College of Medicine, Hallym University, Chuncheon, Korea
Purpose
This study aimed to characterize co-occurrence patterns, statistical associations, and phenotypic clusters of clinical manifestations reported in human toxocariasis case reports indexed in PubMed from 2005 through 2025, inclusive.
Methods
PubMed-indexed case reports from 2005 through 2025 were identified using the query (“Toxocariasis” OR “Toxocara”) AND “Case Reports” [Publication Type] AND 2005:2025[dp] AND ffrft[Filter]. Free full-text articles were retrieved through PubMed Central and other links indexed in PubMed using the ffrft filter. After 7 nonhuman reports and 5 reports without objective diagnostic confirmation were excluded, 139 case reports were included. Antigravity was used to support data extraction, standardization, statistical analysis, and figure generation. Clinical features were mapped to standardized symptom and organ-involvement variables. Pairwise co-occurrence was assessed using 2×2 contingency tables and Fisher’s exact test; association strength was measured using Pearson correlation coefficients; and k-means clustering was used to identify distinct clinical phenotypes.
Results
Four principal clusters were identified: subclinical disease, ocular larva migrans, visceral larva migrans, and neurotoxocariasis/multisystemic disease. The ocular larva migrans and visceral larva migrans clusters were largely distinct, whereas ocular features were frequently co-reported in the neurotoxocariasis/multisystemic cluster. Eosinophilia was strongly associated with systemic manifestations, including fever, dyspnea, cough, pruritus, hepatomegaly, headache, and meningitis, but was not statistically significantly associated with ocular features. Neurotoxocariasis-related features, including headache, meningitis, and seizures, clustered closely.
Conclusion
This artificial intelligence-assisted scoping review identified exploratory clinical association and clustering patterns in human toxocariasis case reports. These findings may support clinical awareness, but they require validation in curated clinical datasets before they are used for diagnosis, risk stratification, or practice guidance.
