Comparison of the sensitivity, specificity, correlation and inter‐assay agreement of eight diagnostic in vitro assays for the detection of African swine fever virus

With the recent spread of African swine fever (ASF) in Europe, Asia and the Caribbean region, after being endemic for decades in Africa, PCR-based commercial kits and various master mixes are increasingly being used in addition to the Office International des Epizooties-recommended protocol from King et al. (World Organization for Animal Health). Often, the availability and cost of commercial kits or master mixes can be a limiting factor for diagnostic laboratories, in addition to the requirements for transportation and storage of temperature-sensitive reagents in remote areas. In such cases, alternatives should be ready to maximize surveillance and mining of ASF. To evaluate alternatives, we tested five commercial quantitative real-time PCR (qPCR) master mixes from Applied Biosystems, Bio-Rad, Biotechrabbit, Promega and Qiagen using the same primers and probe mix derived from the King et al.’s protocol for the sensitivity, specificity, correlation and inter-assay agreement. We further included three ad hoc molecular diagnostic kits (VetMax™ African Swine Fever Virus Detection Kit [Applied Biosystems], ID Gene African Swine Fever Duplex [ID-Vet] and Virotype ASF PCR Kit [Qiagen/Indical]). The limit of detection (LOD) was assessed for each assay. The comparative study panel comprised 83 archived DNA samples from ASF virus (ASFV) clinical samples, belonging to five different genotypes from outbreaks in 16 countries in Asia and Africa. The analytical specificity was assessed against a panel of swine pathogens. The LOD ranged from 13 to 41 gene copies per reaction; VetMax ™ African Swine Fever Virus Detection Kit from Applied Biosystems exhibited the lowest detection limit (13 gene copies per reaction) and iQ Supermix from Bio-Rad the highest detection limit (41 gene copies per reaction). Cq values obtained from the lowest dilution, in which all replicates (n = 25) could still be amplified (50 gene copies per reaction), were not significantly different between kits using Kruskal–Wallis test. Inter-assay agreement was assessed using statistical test Fleiss–Kappa and was shown to be excellent in all cases. Agreement using statistical test Bland–Altman was good for samples with Cq values <25 and moderate for Cq values >25. We conclude that all the assays evaluated in this study can be used for the routine detection of ASFV.