tetano
Editor, Senior Moderator
Water Res
. 2026 Jul 30:306:126596.
doi: 10.1016/j.watres.2026.126596. Online ahead of print.
Prediction of influenza activity in wastewater using oseltamivir biomarkers
Christoforos Bouzoukas[SUP] 1 [/SUP], Triantafyllos-Dimitrios Gerokonstantis[SUP] 1 [/SUP], Reza Aalizadeh[SUP] 1 [/SUP], Anastasia Zafeiriadou[SUP] 1 [/SUP], Anastasios Tsolakidis[SUP] 2 [/SUP], Elpida Fotiadou[SUP] 3 [/SUP], Athina Markou[SUP] 1 [/SUP], Nikolaos S Thomaidis[SUP] 4 [/SUP]
Affiliations
Wastewater-based epidemiology (WBE) has emerged as a powerful approach for population-level surveillance of infectious diseases and pharmaceutical usage; however, integrated modeling frameworks that combine chemical and virological wastewater signals remain underexplored. In this study, we develop and evaluate multivariate models that link antiviral drug biomarkers in wastewater to influenza viral loads, prescription data, and reported case numbers, with the aim of strengthening influenza surveillance and outbreak assessment. Influenza A (IAV) and B (IBV) virus loads were quantified in wastewater samples over three consecutive influenza seasons alongside concurrent measurements of oseltamivir and its primary metabolite, oseltamivir carboxylate (OC). Multivariate regression models were constructed using oseltamivir and OC mass loads as chemical predictors. Model performance was evaluated using classical multiple linear regression, distributed lag models, and an extreme gradient boosting (XGBoost) machine-learning approach. The XGBoost consistently outperformed conventional models, enabling accurate estimation of population-normalized antiviral consumption from wastewater mass loads and demonstrating strong predictive power for both IAV and IBV concentrations. In addition, the developed models accurately estimated the number of medicated influenza patients directly from wastewater-derived biomarker's mass loads, demonstrating the potential of antiviral biomarkers to quantify disease burden at the population level. Correlations between chemical biomarkers and viral loads yielded coefficients up to 0.88, with average prediction errors of approximately ±1.1 log[SUB]2[/SUB] units. Importantly, the models also reliably estimated the number of medicated influenza cases directly from wastewater-derived antiviral signals. These findings demonstrate that oseltamivir and OC serve as robust chemical proxies for influenza activity and machine-learning-based WBE models can effectively translate wastewater measurements into actionable epidemiological indicators. The proposed integrated modeling framework enables retrospective analysis, supports early outbreak detection, and provides a scalable tool for public health surveillance.
Keywords: Influenza A and B; Machine learning; Oseltamivir and oseltamivir carboxylate; Prescription data; Wastewater-based epidemiology.
. 2026 Jul 30:306:126596.
doi: 10.1016/j.watres.2026.126596. Online ahead of print.
Prediction of influenza activity in wastewater using oseltamivir biomarkers
Christoforos Bouzoukas[SUP] 1 [/SUP], Triantafyllos-Dimitrios Gerokonstantis[SUP] 1 [/SUP], Reza Aalizadeh[SUP] 1 [/SUP], Anastasia Zafeiriadou[SUP] 1 [/SUP], Anastasios Tsolakidis[SUP] 2 [/SUP], Elpida Fotiadou[SUP] 3 [/SUP], Athina Markou[SUP] 1 [/SUP], Nikolaos S Thomaidis[SUP] 4 [/SUP]
Affiliations
- PMID: 42566960
- DOI: 10.1016/j.watres.2026.126596
Wastewater-based epidemiology (WBE) has emerged as a powerful approach for population-level surveillance of infectious diseases and pharmaceutical usage; however, integrated modeling frameworks that combine chemical and virological wastewater signals remain underexplored. In this study, we develop and evaluate multivariate models that link antiviral drug biomarkers in wastewater to influenza viral loads, prescription data, and reported case numbers, with the aim of strengthening influenza surveillance and outbreak assessment. Influenza A (IAV) and B (IBV) virus loads were quantified in wastewater samples over three consecutive influenza seasons alongside concurrent measurements of oseltamivir and its primary metabolite, oseltamivir carboxylate (OC). Multivariate regression models were constructed using oseltamivir and OC mass loads as chemical predictors. Model performance was evaluated using classical multiple linear regression, distributed lag models, and an extreme gradient boosting (XGBoost) machine-learning approach. The XGBoost consistently outperformed conventional models, enabling accurate estimation of population-normalized antiviral consumption from wastewater mass loads and demonstrating strong predictive power for both IAV and IBV concentrations. In addition, the developed models accurately estimated the number of medicated influenza patients directly from wastewater-derived biomarker's mass loads, demonstrating the potential of antiviral biomarkers to quantify disease burden at the population level. Correlations between chemical biomarkers and viral loads yielded coefficients up to 0.88, with average prediction errors of approximately ±1.1 log[SUB]2[/SUB] units. Importantly, the models also reliably estimated the number of medicated influenza cases directly from wastewater-derived antiviral signals. These findings demonstrate that oseltamivir and OC serve as robust chemical proxies for influenza activity and machine-learning-based WBE models can effectively translate wastewater measurements into actionable epidemiological indicators. The proposed integrated modeling framework enables retrospective analysis, supports early outbreak detection, and provides a scalable tool for public health surveillance.
Keywords: Influenza A and B; Machine learning; Oseltamivir and oseltamivir carboxylate; Prescription data; Wastewater-based epidemiology.