Background: The T-SPOT.TB can be read by an ELISPOT plate imager as an alternative to a labor-intensive and time-consuming manual reading, but its accuracy has not been sufficiently discussed to date.
Methods: 1,423 test results obtained from manual reading using a microscope and an ELISPOT plate imager were compared. The agreement of qualitative test results was assessed using Cohen's kappa coefficient. The relationship of spot counts was studied using Bland-Altman analysis.
Results: The overall percent agreement of the qualitative test results was 95.43% with a kappa coefficient of 0.91. Positive test results with the maximum net spot count of 8 and borderline test results showed relatively high discordance. The agreement of spot counts in panel A, panel B, and nil control was good, and variability did not increase with higher spot counts. On the basis of study findings, a novel strategy for interpreting the test results by an ELISPOT plate imager was proposed.
Conclusions: To increase diagnostic accuracy, positive test results with the maximum net spot count of 8 and borderline test results should be manually confirmed. Our strategy could be a practical guide for laboratories to build their own strategies for interpreting the test results by an ELISPOT plate imager.