John Metcalfe, MD, MPH Division of Pulmonary and Critical Care Medicine, San Francisco General Hospital, and Curry International Tuberculosis Center

Transcription

1 WITHIN TEST TEST VARIABILITY OF THE QUANTIFERON TB TB GOLD IN TUBE ASSAY IN A LARGE COHORT OF LOW RISK SUBJECTS John Metcalfe, MD, MPH Division of Pulmonary and Critical Care Medicine, San Francisco General Hospital, and Curry International Tuberculosis Center Introduction Methods Overview Results Conclusions and Recommendations

2 Test Precision: Two Basic Components Reproducibility Closeness of agreement among successive measurements obtained under changed conditions Time Operator Calibrator Reagents Laboratory Biologic variability of subjects tested Repeatability Closeness of test results when no conditions of measurement change Extent to which a test will produce the same result when used on the same specimen in identical circumstances Solely related to the random error of measurements Banoo Nat Rev Micro 2006 Why should we care about test variability? High variability can adversely affect accuracy Crucial component of diagnostic test evaluation Required by the US FDA prior to premarket approval Integral part of international reporting standards (STARD) Clinical interpretation of test results borderline zones conversion/reversion

3 Methods: Research Aim Describe QFT GIT repeatability in a large population of low risk individuals living in a low TB incidence setting, with attention to both dichotomous and quantitative test results. Subjects Evaluated for LTBI, Aug 1, 2010 July 31, 2011 (n=3234) Subjects (n=370) with positive, negative with high nil, or borderline QFT GIT (813 total tests) QFT negative, nil<0.34 IU/ml, n=2681 QFT indeterminate, n=182 (5.6%) Missing replicate, n=1 One re test available for analysis (n=326) Two re tests available for analysis (n=17) Three or more re tests available for analysis (n=27)

4 Results: Dichotomous Baseline QFT-GIT Total Subjects Conversion/Reversion, n (%) All subjects (8) Baseline quintiles of IFN-y concentration, (IU/ml) < (0) (15) (19) (3) (1) Results: Quantitative Repeatability of IFN-y Response 0 2 Antigen - nil (IU/mL) Test

5 Repeatability of IFN-y Response Antigen - nil (IU/mL) Test Absolute Variability about Patient-specific Mean 95% Variability Frequency (n) Within-run Difference from Patient-specific Mean (%)

6 Coefficent of Variation (%) Van Zyl Smit PLOSOne 2011; Tuuminen CVI 2010; Cellestis -2 Difference (Test0 - Test ) Average of Test0 and Test1

7 Results: Summary Dichotomous: Conversion/reversion of initial test concentrated in 2 nd and 3 rd quintile of IFNγ response ( IU/ml) 17% (n=25/149) of all subjects Quantitative: Repeatability of test results was marginal to poor 95% variability of IFN γ response occurred within ihi 33% of the mean IFN γ response even when analysis restricted to subjects with a single retest Considerations Our estimates may be biased high Focused on low risk subjectswithunexpected unexpected results Other (likely minor) potential sources of error Plasma storage time, two operators, between lot variability, etc. However, our estimates may be biased low Only looked at repeatability component of total variability Highervariability expected in setting of serial testing Low risk population Higher risk populations may have higher variability Focused on most conservative estimates Sensitivity analyses excluded subjects with low reactivity, and those with greater than 1re test

8 Conclusions and Recommendations As in other immunodiagnostic methods, analytic lti imprecision i i does exist it especially important in test results near cut point Serial IGRA results should be interpreted cautiously Retesting of borderline results in low risk subjects seems prudent before offering preventive chemotherapy Acknowledgements Dr. Edward A. Graviss, The Methodist Hospital Research hinstitute t UCSF Department of Epidemiology and Biostatistics Chuck McCulloch Eric Vittinghoff

9 Supplementary Slides Frequency (n) Untransformed Frequency (n) Log-Transformed Ag-nil (IU/mL) Log (Ag-nil)

10 Krouwer Clin Chem 1984