<![CDATA[This study reviews research on integrating machine learning within Lean Six Sigma for predictive analytics in pharmaceutical manufacturing. It targets three operational priorities reported across the literature, predictive maintenance, quality control, and process optimisation. The review follows a systematic design guided by PRISMA and Saunders’ Onion Model, with Scopus and Web of Science as the primary sources. The search applied 2014 to 2024 publication filter and selected peer reviewed journal articles written in English and aligned to the defined scope. Data analysis used thematic analysis with NVivo 15 to code and synthesize evidence across thirty-four studies. Four themes emerged. First, Lean Six Sigma supports productivity and compliance through waste reduction, variation control, and standardised quality practices in pharmaceutical operations. Second, machine learning strengthens the DMAIC cycle by improving pattern recognition, anomaly detection, forecasting, and monitoring, with clear relevance to Analyse, Improve, and Control activities. Third, three dominant framework strategies appear in the literature, model centric designs that embed specific algorithms, process centric designs that align analytics to DMAIC phase objectives, and platform-based designs that combine IoT or cloud infrastructure with continuous data processing and feedback. Fourth, case evidence reports practical gains in documentation and audit readiness, equipment failure prediction, process and batch optimisation, and sustainability-oriented improvement, with recurring benefits reflected in lower rejection risk, reduced downtime, and stronger traceability. The review also identifies persistent constraints. Studies report weak standardisation in model selection and evaluation, limited empirical validation across settings, and recurring implementation barriers such as fragmented data, legacy IT constraints, limited skilled personnel, and low interpretability in regulated contexts. Future work should develop phase specific guidance for ML embedded LSS, expand real time deployment studies using live manufacturing data, strengthen governance and validation practices aligned to GMP and data integrity, and adapt frameworks to plant level constraints and product specific risk profiles.]]>
