Validation of a hand-held point of care device for lactate in adult and pediatric patients using traditional and locally-smoothed median and maximum absolute difference curves

Background: Lactate is commonly used in septic patients and is a viable biomarker for trauma patients. Its pre-hospital use could assist triaging and managing patients with these conditions. Methods: We evaluated the analytical performance of the point-of-care (POC) StatStrip Xpress Lactate Meter (Nova Biomedical) and compared it to the ABL 800 (Radiometer). We measured lactate in 250 adult and 250 pediatric whole blood samples in 2 laboratories. The performance of the POC meter was assessed by traditional linear regression and Bland-Altman plots, and locally-smoothed (LS) median absolute difference and maximum absolute difference (MAD and MaxAD) curves. Results: The StatStrip was linear with acceptable reproducibility at clinically relevant concentrations. Correlation with the ABL800 showed a negative bias for both populations with slope, bias +/- SD (% bias) of 0.78, -0.4 +/- 0.7 (-14.5%) in children and 0.80-03 +/- 0.6 (- 13.3%) in adults. The proportional bias appeared more significant at concentrations >4 mmol/l (36.0 mg/dl). The StatStrip misclassified 7.6 and 8.8% pediatric and adult samples, respectively, to lower risk categories defined using guidelines driven cut-offs. The LS MAD curves identified one breakout, concentration where the LS MAD exceeds the total allowable error limit of 03 mmol/l (2.7 mg/dl), at lactate concentrations of 3.8 and 3.2 mmol/l (34.2 and 28.8 mg/dl) in the pediatric and adult curves, respectively. Breakthroughs, points at which the LS MaxAD curve exceeds the 95th percentile of MaxADs, occur at concentrations above 7.5 mmol/l (67.6 mg/dl) for both populations where the performance of the POC meter became erratic. We concluded that if serial lactate measurements are performed, the same method should be used for baseline and follow up measurements. The LS MAD and LS MaxAD curves allowed visual and quantitative mapping of the performance of the lactate POC meter over the range of concentrations measured. Conclusions: This approach seems useful for the identification of points at which the performance of a POC meter differs significantly from a comparison method and thresholds of poor analytical performance. (C) 2017 Elsevier B.V. All rights reserved.