Tech Transfer Is Two Jobs, Not One
We use one phrase, tech transfer, to describe two jobs that have almost nothing in common.
A drug product transfer and a drug substance transfer look like the same exercise on paper. Both get a line on the program timeline, a budget, a project lead, and a stack of documents to move from one site to another. Filed under the same heading, they tend to get planned, resourced, and risk-rated as if they are the same kind of work. They are not.
A drug product transfer is, most of the time, a mechanical exercise. You are moving a formulation step, a filtration, a fill into a vial or syringe, and maybe a lyophilization cycle. None of that is trivial, and any of it can go wrong, but the technical risk of moving it from one site to another is usually low. The process is well characterized, the unit operations are familiar, and a competent receiving site has done something very close to it before. You can plan around it with reasonable confidence.
Drug substance is a different animal entirely, because you are not moving a process. You are moving live cells, and live cells are finicky. They do not read your project plan. A single difference between the sending site and the receiving site, something that looks minor on paper, can move titer, shift product quality, or turn a campaign into a string of investigations. The number of things that can affect yield in a single campaign at one established site is already large. The number of things that can affect it during a transfer, when you are changing the equipment, the scale, the materials, and the hands all at once, is exponentially higher.
That is the part the shared label hides. When a drug substance transfer is planned with the same assumptions as a drug product transfer, the timeline, the comparability strategy, and the contingency all end up sized for the wrong job. The biology does not care what the plan says. It will tell you what it thinks of your assumptions when the first runs come back, and by then you are reacting instead of preparing.
What preparing actually looks like
If a drug substance transfer carries this much more risk, then it deserves more than a bigger version of the drug product playbook. A few things have stood out to me as the ones worth getting right early.
The first is collaboration between the sending and receiving sites, and it is the one most often skipped. The more openly the two organizations can work together, the easier the troubleshooting gets. The sending site can share what they have learned about the process, the quirks that never made it into a document, and the receiving site can surface the issues they are seeing in a way that actually helps. Building that kind of forum is not always necessary, and it is not always possible. It depends a great deal on the relationship and on what the future holds between the two sites. If you are transferring away from a site, or the sending site is a competitor of the receiving one, the incentives to help may simply not be there. But where you can build it, it is hugely valuable, and it tends to pay back into the relationship as well as the transfer.
The second is comparability, started as early as you can manage, well before process validation. It is tempting to treat comparability as a late-stage confirmation step, something you run once the process is settled. The trouble is that yield and operational metrics can look fine while the product itself has quietly shifted. Being able to see molecular changes across the two processes, not just titer and operational changes, is what tells you whether you have actually moved the same product. Glycosylation, charge variants, aggregation, the attributes that matter for the molecule are the ones you want eyes on early, while there is still time to adjust, rather than discovering a shift after you have committed to the process.
The third is timeline buffer, when you are fortunate enough to have any. It almost always gets eaten up. That is not a sign the buffer was wrong. It is there to be eaten up. That is the point. A drug substance transfer planned with no slack is planned to fail at the first surprise, and there is almost always a surprise.
Underneath all of it sits the reason any of this matters, which is the sheer number of variables that can move the outcome. Bioreactor scale and geometry play a large role. So does impeller type, and the shear stress that comes with it. So do pH, dissolved oxygen, and CO2 levels. And sometimes the culprit is something as maddening as lot-to-lot variability in the raw materials that go into the media or the feeds, a difference you did not introduce and cannot fully control. In a single campaign at one site, any one of these can shift titer. In a transfer, you are changing several of them at once, which is exactly why the risk does not add up so much as it multiplies.
The point
None of this makes a drug substance transfer safe. It makes it honest. The goal is not to eliminate the risk, because you cannot, but to plan the work as the thing it actually is rather than the thing the label suggests. A drug product transfer you can largely plan around. A drug substance transfer you have to prepare for, with collaboration, early visibility into the molecule, real buffer, and a clear-eyed respect for how many ways live cells can surprise you. The teams that do well are not the ones that avoid the surprises. They are the ones that expected to have them.