Introduction
Autologous Chimeric Antigen Receptor (CAR) T-cell therapy represents the most significant class of approved cell and gene therapies to date. However, regulatory approvals thus far have been in specific hematologic malignancies. The promising and demonstrated clinical efficacy of CAR-T therapies in blood cancers is also driving substantial investment toward expanding their applications to solid tumors and autoimmune diseases – meaning scaling of dose throughput will be mandatory to meet the target patient populations. So, since CAR-T therapies account for over 50% of the preclinical and clinical development pipeline within cell and gene therapies, and the growing number of potential patients with the new indications, the necessity for optimizing manufacturing processes and improving patient accessibility has reached a necessity.
One of the major challenges facing autologous CAR-T therapies is their prohibitive cost, exceeding $1 million USD per treatment (1), which limits access to patients with both socioeconomic means and proximity to advanced medical centers. Integrating machine learning (ML) into patient selection and manufacturing processes offers a promising avenue for cost reduction and enhanced therapeutic efficacy. By leveraging ML algorithms to identify patients most likely to achieve a positive clinical response and implementing scalable, high-throughput manufacturing strategies such as robotics, the industry can significantly enhance the financial feasibility of CAR-T therapies. Specifically, by focusing treatment to patients whose T cells can generate a high-quality drug product and producing CAR-T cells within roboticized cleanroom environments, cost reductions of approximately 50% should achievable alongside improved therapeutic outcomes.
Regulatory agencies in the United States and the European Union recognize the role of advanced technologies in enhancing drug manufacturing efficiency and have initiated efforts to facilitate their integration. In January 2025, the U.S. Food and Drug Administration (FDA) issued Draft Guidance on the use of artificial intelligence (AI) in drug development, while the European Medicines Regulatory Network (EMRN) is aligning drug compliance approaches with the EU AI Act, which was enacted in August 2024. These initiatives provide industry-specific regulatory guidance on AI applications across the drug development and manufacturing lifecycle, paving the way for the broader implementation of AI-driven innovations in CAR-T therapy.
Cost Analysis
The cost and access challenges in autologous CAR-T therapy arise from a combination of interrelated factors. Drug costs are significant, with the average billed cost of a CAR-T drug product reaching approximately $402,500 (2). The production process involves vein-to-vein times of three to six weeks and is constrained by current manufacturing methods, which typically yield only a single dose per batch. Labor and infrastructure contribute heavily to the overall expense, as each dose requires over 200 human labor hours (3) and the use of batch-dedicated equipment in GMP-compliant cleanroom environments. While advances in automation have improved specific stages, such as T-cell isolation, activation, expansion, and quality control – human labor and cleanroom costs remain dominant cost drivers. Preliminary calculations demonstrate that automating certain QC and bioprocessing steps where equipment is engineered in a modular approach amenable to AI controlled robots reduces the overall costs of manufacturing by 50-55%. An additional opportunity for using AI in the logistics is being evaluated as the biological nature of CAR-T therapy imposes stringent cold chain logistics requirements. Any disruptions in these logistics can threaten the viability of raw materials and final drug products, sometimes with life-threatening consequences for patients. We should note that although programs using more cost effective and scalable allogenic methods remain underway, the allogenic T-cell field experienced several setbacks in 2024 that will move the clinical calendar reality even further out.
Automation in Manufacturing
The traditional horizontal manufacturing layout for autologous CAR-T therapies relies on scaling through the addition of equipment and human operators. However, this approach becomes increasingly unsustainable as demand grows. To address this limitation, manufacturing strategies must incorporate automation and digital transformation. Integrated automation systems aim to replace human-dependent processes with robotics and AI to manage raw materials, drug product testing, and all associated processing equipment. The Pharma 4.0 (4) framework provides a roadmap for creating interconnected manufacturing architectures. This transition requires drug developers and contract development and manufacturing organizations (CDMOs) to adopt new skill sets and embrace a forward-thinking approach, which historically has been slow.
Challenge In Scaling
The largest challenges to scaling CAR-T manufacturing can be evaluated through the lenses of cost, time, and risk. Costs are primarily driven by the linear scaling of equipment, cleanroom space, and labor. Production timelines are further constrained by bottlenecks in quality control and testing, while risks are amplified by the complexities of cold chain logistics. These logistics present vulnerabilities for both raw materials and finished drug products, necessitating a reevaluation of processes and methods to improve efficiency and minimize human involvement.
Prioritizing The Solutions
The most impactful cost-reduction strategy centers on minimizing human labor hours and human-equipment interactions wherever feasible. While human involvement has traditionally been perceived as a safeguard, it introduces inefficiencies, contamination risks, and potential errors. Digital transformation through AI and robotics mitigates these issues. Two approaches are being explored to achieve these goals. The first involves developing all-in-one manufacturing systems that integrate all production steps into a single “black box” platform. Although promising, these systems encounter hurdles related to capital investment, engineered device reliability, and regulatory compliance. The second approach emphasizes modular engineering optimization of commercially available equipment, focusing on scale and incremental improvements to specific manufacturing steps and infrastructure that leverages AI-driven analytics and robotic automation. This method is more adaptable, requiring lower initial investment and offering substantial improvements in cost efficiency and risk management. Early assessments suggest that quality control testing and bioprocess automation offer the most significant opportunities for cost reduction.
Focusing on Quality Control
Quality control testing represents a critical opportunity for automation. Currently, QC processes involve multiple tests conducted on various pieces of equipment by specialized scientific staff. Robotics can streamline this workflow by automating repetitive tasks, such as liquid handling and sample loading and unloading. By reducing reliance on human intervention, these systems improve efficiency and consistency. Machine learning algorithms, similar to Process Analytical Technologies ML (5) , further enhance this process by interpreting assay data, expediting the quality control workflow. Preliminary calculations indicate that these advancements could decrease human labor hours by a factor of four, underscoring the transformative potential of automation in this domain.
Focusing on Bioprocessing
The current approach to autologous bioprocessing expands capacity horizontally, adding more equipment and operators in expansive ballroom-style layouts. However, this model faces inherent limitations and, to date, has not demonstrated the ability to exceed 100 concurrent units in process. Through the use of automation and engineering, a vertical stacking design for equipment serviced by robots can triple output within the same GMP cleanroom footprint. This approach integrates advanced robotic systems equipped with movable joints and sensitive gripper touch dexterity, operating alongside multiple machine learning models running in parallel. While the technical challenges involved are non-trivial, the potential efficiency gains justify the effort.
Focusing on Vein-To-Vein Logistics
Machine learning also has a critical role in derisking vein-to-vein logistics, encompassing both the transportation of donor cells and the delivery of final drug products. Lessons can be drawn from the evolution of logistics companies, such as Amazon, which revolutionized supply chain operations with real-time data and optimized delivery routes. Similarly, AI can enhance vein-to-vein logistics by optimizing transportation routes, providing real-time status updates, and presenting actionable data to human operators managing shipping control tower services. Such advancements would ensure both reliability and responsiveness in this vital process.
Builder Skills
The development, deployment, and maintenance of these advanced systems necessitate a multidisciplinary workforce. Essential skills include expertise in machine learning operations, validated ML verification procedures, robust networking capabilities bridging on-premise and cloud environments, robotics engineering, and on-site service engineering. Collaboration between robotics manufacturers, cell and gene therapy developers, and cloud hyperscalers will be key to sustaining these innovations.
Summary
The integration of AI and robotics into autologous CAR-T therapeutic manufacturing has the potential to significantly increase drug product throughput by a factor of three to four times while reducing production costs by more than 50 percent. This profound production intensification and cost reduction can significantly improve affordability and expand access to these transformative therapies.
References and notes
- Cavallo MC et al. Cost of implementing CAR-T activity and managing CAR-T patients: an exploratory study. BMC Health Services Research 24, Article number: 121 (2024)
- Di M. et al. Costs of care during chimeric antigen receptor T-cell therapy in relapsed or refractory B-cell lymphomas. JNCI Cancer Spectrum, Volume 8, Issue 4, August 2024
- Dias J. et al. CAR-T cell manufacturing landscape – Lessons from the past decade and considerations for early clinical development. Molecular Therapy: Methods & Clinical Development Vol. 32 June 2024
- ISPE Baseline Guide Vol 8: Pharma 4.0 1st Edition, December 2023
- Williams T. et al. Machine learning and metabolic modelling assisted implementation of novel process analytical technology in cell and gene therapy manufacturing. Scientific Reports 13, Article number: 834 (2023).
