Link to home

The Use of Latent Class Analysis to Estimate the Sensitivities and Specificities of Diagnostic Tests for Squash vein yellowing virus in Cucurbit Species When There Is No Gold Standard

December 2013 , Volume 103 , Number  12
Pages  1,243 - 1,251

William W. Turechek, Craig G. Webster, Jingyi Duan, Pamela D. Roberts, Chandrasekar S. Kousik, and Scott Adkins

First, second, third, and sixth authors: United States Department of Agriculture–Agricultural Research Service (USDA-ARS), U.S. Horticultural Research Laboratory, 2001 South Rock Road, Fort Pierce, FL 34945; fourth author: University of Florida–IFAS, Southwest Florida Research and Education Center, 2685 State Road 29 North, Immokalee 34142; and fifth author: USDA-ARS, U.S. Vegetable Laboratory, 2700 Savannah Highway, Charleston, SC 29414.

Go to article:
Accepted for publication 27 June 2013.

Squash vein yellowing virus (SqVYV) is the causal agent of viral watermelon vine decline, one of the most serious diseases in watermelon (Citrullus lanatus L.) production in the southeastern United States. At present, there is not a gold standard diagnostic test for determining the true status of SqVYV infection in plants. Current diagnostic methods for identification of SqVYV-infected plants or tissues are based on the reverse-transcription polymerase chain reaction (RT-PCR), tissue blot nucleic acid hybridization assays (TB), and expression of visual symptoms. A quantitative assessment of the performance of these diagnostic tests is lacking, which may lead to an incorrect interpretation of results. In this study, latent class analysis (LCA) was used to estimate the sensitivities and specificities of RT-PCR, TB, and visual assessment of symptoms as diagnostic tests for SqVYV. The LCA model assumes that the observed diagnostic test responses are linked to an underlying latent (nonobserved) disease status of the population, and can be used to estimate sensitivity and specificity of the individual tests, as well as to derive an estimate of the incidence of disease when a gold standard test does not exist. LCA can also be expanded to evaluate the effect of factors and was done here to determine whether diagnostic test performances varied among the type of plant tissue being tested (crown versus vine tissue), where plant samples were taken relative to the position of the crown (i.e., distance from the crown), host (i.e., genus), and habitat (field-grown versus greenhouse-grown plants). Results showed that RT-PCR had the highest sensitivity (0.94) and specificity (0.98) of the three tests. TB had better sensitivity than symptoms for detection of SqVYV infection (0.70 versus 0.32), while the visual assessment of symptoms was more specific than TB and, thus, a better indicator of noninfection (0.98 versus 0.65). With respect to the grouping variables, RT-PCR and TB had better sensitivity but poorer specificity for diagnosing SqVYV infection in crown tissue than it did in vine tissue, whereas symptoms had very poor sensitivity but excellent specificity in both tissues for all cucurbits analyzed in this study. Test performance also varied with habitat and genus but not with distance from the crown. The results given here provide quantitative measurements of test performance for a range of conditions and provide the information needed to interpret test results when tests are used in parallel or serial combination for a diagnosis.

This article is in the public domain and not copyrightable. It may be freely reprinted with customary crediting of the source. The American Phytopathological Society, 2013.