Use of machine learning to unlock cycle performance and future parametric release of terminally sterilized medical devices at scale using low temperature gaseous hydrogen peroxide (VH2O2): A Review

Abstract

Abstract Gaseous hydrogen peroxide (VH2O2) is becoming an increasingly important low-temperature sterilization modality for medical devices and is being positioned as a sustainable alternative to using ethylene oxide that sterilizes approximately 50% of medical devices globally. This review examines VH2O2 from a comparative sustainable process capability perspective and explores existing operational performance and limitations to overcome for scale optimisation including parametric release considerations. A PRISMA style review of the literature indicates that VH2O2 technology is microbiologically effective and industrially promising. Best published studies have improved our understanding of key determinants such as pressure, humidity, temperature, concentration, condensation behaviour, material compatibility, and biological indicator response. However, critical process parameters and their interactions have not been statistically defined that limits optimisation including for full industrial scale cycle implementation. Additionally, continuous process data generated during sterilization cycles is not used for operational assurance. Machine learning (ML) is examined as a complementary approach to address sterilization science opportunities. Evidence from related sterilization modalities shows that ML can reduce experimental burden by decreasing the number of experimental runs required, improving cycle output anomaly detection, supporting optimisation, and enhancing use of continuous sensor data. These findings highlight that sustainable VH2O2 can potentially be further advanced through ML enabled optimisation as a reliable, data-driven, industrial-scale terminal sterilization modality for global deployment.