Author ORCID Identifier

Date Available


Year of Publication


Degree Name

Doctor of Philosophy (PhD)

Document Type

Doctoral Dissertation




Pharmaceutical Sciences

First Advisor

Dr. Robert A. Lodder


Counterfeit, adulterated, and misbranded medicines in the pharmaceutical supply chain (PSC) are a critical problem. Regulators charged with safeguarding the supply chain are facing shrinking resources for inspections while concurrently facing increasing demands posed by new drug products being manufactured at more sites in the US and abroad. To mitigate risk, the University of Kentucky (UK) Central Pharmacy Drug Quality Study (DQS) tests injectable drugs dispensed within the UK hospital. Using FT-NIR spectrometry coupled with machine learning techniques the team identifies and flags potentially contaminated drugs for further testing and possible removal from the pharmacy. Teams like the DQS are always working with limited equipment, time, and staffing resources. Scanning every vial immediately before use is infeasible and drugs must be prioritized for analysis. A risk scoring system coupled with batch sampling techniques is currently used in the DQS. However, a risk scoring system only allows the team to know about the risks to the PSC today. It doesn’t let us predict what the risks will be in the future. To begin bridging this gap in predictive modeling capabilities the authors assert that models must incorporate the human element. A sister project to the DQS, the Drug Quality Game (DGC), enables humans and all of their unpredictability to be inserted into a virtual PSC.

The DQG approach was adopted as a means of capturing human creativity, imagination, and problem-solving skills. Current methods of prioritizing drug scans rely heavily on drug cost, sole-source status, warning letters, equipment and material specifications. However, humans, not machines, commit fraud. Given that even one defective drug product could have catastrophic consequences this project will improve risk-based modeling by equipping future models to identify and incorporate human-induced risks, expanding the overall landscape of risk-based modeling.

This exploratory study tested the following hypotheses (1) a useful game system able to simulate real-life humans and their actions in a pharmaceutical manufacturing process can be designed and deployed, (2) there are variables in the game that are predictive of human-induced risks to the PSC, and (3) the game can identify ways in which bad actors can “game the system” (GTS) to produce counterfeit, adulterated, and misbranded drugs.

A commercial-off-the-shelf (COTS) game, BigPharma, was used as the basis of a game system able to simulate the human subjects and their actions in a pharmaceutical manufacturing process. BigPharma was selected as it provides a low-cost, time-efficient virtual environment that captures the major elements of a pharmaceutical business- research, marketing, and manufacturing/processing. Running Big Pharma with a Python shell enables researchers to implement specific GxP-related tasks (Good x Practice, where x=Manufacturing, Clinical, Research, etc.) not provided in the COTS BigPharma game. Results from players' interaction with the Python shell/Big Pharma environment suggest that the game can identify both variables predictive of human-induced risks to the PSC and ways in which bad actors may GTS. For example, company profitability emerged as one variable predictive of successful GTS. Player's unethical in-game techniques matched well with observations seen within the DQS.

Digital Object Identifier (DOI)

Funding Information

2020-2021: The project described was supported by the National Center for Research Resources and the National Center for Advancing Translational Sciences, National Institutes of Health, through Grant UL1TR001998. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.