Within-subject interlaboratory variability of QuantiFERON-TB gold in-tube tests

• Published in: PloS one 2012-09, Vol.7 (9), p.e43790-e43790
• Main Authors: Whitworth, William C, Hamilton, Lanette R, Goodwin, Donald J, Barrera, Carlos, West, Kevin B, Racster, Laura, Daniels, Laura J, Chuke, Stella O, Campbell, Brandon H, Bohanon, Jamaria, Jaffar, Atheer T, Drane, Wanzer, Maserang, David, Mazurek, Gerald H
• Format: Article
• Language: English
• Subjects: Adult; Aerospace medicine; Aged; Analysis; Anticoagulants; Antigens; Biological response modifiers; Calibration; Clinical Laboratory Techniques - methods; Data processing; Data transfer (computers); Disease control; Disease prevention; Enzyme-Linked Immunosorbent Assay; Enzymes; Epidemiology; Error analysis; Female; Health aspects; Health care; Humans; Immunoassays; Infections; Interferon; Interferon-gamma - blood; Laboratories; Male; Mathematics; Medical personnel; Medical tests; Medicine; Middle Aged; Mycobacterium; Observer Variation; Reagent Kits, Diagnostic; Skin tests; Studies; Tuberculin; Tuberculosis; Tuberculosis - blood; Tuberculosis - diagnosis; Tuberculosis - epidemiology; Variability; Young Adult; γ-Interferon
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0043790

The QuantiFERON®-TB Gold In-Tube test (QFT-GIT) is a viable alternative to the tuberculin skin test (TST) for detecting Mycobacterium tuberculosis infection. However, within-subject variability may limit test utility. To assess variability, we compared results from the same subjects when QFT-GIT enzyme-linked immunosorbent assays (ELISAs) were performed in different laboratories. Subjects were recruited at two sites and blood was tested in three labs. Two labs used the same type of automated ELISA workstation, 8-point calibration curves, and electronic data transfer. The third lab used a different automated ELISA workstation, 4-point calibration curves, and manual data entry. Variability was assessed by interpretation agreement and comparison of interferon-γ (IFN-γ) measurements. Data for subjects with discordant interpretations or discrepancies in TB Response >0.05 IU/mL were verified or corrected, and variability was reassessed using a reconciled dataset. Ninety-seven subjects had results from three labs. Eleven (11.3%) had discordant interpretations and 72 (74.2%) had discrepancies >0.05 IU/mL using unreconciled results. After correction of manual data entry errors for 9 subjects, and exclusion of 6 subjects due to methodological errors, 7 (7.7%) subjects were discordant. Of these, 6 (85.7%) had all TB Responses within 0.25 IU/mL of the manufacturer's recommended cutoff. Non-uniform error of measurement was observed, with greater variation in higher IFN-γ measurements. Within-subject standard deviation for TB Response was as high as 0.16 IU/mL, and limits of agreement ranged from -0.46 to 0.43 IU/mL for subjects with mean TB Response within 0.25 IU/mL of the cutoff. Greater interlaboratory variability was associated with manual data entry and higher IFN-γ measurements. Manual data entry should be avoided. Because variability in measuring TB Response may affect interpretation, especially near the cutoff, consideration should be given to developing a range of values near the cutoff to be interpreted as "borderline," rather than negative or positive.