Performance of an optimized paper-based test for rapid visual measurement of alanine aminotransferase (ALT) in fingerstick and venipuncture samples

Abstract

Background: A paper-based, multiplexed, microfluidic assay has been developed to visually measure alanine aminotransferase (ALT) in a fingerstick sample, generating rapid, semi-quantitative results. Prior studies indicated a need for improved accuracy; the device was subsequently optimized using an FDA-approved automated platform (Abaxis Piccolo Xpress) as a comparator. Here, we evaluated the performance of the optimized paper test for measurement of ALT in fingerstick blood and serum, as compared to Abaxis and Roche/Hitachi platforms. To evaluate feasibility of remote results interpretation, we also compared reading cell phone camera images of completed tests to reading the device in real time. Methods: 96 ambulatory patients with varied baseline ALT concentration underwent fingerstick testing using the paper device; cell phone images of completed devices were taken and texted to a blinded off-site reader. Venipuncture serum was obtained from 93/96 participants for routine clinical testing (Roche/Hitachi); subsequently, 88/93 serum samples were captured and applied to paper and Abaxis platforms. Paper test and reference standard results were compared by Bland-Altman analysis. Findings: For serum, there was excellent agreement between paper test and Abaxis results, with negligible bias (+4.5 U/L). Abaxis results were systematically 8.6% lower than Roche/Hitachi results. ALT values in fingerstick samples tested on paper were systematically lower than values in paired serum tested on paper (bias -23.6 U/L) or Abaxis (bias -18.4 U/L); a correction factor was developed for the paper device to match fingerstick blood to serum. Visual reads of cell phone images closely matched reads made in real time (bias +5.5 U/L). Conclusions: The paper ALT test is highly accurate for serum testing, matching the reference method against which it was optimized better than the reference methods matched each other. A systematic difference exists between ALT values in fingerstick and paired serum samples, and can be addressed by application of a correction factor to fingerstick values. Remote reading of this device is feasible