The GLP-1 Shortage Didn't End. It Moved to the Autoinjector.

On April 30, 2026, the FDA proposed to leave semaglutide, tirzepatide, and liraglutide off the section 503B list of bulk drug substances. The proposal hit the Federal Register the next day (FR doc 2026-08552, published May 1), and the comment period closes on June 29, 2026. Strip out the procedural language and the message is blunt: the agency found no clinical need for outsourcing facilities to compound these three drugs from bulk. The only other legal route had already closed, since all three were cleared from the FDA shortage list (tirzepatide in October 2024, semaglutide on February 21, 2025). Lose both pathways and large-scale 503B compounding of GLP-1s is finished.

Compounding had been the pressure-relief valve for three years. Branded supply could not keep up, so outsourcing facilities filled the gap. The FDA is now welding that valve shut at the exact moment demand climbs again. Oral Wegovy launched on January 5, 2026, and wrote roughly 1.3 million prescriptions in its first quarter, the strongest US GLP-1 launch on record. Wegovy HD, the 7.2 mg injectable, was approved on March 19 and went nationwide in April.

Then comes the part the headline misses. Every patient pushed off compounded supply who still needs an injectable now depends on branded fill-finish and autoinjector assembly, and that is the one thing in this market that will not scale on a quarterly clock. You can make the peptide. Filling and assembling the device around it is where the line backs up.

The constraint moved downstream

For most of the GLP-1 era the open question was whether anyone could make enough drug substance, and peptide synthesis was the visible bottleneck. The industry has largely put that behind it. The bottleneck now sits one stage later, in sterile fill-finish and final device assembly, and the largest players have already voted with their balance sheets.

• Novo Nordisk chose vertical integration: Inside the $16.5 billion Novo Holdings acquisition of Catalent (completed December 2024), Novo paid roughly $11 billion to take direct control of three fill-finish sites: Anagni, Bloomington, and Brussels. Nobody spends $11 billion to own filling capacity unless filling capacity is what stands between them and their market.
• Lilly routed around the device: Single-dose vials scale far more easily than pens and autoinjectors, so the company pushed out vial doses of tirzepatide to relieve pressure while autoinjector assembly stayed the limiter. Lilly’s Concord, North Carolina site is dedicated to autoinjector fill-finish.
• Oral Wegovy proves the point: A 25 mg once-daily pill sidesteps sterile fill-finish and autoinjector assembly entirely, which is precisely why it reached roughly 1.3 million scripts in a single quarter. The injectable cannot copy that move. It is bolted to a primary container and a delivery device, and the two have to be qualified together.

Both companies read the situation the same way: the finished combination product is the constraint, not the molecule. Synthesize more peptide in a quarter and you have solved nothing, because a new aseptic filling line or a new autoinjector assembly cell takes far longer than a quarter to stand up. The constraint did not go away. It moved from the molecule to the machine that fills and assembles the device around it.

Why the device is harder than the molecule

An injectable GLP-1 is a combination product, not a standalone drug. It pairs the drug constituent with a primary container (the syringe barrel or cartridge) and a delivery device (the pen or autoinjector), all regulated together under 21 CFR Part 4. The drug and the device get reviewed and qualified as one system rather than as separate parts bolted together at the end.

That coupling is where the timeline goes. Aseptic processing runs under EU Annex 1 (revised) and 21 CFR 211 cGMP, and the rate-limiting assets are isolator-based filling lines and final autoinjector assembly equipment, which carry the 18-to-24-month lead times that anchor this briefing. New commercial prefilled-syringe and cartridge isolator lines are scheduled through mid-2027 and 2028. That schedule is the real clock, and it does not bend because a comment period closed or because a launch went well.

When 503B supply disappears on June 29, the demand it was absorbing does not evaporate. It transfers onto branded device-delivered product, against a fill-finish and assembly base that was already the constraint and is booked years out. Demand moves at the speed of prescriptions. Device capacity moves at the speed of capital equipment. The two will not meet on time.

The container/closure trap: silicone oil and glass

Even with a filling line in hand, the primary container resists the formulation. High-concentration subcutaneous biologic and high-concentration drug-product formulations meet a glass syringe that was never designed to be gentle with them, and the failure modes are well documented.

• Silicone oil: A glass-barrel syringe needs silicone oil so the plunger can glide, and that lubricant does not stay put. It sheds into the product, migrates to the oil-water interface, and can interact with protein to nucleate aggregation and sub-visible particulates. For a sensitive biologic, the lubricant that makes the syringe work doubles as a particle and immunogenicity risk.
• Tungsten: The pin that forms the needle bore in a glass syringe leaves tungsten residue behind, and tungsten is a known nucleator of protein aggregation for sensitive molecules. The manufacturing of the container introduces a contaminant into the container.
• Glass delamination: High-pH and high-ionic-strength solutions can leach lamellae, thin glass flakes, off the interior surface of the barrel. That is a recall-grade defect, it has already driven recalls in high-pH biologics, and it is one of the main reasons the industry is walking away from glass.

None of these are edge cases. They make up the standard hazard list for any high-concentration, self-injected product in a glass primary container, and they are exactly the risks a CMC team has to flag before a formulation ever reaches a filling line.

The COP pivot and its hidden cost

The industry's response is a material change, from siliconized glass to cyclic olefin polymer (COP) and related cyclic olefin copolymer (COC) barrels. The case is strong, but the spec sheet leaves out a cost: every polymer barrel and every new elastomer closure resets the extractables and leachables (E&L) data package.

The market is moving with that logic. One market analysis (GlobeNewswire, February 2026) projects the overall prefilled-syringe market growing from $9.71 billion in 2025 to $18.08 billion by 2031, a 10.93% compound annual growth rate, with the plastic and polymer (COP) sub-segment growing faster at an estimated 11.71% CAGR. The same report projects global GLP-1 sales rising from roughly $40 billion in 2023 to about $150 billion by 2032, the demand pull underneath the whole container shift. Those market figures are projections from a single published analysis, not booked results, and should be read that way.

Here is the cost that pivot hides. Glass was risky because of its interaction chemistry, and that chemistry does not disappear with the switch to COP. It changes, and a changed interaction profile means qualifying the new material and the new closure from scratch. That E&L work was always required. What is new is that it now sits squarely on the critical path. Once the molecule is settled and a filling line is available, the container/closure qualification is the gate. The regulatory map for that work is specific:

• USP <1663>: governs the assessment of extractables: what comes off the packaging and delivery system under worst-case stress extraction.
• USP <1664>: governs the assessment of leachables: what actually migrates into the product under in-use conditions.
• ICH Q3E: is the harmonized guideline for assessing and controlling extractables and leachables in drug products.
• PQRI PODP: sets the analytical evaluation threshold (AET) for parenterals: 1.5 µg/day for non-cohort-of-concern substances and 5 µg/day for sensitizers and irritants.

Any switch from siliconized glass to a COP barrel, or any new rubber or elastomer closure, resets that study and its threshold derivations. This is the qualification that gates a new container/closure system.

High-concentration subcutaneous is a viscosity-and-force problem

One more coupling tends to ambush teams that treat formulation, container, and device as sequential handoffs. To deliver a meaningful dose in a small subcutaneous volume, concentration has to rise, and viscosity rises with it on an exponential curve rather than a straight line. High-concentration biologic formulations can reach viscosities as high as roughly 1000 cP (per the peer-reviewed literature on high-dose subcutaneous delivery).

Now put that fluid in a device. Subcutaneous volume has historically been capped near 1 to 2 mL (recently extended toward 3 mL), and autoinjectors target roughly 1.0 mL delivered in 10 to 15 seconds. Pushing a viscous fluid through a fine needle on that schedule forces the device to apply more force, and excess force can stress or damage the primary container itself. Formulation viscosity, container material, and device spring force are not three separate problems. They are one coupled design problem, and solving them in sequence is how programs discover at the device stage that the formulation was never deliverable.

The system, not the molecule, is the hard part: a high-concentration formulation that resists aggregation at the silicone-oil and air-liquid interfaces, a container that does not shed particles or delaminate or leach above threshold, and a device that delivers the dose without damaging the container. Get any one of those wrong and the combination product does not ship.

Where DeepC fits

DeepC does not synthesize your peptide, and it should not claim to. GLP-1s are peptides, not monoclonal antibodies. What DeepC brings is the discipline the device crunch now demands: the high-concentration subcutaneous CMC reasoning, the container/closure extractables qualification, and the in-use leachables safety case that gate every prefilled syringe and autoinjector, whatever molecule sits inside. Three named agents map onto the three problems this briefing lays out.

Biologics Formulation & CMC Agent

This is the agent for the formulation-meets-device problem. It produces structured CMC memos (BF01, BF02) that stream into the side panel, and its stated scope covers high-concentration subcutaneous formulations directly: viscosity management, aggregation pathways, immunogenicity flags, and fill-finish considerations. That is the exact list from the viscosity-and-force and silicone-oil sections above.

It works against real precedent rather than training-data recall. It cross-references excipients against the unified IID and PharmaExcipients surface for maximum permissible levels per route, which is how you justify the buffer and surfactant choices that suppress aggregation at the silicone-oil and air-liquid interface. It extracts precedent from approved-product labels via DailyMed (BLA to SETID to SmPC section 6.1), so you can see what excipient and container systems approved high-concentration subcutaneous products actually used. Reports follow the ICH framework: Q5C (stability of biological products), Q6B (specifications), Q2(R2) and Q14 (analytical procedure validation and development), and Q5A(R2) where viral safety applies. The structure is fixed: Overview, Excipient Rationale, Regulatory Framework, CMC Red Flags with severity, Precedent Products, and Recommended Next Steps.

The CMC Red Flags section earns the agent its place on a device program. Silicone-oil sensitivity, the viscosity-versus-device-force trade, and container-interaction risk all get flagged with severity before they reach a filling line, not after. A representative request from the product is a CMC memo for a high-concentration IgG4 mAb intended for autoinjector subcutaneous delivery: the same high-concentration-SC and fill-finish reasoning a GLP-1 device team needs, applied to the modality the agent is built for. See the biologics agents for the full scope.

Extractables Agent

Switch from siliconized glass to a COP barrel and a new elastomer closure, and this is the agent that regenerates the extractables qualification the change forces. It qualifies substances from container/closure stress-extraction studies under USP <1663> and produces ICH Q3E and USP <1664> reports anchored to PQRI PODP, with the AET set at 1.5 µg/day for non-cohort-of-concern substances and 5 µg/day for sensitizers and irritants in parenteral products.

It captures the study context that decides whether a result is defensible: solvent system, temperature, duration, surface-area-to-volume ratio, and detection method. It computes per-substance worst-case daily exposure from extractable mass per unit and units per day. When direct toxicology data are missing, it runs read-across via structural similarity against named class analogs, including bromobutyl and chlorobutyl rubber analogs by CAS, which is exactly the chemistry of the elastomer plunger and closure on a prefilled syringe. It runs in two modes: a full acceptance argument when concentrations are in hand, or a threshold-only derivation while you are still waiting on stress-extraction analytical output. One of its listed example requests is to qualify the extractables from a prefilled-syringe stopper study against an IV dose, which is this exact problem.

Leachables Agent

Extractables tell you the worst case. Leachables tell you what actually migrates into the product across a chronic, self-injected dosing schedule, and for a weekly or daily GLP-1 that distinction matters. This agent derives per-substance in-use migration thresholds: total daily intake from concentration times maximum daily dose volume, the ICH M7(R2) less-than-lifetime brackets quoted explicitly (120 µg/day at one month or less, 20 µg/day for one to twelve months, 10 µg/day for one to ten years, 1.5 µg/day above ten years), the PQRI PODP 5 µg/day floor for sensitizers and irritants, and a margin-of-safety calculation against the lower of the two. It runs the EMA Appendix 1 nitrosamine lookup when a leachable falls into the cohort of concern.

The chemistries it handles explicitly are the ones a siliconized or COP prefilled-syringe and autoinjector system actually generates: silicone oil, aldehydes from rubber-stopper degradation, antioxidants (BHT, BHA, Irganox), vulcanization residues (MBT, ZDC), and polymer oligomers. That is the in-use migration profile of a chronic GLP-1 dosing schedule, assessed against the right thresholds.