Comment on Wang et al. Development of a Novel Double Antibody Sandwich ELISA for Quantitative Detection of Porcine Deltacoronavirus Antigen. Viruses 2021, 13, 2403

We were interested in reading an article published by Wang et al. [1] in Viruses on 30 November 2021. The authors developed a double-antibody sandwich enzyme-linked immunosorbent assay (DAS-ELISA) for porcine deltacoronavirus (PDCoV) detection using a monoclonal antibody against the PDCoV N protein and an anti-PDCoV rabbit polyclonal antibody. They used kappa analysis to determine the consistency between the DAS-ELISA and reverse transcriptase real-time PCR (RT-qPCR). The kappa value obtained was 0.827, indicating an almost perfect agreement between the two methods.

Although this article has provided valuable information, some substantial points that may cause misinterpretation of the results need to be clarified. Kappa analysis includes Cohen’s and Fleiss’ kappa analyses. Generally, Cohen’s and Fleiss’ kappa analyses are used to analyze intra- and inter-rater agreements, respectively [2]. Cohen’s kappa analysis is suitable for evaluating two raters, whereas Fleiss’ kappa analysis is suitable for evaluating more than two raters. In Cohen’s kappa analysis, the weighted kappa should be used to calculate the agreement in the presence of more than two categories [3]. According to this article, Cohen’s kappa was applicable based on this situation. Cohen’s kappa was calculated as follows:

The value of $p$ and $q$ are the sample frequency.

The authors compared the two detection methods using two types of clinical samples, namely, 205 fecal and 59 intestinal samples. Unlike the authors, we calculated the agreement between the RT-qPCR and DAS-ELISA by using the two types of samples with SPSS 18 statistical package (SPSS 18 Inc., Chicago, IL, USA) software. The kappa values in the fecal and intestinal samples were 0.807 and 0.645, respectively. Furthermore, a simple sum of the data was performed, and the kappa value obtained was 0.781 (Table 1). The three kappa values were significantly different from the authors’ kappa value of 0.827. We would be grateful if the authors could explain their results in detail and clarify the misunderstanding.

Considering the applicability of Cohen’s kappa analysis, we suggest that the kappa values should be calculated in the presence of two or more types of samples. In our opinion, any agreed conclusions must be supported by methodological and statistical methods. We emphasize the importance of rigor and using the correct statistical approach in any scientific publication. Otherwise, misinterpretation cannot be avoided.