Abstract
Establishing how to effectively manufacture cell therapies is an industry-level problem. Decentralised manufacturing is of increasing importance, and its challenges are recognised by healthcare regulators with deviations and comparability issues receiving specific attention from them. This paper is the first to report the deviations and other risks encountered when implementing the expansion of human pluripotent stem cells (hPSCs) in an automated three international site-decentralised manufacturing setting. An experimental demonstrator project expanded a human embryonal carcinoma cell line (2102Ep) at three development sites in France, Germany and the UK using the CompacT SelecT (Sartorius Stedim, Royston, UK) automated cell culture platform. Anticipated variations between sites spanned material input, features of the process itself and production system details including different quality management systems and personnel. Where possible, these were pre-addressed by implementing strategies including standardisation, cell bank mycoplasma testing and specific engineering and process improvements. However, despite such measures, unexpected deviations occurred between sites including software incompatibility and machine/process errors together with uncharacteristic contaminations. Many only became apparent during process proving or during the process run. Further, parameters including growth rate and viability discrepancies could only be determined post-run, preventing 'live' corrective measures. The work confirms the critical nature of approaches usually taken in Good Manufacturing Practice (GMP) manufacturing settings and especially emphasises the requirement for monitoring steps to be included within the production system. Real-time process monitoring coupled with carefully structured quality systems is essential for multiple site working including clarity of decision-making roles. Additionally, an over-reliance upon post-process visual microscopic comparisons has major limitations; it is difficult for non-experts to detect deleterious culture changes and such detection is slow.