A MACO Is Only as Good as the Assumptions Behind It

A maximum allowable carry-over value is not a toxicological number. It is a toxicological number multiplied by a set of process assumptions — the batch size it distributes into, the dose of the next product, the share of surface area that carries residue, the recovery of the method that measures it, and the cleaning step that was supposed to run. Read as a body of evidence, the enforcement record of 2024 to 2026 is rarely a story about toxicity assessment. It is a story about the assumptions wrapped around it.

From PDE to MACO: the calculation chain

The regulatory logic is settled. A health-based exposure limit — a permitted daily exposure (PDE), or a health-based exposure limit (HBEL) in EMA and PIC/S vocabulary — is derived from a point of departure divided by adjustment factors F1 to F5. ASTM E3219-25, the current edition of the HBEL guide, sets out the sequence in its publicly visible section 4.5: hazard characterisation; critical effect(s) with dose-response; point(s) of departure; PoD-specific adjustment factors; and calculation and justification of the final value. Its scope clause 1.3 covers APIs including biologics, intermediates, cleaning agents and excipients. The EMA guideline that made HBELs the expected basis for cross-contamination control has been effective since 1 June 2015; the PIC/S documents based on its 2018 Q&A have applied since 1 June 2020. There is no new science here.

MACO is the next link, and the character of the number changes there. The PDE is converted into a carry-over limit by dividing the PDE as a daily mass by the maximum daily dose of the subsequently manufactured product and multiplying by that product's minimum batch size:

PDE (mass/day) ÷ MDD of next product (mass/day) × minimum batch size of next product (mass) = MACO (mass)

That MACO is then distributed across shared product-contact surface area, corrected for recovery, and compared with a measured residue per unit area. Each term after the PDE is a claim about the physical world: that the next product really is the smallest-batch, lowest-dose product in the train; that batch size is a denominator the vessels actually support; that a known fraction of surface area carries the residue; that the swab or rinse recovers a known fraction of what is there; and that a cleaning procedure, at a defined frequency, produced that residue. The arithmetic cannot validate any of them. It only inherits them.

Where the enforcement record shows assumptions breaking (2024–2026)

The 2025 and 2026 warning letters do not attack the health-based logic of the limit — they attack the correspondence between the limit and the plant.

At Granules India Limited (Warning Letter 320-25-48, 26 February 2025, following an inspection from 26 August to 6 September 2024), FDA found significant contamination in multiple ducts of non-dedicated equipment, with swab samples downstream of HEPA filters carrying residues of several drug products and microbial contamination too numerous to count. The firm argued low risk from retrospective MACO calculations. FDA rejected that: "Cross-contamination is not uniform, and testing of limited reserve samples alone cannot ensure products are contaminant-free. While MACO is helpful in validated cleaning scenarios, it cannot apply where cleaning has not been performed, or exposure is inconsistent." FDA added that the batch-size denominators were inconsistent with the size of the relevant vessels, and that the limits were not adequately justified. Both failures are assumptions, not toxicology: one about whether the modelled cleaning happened, one about whether the denominator described the equipment.

At Cohance Lifesciences Limited (320-26-40, 30 January 2026, inspection 4–12 August 2025), FDA identified a 2018 cleaning validation study that had not used the identified worst-case product, a programme never re-evaluated as new products entered shared equipment, and no routine clean verification — and stated: "Your cleaning verification lacks the ability to detect cross-contamination from carryover." Analytical testing confirmed multiple APIs on equipment documented as clean, above the firm's allowable limit.

At Harbin Jixianglong Biotech Co., Ltd. (320-26-73, 1 May 2026), the firm's own cleaning procedure required a carry-over limit calculated from API toxicity; the validation study for equipment making peptide APIs including semaglutide and tirzepatide established none, justified no swab locations, and contained no swab results. An assumption chain with no first term.

The pattern repeats with variation. ABBE Laboratories (320-25-98, 6 August 2025) held a 2009 study that did not determine the worst case or the product cleaned, never revisited it when a titanium-dioxide lotion the firm judged hardest to clean entered shared equipment, and lacked API residue analysis. MediNatura New Mexico (320-26-27, 10 December 2025) had a last official validation study from 2007 and a programme FDA found not reflective of current operations. Dabur India (320-26-105, 24 July 2026) relied exclusively on visual inspection, with no scientifically determined MACO limits and no surface sampling of difficult-to-clean areas. Cosmetic Manufacturers (Aust) Pty Ltd (320-26-43, 6 February 2026) ran a protocol requiring swab samples from different locations in each qualification run, so reproducibility of the cleaning process was not demonstrated at all. Chemco Corporation (320-26-60, 7 April 2026) had no cleaning validation studies for non-dedicated equipment, and white residue remained in a tank whose logbook recorded it as clean.

Recovery and detectability

A MACO that no method can measure is a paper control. Three links in the chain are routinely understated.

First, recovery. An acceptance limit expressed as a residue per unit area means nothing without a recovery determination made on the same material, surface and technique as routine sampling, with the correction factor applied transparently. Where recovery is reported: a published laboratory study on cleaning validation in QC laboratories reported swab recovery above 70 percent and rinse recovery above 65 percent on stainless steel as achievable targets. A limit that silently assumes 100 percent recovery is set roughly 40 percent too loose.

Second, detectability. PIC/S PI 006-4, in force since 1 October 2026, makes the consequence explicit in section 7.13.2: products whose maximum allowable carry-over limits fall below the limit of visual detection — incorporating a safety factor — should be subject to cleaning verification after every change-over. Section 7.8.3 states that visual inspection will likely not be appropriate for lower-HBEL products or where visual thresholds sit near the required safety limit. The letter to Cohance is the enforcement form of the same point. The practical question is whether the limit of quantification, after recovery correction, sits below the carry-over limit for the lowest-HBEL product on the train — and what control exists where it does not.

Third, method character. Non-specific methods such as TOC and UV cannot distinguish residues, so all signal must be attributed to the worst-case component — a conservative bias that must nonetheless be documented. Rinse sampling averages results across the surface, so high-residue areas can be missed where swabbing would have found them. For biologics the residue that survives cleaning is degraded, and the relevant PDE is the one for the degradant rather than the active molecule: a bench-scale study of cleaning-induced degradants from a half-life-extended BiTE molecule and a monoclonal antibody, spiked at 650 µg, found no increased immunogenicity in human cell assays — which is why the resulting limits are analytically achievable at all.

Assumption governance

Standards now exist for the method layer. ASTM E3470-25 applies FMECA and HACCP/HARPC to cleaning risk identification, analysis and control, linking criticality to HBEL-derived specifications rather than workshop opinion — while noting in its significance section that microbiological concerns are not addressed by the HBEL approach and need a separate route. A 2026 review in the PDA Journal argues that the choice between worst-case matrix and product-specific validation is itself a risk-based decision to be documented, with visible residue limits as the bridge. The EMA/PIC/S concept paper on the Annex 15 revision proposes extending the annex to chemical and biological active substance manufacturers and aligning it with ICH Q9(R1), with publication now planned for April 2028.

The requirement is that each assumption has an owner, a source, a validity window, and a trigger. A named assumptions register — subsequent product, batch size, maximum daily dose, surface area, recovery factor, dirty and clean hold times, maximum campaign length, method LOQ — converts a MACO document into something change control can act on. The trigger list matters more than the register: a new or more potent product entering a shared train, a batch size or vessel change, a formulation change altering solubility or cleanability, a change of cleaning agent or cycle, or a new toxicological data point on the HBEL.

A practical checklist

• For every MACO on file, can we name the cleaning procedure it presumes, and show it was executed at the assumed frequency?
• Do our batch-size denominators match the vessels actually used?
• Is the worst-case selection current, documented, and formally reopened when a product or equipment change enters the shared train?
• Do the acceptance limits state recovery-adjusted values traceable to a recovery study on the same material and method as routine sampling?
• Does the method LOQ, after recovery correction, sit below the carry-over limit for the lowest-HBEL product in the train? If not, what independent control covers the gap?
• Do we hold data, not convention, for dirty hold time, clean hold time and maximum campaign length?
• Does the validation report contain the actual swab results, with a defined treatment of results below the limit of quantification?
• Are HBELs, including those for intermediates and cleaning agents, assessed against the current ASTM edition and re-opened when the toxicological database changes?

Fazit

A MACO without a documented cleaning step, a defensible worst-case basis, a matching batch-size denominator and a method that can detect the residue is not a weak control — it is not a control at all. The 2025–2026 record shows the failure mode consistently: not a wrong PDE, but an assumption that was never stated, never verified in execution, and never reopened when the plant changed. Teams that keep their assumption chain as explicit, versioned and testable as the toxicological derivation itself are the ones whose limits can be defended under inspection — and, more to the point, the ones whose cleaning actually protects patients.

Sources

Primary: FDA Warning Letter 320-25-48 to Granules India Limited, MARCS-CMS 697115, 26 February 2025. Supporting: FDA Warning Letters 320-25-98 (ABBE Laboratories), 320-26-27 (MediNatura New Mexico), 320-26-40 (Cohance Lifesciences), 320-26-43 (Cosmetic Manufacturers (Aust) Pty Ltd), 320-26-60 (Chemco Corporation), 320-26-73 (Harbin Jixianglong Biotech) and 320-26-105 (Dabur India); ASTM E3219-25 and E3470-25; PIC/S PI 006-4, in force 1 October 2026; EMA/CHMP/CVMP/SWP/169430/2012 and its 2018 Q&A; EMA/PIC/S Annex 15 concept paper EMA/INS/GMP/163466/2026; Cohen et al., J Pharm Sci 114(2):866–877, DOI 10.1016/j.xphs.2024.10.041; Boyapally, PDA J Pharm Sci Technol, DOI 10.5731/pdajpst.2026-000007.1; Moura et al., DARU J Pharm Sci 33:20 (2025), DOI 10.1007/s40199-025-00566-x.