| Résumé | Objectives: To determine the feasibility of near-infrared analysis for quantitating urea, creatinine, and protein in urine. Practical advantages of this method include ease of sample presentation and the absence of reagents or disposables. Design and Methods: The near-infrared methods were developed by first measuring the spectra of 123 different urine samples and, using independent clinical analyses, determining the protein, creatinine, and urea levels in each. Calibration models relating near-infrared spectroscopic features to those independently determined concentrations were optimized, and each model then validated using a set of 50 additional samples. Results: Standard errors of calibration were 14.4 mmol/L, 0.66 mmol/L, and 0.20 g/L, and standard errors of prediction 16.6 mmol/L, 0.79 mmol/L, and 0.23 g/L, respectively, for urea, creatinine, and protein. Conclusions: Near-infrared urea quantitation is as accurate as the reference method, enzymatic (urease) conductivity, used here for calibration. Creatinine analysis is slightly less accurate relative to the reference (Jaffé rate) method; however, these errors can be minimized by careful attention to factors affecting precision. The accuracy of the near-infrared protein analysis cannot approach that of the reference method; nevertheless, the technique is potentially useful for coarse screening and for quantifying protein levels above 0.3 g/L. |
|---|
