AAV at Launch: The Hidden Work of Scaling, Supplying, and Proving Sameness
Moving an AAV program from pivotal into the real world exposes a different class of problems. Hospitals want backup vials, regulators want harmonized methods, and finance wants costs that hold under tariffs and dual-sourcing debates. The panel shared what they wish more teams modeled earlier, from demand peaks to compendial surprises and when to add a second site.
Captured live at Advanced Therapies Europe 2025, this session distills how to model demand, harden supply chains, align EU and US expectations, and stage scale for late-stage and commercial AAV launches.
TL;DR In a hurry? Here are the essentials at a glance:
Start with supply, not steel. Model country-by-country access, hospital overage, and where to place stock. Expect backup vial requests to double perceived output needs.
Demand is not flat. Many inherited indications create an early peak, then settle toward incidence. Decide now whether you will buffer with stock, validate multiple scales, or both.
EU vs US needs diverge in practice. Plan for EU QP release, local testing, and compendial differences. Potency often sails through; classic assays like mycoplasma trigger more debate.
Do not over-or undershoot scale. Push commercial forecasting early so your validation scale fits launch, not a guess.
Dual source with intent. Launching with two sites adds approval risk. Many teams add the second site post-approval when revenue supports it.
Comparability is a sampling problem first. Bank process intermediates and DS from early runs so you can bridge methods and standards later.
AI is useful, but targeted. Today it helps with clinical selection and mechanistic modeling; use digital twins to limit robustness burden and to justify what you do and do not test.
Model demand and supply early
Hospitals, payers, and ministries will shape your inventory more than your reactor size. Rare disease launches often need local backup vials, cold-chain staging, and just-in-time replacement plans. Build that overage into both COGS and capacity.
“Every hospital will want to have a small stash of things, and that typically requires a lot more vials than people expect.” — Bastiaan Leewis, MeiraGTx
Do not assume steady state. Many inherited indications spike at launch, then narrow to newborn incidence. Decide whether you will validate multiple scales for the peak or build stock and run down. Both have regulatory and cash impacts.
Choose where to make and test, on purpose
Manufacturing site location is more than tax and tariffs. Talent access, university links, and clinical network proximity all matter. In the EU, a QP will release your lots, and many teams still need some EU testing even if they make in the US.
“One consideration for Europe is always that you have to have a QP release. A testing site in Europe is typically needed even if you manufacture somewhere else.” — Bastiaan Leewis, MeiraGTx
Tariffs in the mid-teens may be absorbable in total COGS. If they climb, re-run the model. Keep commercial options open for fill-finish close to market.
Harmonize analytics, expect compendial surprises
Potency often gets a fair hearing when the scientific rationale is clear. The unexpected friction comes from classic compendial methods where countries differ on controls and replicates.
“We were expecting more issues with potency. The biggest comments came on traditional compendial assays like mycoplasma, and country expectations varied.” — Bastiaan Leewis, MeiraGTx
Identify country-specific twists early. Where possible, pick methods and acceptance criteria that will survive across regions, then document the scientific bridge.
Scale that fits validation and launch
Lock a validation scale that matches your first 12 to 24 months of commercial demand, not an optimistic midpoint. Push commercial teams for real forecasts so you do not validate at 50 L when you need 500 L, or the reverse.
“You need to think about your end point very early. Get the scale right so validation fits supply later on.” — Bernd Schmidt, Sensorion
Dual sourcing and second sites
Two launch sites can slow you down if one misses. Many teams validate a single site for approval, then bring a second site online post-launch to add resilience and regional flexibility.
“Launching with two sites adds approval risk. We prefer adding the second site post-approval when revenue supports it.” — Bastiaan Leewis, MeiraGTx
If you must dual source for volume, budget the extra comparability and align your filings so one site’s stumble does not stall the program.
Make-or-buy, and the real overhead line
In-house can remove real bottlenecks. Teams have internalized plasmids and QC when third-party queues stretched to 18 months. The tipping point is pipeline width and utilization, since overhead often outweighs unit COGS until capacity fills.
“We were waiting 18 months for an HPLC result. Analytical capability in-house becomes worthwhile sooner than you think.” — Bastiaan Leewis, MeiraGTx
What to want from a CDMO
Treat the selection as a partnership test, not a pricing exercise. Look for pockets of excellence that match your phase, low staff turnover, line of sight to commercial, and honest program governance when things go wrong.
“I want open conversations and a partner who has felt the late-stage pain before. Cost matters, but line of sight to commercialization matters more.” — Bernd Schmidt, Sensorion
You may pair a manufacturing specialist with a separate distribution expert. Few providers excel equally at both.
Comparability by design
Comparability fails without banked material. From your earliest GLP and clinical runs, archive process intermediates and drug substance in volumes that support future bridges across sites, methods, and reference standards.
“It is counterintuitive to take product away from patients early on, but without retained samples you cannot bridge later.” — Bastiaan Leewis, MeiraGTx
Document mechanism-based justifications and use them consistently across regions.
AI, digital twins, and what is ready now
AI is not a catch-all for CMC. Today, teams see value in digital twins and mechanistic models to focus robustness studies and defend design spaces. On the clinical side, AI helps with patient identification and secondary endpoints.
“We are using mechanistic models now. AI on CMC is still early, but digital twins can reduce the robustness burden you need to run.” — Bastiaan Leewis, MeiraGTx
Adopt targeted tools rather than broad promises. Use them to justify the testing you run, and the testing you do not.
Geopolitics, China, and risk
China’s capabilities are real, timelines can be fast, and trial models have been flexible. The political risk is also real. If you consider China, many teams limit scope to Asian supply within a broader global strategy and maintain non-Chinese options.
“Purely commercially, China is very attractive. The challenge is long-term security of supply if politics shift.” — Bernd Schmidt, Sensorion
Final word
Commercial AAV lives or dies on details that do not fit on a bioreactor slide. Model demand honestly, design supply with overage, harmonize the analytics that regulators actually question, and stage your second site when it helps rather than hurts. Archive early samples so you can prove sameness later. Do this well and AAV moves from a pivotal dataset to reliable patient access.
Key challenges include ensuring a robust supply chain, navigating regulatory requirements, securing manufacturing capacity, and addressing patient access issues.
Differences in regulatory frameworks lead to variances in analytical method acceptance and market access strategies, which can influence where to focus efforts in manufacturing.
CDMOs can help bridge gaps in manufacturing expertise and resources, streamline processes, and enhance flexibility as companies scale from clinical trials to commercialization.
Yes, technologies like liquid nanoparticles show potential; however, they need further scientific validation and consistency to be considered viable alternatives in clinical applications.
AI is being explored primarily for discovery phases; its application in manufacturing and comparability studies is still in early stages but presents opportunities for optimization in data handling.
)