Peak integration errors and analyst interaction in pharmaceutical quality control using LC-UV

Peak integration is a major and often underestimated source of uncertainty in pharmaceutical LC-UV quality control (QC), particularly when minor impurity peaks appear as shoulders or rider peaks on the tails of overloaded active pharmaceutical ingredient (API) peaks. In routine QC, APIs are frequently injected at high concentrations to ensure sufficient sensitivity for impurity analysis, which can result in partial column overloading and pronounced peak tailing. Here, we systematically evaluated integration accuracy and analyst interaction for commonly used rule-based integration algorithms implemented in commercial chromatographic data systems (Agilent OpenLab and Waters Empower) and compared them with an in-house independent integration system (IIS). The assessment included synthetic chromatograms with known areas as well as experimental LC-UV data for metoprolol and six related impurities. For baseline-resolved symmetric peaks, relative integration errors remained within approximately ±1.5% across all systems. For partially overlapping symmetric peaks, integration errors depended on peak size, noise level, and the integration method applied. In several rider-peak scenarios, errors for the perpendicular drop method (PDM) exceeded 100%. Application of the exponential skim method (ESM) reduced errors considerably, typically to 10-30% and in some cases to below ±5%, although not consistently. In the selected complex cases, greater analyst interaction was required in the commercial systems, increasing the potential for user-dependent variability. Overall, in the investigated cases, integration accuracy was strongly dependent on peak shape and overlap, with larger errors observed in asymmetric, shoulder, and rider-peak scenarios. These findings highlight the importance of integration-method selection and standardized integration practice in pharmaceutical LC-UV impurity analysis.