Bayesian securing of the pharmaceutical supply chain

Introduced in 2011, the European Union Falsified Medicines Directive asks for an enhanced security of the pharmaceutical supply chain. In this frame, manufacturers must apply safety measures to enable the verification of authenticity and identification of individual packs. This is the so-called serialization of the pharmaceutical supply chain. However, these measures are only dedicated to the analysis of the secondary packaging and do not enable the analysis of the product’s quality.
Therefore, we propose an end-to-end strategy based on spectroscopic fingerprints and risk-oriented statistical models for the verification of the quality of medicines along the supply chain. They can provide a precise description of the chemical composition of samples, and hence can be used to fingerprint a pharmaceutical product. A representative set of these spectra is sampled from each batch at release using appropriate devices. This sample is used to build a statistical tolerance band, that is assumed to contain a high proportion, say at least 90% of the future spectra of the product. The construction of such a band relies on the newly emerging chemometrics techniques such as functional data analysis (e.g. Bayesian wavelet or splines regressions). The upper and lower limits of the tolerance band are used as threshold or reference spectra for the conformity of a new spectrum. This allows us to declare the conformity of a product with a certain probability confidence.
Compared to classical measures (p-values, Hit Quality Indexes), this functional data analysis and risk-oriented approach enables to detect very small perturbations of the spectrum caused, for example by degradation, batch inversion, etc. The computed tolerance band may be stored in a cloud server and accessed throughout the supply chain to check the conformity of the product itself.
The spectral serialization of pharmaceutical batches is another brick in the wall of pharmaceutical supply chain securing.