Accuracy of ELISA for the detection of paratuberculosis in cattle: A systematic review and meta-analysis

Paratuberculosis is a chronic disease that poses a threat to the global dairy industry. In the absence of effective treatments and ideal vaccines, screening and culling strategies are the primary means of control. Enzyme-linked immunosorbent assay is the most commonly used method for paratuberculosis screening, and its results have been widely used to estimate the prevalence of paratuberculosis. However, no comprehensive meta-analysis has evaluated the detection performance of ELISA for paratuberculosis in cattle. This systematic review and meta-analysis aimed to fill this gap. We conducted a literature search using "paratuberculosis" and "ELISA" as key words in the PubMed, Web of Science, Elsevier ScienceDirect, and CAB Abstracts databases for relevant articles published in English before March 18, 2025. A total of 4,535 articles were retrieved. After screening articles that included extractable data on individual cattle samples and a reference standard (quantitative PCR or culture), 75 research articles were included in the systematic review. The quality of all 75 articles was then assessed using the VETQUADAS tool. Following the inclusion of all articles in the subsequent analysis, 101 studies were extracted from these articles spanning 42 yr (1983-2024) for the meta-analysis. This included the use of the R software to pool the results, as well as sensitivity analysis, detection of publication bias, assessment of heterogeneity, subgroup analysis, and meta-regression. The pooled results showed that ELISA had a specificity of 0.93 (95% credible interval, 0.91-0.94) and a sensitivity of 0.56 (95% credible interval, 0.50-0.62) for detecting bovine paratuberculosis, with a summary receiver operating characteristic area under the curve of 0.75 (95% credible interval, 0.65-0.83). Substantial heterogeneity was observed across studies (bivariate I² = 75.75%). The sensitivity analysis results demonstrated that the model was generally stable. The studies exhibited substantial heterogeneity and a notable threshold effect, but no significant publication bias was detected. The multivariable meta-regression indicated that number of tests, regional prevalence, ELISA method, and quality grade showed associations with sensitivity (or specificity, or both) after mutual adjustment. Based on the subgroup analysis, meta-regression, and quality assessment, we recommend consistent use of the same ELISA platform over time, interpretation of thresholds according to herd prevalence and disease control objectives rather than default values, and confirmatory follow-up for suspect results. Future studies should employ rigorous sampling methods with clearly described selection criteria to avoid overestimating sensitivity due to spectrum bias, and provide detailed reporting on sample preservation and handling of suspect results. These results offer a scientific basis for the development and interpretation of ELISA results, thereby enhancing the utility of ELISA methods in the surveillance and control of paratuberculosis.