Development of Analytical Methods and Machine Learning Platforms for Assessing the Quality of Pharmaceuticals and Drugs of Abuse

In low and middle-income countries (LMICs), the prevalence of counterfeit medicines, estimated at 1 in 10 pharmaceuticals, underscores the urgent need for quick and affordable presumptive analytical screening tools to mitigate the dangers associated with counterfeit drugs.
This research aims to tackle the challenges in pharmaceutical quality control and post-market surveillance by addressing the limitations of conventional predictive models and exploring the potential of non-brand-specific strategies. Traditional predictive models, often trained on libraries of authentic products, encounter difficulties in identifying quality issues in products beyond their training set. To overcome this, the research employs non-brand-specific strategies, utilizing lab-formulated samples correlated with high-performance liquid chromatography (HPLC) standards as a presumptive analytical method for post-market pharmaceutical screening, assessed through systematic metrics.
The study probes the effectiveness of Paper Analytical Devices (PADs) in discerning the active pharmaceutical ingredient (API) in over 20 antibiotic medicines. Leveraging chemical reactions to produce a unique color pattern, PADs are trained through machine learning predictive models to detect substandard formulations in the field. Despite their high sensitivity, PADs face challenges with specificity when dealing with APIs sharing identical functional groups. In response, the research investigates the integration of PADs with spectroscopic analytical techniques, creating an orthogonal analytical approach aimed at developing a robust technology for reliably detecting substandard and falsified medicines.
External factors, including variations in formulations, excipients, capsule type, color, and instrumental design/device fabrication, significantly influence pharmaceutical analysis. The research highlights the impact of capsule composition on spectroscopic analytical techniques, such as Near-Infrared (NIR) modeling. Factors like capsule type, color, or thickness introduce variance in NIR spectra comparable to the variance resulting from the type of API stored within the capsules. Recognizing and accounting for these external factors are essential for accurate and reliable analysis.
Furthermore, the research addresses a critical financial challenge in low and middle-income countries (LMICs) by introducing a cost-saving technique for achieving a more robust post-market surveillance. HPLC, while a powerful analytical method, demands specialized expertise and incurs substantial operational and maintenance costs, making it economically burdensome in resource-limited settings.