Characteristics of diagnostic tests used in the 2002 low-pathogenicity avian influenza H7N2 outbreak in Virginia.

An outbreak of low-pathogenicity avian influenza (LPAI) H7N2 occurred in 2002 in the Shenandoah Valley, a high-density poultry production region in Virginia. Infected flocks were identified through a combination of observation of clinical signs and laboratory diagnostic tests designed to detect avian influenza (AI) antibodies, virus, or H7-specific RNA. In this report, fitness for purpose of 3 virus/RNA detection assays used during the outbreak was examined: 1) antigen capture enzyme immunoassay (AC-EIA), 2) real-time reverse transcription polymerase chain reaction (RRT-PCR), and 3) virus isolation (VI). Results from testing 762 turkey and 2,216 chicken tracheal swab pooled specimens were analyzed to determine diagnostic sensitivities and specificities of these tests under field conditions using Bayesian techniques for validation of diagnostic tests in the absence of a "gold standard." Diagnostic sensitivities (with 95% probability intervals) in turkeys of AC-EIA and RRT-PCR, in reference to VI, were 65.9 (50.6; 81.3)% and 85.1 (71.9; 95.7)% and of VI 92.9 (78.0; 98.8)% in reference to AC-EIA or 88.7 (76.0; 97.2)% in reference to RRT-PCR; in chickens, diagnostic sensitivities were 75.1 (45.6; 94.2)%, 86.3 (65.9; 97.1)%, and 86.2 (65.8; 97.1)% or 86.3 (66.4; 97.2)%, respectively. Specificities were 99.1 (97.9; 99.8)%, 98.9 (98.0; 99.5)%, and 98.6 (97.4; 99.4)% or 98.8 (97.8; 99.5)% in turkeys and between 99.25% and 99.27% with probability intervals of approximately +/-0.4% for all tests in chickens. Simultaneous use of AC-EIA and RRT-PCR contributed significantly to the rapid control of the outbreak, but the AI RRT-PCR assay with >85% sensitivity and approximately 99% specificity, combined with relatively low cost and fast turnaround, could be used as the sole diagnostic test in outbreaks of LPAI.