Oncology Clinical Trial Supply: Forecasting a Dose Nobody Knows in Advance

In most therapeutic areas, a patient's supply need is fairly predictable: one kit per visit, a fixed dose, a known treatment duration. Oncology breaks every one of those assumptions at once. The dose depends on the patient's body size and changes when they lose weight. It gets reduced for toxicity. It's given alongside two other drugs, one of which you may have to buy commercially in twenty countries. And the treatment continues until the disease progresses — which, for your best responders, may be years longer than anyone planned.

Oncology is the largest therapeutic area in clinical development, and it's also where supply forecasts are most often wrong. Not because oncology supply teams are less capable, but because the variables compound. Here's how to plan for each of them.

Why is oncology trial supply so different?

Five features, usually present together:

1. Body-size-based dosing. Many oncology drugs are dosed by body surface area (mg/m²) or body weight (mg/kg), so the number of vials per dose varies from patient to patient.
2. Site-prepared IV administration. Doses are often prepared by a site pharmacist from multiple vials, introducing preparation, in-use stability, and handling constraints.
3. Combination regimens. The investigational drug is frequently given with standard-of-care chemotherapy or other agents that must be sourced, supplied, or reimbursed.
4. Dose modifications. Reductions, delays, and interruptions for toxicity are routine and alter consumption mid-treatment.
5. Treat-until-progression. Treatment duration is open-ended, creating a long and highly uncertain demand tail.

How does BSA- or weight-based dosing affect forecasting?

With fixed dosing, you forecast patients and visits. With body-size dosing, you also have to forecast vials per dose — and that depends on the distribution of patient size in your population, the vial strengths available, and whether vial sharing is permitted (typically it isn't in trials, so every partial vial is waste).

A practical approach:

• Model the population's size distribution, not an average patient. A trial in a population with significant weight loss behaves very differently from one in early-stage disease. Regional differences matter too — a global trial's average body size may differ markedly between countries.
• Map dose bands to vial counts. For each dose level, calculate how many vials a patient in each size band consumes, including the wasted remainder.
• Test vial strength options early. Offering two vial strengths can materially reduce waste for high-dose patients — but it complicates packaging and IRT. This is a decision for the protocol stage, not the first resupply.
• Recalculate as data arrives. Actual patient size distributions from the first cohorts should replace assumptions quickly.

This is exactly where a probabilistic forecasting approach earns its keep: you're forecasting a distribution of distributions.

How should you handle combination therapy and standard of care?

Combination regimens create a sourcing problem that often dwarfs the IMP problem. The background agent may be:

• Supplied centrally by the sponsor, sourced commercially and relabeled — maximum control and consistency, maximum cost, and exposure to commercial shortages.
• Sourced locally by sites from hospital stock — faster and cheaper, but brand, formulation, and availability vary by country, and reimbursement rules differ.
• A mix, depending on country, blinding needs, and whether the agent is part of the comparison.

The decision should be made country by country and written down, because the answer in Germany is rarely the answer in Brazil. If the combination agent is also your comparator, the full set of comparator sourcing risks — shortages, batch consistency, expiry — applies. And remember the premedications, diluents, and infusion materials: in oncology, ancillary supplies are unusually numerous and unusually likely to stop a visit.

How do dose modifications change consumption?

Dose reductions reduce vial consumption for body-size dosing but may increase kit consumption for fixed-strength oral drugs if a lower strength is needed. Dose interruptions shift demand later in time. Both are common enough in oncology that ignoring them produces systematic errors.

Build them into the forecast explicitly: an assumed proportion of patients reducing at each dose level, informed by earlier-phase safety data, and a plan for the lower strengths to exist at sites before the first reduction is needed. A patient whose reduction dose isn't on the shelf is a patient who either misses treatment or receives the wrong dose.

How do you forecast a treat-until-progression tail?

This is oncology's defining supply challenge. Median treatment duration is only half the story; supply risk lives in the tail. A handful of long-term responders can remain on treatment for years — often after the main study has read out, in an extension or rollover protocol.

Plan for it deliberately:

• Model duration as a distribution, using earlier-phase or class data, not a single median.
• Separate the "core" and "tail" supply problems. Core demand is large and fairly predictable; tail demand is small, long, and demands continuity — shelf life, re-supply of a few patients across many sites, and eventually transition to post-trial access.
• Plan manufacturing for the tail. Late small batches are expensive; one well-timed campaign with adequate stability may be better than several tiny ones.
• Decide early what happens after the study ends — extension, rollover, post-trial access, or commercial supply.

What about preparation, handling, and blinding?

Site-prepared IV doses add their own constraints: in-use stability of the prepared infusion, compatibility with specific infusion materials, and handling requirements for cytotoxic or hazardous drugs. Where the study is blinded, preparation is often done by an unblinded pharmacist who must be protected from revealing the assignment — with matched infusion bags, overlays, or tubing where the products differ in appearance. Our blinding post covers where these designs typically fail.

And if the program uses adaptive and basket designs — increasingly common in oncology — every one of these variables is multiplied across arms that may open or close mid-study.