FDA's 1 May 2026 warning letter to Harbin Jixianglong Biotech found a cleaning procedure that required a toxicity-based carryover limit, and a validation study that established none — with no justification for swab locations and no swab results.
Standfirst
A warning letter issued on 1 May 2026 to a Chinese peptide API manufacturer describes a cleaning validation package missing the three things an inspector checks first: a numeric carryover limit, a rationale for the swab locations, and the swab results.
What happened
FDA inspected Harbin Jixianglong Biotech Co., Ltd. (FEI 3024038751) in Harbin, Heilongjiang, from 3 to 7 November 2025. Warning Letter 320-26-73, MARCS-CMS 723330, dated 1 May 2026, was issued by CDER under the subject "CGMP/API/Adulterated and Misbranded Drugs," and identifies the firm as a manufacturer of peptide APIs, including Glucagon-Like Peptide-1 (GLP-1) receptor agonist APIs.
On cleaning, the letter states that the firm "lacked adequate cleaning validation studies that demonstrate your cleaning procedures for your manufacturing equipment are effective." Three specifics follow. The firm's own cleaning procedure requires calculation of a maximum allowed carryover limit based on API toxicity, yet the firm "did not establish a maximum allowed carryover limit in your cleaning validation study for the equipment used to manufacture peptide APIs, including semaglutide and tirzepatide." The study "did not provide justification for the swab sampling locations," and it "did not contain results from the swab sample testing as required by the same cleaning procedure."
The firm replied that it would run a new cleaning validation protocol during its next campaign, with chemical residue testing on direct and indirect product surfaces. FDA found the response inadequate: it lacked a systemic corrective and preventive action for the "misalignment between your cleaning procedure requirements and your cleaning validation study."
The cleaning findings sit inside a broader package: relabelled semaglutide bought from suppliers not on the approved list, with manufacturing and retest dates changed without supporting data; two semaglutide batches shipped to the U.S. without process validation; tirzepatide released before stability data supported its retest date; and analytical method gaps for assay, related substances and bacterial endotoxin. The letter records a 10 February 2026 teleconference at which the firm agreed to voluntarily recall the two relabelled semaglutide batches in U.S. distribution.
Why it matters
A maximum allowable carryover limit is the quantitative bridge between a cleaning procedure and a sampling result: it converts toxicological and dose information for the previous product into a residue concentration on a defined surface area, against which swab and rinse results can be judged. Without the limit, a swab result has nothing to compare against and the procedure cannot be shown effective. Likewise for location rationale: without a documented worst-case argument, clean swabs prove only that the sampled points were clean.
What is particularly interesting
The gap is not a missing procedure. The procedure existed and demanded a toxicity-based carryover limit; the study did not deliver one. That is an internal contradiction a quality unit can find before an inspector does, by reading the SOP and the validation report side by side.
The peptide context sharpens the arithmetic. Semaglutide and tirzepatide are high-potency molecules run on shared equipment — precisely the situation health-based limits were designed for, since low doses drive low permissible carryover. A toxicity-based approach is defensible, but it needs a derived, documented number.
Residue methods for peptides on stainless steel must be validated for recovery and specificity, and the analytical weaknesses the letter describes for related substances (poor resolution, overlapping peaks) tend to surface in cleaning support methods too.
The response critique is instructive: committing to a protocol in a future campaign is a plan, not a systemic correction, and FDA's objection concerned the system that allowed the misalignment.
How such a gap can realistically develop
The trajectory is ordinary. A procedure is adopted from a template and includes a carryover-limit requirement. The limit is never calculated for the actual equipment train and product mix, because the calculation needs PDE/ADE inputs, batch sizes and surface areas held by different functions. Sampling locations are inherited from an earlier equipment set. Validation work is scheduled around production campaigns, so under capacity pressure the study becomes a documentation package for whatever campaign ran rather than a scientific exercise. Each product added to the shared train should add a row to a cleaning matrix, and sometimes does not. Over time the study's purpose drifts from demonstrating effectiveness to satisfying an SOP requirement, and the two documents stop agreeing. None of this is a single deliberate decision; it is the product of deferral.
Questions worth asking in our own programme
- For every procedure requiring a numeric carryover limit, can we point to the derivation — formula, PDE/ADE inputs, worst-case product, surface area, batch size?
- Do our swab locations have a written rationale tied to equipment drawings and hardest-to-clean geometry, or are they inherited from a previous equipment generation?
- Does the validation report contain the actual swab results, with a defined treatment of results below the limit of quantification?
- Are recovery determinations documented per residue, surface material and sampling technique, with any correction factor applied transparently?
- When a new product enters a shared equipment train, what triggers recalculation of the limit and revision of the cleaning matrix?
- If we received this observation, could we show systemic correction — procedure, rationale, training, QA approval — or only a single campaign protocol?
Wider context
Peptide GLP-1 APIs draw heightened attention on supply and documentation grounds alike, and the potency that makes them valuable is what drives low carryover limits on shared equipment. Reading a cleaning procedure against the study that validates it is not new FDA practice, but this letter states the mismatch explicitly.
Validation takeaway
A cleaning SOP that requires a carryover limit obliges the validation study to derive one. Where the limit, the rationale and the swab results are all absent, the site does not have a documentation lapse — it has no demonstrated cleaning process for that equipment.
Sources
- FDA Warning Letter 320-26-73, Harbin Jixianglong Biotech Co., Ltd., MARCS-CMS 723330, 1 May 2026 (primary source).
- Secondary reporting on the same letter, used only to corroborate facts already cited: BioSpace and NTD coverage of the Jixianglong warning letter.
Validation takeaway
A cleaning SOP that requires a carryover limit obliges the validation study to derive one; without the limit, the location rationale and the swab results, the cleaning process is not demonstrated for that equipment.
Questions worth asking in our own programme
- For every cleaning procedure that requires a numeric carryover limit, can we point to the derivation in the study — formula, PDE/ADE inputs, worst-case product, surface area and batch size?
- Do our swab locations have a written rationale tied to equipment drawings and hardest-to-clean geometry, or are they inherited from a previous equipment generation?
- Does the validation report contain the actual swab results, with a defined treatment of results below the limit of quantification?
- Are recovery determinations documented per residue, surface material and sampling technique, with a defined acceptable range and any correction factor applied transparently?
- When a new product enters a shared equipment train, what triggers recalculation of the limit, revision of the cleaning matrix and revalidation?
- If we received this observation, could we show systemic correction — procedure, rationale, training, QA approval and matrix update — or only a single campaign protocol?
