Rapid Detection and Quantification of Falsified Viagra

Using Cloud-Based Portable NIR Technology and Machine Learning

Reference: Hervé Rais, Pierre Esseiva, Olivier Delémont, Cédric Schelling, Stefan Stanojevic, Serge Rudaz, Florentin Coppey, Rapid detection and quantification of falsified Viagra using cloud-based portable NIR technology and machine learning, Journal of Pharmaceutical and Biomedical Analysis, Volume 263, 2025, 116940, ISSN 0731-7085, https://doi.org/10.1016/j.jpba.2025.116940.

The Challenge of Counterfeit Drugs

Counterfeit pharmaceuticals pose a significant threat to public health worldwide. These fake products often contain incorrect active ingredients, harmful contaminants, or no API at all, leading to ineffective treatment and potential health risks. Traditional laboratory methods like HPLC are reliable but time-consuming and require specialized equipment, limiting their use in field settings like customs borders or remote pharmacies.

In response to this growing issue, innovative technologies such as portable NIR spectroscopy combined with machine learning are revolutionizing the way we detect and quantify falsified medicines.

The Set-Up

This groundbreaking study, conducted by Hervé Rais in collaboration with the University of Lausanne and the University of Geneva, utilized a portable NIR spectrometer to analyze authentic and falsified Viagra samples. The dataset included 30 authentic Viagra tablets and 122 falsified ones, scanned both through their blister packaging (non-contact) and in direct contact mode. 

Machine learning models were trained on these spectra to classify samples as authentic or falsified and to quantify the sildenafil citrate (the active ingredient in Viagra) content. The models were validated through bootstrapping iterations (100 times) to ensure robustness. This approach leverages NIRLAB’s cloud-based platform, allowing for real-time data upload, analysis, and model deployment, making it ideal for on-site screening without the need for laboratory infrastructure.

Results

High Sensitivity and Accuracy in Detection

The NIR models demonstrated exceptional performance in distinguishing authentic from falsified Viagra. Principal Component Analysis visualizations clearly separated the two classes, with authentic samples clustering tightly in blue and falsified ones in red. Measurements through the blister packaging performed comparably to direct contact scans, proving the technology’s practicality for non-invasive testing.

Classification accuracy reached near-perfect levels, with no overlap between groups after preprocessing. No false positives were reported, ensuring reliable identification of counterfeits. This high sensitivity is crucial for field applications where quick decisions can prevent dangerous fakes from entering the market.

Precise Quantification of Active Ingredients

Beyond detection, the models accurately quantified sildenafil citrate levels in all samples, including falsified ones. For authentic Viagra, predicted concentrations aligned closely with the declared values. For falsified samples, the models correctly identified low API levels, which naturally deviated from the expected authentic concentrations due to their counterfeit nature—often lacking the active ingredient entirely or containing substitutes. The quantification results were within acceptable uncertainty margins and demonstrated excellent accuracy when validated against reference methods like HPLC.

These outcomes highlight how integration of portable device with advanced machine learning provides not just qualitative but quantitative insights, enabling authorities to assess the severity of falsification—whether it’s under-dosed, over-dosed, or completely inert.

Real-World Applications and Impact

This technology has immense potential for deployment in vulnerable regions where counterfeit drugs are prevalent. Customs checkpoints, airports, and markets can use portable NIR devices for immediate screening, preventing dangerous products from entering supply chains. 

By scanning products on-site, users can obtain immediate, actionable data, enhancing the efficiency of screening processes. The cloud-based system allows for continuous model updates based on new field data, improving intelligence on counterfeit trends and regional variations.

This study underscores the potential for portable NIR in vulnerable regions, where traditional lab-based methods are time-consuming and resource-intensive. Future enhancements could include adapting models for different packaging types and environmental conditions, further optimizing detection capabilities.

To explore more about spectroscopy and discover the full capabilities of NIRLAB’s technology, we invite you to read our other informative articles, here. Additionally, for personalized inquiries, please don’t hesitate to reach out to us at contact@nirlab.com.

Access the Full Academic Paper

For those interested in a deeper dive into the methodologies, results, and implications of this research, the full academic paper is available below or read the full study here.¹