21 ADCs Are Approved. Your Plant Still Cannot Safely Weigh the Payload.

Two Approvals, One Payload Class

On 14 May 2025, the FDA cleared Emrelis (telisotuzumab vedotin-tllv) for c-Met-high non-squamous lung cancer. Strip away the antibody and the linker and what remains is the part that does the killing: MMAE, a microtubule inhibitor too cytotoxic to have ever been developed as a standalone drug, because the therapeutic window did not exist. Four months earlier, on 17 January 2025, Datroway (datopotamab deruxtecan-dlnk) reached the market for HR-positive, HER2-negative breast cancer, carrying a topoisomerase I inhibitor of the deruxtecan class. The logic is identical. You bolt a molecule too toxic to give a patient on its own to a targeting antibody, and that is the whole point of the format.

Two approvals in one year, both built on payloads in the same potency class. One question decides whether you can make either of them. When a technician walks up to a jar of that bulk payload and unscrews the lid, what airborne concentration are they allowed to breathe over an eight-hour shift?

For these compounds the answer is a number most people outside industrial hygiene have never had to think about, and it is not in milligrams. It runs in nanograms per cubic meter, sometimes in the low single digits. That number is the Occupational Exposure Limit, and it is the binding constraint on a highly potent program. Containment, not chemistry and not dose, is what governs.

Three numbers that set the terms

There are 21 antibody-drug conjugates approved worldwide across the FDA, EMA, China's NMPA, and Japan's MHLW as of 2025, roughly 15 of them by the FDA cumulatively, per Huateng Pharmaceutical and the ADC Directory. The programs are real, the regulatory path is proven, and the pipeline behind them is large.

Under 1 microgram per cubic meter is the OEL that defines an OEB5 compound. That is the band where "highly potent" carries an engineering specification rather than a marketing adjective, and below the line dedicated closed isolators are no longer optional. Under 50 nanograms per cubic meter is the typical OEL for an ADC payload, and many fall to single-digit nanograms or near 1 ng/m3, per FPS Pharma. That is fifty to a thousand times below the OEB5 floor, and it is the number that runs your building.

The Number That Runs the Building

An Occupational Exposure Limit is the airborne concentration, expressed as an eight-hour time-weighted average in micrograms per cubic meter, that a worker can breathe across a working lifetime without harm. It reads like a safety footnote and behaves like the master variable for your facility.

The mechanism is straightforward. The OEL drives the OEB band (the Occupational Exposure Band), the OEB band drives the engineering controls, and the engineering controls are capital. The gap between a general-ventilation room and a glovebox-grade containment isolator shows up in your capex line, your floor plan, your validation timeline, and your insurance. One number, derived from a toxicology dataset, propagates all the way to the concrete.

There are two ways to get that number wrong, and both are expensive.

The right OEL is the defensible one: the number an industrial hygienist derived from the actual point of departure, every uncertainty factor justified, that a Qualified Person and an auditor can both re-derive and sign off on. For most new oncology payloads there is no published exposure limit to lean on, which is exactly why the in-house derivation has to be right.

What an ADC Payload Actually Demands

The OEB framework stratifies compounds by potency, and each band carries an implied containment expectation. The boundaries below are an industry convention rather than a single regulatory standard, but Adragos, PharmaSource, Qualia, and Pharmaceutical Technology converge on the same lines.

Now place an ADC payload on that table. With a typical OEL under 50 ng/m3, it lands in neither OEB4 nor OEB5. It sits in OEB6 territory, the ultra-potent tier, and sometimes near the floor of it. A documented European case study put estimated OELs around 1 ng/m3, close to the limit of what current containment can even verify, and that is why some manufacturers have started floating an OEB6 or OEB7 designation just to have language for it, per Pharmaceutical Technology.

Where you land dictates the containment, and the progression is not a menu:

• Open handling: for the least hazardous compounds.
• Local exhaust ventilation and ventilated enclosures: as potency climbs.
• Split-butterfly valves and contained transfer: to move powder without breaking containment.
• Fully closed pharmaceutical isolators: under negative pressure for OEB5.
• Gloveboxes and high-containment isolators: for OEB6 and the single-digit-nanogram payloads.

An OEB5 isolator is typically expected to demonstrate containment performance below roughly 0.1 µg/m3 in surrogate testing, with a containment factor on the order of a million to one, per Qualia. An ADC payload sitting below that line needs the upper end of that capability, verified, before a single gram of real compound enters the suite.

The 2026 Procurement Reality

For years containment was a downstream engineering detail: pick the CDMO, win the program, then sort out the suite. That sequence has flipped.

On 22 May 2026, the CDMO Adragos Pharma published a containment checklist for HPAPI sourcing that states the case plainly: containment verification is the mandatory first gate in any highly potent CDMO selection. Before a sponsor evaluates capacity, analytical capability, price, or timeline, they confirm one thing. Does the facility's containment infrastructure match the OEB band of this specific compound, demonstrated with facility-specific surrogate Containment Performance Test data?

That reorders the entire conversation. The OEB band stops being something your engineering team derives after the deal closes. It becomes the first line on the RFP, and the OEL that drives it becomes the first number a sponsor has to defend. The practical test that now separates real capability from marketing is blunt: show me your SMEPAC-equivalent surrogate-powder data, at this OEB level, for this equipment train.

Underneath that procurement gate sits a stack of standards. ISPE's Risk-MaPP Baseline Guide (Volume 7), aligned with ICH Q9 quality risk management, is the framework for containment and cross-contamination strategy. ASTM E2476 supports the risk-based assessment. SMEPAC, the Standardized Measurement of Equipment Particulate Airborne Concentration, is the ISPE method that verifies a given piece of equipment actually achieves the target containment with a surrogate powder before you trust it with the real one.

Get the OEB classification wrong here and the cost is immediate. Classify too aggressively and you are sourcing a glovebox-grade suite you do not need, pricing yourself out. Classify too loosely and you become a sourcing dead end the moment a competent sponsor asks for the SMEPAC data, because the facility you picked cannot meet a band you under-called. The derivation is load-bearing at the procurement stage now, not just the engineering stage.

This is also where the demand comes from. Oncology accounts for roughly 72.53% of highly potent API market spend, per Mordor Intelligence, driven by cytotoxic dosing regimens and the ADC pipeline. PharmaSource reports that about 28% of FDA new molecular entity approvals in 2024 were highly potent. Mordor Intelligence projects the HPAPI market growing from about $29.34B in 2025 to roughly $32.02B in 2026 and $49.59B by 2031 (about a 9.14% CAGR); Precedence Research, working from a separate base, projects roughly $69.13B by 2035. Both are projections from market-research firms rather than booked revenue, and the two bases differ, so treat them as directional. The direction itself is not in doubt. More than 100 ADCs are in clinical development worldwide as of late 2025, per Precision for Medicine, and every one that succeeds becomes a containment problem somebody has to solve before the first commercial batch.

The Hardest Case: Genotoxic Payloads With No Safe Threshold

So far this has read like a derivation problem with a clean answer. For ADC payloads, what these molecules do makes it worse.

The classic ADC payloads are potent and genotoxic or anti-mitotic by design. The auristatins (MMAE, the Emrelis payload) and maytansinoids are tubulin inhibitors. The deruxtecan and exatecan class (the Datroway payload) are DNA-damaging topoisomerase inhibitors. You reach for these agents precisely because they disrupt cell division or damage DNA, and they stay good at it when the cell belongs to a manufacturing operator instead of a tumor.

For a standard toxicant, you derive an OEL from a No Observed Adverse Effect Level: a dose below which nothing bad happens, scaled down with uncertainty factors. A genotoxic compound presumed to act without a threshold gives you no NOAEL to anchor to, because the premise is that any exposure carries some risk. So the derivation changes. You either extrapolate linearly from a carcinogenic-potency measure (a TD50), or, when potency data are absent, you fall back to the ICH M7 Threshold of Toxicological Concern.

Watch where that lands. The M7 systemic TTC is 1.5 µg/day. Convert it to an airborne limit using the default worker breathing volume of 10 m3 per shift, and you get:

1.5 µg/day divided by 10 m3/shift = 0.15 µg/m3 airborne, as a conservative starting OEL.

Source: DeepC's own M7-TTC derivation, not an external citation.

That arithmetic, which is DeepC's own derivation rather than an external citation, puts a genotoxic payload at 150 ng/m3 before you have looked at a single compound-specific study, and real ADC payloads with potency data routinely come in an order of magnitude below that, into the picogram-adjacent range. If the molecule is also a respiratory sensitizer, a separate hard floor applies: below 0.01 µg/m3, regardless of what the threshold math says, because sensitization is a different kind of hazard that potency-based derivation does not capture.

These derived limits get sanity-checked against published NIOSH Recommended Exposure Limits and OSHA Permissible Exposure Limits (29 CFR 1910.1000) wherever the API or a close analogue appears. For most new oncology payloads, nothing appears. No PEL, no REL, no precedent to point an auditor at. The only number available is the one you derive, and that is why a defensible, fully shown derivation is not a nicety. It is the only thing standing between your operators and an exposure no regulator has set a limit for.

Two More Traps in a Shared Facility

Containment of the bulk powder is the headline problem, and it is not the only one. The same potency that drives the OEL drives two adjacent failure modes.

Neither of these replaces the containment problem; they compound it. A potent ADC payload makes you solve all three at once: how to handle it, how to clean after it, and how to distribute milligram doses of it uniformly.

Where DeepC Fits

Every problem in this briefing reduces to the same thing. You need a defensible number, derived from real data, that a Qualified Person and an auditor can re-derive line by line, produced before you commit capital. That is what DeepC's regulatory-safety agents are built to generate, and three of them map directly onto the three problems above.

The OEL Agent derives the number that runs the building

The OEL Agent derives Occupational Exposure Limits and OEB containment bands following Sussman/Naumann/Pfister (Tox Sci 2016) and ECETOC TR 101, with output in µg/m3 as an eight-hour TWA. It is built to answer the exact request a CMC or EHS team brings to it: set an OEL and OEB for a highly potent oncology API for tablet manufacturing operations. Mapped to the problems in this briefing:

The PDE Agent sets the cross-contamination limit

The PDE Agent derives Permitted Daily Exposure values for cross-contamination control in shared facilities per EMA/CHMP/CVMP/SWP/169430/2012, the basis for cleaning validation and MACO. It computes PDE as (NOAEL x 50 kg) divided by the F1 through F5 factor stack, each factor justified, and returns worked MACO examples expressed in mg per equipment area and ppm in the next-product batch. It handles the cases that matter most for potent compounds: highly sensitizing APIs get a dedicated-facility recommendation per GMP Section 3.6, and genotoxicants without a threshold drop to the TTC fallback. This is the agent that answers the real question behind the residue limit. Can I clean between campaigns, or do I need to dedicate the suite?

The Formulation Agent designs for the low dose

The Formulation Agent designs formulations to ICH Q8(R2) and CTD Module 3.2.P.2: a Quality Target Product Profile, Critical Quality Attributes, Critical Material Attributes and Process Parameters, an ICH Q9 risk assessment, a control strategy, and a multi-criteria comparison of three to five candidate formulations from distinct strategic angles. Its documented example requests include the approach for a highly potent, moisture-sensitive API. For an ADC or ultra-potent finished product, that means treating content uniformity as the primary CQA it is, and reasoning through the carrier and diluent strategy, blending approach, and the trade-offs that come with distributing milligram-and-below doses uniformly, with each candidate's rationale and risk made explicit.