Diagnostic accuracy of the interferon-gamma release assay (IGRA) for cases of feline mycobacteriosis

The aim of this study was to evaluate the sensitivity and specificity of the interferon-gamma release assay (IGRA) for diagnosing infections with members of the Mycobacterium (M.) tuberculosis-complex (MTBC) and non-tuberculous mycobacteria (NTM) in domestic cats, and to generate defined feline-specific cut-off values using receiver operating characteristic (ROC) curve analysis to improve test performance. Records of 594 cats that had been tested by IGRA were explored to identify individuals that had a culture and/ or polymerase chain reaction (PCR)-confirmed case of mycobacterial disease, and those that had a final diagnosis of non-mycobacterial disease. A total of 117 cats - 80 with mycobacterial disease and 37 diagnosed with a condition other than mycobacteriosis - were identified for further detailed analysis. This population was used to estimate test sensitivity and specificity, as well as likelihood ratios for the IGRA to correctly identify a cat with or without mycobacterial disease. Agreement between IGRA results and culture/PCR using current and proposed new cut-off values was also determined. ROC analysis of defined confirmed infected and non-mycobacterial disease control cats allowed an adjustment of current test cut-offs that increased the overall test sensitivity for MTBC infections from 83.1 % (95 % confidence interval [CI]: 71.5-90.5 %) to 90.2 % (95 % CI: 80.2 95.4%), and M. bovis infection from 43 % (95 % CI: 28.2 60.7%) to 68 % (95 % CI: 51.4 82.1%) while maintaining high test specificity (100 % in both cases). Overall agreement between IGRA results and culture/PCR, while recognising that neither culture nor PCR tests have perfect sensitivity, improved from weak (kappa = 0.57) to moderate (kappa = 0.71) using new proposed IGRA test cut-off values. Application of these results, based upon the statistical analysis of accumulated test data, can improve the diagnostic performance of the feline IGRA, particularly for identifying infections with M. bovis, without compromising specificity.