Wastewater remediation of pharmaceuticals with ozone and granular active carbon: a risk-driven approach

This study aimed to investigate the removal efficiency of (psycho)pharmaceuticals by ozonation and granular active carbon (GAC) in wastewater effluent, using risk as the metric for adequate removal instead of aqueous concentrations. Conventionally treated effluent was further treated with ozone or GAC until there was a 25% reduced UVA₂₅₄ absorbance, to allow for a direct comparison of the two treatment types. Samples were analysed using Ultra High-Performance Liquid Chromatography-Quadrupole Time of Flight-High Resolution Mass Spectrometry (UHPLC-qTOF-HRMS), where 20 (psycho)pharmaceuticals were quantified, and their risk was assessed using Predicted No Effect Concentrations (PNECs). A further assessment was performed using Quantitative Structural Activity Relationships (QSARs) for both parent compounds and their Oxidation Transformation Products (OTPs) to compare the relative toxicity of new species being formed by the ozone treatment. The total median removal efficiency across all compounds was 60 ± 3% for ozone in terms of concentration, yielding a 73 ± 2% reduction in terms of risk for the parent compounds, while the median removal efficiency for GAC is 57 ± 9% as expressed in concentration, and 46 ± 11% in terms of risk reduction. When factoring in the OTP toxicity, the median risk reduction for ozone flips to -274 ± 124%, indicating that there may be an increase in risk during ozonation. Pearson correlations on molecular descriptors indicated that ozone removal most strongly correlated with the number of activated aromatic rings (<i>r</i> = 0.65), while for GAC the topological polar surface area correlated strongest with removal (<i>r</i> = 0.54), therefore indicating that ozone and GAC target different types of molecules. The study demonstrates the merits of a risk-driven approach over concentration-based removal targets in current legislation, but also highlighted some drawbacks, especially with regards to data gaps and model accuracies.