were within CLIA acceptance limits except for non-relevant aberrations at the highest and lowest creatinine levels (Figure 4A). In addition, the effect of spotted blood volume was studied. Notably, all volumes were within CLIA limits except for 25 μL at level 2 and 20 and 25 μL at level 3 (Figure 4B). Finally, creatinine levels remained consistent regardless of the number of DBS spots used (Figure 4C). Thus, spotting effects did not affect creatinine measurement from DBS except for double spotting at a high or low creatinine level and for 20–25 µL spots at median and low creatinine levels.

Figure 3:

Stability of creatinine from DBS. DBS cards with venous EDTA whole blood were stored for a varying number of days (0–31 days, x-axis) under varying storage temperatures. (A) −20 °C, (B) 4 °C, (C) 20 °C and (D) +37 °C per condition n=20. Under each graph, a table testing the results against the acceptance limit of CLIA for proficiency testing is shown (levels exceeding the limit are shown in italics). TV; target value.

Figure 4:

Spot effects on the concentration of creatinine from DBS. (A) Effect of double spotting. Dotted lines show the CLIA criterium of 10 % compared to the target value of the regular conditions (black bars). (B) Effect of volume of blood spotted. Five different blood volumes (20–50 µL) were spotted for three patients across the creatinine spectrum. (C) Effect of the number of spots with proportionate extraction buffer volume. 1, 2, 3, or 4 spots with 150 µL extraction buffer (PBS) per spot were extracted from three patient samples.