Quality-by-Design (QbD) in Device Manufacturing

Abstract

Quality-by-Design (QbD) -- the systematic approach of defining quality targets, understanding process-quality relationships through design of experiments, and establishing design spaces within which process parameters can vary without compromising product quality -- has transformed pharmaceutical manufacturing since ICH Q8-Q12 but remains incompletely adopted in medical device manufacturing where quality assurance still relies predominantly on end-product inspection rather than process understanding. Device manufacturing complexity -- multi-material assemblies, micrometre-tolerance machining, sterile packaging, and software-hardware integration -- creates quality challenges that inspection-based approaches address reactively rather than preventing proactively. This study presents the QbD Device Manufacturing Assessment Framework (QBDMAF), evaluating five QbD implementation strategies -- critical quality attribute (CQA) identification through failure mode analysis, design of experiments for process understanding, multivariate statistical process control, real-time process analytical technology (PAT), and AI-driven predictive quality modelling -- across four device manufacturing domains: metallic implant machining, polymer moulding for disposables, electronic assembly for active devices, and drug-device combination product manufacturing. Our QbD Manufacturing Effectiveness Score (QMES) integrates defect reduction, process capability improvement, design space establishment, real-time quality assurance, and regulatory compliance enhancement. AI-driven predictive quality modelling achieved the highest QMES (0.928) through gradient-boosted models trained on 2.4 million manufacturing data records that predicted lot-level quality outcomes with 96% accuracy before final inspection, enabling real-time process correction that reduced defect rates by 62%, while PAT achieved the highest real-time quality assurance (0.955) through in-line spectroscopic and dimensional monitoring that eliminated 88% of inspection-related production delays.