ISKNV, a megalocytivirus, produced the largest testing base and the highest cumulative positivity of the extensively sampled agents: 23.3% across 30,917 tests. Lactococcus spp. reached 18.4% across 8,589 tests, while Streptococcus spp. reached 15.2% across 12,629 tests. The network covered 280 companies operating at 154 culture sites in 16 states. Most samples came from the Southeast, which contributed 75.6% of all diagnostic records, compared with 13.3% from the Northeast and 11.1% from the Central-West. The Southeast consequently remained the principal production core for detection, while expanding foci in the Northeast provided the clearest evidence of geographic change.

Several patterns in the Northeast were compatible with genuine northward expansion rather than simply more intensive testing. ISKNV positivity rose from 0.2% in 2023, based on 1,484 tests, to 14.6% in 2024 and 17.8% in 2025. Streptococcus spp. detection increased from zero in 2023 to 14.8% in 2024 and 15.5% in 2025, while Lactococcus spp. rose from zero to 1.3% and then 5.9%. Georeferenced records showed the number and northern reach of positive localities increasing for these three agents between 2023 and 2025. By contrast, Edwardsiella spp. remained largely anchored to the Southeast, and Francisella orientalis was absent from the Northeast in all three years. Because regional sampling also expanded, the authors treated the Northeast signal as an inference requiring prospective confirmation rather than a demonstrated causal pathway.

Time-series analysis provided a more qualified view of endemicity and change. Augmented Dickey-Fuller tests classified the series for Aeromonas spp., Edwardsiella spp., F. orientalis, ISKNV and Lactococcus spp. as stationary, a property compatible with broadly stable circulation but not proof that a pathogen was endemic. Streptococcus spp. was non-stationary, with an ADF p-value of 0.267, and its positivity rose steadily from 10.7% in 2023 to 13.5% in 2024 and 18.6% in 2025 while the testing base expanded. This combination was interpreted cautiously as consistent with directional geographic expansion. Flavobacterium spp. was also non-stationary, but its abrupt changes occurred against very small period-specific denominators, making sampling instability a more plausible explanation than sustained epidemiological change.

Quarterly aggregation produced the lowest prediction error for every pathogen among the monthly, four-monthly, weekly and quarterly intervals evaluated, with mean absolute errors ranging from 0.068 for Edwardsiella spp. to 0.181 for Lactococcus spp. The result identifies quarterly sampling as a reasonable candidate for surveillance design, rather than a formally optimized standard that has already been validated prospectively. Residual distributions were non-normal for most pathogens and frequencies, indicating that uncertainty intervals derived from these models require caution. The surveillance data also showed that host developmental stage had heterogeneous effects on pathogen detection. Stage improved pathogen-specific model fit for Aeromonas spp., Streptococcus spp. and Lactococcus spp., but not for the other four agents; across the pooled pathogen assemblage, its influence was weak.

Mixed-effects models accounted for clustering of records within culture sites, epidemiological weeks and individual observations. A model with all three random effects was best supported by information criteria, and fixed-effect estimates remained directionally consistent with the simpler generalized linear models, indicating that farm-level clustering did not create the principal patterns. Between-site variation was substantial for every pathogen. Developmental stage was especially influential in models for Lactococcus spp. and ISKNV, while year was significant for most agents and macro-region generally was not after site was included. These results show why simple national averages can conceal the local variation that matters in farm health surveillance, while also indicating that estimates for sparsely sampled regions and production stages remain less precise.

Pathogens were usually detected independently rather than in strong combinations. Pairwise correlations ranged from −0.02 to 0.12. The only multi-pathogen network module linked ISKNV, Streptococcus spp. and Lactococcus spp., the three most widely detected agents, through weak associations. These values are compatible with shared environmental and production pressures but do not demonstrate direct biological interaction, pathogen facilitation or a stable co-infection syndrome. The cross-sectional presence-or-absence design also could not establish which agent appeared first. The signal is therefore hypothesis-generating and supports longitudinal work designed to order viral and bacterial detections in time rather than treating the network module as evidence of a causal pathway.

Antimicrobial susceptibility testing showed that resistance patterns were structured by pathogen, antibiotic and their interaction, with developmental stage also affecting minimum inhibitory concentrations (MICs). The analysis included 4,318 isolates across Aeromonas, Edwardsiella, Flavobacterium, Lactococcus and Streptococcus. A Gamma generalized linear model fitted to 160 median-MIC combinations found higher MICs in fry than adults, while juveniles did not differ significantly from adults. Aeromonas spp. had the highest baseline MICs relative to the other modeled genera. Resistant isolates also had significantly higher MICs than susceptible isolates, supporting internal consistency between categorical antibiograms and quantitative measurements.

Pathogen-by-antibiotic interactions showed why susceptibility cannot be generalized across bacterial agents. Edwardsiella spp. and Lactococcus spp. had elevated MICs in combination with enrofloxacin and florfenicol relative to the model’s reference structure, while Streptococcus spp. also showed elevated MICs with enrofloxacin. These are combination-level associations because the model was fitted to median values rather than individual isolates and did not directly model farm-level clustering of isolates. Moreover, aquatic-animal breakpoints were unavailable for some pathogen-drug pairings, so the study reported those MICs numerically without assigning susceptible, intermediate or resistant categories. The results do not establish clinical treatment outcomes, but they show that antimicrobial susceptibility data need to be interpreted separately for each pathogen and compound.

Taken together, the surveillance data challenge a historically Streptococcus-centered view of Brazilian tilapia health and provide a baseline for a more pathogen-specific national system. The clearest priorities include resolving infections to species, serotype and lineage, maintaining surveillance for S. agalactiae serotype Ia, balancing production-stage sampling, and improving coverage outside the Southeast. No serotype Ia was detected by serotype-specific qPCR during the three years, although absence from a non-probabilistic sample cannot establish national absence. Future studies would also need probabilistic recruitment, correction for diagnostic performance, standardized observation units and prospective validation of alert thresholds. Until those steps are completed, the results are most useful as a descriptive map of pathogen circulation, temporal variability and antimicrobial susceptibility in the farms that submitted samples.

Subject of Research: Multi-pathogen epidemiological surveillance of Brazilian Nile tilapia aquaculture

Article Title: Epidemiological surveillance reveals major pathogen dynamics in Brazilian Nile tilapia aquaculture from 2023 to 2025

Article References: Cardoso-Guimarães, M., Fernandez-Alarcon, M. F., Correa, R., Miyoko-Natori, M., Galetti, R., Brito de Souza, A., Squizatto, A. C., Silva, R., Araújo de Almeida, L., Mota Sátiro, T., Ildefonso, R., & Rozas-Serri, M. (2026). Epidemiological surveillance reveals major pathogen dynamics in Brazilian Nile tilapia aquaculture from 2023 to 2025. Frontiers in Veterinary Science, 13, Article 1944605. https://doi.org/10.3389/fvets.2026.1944605

Image Credits: AI Generated

DOI: 10.3389/fvets.2026.1944605

Keywords: Nile tilapia, Brazilian aquaculture, ISKNV, Lactococcus, Streptococcus, epidemiological surveillance, antimicrobial susceptibility, pathogen co-detection, disease ecology, Epidemiological, surveillance, reveals