Regulatory Compliance of Pharmaceutical Supply Chains
by Eleni Pratsini and Doug Dea
The U.S. Food and Drug Administration (FDA) launched a major initiative to modernize the regulation of drug manufacturing and product quality in 2002. This program calls for the application of modern risk and quality management techniques, the use of engineering knowledge and new manufacturing technologies, and the application and demonstration of cutting-edge science throughout the entire production process. The IBM Zurich Research Lab in collaboration with IBM's Business Consulting Services in Life Sciences have developed a tool to assist pharmaceutical companies quantify their risk exposure to the new regulations and subsequently restructure their supply chains and manufacturing assets to minimize exposure to regulatory risk and at the same time maximize their revenue.
Driven by the increase in the number of adverse events and drug recalls in recent years, FDA has changed its method of monitoring drug manufacturing. They introduced systems thinking, quality by design and related processes that assure the quality of any product in manufacturing. This resulted in a new FDA initiative 'Good Manufacturing Practices (GMP's) for the 21st century'. The initiative requires pharmaceutical companies to comprehensively manage patient risk, base new drug submissions and manufacturing approaches on demonstrable scientific principles, and simultaneously implement inspection-ready 'GMP Systems' which embed compliance and quality in their operations. Of particular significance is the risk transfer among products resulting from the new approach of GMP Systems inspections. All products produced at a facility may be considered 'adulterated' if any one GMP system fails inspection. It is possible that a single low-revenue high-risk product may wipe out an entire facility's revenue. The assets and infrastructure of pharmaceutical companies were designed to meet the 25-year old GMP regulations and are now exposed to the new risks brought up by the FDA programs related to quality by design.
One of the practical implications of the FDA's challenge is the restructuring of a company's supply chain. A company will need to manage its risks across products, technologies, and sites, and ensure that its drugs are safe at the point-of-use. Using a novel statistical approach, we developed a model to quantitatively measure a company's current exposure to compliance hazards for each of the risk sources based on historical performance. It is based on a retrospective analysis of non-conformance issues and their root causes of failure, leading to a view on their exposure to risk. These risk indices are then fed into a mathematical programming model that determines the sequence of corrective actions the company can take in order to minimize its exposure to risk, minimize operational and action-induced costs and maximize its revenue. The solution is, of course, constrained by budgetary considerations, physical restrictions and a defined rate of change. The problem is represented as a mathematical model with the combination and interaction of risk indices resulting in a nonconvex, nonlinear integer formulation. Through variable reformulation and logarithmic transformation, a mixed integer convex nonlinear model was obtained which was then linearized. Typical corrective actions are risk mitigation for any of the sources, closing down a technology or site, delisting a product, introducing a new technology, and others. Both company-internal and company external actions can be considered.
Based on examples generated by real data, the model was tested and the output analyzed. The risk interactions generated solutions that were not intuitively obvious. The sequence and timing of actions depended on budgetary as well as risk considerations. Products were moved to other technologies if their technology was aging and thus became too risky. In certain instances products were moved to other sites which in turn exposed all products on the new site to new levels of risk. This 'adulteration'produced the subsequent transfer of other products to other sites (see Figure for an example). By considering the risk transfer in the optimization model, a company can reduce its exposure to business risk by 40% while increasing its secured profitability by 30%.
The model can be used as a simulation engine for testing various alternative solutions, or for partial optimization when the end state is fixed but the sequence of actions to get to that state is optimized.
The mathematical model has a large number of binary variables resulting in an NP hard problem. The complexity of the optimization model is due in part to the infinite combination of corrective actions that can take place. Furthermore, considering multiple periods in a planning horizon further complicates the model. Current research investigates heuristic procedures for obtaining near optimal solutions for very large problems in minimal computation time.
IBM Zurich Research Lab./SARIT, Switzerland
Doug Dean, IBM Business Consulting