A single dominant signal integrates pharmaceutical mixtures across urban wastewater sources

Urban wastewater contains complex mixtures of pharmaceuticals whose interpretation is often limited by dilution effects and site-specific variability. Despite extensive monitoring efforts, few approaches provide concise indicators capable of summarizing system-level contamination patterns across heterogeneous wastewater sources. In this study, we propose a dilution-adjusted multivariate framework that uses caffeine as an anthropogenic reference signal to reveal dominant system-level contamination structures in urban wastewater. We analyzed wastewater from five municipal wastewater treatment plant influents and one hospital discharge point in the Curitiba Metropolitan Region, Brazil, based on 61 composite samples. Twenty-three target compounds were measured using LC-MS/MS; caffeine was used as an anthropogenic dilution normalizer, and eleven pharmaceuticals and hormones were retained for multivariate profiling following screening and preprocessing. Dimensionality reduction revealed a highly concentrated variance structure, with the first principal component explaining 87.4% of the total variance, indicating a dominant metropolitan-scale pharmaceutical contamination gradient. Exploratory factor analysis (KMO = 0.841) yielded a two-factor solution explaining 98.4% of the common variance, consisting of (i) a dominant pharmaceutical burden factor reflecting baseline population consumption across multiple therapeutic classes and (ii) a partially separable anti-infective/tuberculosis-treatment signal, despite substantial cross-loadings. Bootstrap-validated hierarchical clustering identified two clusters: a dominant cluster comprising 59 of 61 samples (Jaccard = 0.98) and a minor cluster of two atypically elevated observations (Jaccard = 0.87), demonstrating the capacity to flag system-relevant deviations from background conditions. Hospital wastewater largely overlapped with community wastewater in multivariate space, indicating strong integration at this spatial scale; however, one hospital observation coincided with an atypical high-profile event, suggesting that differentiation is event- and mixture-dependent rather than structural. Ordination patterns were robust to alternative censored-data imputation strategies (|r| = 0.998). Overall, caffeine-normalized multivariate profiling reveals that pharmaceutical contamination in urban wastewater systems is governed by a small number of dominant signals, providing an integrated and interpretable framework for analyzing complex mixtures across heterogeneous urban sources.