A Clinical Perspective on Supply Decisions
Author: Dr Angeliki Kotsianti, physician-scientist and molecular pathologist. She held the final medical judgment on the medicines of a major pharmaceutical company, including decisions on reformulation, recall and supply interruption, and built its global strategy for medically necessary products. She now advises on medical governance and regulatory strategy with PharmaFlow Ltd.
Every structural fault in the way medicines are developed and supplied arrives, eventually, as a decision that a doctor has to make about a named patient.
Supply interruption is usually discussed as a question of pedigree, procurement, and track and trace. At the point of care, it is none of those things. It is whether to begin a course that may not be completed, whether to move a stable patient onto an alternative agent, and whether the evidence behind that alternative is strong enough to justify moving them. Those are medical judgments. The clinician making them has usually had no part in the decisions that produced the constraint, and rarely has time to establish what those decisions were.
The track and trace measures introduced after the Heparin contamination would not prevent a recurrence, and the reason has to do with sequence. The harm in that episode was clinical before anyone recognised it as a supply failure. Patients were injured and died while the system still believed the material was genuine, and the clinicians treating them had no route by which what they were seeing could be raised as a question about supply. Reporting runs in one direction. Until that changes, failures of this kind will keep being found out the way this one was, which is late.
The penicillin story is usually told as evidence about manufacturing capacity, and it is. From the clinical side it is also a story about consent. Albert Alexander was started on a course the Oxford team knew might not be finished. Someone weighed an incomplete course against no course at all, for a patient whose name they knew, on evidence that did not yet exist. That was a medical judgment, taken in conditions no protocol covered. The production figures from those months are recorded in some detail. The reasoning behind the decision to treat him is not.
In most companies benefit-risk is a named function with a named owner, and supply is a named function with a named owner. What a supply constraint means for patients belongs to neither. It gets answered late, informally, and often by whoever happens to be nearest to it. In my own experience the judgments that reached me on reformulation, recall and supply interruption had in substance been made already, months or years earlier, when a formulation or a presentation or a dose was chosen and nobody asked what it would mean at the bedside. What arrived on my desk was a choice between poor options.
A class of therapies has now arrived that makes the problem literal. Where a medicine is manufactured from a patient and returned to them, the decisions include whether that patient is eligible at all given the distance to the apheresis site, whether their bridging therapy will hold through the manufacturing window, and whether an out-of-specification product should be given to someone who has no alternative. Manufacturing failure is not a rare event. The UK National CAR T Panel reported a failure rate of just under four per cent after the first manufacturing attempt in large B-cell lymphoma patients approved for commercial autologous therapy, with further patients affected by out-of-specification product or by delay. Each of those is a patient who has already been leukapheresed and bridged. Each is a medical judgment carrying regulatory accountability, and the operating model has nowhere to put it, so it is handled as logistics at one end of the chain or the other. The batch here is a single patient. Scale has become a question about one person at a time, and it is still nobody’s to answer.
Geography is itself a clinical outcome. An analysis of SEER-Medicare data covering patients treated for third-line or later diffuse large B-cell lymphoma found that those who received CAR-T were less likely to carry multiple comorbidities and more likely to live in higher income areas. The same study modelled the effect of distance. Reducing the average journey to an authorised treatment centre in poor-access states from 104 miles to 34 would raise the proportion receiving CAR-T from 6.6 to 9.1 per cent. The relative gain is large. The absolute position is that more than nine in ten of these patients would still not receive the therapy. That is not a logistics statistic. It describes patients accepting less effective treatment because of where they live and what they earn, and it is the same pattern as any therapy that needs infrastructure the local hospital does not have. Decentralisation is a governance problem before it is a manufacturing one.
Those figures are American, and a single-payer system removes the cost of the medicine itself. It does not remove the pattern. UK data show that referral for CAR-T is broadly even across the country, but that patients from the most deprived areas are infused at 73 per cent against 86 per cent for those from the least deprived. Once infused, response, progression-free survival and toxicity are similar. The therapy does not work less well for poorer patients. They are less likely to receive it. Modelling of the English network finds that concentrating advanced cell therapies in a small number of centres increases travel time in every scenario tested, with the burden falling hardest on lower socioeconomic groups and up to 40 per cent of patients left more than an hour from a treating centre. Centralisation is a reasonable answer to a manufacturing problem. It is also a decision about who will find it hard to be treated, and it is not currently made by anyone who has to answer for that.
Pharmacovigilance carries substantial governance because harm caused by a medicine is understood to need a named owner, defined thresholds and mandatory escalation. Harm caused by the absence of a medicine has nothing equivalent. A patient who deteriorates because treatment was interrupted has been harmed as surely as one who suffers an adverse reaction, but no system requires that this is recorded, investigated or reported. Companies supplying medically necessary products should be expected to hold a clinical continuity plan for each one, owned by a medical officer, with escalation thresholds and a documented position on therapeutic alternatives worked out before a shortage rather than during one.
It is well established that a substantial proportion of published findings cannot be reproduced. A newer problem now sits alongside it. Evidence generated or filtered by machine learning is entering development decisions, including decisions on formulation, stability and demand. I have worked on these methods and they can be sound. They can also produce confident output from data that would not survive inspection. Any reform of the medicines system should require that evidence used to justify a development or supply decision is traceable to its source and validated by someone competent to judge it clinically.
Reform of this kind will not be delivered by supply chain professionals alone, nor by clinicians alone. At present the two describe the same failure in vocabularies that do not meet. My proposal is a narrow one. Wherever a supply decision is to be taken earlier or differently, require also that a named medical owner is accountable for what that decision will mean for the patient who receives the medicine, or does not.
Written in 2026 as an appendix to Medicines for the 21st Century, the paper presented to the House of Commons Health and Social Care Committee in July 2019 by Friends of Medicines Modernisation with PharmaFlow Ltd.