FDA Handling of Bacterial Endotoxin Test Method Issues in NDA and BLA Reviews

How FDA handles bacterial endotoxin test method issues in NDA and BLA reviews: low endotoxin recovery, recombinant factor C, and method suitability

Bacterial endotoxin testing (BET) is a routine release test, but FDA review records show it is also a recurring source of information requests, postmarketing commitments (PMCs), and occasionally Complete Response (CR) actions. Across Drugs@FDA review documents, three themes dominate: low endotoxin recovery (LER, also called endotoxin masking) in biologic formulations, the acceptability of recombinant factor C (rFC) and the monocyte activation test (MAT), and basic method-suitability gaps such as missing inhibition/enhancement data or an unstated routine test dilution.

Key takeaways

• FDA does not accept a LAL method affected by LER as sufficient for release on its own. The recurring pattern is to investigate or qualify a method that mitigates LER, use the rabbit pyrogen test (RPT) as an interim safeguard where needed, and allow MAT or other alternatives only with feasibility evidence, full validation, and often a fallback method ^{3033057370308309317}.
• LER is most often resolved through PMCs rather than CR letters. Examples include Entyvio, Mvasi, Takhzyro, Lumoxiti, Ultomiris, Ebanga, Xenpozyme, Wezlana, Imdelltra, Tremfya, Imaavy, and Pyzchiva ^{657871576263737470676459}.
• When endotoxin control is missing entirely, CRs do follow. Merilog (insulin aspart-szjj) received a CR in September 2023 because neither the LAL method nor the proposed RPT was shown to reliably detect endotoxin ⁴²⁰⁸. Lymphir (denileukin diftitox-cxdl) received a first-cycle CR citing the absence of an appropriate endotoxin assay ¹²²⁹⁴.
• The clearest rFC precedent in the review record is Lymphir, where FDA accepted an EndoLISA d-rFC method as validated for drug-product release in the resubmission, and the BLA was approved on August 7, 2024 ⁸³.
• Method-suitability deficiencies are usually procedural. FDA asks for missing inhibition/enhancement reports, three-lot data, MVD calculations, the routine commercial test dilution, or an endotoxin limit consistent with USP <85> at the maximum dose ^{137131138146149147}.

The guidance framework FDA reviewers apply

FDA's Pyrogen and Endotoxins Testing: Questions and Answers guidance was first issued in June 2012 and revised in March 2026 ²⁶¹²⁶². The current version sets out the following expectations:

• Alternative methods. Firms may use alternatives to USP procedures when they offer advantages in accuracy, sensitivity, precision, selectivity, or automation. Alternatives should be validated under USP <1225> and shown to give equivalent or better results. If results conflict, the USP gel-clot method is controlling unless the monograph states otherwise ¹⁰⁷.
• Switching between BET methods. When moving from one method to another that measures the same entity, equivalence should be shown by direct comparison. Limit of detection and inhibition/enhancement are the fundamental comparisons, using spiked product and, where available, positive field samples ¹⁰⁶.
• Interference. Interference should be mitigated by dilution up to the MVD or by other validated sample preparation. If that fails, the rabbit pyrogen test should be used ¹⁶¹.
• MVD versus routine dilution. The MVD should not be the routine test dilution. Sponsors should find the lowest dilution that neutralizes interference and screen just above it. FDA's example: with an MVD of 1:100 and inhibition at 1:10 but not 1:20, screening at 1:30 is supported ¹⁶².
• Pooling. For pooled small-volume parenterals, the MVD is divided by the number of units pooled, and FDA suggests no more than three units per composite ¹⁷⁰.
• Hold times and assayable endotoxin. Sampling plans should account for in-process hold times ¹⁰⁵. Storage and handling procedures should rest on data showing the stability of assayable endotoxin, and purified endotoxin may behave differently from native endotoxin ¹⁷⁰. The retrieved guidance text does not use the terms "masking" or "low endotoxin recovery" explicitly. These hold-time and assayable-endotoxin statements are the closest guidance-level hooks ¹⁷⁰¹⁶¹.

On rFC specifically, the 2012 Q&A stated that a manufacturer electing to use a recombinant factor C assay should validate it according to USP <85> requirements for photometric quantitative techniques and USP <1225> ¹⁰⁴. That paragraph is not in the March 2026 edition. USP <86> (Bacterial Endotoxins Test Using Recombinant Reagents) became official in May 2025, so rFC is no longer accurately described as non-compendial. The March 2026 Q&A does not discuss <86>.

For endotoxin limits, FDA's 2020 draft guidance on setting endotoxin limits for investigational oncology products restates the USP <85> K/M framework ¹¹⁵. K is 5 EU/kg/hour (or 100 EU/m²/hour) for non-intrathecal injections and 0.2 EU/kg/hour for intrathecal injections. These are upper limits, and stricter limits can be warranted ¹⁶⁵.

Low endotoxin recovery: what FDA asks for

Study design expectations

Across BLA reviews, FDA's LER information requests follow a consistent template:

• Undiluted product, maximum hold, representative containers. For Sylvant (siltuximab), FDA found that the initial proposal did not address LER over the maximum hold of formulated drug substance and drug product. FDA required spiking studies in undiluted material, over maximum hold times, in representative containers ¹⁸¹¹⁸⁴.
• Method match. For Keytruda (pembrolizumab), FDA rejected LER data generated by kinetic turbidimetric testing because the drug substance release method was kinetic chromogenic. FDA asked for a repeat study with the release method, concentrated CSE or RSE, and the maximum hold time including the interval between sampling and testing ¹⁹¹.
• Process-relevant temperature. For Ocrevus (ocrelizumab), FDA questioned a recovery study held at 2 to 8°C when manufacturing occurred at a different temperature, because temperature can affect masking kinetics ¹⁸⁷. For Jeuveau, FDA noted the same kind of limitation but did not require another study because the formulation lacked polysorbate and LER risk was low ¹⁸⁹.
• Appropriate controls. For Benlysta (belimumab), FDA requested a drug product LER study with spiked LAL Reagent Water as a control at at least two time points after the initial spike, such as 24 and 48 hours ¹⁹⁸.
• Multiple lots. For Zinbryta (daclizumab), one lot showed adequate recovery. FDA still considered two more lots necessary and handled this as a PMC because the risk assessment indicated low contamination risk and the specifications sat well below the safety threshold ¹⁹⁹.
• Repeat studies when the first fails. For Takhzyro (lanadelumab-flyo), the original LER study did not support the claimed hold time. After multiple information requests, a repeat study using kinetic turbidimetric LAL under USP <85> showed no LER in drug substance or drug product ¹⁸⁶.

A 50% recovery threshold appears repeatedly. The Ultomiris PMC required an alternate method if recovery fell below 50% ⁶². For Adakveo, FDA concluded that USP <85> BET could not detect endotoxin at release because spike recovery was below 50% ³¹⁷. For Retacrit, FDA required at least 50% recovery using gel-clot endpoint titration with time-zero controls ³²⁴.

How FDA resolves confirmed LER

When LER is confirmed, the review record shows several resolution pathways:

The Vimizim (elosulfase alfa) record adds a mechanistic point. FDA stated that the release method under-reported spiked endotoxin in undiluted product and that the spike appeared to become non-pyrogenic over time. FDA requested comparison with RPT, investigation of the LER mechanism, and evaluation of alternative methods ³²⁸.

Representative LER postmarketing commitments

For Pombiliti, FDA also referenced PDA Technical Report No. 82. FDA noted that an endotoxin-spiked-product rabbit study could show whether RPT is unnecessary as an interim release method ³³³⁴.

Recombinant factor C: limited but instructive precedent

The review record contains few product-specific rFC decisions. The clearest is Lymphir (denileukin diftitox-cxdl), BLA 761312:

• First cycle. FDA found no appropriate assay to measure endotoxin and inadequate controls for drug substance and drug product release testing. This was a principal product-quality basis for the CR ¹²²⁹⁴. The record describes repeated information requests about LER and states that no method had been successfully validated to detect endotoxin ²⁹³²⁹⁷.
• Resubmission. Citius introduced an EndoLISA d-rFC method. FDA found it acceptable and validated for drug product release testing ⁸³. FDA also found the new testing facility for the EndoLISA assay, which was non-compendial at the time of that 2024 review, acceptable after inspection ²⁹⁵.
• Residual commitment. The approval included a PMC to add endotoxin monitoring to the drug substance control strategy and submit validation data from three independent drug substance batches, with a final report due March 2026 ³⁰².

The retrieved Lymphir records do not describe the rFC validation parameters, spike-recovery data, or any head-to-head comparison with LAL ⁸³. Those records predate USP <86> (official May 2025). Sponsors should not treat the 2012 photometric and USP <1225> path as the only current standard. The March 2026 Q&A still expects a direct comparison of limit of detection and inhibition/enhancement when switching BET methods that measure the same entity ¹⁰⁶, and 21 CFR 610.9 remains the equivalence basis FDA has cited for biologics ³⁰³³⁰⁵. For biologics, 21 CFR 610.9 equivalence is the regulatory basis FDA cites when a sponsor replaces a required test ³⁰³³⁰⁵.

Method suitability: common deficiencies

Most method-suitability findings in NDA and BLA reviews are documentation gaps, not scientific failures:

• Missing suitability report. For Exondys 51 (eteplirsen), the suitability study was done but not submitted. Once provided, it showed three batches tested at four dilutions within the MVD, with acceptable spike recovery ¹³⁷. The Gallium Ga 68 gozetotide application initially lacked an inhibition/enhancement study ¹³⁹.
• Routine test dilution not stated. FDA issued requests for pantoprazole sodium, Bydureon BCise, and Akovaz because the dilution for routine commercial testing was not identified, even where suitability testing showed no interference ^131138146.
• Validation up to the MVD. For Fetroja (cefiderocol), FDA requested validation across dilutions up to the MVD and identification of the routine dilution. The applicant showed that both 1 mg/mL and the MVD concentration were non-interfering and committed to 1 mg/mL for routine testing ¹⁵¹.
• Three-lot data. Yervoy (ipilimumab) and docetaxel reviews requested MVD and inhibition/enhancement data from three lots ¹⁴⁷¹⁴⁹.
• Limits exceeding USP <85>. For docetaxel, FDA found the proposed specification exceeded the USP <85> recommendation at the maximum dose and required a revised limit and revised suitability testing ¹⁴⁹.
• Legacy information requests. Accretropin, Prolensa, and Kalbitor reviews show FDA asking for the assay type, limit derivation, MVD calculation, lysate sensitivity, pooling practices, and inhibition/enhancement results ^140150152.

Where dilution alone resolves interference, FDA accepts it. For Acuvail (ketorolac ophthalmic solution), kinetic turbidimetric LAL at a validated dilution met acceptance criteria, and the review listed no microbiology deficiencies ⁴⁶.

When endotoxin issues reach a Complete Response

The review records identify the following CR actions and major deficiencies tied to endotoxin testing:

In the retrieved records, no CR or major deficiency centered on rFC itself ¹²⁸³. Confirmed LER in a product with an otherwise workable control strategy, such as an interim RPT, was handled through PMCs. Adakveo is an example: FDA found it approvable because endotoxin-spiked product was pyrogenic in rabbits, which supported RPT as the interim release test ³¹⁷.

Practical implications for sponsors

1. Run LER studies early, and run them to FDA's template. That means undiluted product, maximum process hold, representative containers, process-relevant temperature, the actual release method, RSE or CSE spikes, and appropriate controls at multiple time points ^181191187198. FDA has asked for repeat studies when any of these elements was missing ¹⁸⁶¹⁹¹.
2. Have an LER mitigation path ready. Options seen in approvals include gel-clot, demasking buffers, MAT, rFC, and interim RPT supported by spiked-product rabbit data ^16577483317.
3. Do not rely on RPT without spiking data. Merilog shows that RPT can fail as a fallback when product pharmacology may confound the rabbit response and spiking studies are missing ²⁰⁸.
4. Treat rFC status as time-dependent. Lymphir (approved August 7, 2024) predates USP <86>, which became official in May 2025. The March 2026 Q&A still does not discuss <86>. For a pre-<86> switch, FDA's stated expectation was USP <85> photometric requirements, USP <1225>, a comparison of limit of detection and inhibition/enhancement, and 21 CFR 610.9 equivalence for biologics ^104106303. Post-May 2025 filings should address <86> with the review division rather than rely only on the 2012 alternative-method path.
5. Close out suitability basics in Module 3. Submit the full inhibition/enhancement report with three lots, show the MVD calculation, state the routine test dilution (which should not be the MVD), and justify the limit against K/M at the maximum hourly dose ^137147162115.

Open questions that merit product-specific follow-up include how FDA has evaluated rFC comparability data beyond Lymphir, which formulation components were implicated in redacted LER cases, and whether completed LER PMCs led to method changes through supplements.