When Clinical Supply Becomes A Clinical Decision

By Hedley Rees and Angeliki Kotsianti, MD

The authors write jointly throughout. Where the article speaks in the first-person singular, the voice is that of the clinician co-author.

In the first part of this series, we argued that clinical supply should be involved earlier in development, while product, process, and protocol decisions can still be changed. But bringing supply into the room earlier is only part of the solution. The harder question comes when supply intelligence reveals a constraint that affects a participant, the study, or the meaning of the data being generated.

At that point, clinical supply is no longer simply informing a development decision. Someone has to decide what the signal means, what happens next, and who owns the consequence.

There is a second reading of the same data, and it is the one that reaches a board. An excursion trend across a cluster of sites tells you the pack and the distribution model were badly matched. It also tells you something about the exposure data coming out of those sites, and whether it will hold when an inspector asks how you know what the participants actually received. A run of dosing queries tells you the instructions are unclear. It also asks whether administration was consistent enough for the dose-response relationship to mean what the protocol assumes it means. I have watched this play out across more than a hundred global regulatory submissions. Programs almost never stall because a signal was missing. They stall because somebody closed the signal at the operational level, in good faith, and nobody read it at the level where it changed what the data meant.

As we explored in our first article, the penicillin story is ultimately a lesson in integrating product, process, supply, and clinical use. Eighty-five years later the decision is unchanged. What has changed is who makes it. It is made when a resupply gap forces a dosing hold, when an excursion puts a batch in quarantine and a participant is due on Thursday, when a site cannot obtain comparator, when there is enough material for one cohort and two are enrolling. In 1941 the people making that judgement were the people making the medicine, and they were in the same building. Today they are in different organizations, often on different continents, and frequently nobody has been designated to make it at all. So, it gets absorbed. An investigator at a site decides, or a supply manager picks between two imperfect allocations, and neither of them was given a standard to apply or a place to record what they did.

Modern development is more regulated and technically complex, and the lesson is not to imitate wartime working practices. It is to apply the same discipline: a medicine is not fully designed until the product, process, clinical use, and supply system can work together, and until the clinical consequences of constraint have an owner.

What An Integrated Review Should Ask

Before protocol lock, a cross-functional product team should be able to answer:

1. Patient and site use: Can the planned presentation and administration model be used consistently across the proposed sites and patient population?
2. Product and process: Which product or process attributes are most likely to affect stability, release, comparability, or pack design?
3. Demand and distribution: What assumptions drive the supply strategy, and how would amendments, country activation, or patient behavior change them?
4. Operational learning: Which supply signals will be reviewed as evidence that the product, pack, protocol, or training should change, and which will also be read as evidence about the integrity of the data?
5. Future scale: What manufacturing, material, or distribution constraints could limit the transition to later-stage studies or commercial supply?
6. Clinical accountability under constraint: If supply is short, delayed, or quarantined, who decides what happens to the participants affected, on what clinical standard, and how is that decision recorded?

The sixth question is mine, and in my experience it is almost never answered in advance. It is not a hypothetical. Shortages, excursions, allocation decisions, and dosing holds are ordinary features of clinical programs. Each one turns an operational fact into a clinical judgement about identifiable people, and each one is currently answered by whoever happens to be holding it when it arrives.

The Authority The Review Needs

The review does not need another committee. It needs governance in the only sense that matters: a common decision record connecting the target product profile, protocol, CMC plan, and clinical supply strategy, with a named owner who has the authority to hold protocol lock when the evidence does not support it. Without that authority the review is a meeting. Dependencies get logged and the program proceeds as planned.

In 2009, sustained shortages of oncology medicines put physicians in the position of deciding which patients started treatment and which waited. I built the framework they used to decide, across nearly 200 markets. Clinical judgement was never the thing that was missing. Those physicians knew precisely what they were weighing. What they had no basis for was defending the decision afterward or recording it in terms that meant anything to anyone who reviewed it later. The framework produced no additional supply. What it produced was a decision that could be owned and audited, and that turned an operational failure into something the organization could account for rather than something it absorbed.

The same pattern appeared later and at far larger scale in the industry response to nitrosamine and related impurity findings. No single function could own the answer. Supply chain, manufacturing, regulatory affairs, and medical governance had to move as one, because the problem originated in process and supplier decisions and surfaced as a patient safety and regulatory question. Where the gaps between those functions had an owner, the work moved. Where they did not, it stalled.

The lesson for clinical development is direct. Integration of the kind the Penicillin Protocol describes will not survive contact with a program timeline unless somebody holds the authority to act on what integration reveals. Shared understanding without decision rights produces well-informed teams that proceed exactly as they would have anyway.

The Role Clinical Supply Leaders Can Play

Clinical supply leaders are well placed to sponsor this change because they operate at the interface of product, protocol, site, and patient. Their contribution is not simply executional. It is to bring forward the operational knowledge that shows whether a development plan will work in practice and to insist that the clinical consequences of supply constraint are decided deliberately rather than by default.

Four actions can start the shift:

1. Join candidate and protocol design discussions early. Ask targeted questions about stability, administration, site handling, labeling, distribution, and the feasibility of the intended patient experience.
2. Use an integrated product review, with medical in the room. Bring CMC, clinical development, medical, operations, quality, manufacturing, and clinical supply around the same target product profile and supply assumptions. A review convened to discuss the patient experience without a clinician present will produce operational answers to clinical questions.
3. Close the learning loop twice. Review supply data during the study for what they reveal about the product and protocol and for what they reveal about the integrity of the data being generated.
4. Name the owner before the constraint arrives. Agree at protocol lock who decides what happens to participants when supply is short or held, and on what clinical basis. That decision is far better made in advance, in writing, by someone with the standing to make it than at speed by whoever is nearest to it.

A Practical Definition Of Success

The traditional measures of clinical supply performance remain essential: the right product, at the right site, for the right participant, at the right time. An integrated model adds two further tests. Did supply intelligence change a decision early enough to make the study more robust? And when supply could not be delivered as planned, was the clinical consequence decided by someone accountable, against a standard, and on the record?

That is the practical value of the Penicillin Protocol. It treats supply as part of the system that creates the medicine rather than the final function charged with delivering a plan designed elsewhere. Penicillin was not made by a supply chain serving a scientific program. It was made by people who could not afford to treat the science, the process, the supply, and the patient in front of them as four separate problems. Modern development can afford to treat them separately, and mostly does. The programs that get into trouble are rarely the ones that failed to see the connection. They are the ones where seeing it changed nothing.

About The Authors:

Hedley Rees is a pharmaceutical supply chain consultant and author specializing in drug development, CMC, and supply-chain strategy. He is the author of Supply Chain Management in the Drug Industry (Wiley, 2011) and Transforming the Pharmaceutical Supply Chain (Wiley, 2025). The accompanying explainer, The Penicillin Protocol: Taking Medicines Back to the Future, introduces the integrated development model behind this article.

Angeliki Kotsianti is a physician-scientist and molecular pathologist. She spent nineteen years at Pfizer, where she chaired benefit-risk and held the final medical judgement on the company's medicines, including decisions on manufacturing, reformulation, recall and supply. She built Pfizer's global Medically Necessary Products strategy and co-led the company's representation in the EU Critical Medicines Alliance, whose work informed the Critical Medicines Act. She now advises on medical governance and regulatory strategy as a consultant with PharmaFlow Ltd, working on how decisions taken early in development determine whether medicines reach patients. Her specialist interest is the governance of evidence, including the validation of AI-derived findings. She is an inventor on eight patents in machine-learning diagnostics and co-founder of Oncolyze Inc.