Bioburden and Bacterial Endotoxin Testing: USP <61>/<62> and <85> LAL Methods

Two microbiological numbers gate the release of almost every parenteral product: how many viable organisms are present (bioburden) and how much bacterial endotoxin is present. Neither test proves sterility on its own. Together with the sterility test and the contamination control strategy, they form the evidence a Qualified Person or release-responsible analyst relies on to say a batch is safe to inject. This page covers the compendial chapters behind both tests, the actual procedures, the math that trips people up (endotoxin limits, maximum valid dilution), and the inspection findings that recur in microbiology QC labs.

If you can explain why a positive product control is your single most important LAL control, why bioburden timing matters more than the count itself, and how you would defend an endotoxin limit calculation to an inspector, you understand this topic at the level a release decision demands.

The two questions and why they are separate

Bioburden and endotoxin answer different questions about different hazards.

Bioburden is the count of viable, culturable microorganisms in or on a material. It tells you the microbial load entering a process step. For a terminally sterilized product, pre-sterilization bioburden defines the challenge the sterilizing step has to overcome and feeds the sterility assurance calculation. For an aseptically filled product, bioburden of the bulk before filtration defines the load on the sterilizing-grade filter. Bioburden is a living-organism count.

Bacterial endotoxin is a measure of lipopolysaccharide (LPS), a structural component of the outer membrane of Gram-negative bacteria. Endotoxin is a pyrogen: injected into the bloodstream it causes fever, and at higher exposure it drives hypotension, disseminated intravascular coagulation, and septic shock. Endotoxin is heat-stable and survives the conditions that kill the organism. A product can be sterile and still fail for endotoxin, because the dead Gram-negative cells left their LPS behind. Endotoxin is a chemical-structure measurement, not a viable count.

Keep the distinction sharp in interviews. Killing organisms removes bioburden but does not remove endotoxin. Removing endotoxin requires depyrogenation (dry heat) or removal/dilution, not sterilization.

Regulatory and compendial basis

The governing documents are compendial chapters and the cGMP regulation that makes them enforceable.

The cGMP hook is 21 CFR 211.167 (special testing requirements), which requires that each batch of a product purporting to be sterile and pyrogen-free be tested by appropriate laboratory methods, and 21 CFR 211.165 (testing and release for distribution). USP <61>, <62>, and <85> are harmonized across USP, Ph. Eur., and JP under the ICH Q4B process, so a method validated to one is generally acceptable to the others, with documented confirmation.

21 CFR 211.167(a): “For each batch of drug product purporting to be sterile and/or pyrogen-free, there shall be appropriate laboratory testing to determine conformance to such requirements.”

A useful interview line: bioburden and endotoxin testing are not optional add-ons; they are the laboratory evidence 211.167 demands for the “pyrogen-free” claim, and for many products the bioburden data is also the input that justifies the sterilization cycle.

Bioburden: USP <61> microbial enumeration

What the test produces

USP <61> gives you two numbers per sample:

• TAMC (Total Aerobic Microbial Count): viable aerobic bacteria, reported in CFU per gram, per milliliter, or per device/unit, recovered on a general soybean-casein digest agar (TSA, also called SCDA).
• TYMC (Total combined Yeasts and Molds Count): viable fungi, recovered on Sabouraud dextrose agar (SDA).

How to run it, step by step

1. Choose the recovery method. Three are described:

   • Membrane filtration for filterable liquids. Preferred when the product is antimicrobial, because you filter the organisms onto a membrane and rinse the inhibitory product away before placing the membrane on agar.
   • Pour plate or spread plate (plate-count method) for non-inhibitory samples.
   • Most Probable Number (MPN) for low counts in difficult matrices; least precise, used only when the other two will not work.

2. Prepare the sample. Dissolve, dilute, or suspend in a validated diluent (commonly pH 7.0 phosphate buffer or peptone saline). Neutralize antimicrobial activity if present, using validated neutralizers (lecithin and polysorbate are typical) or dilution.

3. Plate and incubate (USP <61> conditions).

   • For TAMC, hold SCDA plates at 30 to 35 C, reading at 3 to 5 days.
   • For TYMC, hold SDA plates at 20 to 25 C, reading at 5 to 7 days.

4. Count and calculate. Count colonies on plates yielding a countable range (for the plate method, typically not more than 250 colonies for TAMC and not more than 50 for TYMC, on the most appropriate dilution). Multiply by the dilution factor to get CFU per gram or per mL.

5. Report TAMC and TYMC against the <1111> acceptance criteria for the route of administration.

Method suitability (the step people skip and get cited for)

Before the method counts for anything, you must show it can actually recover organisms from your specific product. This is method suitability testing (sometimes called the growth-promotion / recovery validation), grounded in USP <61> itself and USP <1227>.

You inoculate the sample preparation with a small dose of each challenge organism, no more than about 100 CFU, and show that the count you recover stays within a factor of two of a parallel inoculum control run without product. The standard panel:

If recovery is below the factor-of-2 criterion, the product is inhibitory and you must modify the method (increase neutralizer, increase dilution, or switch to membrane filtration with rinsing) and re-validate. Method suitability is product-specific and must be repeated when the formulation or manufacturing process changes materially.

Worked example, bioburden calculation

A 10 g sample of a topical cream is suspended in diluent to 100 mL (1:10 dilution). 1 mL is pour-plated on SCDA. After incubation, the plate shows 48 colonies.

CFU/g = colonies x dilution factor / sample volume plated = 48 x 10 / 1 = 480 CFU/g TAMC.

Under USP <1111>, a stated limit of 10^n is read as allowing a count up to 2 x 10^n before it is exceeded. In practice that puts the ceilings at 20 CFU/g for a 10^1 limit, 200 CFU/g for 10^2, and 2000 CFU/g for 10^3. So if <1111> sets TAMC NMT 10^2 CFU/g for a topical preparation, a result of 200 CFU/g still passes, 250 CFU/g fails the 10^2 limit, and a count of 480 CFU/g would only meet a 10^3 limit.

Acceptance criteria, USP <1111>

USP <1111> does not set one count limit for all nonsterile products. It assigns acceptance limits by dosage form and route of administration, and the logic behind the assignment is what you need to carry into a release decision or an interview. Two principles drive it:

• The limits tighten as the route gets riskier. A non-aqueous oral solid, which is a dry, low-water-activity environment that resists microbial growth and is taken by the most tolerant route, carries the most permissive TAMC and TYMC limits. As you move to aqueous oral, then topical and transdermal, then routes that approach mucosal or parenteral-adjacent exposure (nasal, inhalation, vaginal, urethral), the allowed counts drop, because water supports growth and the route gives organisms a more direct path into the patient. Inhalation and mucosal routes are held to the strictest nonsterile limits for this reason.
• The specified-organism requirement is route-driven, not count-driven. Which named organisms must be absent depends on where the product goes, not on how high the total count is. Oral products are screened for the enteric indicators because the hazard is fecal-origin pathogens; skin-contact and water-exposed routes are screened for the skin and water opportunists; mucosal routes add the yeast that colonizes those surfaces. The route picks the organism list, which is then tested by the methods in USP <62>.

Worked example for teaching, one route only. Suppose your product is a non-aqueous oral preparation, a dry capsule. Under the <1111> framework, this is the most permissive category, so it carries the highest allowable TAMC and TYMC of the nonsterile forms, and the specified-organism screen for an oral product targets the enteric indicator (the test is for absence of E. coli in the stated quantity). Read the current TAMC and TYMC figures for the “non-aqueous preparations for oral use” row directly from USP <1111> and apply the 2 x 10^n reading rule from the worked calculation above: a stated limit of 10^n is interpreted as a maximum acceptable count of 2 x 10^n. So if you read a TAMC limit of 10^3 CFU/g for this row, the maximum acceptable count is 2000 CFU/g, meaning a result of 1500 CFU/g passes and 2000 CFU/g is still within the limit, while 2500 CFU/g fails. Do the same lookup for your own dosage form and route, because the numbers and the organism list change with the category, and the compendial chapter is the only authoritative source for the current values.

Look up the current acceptance limits and the route-specific organism requirements for your actual product in USP <1111>, or in Ph. Eur. 2.6.12 for the European harmonized equivalent. Do not carry a remembered number into a release decision; the category that applies to your formulation, and the figure attached to it, must come from the current chapter.

Sterile products are not tested to <1111>; they are tested by USP <71> sterility and must show no recovery. <1111> applies to nonsterile products and to in-process bioburden where the firm sets its own action/alert limits based on process capability, not <1111>.

In-process bioburden limits: where the numbers actually come from

USP <1111> answers the question “what may a finished nonsterile product contain.” It does not answer the question most manufacturing sites actually face, which is “what may the bulk contain at this step.” In-process bioburden limits come from a different place, and the distinction is worth getting right because applying <1111> to an in-process control is a category error.

The sterile filtration control point

For an aseptically filled product, the single most consequential bioburden number is the load presented to the sterilising-grade filter. The EMA guideline on the sterilisation of the medicinal product, active substance, excipient and primary container (EMA/CHMP/CVMP/QWP/850374/2015, which came into effect on 1 October 2019) sets out the expectations that most firms build against:

• Bioburden testing should be performed on the bulk solution immediately before sterile filtration for routine commercial manufacturing.
• A limit of not more than 10 CFU per 100 mL (TAMC) is described as acceptable in most situations.
• The sample should be 100 mL, so the method has the sensitivity to detect at that limit. Testing 1 mL and multiplying is not the same test.
• Where a pre-filter is present only as a precaution rather than because the unfiltered bulk carries a higher load, the same limit applies before the pre-filter as well.
• A higher limit before pre-filtration may be acceptable where a starting material is known to carry inherent microbial contamination, provided it is demonstrated that the first filter brings the load to not more than 10 CFU per 100 mL before the final filtration.

The same guideline addresses pre-sterilisation bioburden at a different control point, the load presented to a terminal sterilisation step rather than to a sterilising filter: a maximum of 100 CFU per 100 g or 100 CFU per 100 mL is described as acceptable without further justification for parenteral finished product formulations, and 10 CFU per gram or per mL for active substances and finished products not used parenterally. Keep the two apart when you quote them, because the 10 CFU per 100 mL figure belongs to the sterile filtration section and the 100 CFU figures belong to the terminal sterilisation discussion, and pairing the wrong number with the wrong control point is an easy way to set a limit that does not match your process. Read the current text for the operative wording and confirm which section applies to your product and process.

Two things follow that people frequently get wrong. First, the 100 mL sample volume is part of the limit, not an incidental detail. Second, the limit is a specification at a control point, not a target to trend toward. A process routinely delivering 8 CFU per 100 mL is not comfortably passing, it is telling you something about upstream control.

Alert and action limits are yours to set

Beyond the compendial and guideline limits, in-process bioburden control runs on alert and action limits that the firm derives from its own process capability. The method:

1. Collect a baseline of results from a period when the process was in control, using the same method and sample size you will use going forward.
2. Derive limits statistically from that distribution rather than choosing round numbers. Microbial count data is skewed and frequently contains many zeros, so ordinary normal-distribution limits fit poorly. Non-parametric approaches, such as setting the alert at a high percentile of the historical distribution, generally behave better.
3. Set the alert limit as an early warning that the process is drifting, and the action limit as the point requiring a defined response. Neither may exceed the regulatory or specification limit that applies at that point.
4. Define the response to each before you need it: what is investigated, who is notified, what happens to the batch, and when the excursion becomes a deviation.
5. Review the limits periodically against accumulated data, and re-derive when the process changes. Document the basis every time, because a limit that moved with no recorded rationale is indistinguishable from a limit that moved to stop an excursion being reported.

The last point deserves emphasis. Recalculating a limit so that a signal disappears converts monitoring into a mechanism for not noticing things. A defensible test: the reason for changing a limit should be one that would have applied whether or not a signal was sitting in the data.

Growth promotion and reagent qualification, the controls under the controls

A count is only as trustworthy as the media and reagents that produced it. Two qualification activities sit underneath every result in this article:

• Growth promotion testing of media. Each lot of prepared or purchased media is challenged with the relevant panel of organisms and shown to support recovery before, or contemporaneously with, its use in testing. Using a media lot whose growth-promoting ability was never confirmed leaves every negative result on that lot unsupported, and this is a recurring citation.
• Lysate and standard qualification for BET. Confirm the labelled lysate sensitivity on receipt of each lot, maintain traceability from the control standard endotoxin back to the reference standard endotoxin, and treat LAL reagent water as a critical reagent with its own acceptance and monitoring. Track negative-control behaviour over time rather than only pass or fail, because a drift in negative controls is an early sign of a reagent, water, or labware problem.

Neither activity is glamorous, and both are where an inspector goes when the results look too clean.

USP <62>: tests for specified (objectionable) microorganisms

What and why

Counting total organisms is not enough. A modest total count can still contain an organism that is dangerous for the product’s route of administration. USP <62> provides qualitative tests for named organisms, and the route-specific list in <1111> tells you which ones apply.

The classic specified organisms:

• Escherichia coli (fecal contamination indicator, oral products)
• Salmonella species (oral products, especially of natural origin)
• Pseudomonas aeruginosa (water-borne, opportunistic, topical and inhalation)
• Staphylococcus aureus (topical, skin contact)
• Candida albicans and Clostridia for specific routes
• Bile-tolerant Gram-negative bacteria, which <62> handles as a separate quantitative (enumeration) category with its own counting limits rather than as one of the absence-based named organisms above, used as a broader hygiene indicator

How it works

Each test is an enrichment followed by selective/differential plating:

1. Pre-incubate the sample in a non-selective enrichment broth (soybean-casein digest broth) to resuscitate stressed organisms.
2. Transfer to a selective enrichment broth (for example, enrichment for Salmonella uses Rappaport Vassiliadis broth).
3. Streak onto selective/differential agar (MacConkey agar for E. coli and bile-tolerant Gram-negatives, cetrimide agar for P. aeruginosa, mannitol salt or Baird-Parker for S. aureus, xylose lysine deoxycholate for Salmonella).
4. Confirm presumptive colonies by identification.

Result is reported as absence in the specified quantity (typically 1 g or 1 mL, or 10 g for Salmonella). USP <62> also requires its own suitability test: you spike the product with the target organism and show the method detects it in the presence of product. Negative suitability means the product suppresses growth and the absence result is meaningless.

Beyond the named list: objectionable organisms

A subtle point inspectors probe: USP <62> names organisms, but 21 CFR 211.113 and the concept of “objectionable organisms” go further. An organism not on the <62> list can still be objectionable if it is pathogenic for the route, can proliferate in the product, or compromises the dosage form. Burkholderia cepacia complex in aqueous, non-sterile products is the textbook example and has driven multiple recalls. The firm’s microbiology program must define objectionable organisms by risk assessment, identify recovered isolates to a sufficient taxonomic level, and assess each for objectionability. “It met the <1111> count limit” is not a complete answer if a recovered organism is objectionable for the route.

Bacterial Endotoxin Test (BET): USP <85>

What endotoxin testing replaced and why

Historically pyrogenicity was assessed by the rabbit pyrogen test (USP <151>): inject the product into rabbits, measure fever. The Limulus Amebocyte Lysate (LAL) test, codified in USP <85>, replaced the rabbit test for most products because it is more sensitive, quantitative, faster, cheaper, and avoids animals. LAL detects bacterial endotoxin specifically; the rabbit test detects pyrogens broadly. For most parenterals endotoxin is the pyrogen of concern, so BET is the standard release test, with the rabbit test reserved for products incompatible with LAL or where a non-endotoxin pyrogen is suspected.

LAL is derived from the blood cells (amebocytes) of the horseshoe crab. Endotoxin triggers an enzyme cascade in the lysate that ends in a measurable clot or color change. Modern recombinant Factor C (rFC) reagents reproduce the first enzyme of that cascade without harvesting crabs, and are recognized in USP <1085>, in Ph. Eur. 2.6.32 (whose content is moving into 2.6.14 as Method G, described later in this article), and in a dedicated USP chapter for recombinant reagents.

Units and key terms

• EU = Endotoxin Unit, the activity unit. Defined against the reference standard endotoxin (RSE) and traceable through control standard endotoxin (CSE).
• lambda (the reagent label claim / lysate sensitivity), in EU/mL: the lowest endotoxin concentration that gives a positive result for gel-clot, or the lowest point on the standard curve for photometric methods.
• MVD = Maximum Valid Dilution, as defined in USP <85>: the most you can dilute the sample and still be capable of reading the endotoxin limit.
• MVC = Maximum Valid Concentration: the corresponding least dilution / highest product concentration testable.

The three methods

Gel-clot is the referee method: if a quantitative method result is disputed, USP <85> says the gel-clot result governs. Recombinant Factor C assays are typically read fluorometrically and are quantitative.

Endotoxin limit calculation

The endotoxin limit is product-specific and route-specific. The formula:

Endotoxin limit = K / M

where (per USP <85>):

• K = the threshold pyrogenic dose, expressed as endotoxin per kg of body weight per hour. The conventional values are 5 EU/kg for parenteral products (intravenous, intramuscular, subcutaneous) and 0.2 EU/kg for intrathecal products. For a combination product with a device path, the device limit is set differently: under USP <161> it is 20 EU/device (0.5 EU/mL of rinse or extract) for general and cardiovascular contact, and 2.15 EU/device (0.06 EU/mL) for a device contacting cerebrospinal fluid, on a per-device rinse basis rather than as a value of K. (Note: 350 EU is sometimes quoted, but that is the maximum endotoxin a person may receive per hour for a drug, 5 EU/kg x 70 kg, not a per-device limit.)
• M = the maximum recommended human dose per kg of body weight per hour, with body weight taken as 70 kg unless the product is intended for a defined smaller population.

Worked example, a small-volume injectable:

A drug is dosed at a maximum of 5 mg/kg/hour and the product strength is 10 mg/mL.

1. M (dose) = 5 mg/kg.
2. Endotoxin limit per mg = K / dose = 5 EU/kg / 5 mg/kg = 1.0 EU/mg.
3. Convert to per mL using strength: 1.0 EU/mg x 10 mg/mL = 10 EU/mL is the endotoxin limit for the product.

For large-volume parenterals (LVPs) administered by volume, the limit is often expressed directly: the compendial limit for LVPs is 0.5 EU/mL, and Water for Injection per USP <1231> is 0.25 EU/mL.

Maximum Valid Dilution (MVD)

You cannot test product neat if it interferes with the reagent, but you also cannot dilute past the point where you would miss the limit. MVD sets the ceiling, and USP <85> derives it from the endotoxin limit, the product concentration, and the reagent sensitivity:

MVD = (Endotoxin limit x Concentration of product) / lambda

Worked example continuing from above:

• Endotoxin limit = 10 EU/mL (or 1.0 EU/mg expressed per mg).
• Using the per-mg form: endotoxin limit = 1.0 EU/mg, product concentration = 10 mg/mL.
• lambda of the gel-clot lysate = 0.125 EU/mL.

MVD = (1.0 EU/mg x 10 mg/mL) / 0.125 EU/mL = 10 / 0.125 = 80.

So you may dilute the product up to 1:80 and still be able to detect the endotoxin limit. Any dilution greater than 1:80 is invalid because endotoxin at the limit would fall below the lysate’s detection. You typically test at a dilution at or below MVD that also overcomes interference, often the smallest dilution that passes the inhibition/enhancement test.

Test validation: inhibition / enhancement (the heart of BET)

Before any BET result is meaningful, you must prove the product itself does not inhibit or enhance the LAL reaction. This is the interference test (also positive product control logic), required by USP <85>.

Two controls dominate:

• Positive Product Control (PPC): product spiked with a known amount of endotoxin (commonly 2 lambda for gel-clot, or a mid-curve spike for photometric). For photometric methods, USP <85> expects the spike to be recovered in the range of 50 to 200 percent; recovery outside that range means the product is interfering and the result is invalid.
• Standard curve / standard series: for photometric methods, USP <85> calls for a curve of at least three concentrations whose correlation coefficient |r| is at least 0.980. For gel-clot, you confirm the labeled lysate sensitivity with a standard series, and the geometric mean endpoint should land between 0.5 lambda and 2 lambda.

If interference is found, you remove it by dilution (up to MVD), by neutralization, by pH adjustment to 6 to 8, by using endotoxin-specific buffer, or by treating divalent cation effects. You then re-validate.

Gel-clot procedure, step by step

1. Reconstitute lysate of known lambda. Prepare LAL reagent water (LRW), endotoxin-free.
2. Confirm lysate sensitivity with a replicate standard series (lambda, 2 lambda, 0.5 lambda, 0.25 lambda); the geometric mean of the endpoints should fall between 0.5 lambda and 2 lambda.
3. Prepare product at the chosen dilution (at or below MVD).
4. Set up four tube types: negative control (LRW), positive control (endotoxin at 2 lambda in LRW), product sample, positive product control (product spiked at 2 lambda).
5. Hold the tubes undisturbed at 37 +/- 1 C for 60 +/- 2 minutes.
6. Turn each tube over through a full 180 degrees in a single steady movement. If a solid gel stays in place, the result is positive; if there is no gel or the gel collapses, it is negative.
7. Interpret:
   • Negative control must be negative.
   • Positive control must be positive.
   • Positive product control must be positive (else product inhibits, test invalid).
   • Sample negative at a dilution within MVD = product passes (endotoxin below limit).

Kinetic chromogenic / turbidimetric procedure

1. Build a standard curve across the working range (for example 0.005 to 50 EU/mL), |r| >= 0.980.
2. Run negative control (must read below the lowest standard), each sample dilution in replicate, and a PPC for each sample.
3. The instrument reads onset time (kinetic) and back-calculates EU/mL against the curve.
4. PPC recovery 50 to 200 percent confirms no interference.
5. Compare measured EU/mL (corrected for dilution) against the endotoxin limit.

Worked endotoxin pass/fail example

Measured result for the injectable above, tested at 1:20, kinetic chromogenic: 0.18 EU/mL on the assay.

Corrected for dilution: 0.18 x 20 = 3.6 EU/mL in the undiluted product.

Limit = 10 EU/mL. 3.6 < 10, so the batch passes for endotoxin. PPC recovery was 92 percent (within 50 to 200), so the result is valid.

Low Endotoxin Recovery (LER): the failure mode that hides endotoxin

Everything above assumes that if endotoxin is present, the assay can see it. For a large class of biologic formulations that assumption is wrong, and the phenomenon has a name.

What LER is

Low Endotoxin Recovery, also called endotoxin masking, is the progressive loss of detectable endotoxin activity when a known amount of endotoxin sits in an undiluted product over time. Spike the product on day zero and you recover the spike. Spike it and hold it, and recovery falls, sometimes to near zero, while the endotoxin is still physically present. The assay reports a clean result for a contaminated sample.

The mechanism is not completely settled, but the working understanding is that the formulation disassembles the supramolecular aggregates that LPS normally forms in solution. Endotoxin activity in the LAL cascade depends on that aggregation state, so LPS dispersed into smaller units becomes invisible to the assay without becoming any less real biologically.

Which formulations are at risk

The classic at-risk combination is a surfactant plus a chelator: a polysorbate (commonly polysorbate 20 or 80) together with a chelating buffer species such as citrate, phosphate, EDTA, or histidine. That description covers a very large share of monoclonal antibody and recombinant protein formulations, which is exactly why LER became a regulatory focus for biologics rather than for small molecules.

The important consequence: LER is a property of the formulation, not of the contamination. A product can be perfectly well controlled and still be LER-susceptible, and the susceptibility has to be characterised rather than assumed away.

The hold-time study

The test for LER is a spike-recovery study run over time, not a single-point interference check. The design that regulators expect:

1. Spike undiluted product. This is the step people get wrong. The standard interference test (positive product control) is run at the tested dilution, and dilution is precisely what breaks the masking effect. A product that passes the routine PPC can still be strongly LER-positive, because the masking never had a chance to act at that dilution.
2. Use a known endotoxin spike and monitor recovery across a series of time points rather than at a single end point.
3. Evaluate enough time points to see the trajectory. EMA’s position, set out in its questions and answers for biological medicinal products, is that at least four time points should be evaluated for the results to support a valid conclusion.
4. Match the study conditions to the process. Temperatures and hold durations should reflect what actually happens in manufacturing, including realistic hold times, so the study answers a question about the product as made rather than about an artificial storage condition.
5. Apply a defined criterion. EMA treats two consecutive time points falling below 50 percent recovery, the lower compendial recovery limit, as demonstrating LER.

What to do when LER is found

Finding LER is not a product failure, it is a method problem that has to be solved before the method can be trusted:

• Optimise the compendial method or develop an alternative that recovers endotoxin in that matrix. Common approaches include demasking treatments, dispersing agents, alternative diluents, and reagent systems that behave differently in the matrix.
• Set the finished product endotoxin specification as low as reasonably achievable on the basis of actual manufacturing data. EMA recommends this specifically for LER-exhibiting products, and the logic is straightforward: if the assay’s sensitivity in that matrix is partly compromised, a tighter specification rebuilds some of the safety margin that masking took away.
• Place the data where reviewers expect it. EMA points to CTD section 3.2.P.5.3 (validation of analytical procedures) or 3.2.P.2.3 (manufacturing process development) with a cross-reference, and expects the data in Module 3 for new marketing authorisation applications.

Two distinctions worth getting right

Finished product versus active substance. LER investigation focuses on the finished product as administered to the patient. Where the active substance matrix is similar to the finished product matrix, a separate active substance study may not add anything; where the matrices differ, a risk assessment should establish whether the finished product study is representative, rather than assuming it is.

LER hold time is not QC sample hold time. These are different studies answering different questions and they are routinely confused. A QC sample hold time is the maximum allowable interval between sampling and testing under normal laboratory handling, where the sample may well be diluted or frozen first. An LER study deliberately holds the product undiluted and unfrozen, because dilution and freezing are among the things that disrupt masking. Establishing one does not establish the other.

For study design, PDA Technical Report No. 82 addresses LER specifically and is recognised by EMA as a relevant reference for designing these studies. It focuses on protein-based products, and in the absence of anything more specific it can inform study design for vaccines and gene therapy products too. Consult the technical report itself for its methodology rather than working from summaries of it.

Currency note: EMA maintains the LER and endotoxin masking expectations described above in its questions and answers for biological medicinal products, which is a living page rather than a versioned guideline. The LER entries have been in place since an earlier revision; the page was most recently updated on 20 July 2026, that update adding entries including pre-use filter integrity testing, viral filtration, and the endotoxin LAL test. Read the current published version for the operative wording before relying on any of it, because a Q&A page of this kind is revised in place without the version history a guideline carries, and an entry can change without any announcement.

The interview version

If asked about LER, the answer that shows you understand it: LER is a formulation-driven loss of detectable endotoxin over time, typical of surfactant plus chelator biologic formulations, and the reason it matters is that a routine positive product control run at the working dilution can pass while the product is strongly masking at full strength. You detect it with an undiluted spike-recovery study across several time points under process-relevant conditions, treat two consecutive points below 50 percent recovery as LER, and respond by fixing the method and tightening the specification, not by explaining the result away.

Recombinant Factor C and the animal-reduction shift

rFC reagents reproduce the LAL cascade’s first enzyme without harvesting horseshoe crab blood. They are endotoxin-specific (they do not react to beta-glucans, which can cause false positives in LAL), reduce animal use, and are recognized in USP <1085>, in Ph. Eur. 2.6.32 (now being consolidated into 2.6.14, see below), and in the dedicated compendial chapter named below.

USP <86>, the chapter that changed the status of recombinant reagents

The relevant USP chapter is USP <86>, Bacterial Endotoxins Test Using Recombinant Reagents, which covers both recombinant cascade reagents and recombinant Factor C. USP published the final text for early adoption on 1 November 2024 and the chapter became official on 1 May 2025.

The practical significance is the change in regulatory posture, and it is worth stating precisely because it is often overstated. Before a dedicated chapter existed, a firm using recombinant reagents was using a non-compendial alternative method and carried the full burden of justifying that choice as an alternative to the compendial test. With <86> in place, recombinant reagent methods are described in a compendial chapter of their own, which removes that framing. It does not remove the validation work. A firm adopting recombinant reagents still performs product-specific method suitability, still demonstrates that the method performs in its own matrices, still routes the change through change control, and should still hold a comparability rationale against the method being replaced. What changed is the starting position of the conversation, not the evidence required.

Note also that <86> sits alongside <85> rather than replacing it. Choosing between them is a documented decision, not a default.

Europe is consolidating rather than adding a chapter

The European Pharmacopoeia took a different structural route to the same destination, and anyone writing procedures against Ph. Eur. references should plan for it now. Rather than keeping a separate recombinant chapter, EDQM has revised Ph. Eur. 2.6.14, Bacterial endotoxins, to incorporate the fluorimetric end-point method using recombinant Factor C as a new Method G within that chapter, and the standalone chapter 2.6.32 is being suppressed, with its technical content transferred into 2.6.14. The suppression takes effect with Issue 13.1, and the implementation date for the consolidated arrangement is 1 January 2027.

The practical consequence is small but easy to miss: any SOP, specification, method, or filing that cites 2.6.32 as the controlling chapter for a recombinant Factor C method will be citing a suppressed chapter from 2027, and will need to point at 2.6.14 Method G instead. That is a document-control exercise across potentially many controlled documents, so it is worth scoping while there is time rather than discovering it during an inspection. Note the asymmetry with the US: USP created a separate chapter, <86>, while Ph. Eur. is folding the same technology into its existing endotoxin chapter. Both recognise recombinant reagents; they simply organise the text differently, and a global procedure has to cite both correctly.

Expect interviewers to ask whether your laboratory has evaluated recombinant reagents. A strong answer covers specificity (no beta-glucan interference, which removes a real source of false positives), the animal-sourcing and supply-continuity argument, the current compendial status under USP <86> and, in Europe, the move of the rFC method into Ph. Eur. 2.6.14 as Method G with 2.6.32 suppressed, and the validation and change-control work that still has to be done regardless of compendial status.

Choosing a pyrogen testing strategy, and the end of the rabbit test in Europe

Endotoxin is the pyrogen that matters for most parenteral products, but it is not the only one. Gram-positive cell wall components, certain viral and fungal materials, and some process-related and material-related contaminants are pyrogenic without being endotoxin. The BET cannot see any of them. Deciding which test answers your question is a documented risk decision, and the European framework for making it changed recently.

What changed in Europe

The European Pharmacopoeia has removed the rabbit pyrogen test. The mechanics:

• A new general chapter, Ph. Eur. 5.1.13, Pyrogenicity, entered into force on 1 July 2025, published in Supplement 11.8 together with revisions removing the rabbit pyrogen test from a large number of texts and monographs.
• The dedicated rabbit test chapter, Ph. Eur. 2.6.8, Pyrogens, is suppressed, with implementation from 1 January 2026.
• Rather than naming one replacement test, 5.1.13 places the burden on the user to select a pyrogenicity testing strategy that is justified by a risk assessment of the specific substance or product.
• The available methods are the bacterial endotoxins test (Ph. Eur. 2.6.14, which now carries the recombinant Factor C method as Method G, with the standalone chapter 2.6.32 being suppressed) where endotoxin is the pyrogen of concern, and the monocyte-activation test, Ph. Eur. 2.6.30, where non-endotoxin pyrogens cannot be excluded.

This is a genuine change in obligation, not a preference. A European product whose control strategy still rests on the rabbit test needs a different strategy and a justified one.

The US position, which has moved the other way

Do not read the European change as a global deletion of the rabbit test. The US framework retains it, and FDA restated that position recently.

FDA issued Pyrogen and Endotoxins Testing: Questions and Answers (Edition 2) in March 2026 as final guidance from CDER, CBER, CVM, CDRH, and ORA. Version 1 of that guidance was published in 2012, and FDA has withdrawn its 1987 guidance on validating the LAL test as an end-product endotoxin test, on the basis that it no longer reflects current thinking. The Edition 2 document explains where FDA’s expectations go beyond what the compendial chapters cover.

Several points in it bear directly on test selection:

• A rabbit pyrogen test is still required by regulation for certain biological products. 21 CFR 610.13(b) carries that requirement, and 21 CFR 610.9 provides the route to waive it where a method equivalent to the rabbit pyrogen test is demonstrated. For biologics this is a regulatory requirement with a defined waiver path, not merely a compendial preference.
• Some USP monographs still call for the rabbit test. FDA’s stated position is that a firm may substitute an endotoxins test or an alternative cell-based test where it can demonstrate equivalent pyrogen detection, with the relevant review division considering alternatives such as monocyte activation case by case.
• Alternative methods are permitted with conditions. FDA points to the USP General Notices provision on alternative and harmonised methods, expects validation along the lines of USP <1225>, Validation of Compendial Procedures, and expects the alternative to show equivalent or better results. Where methods conflict, the gel-clot result governs unless the product monograph says otherwise, which is the same referee logic described earlier in this article.
• Two situations where FDA points back toward the rabbit test. Where a risk assessment indicates non-endotoxin pyrogens may be present, and where assay interference cannot be resolved by dilution up to the MVD or by other validated sample preparation, the guidance indicates the rabbit pyrogen test may be the more appropriate choice.

Note also what Edition 2 does not do: it anchors on USP <85> and does not adopt the recombinant reagent chapter <86> as an equivalent starting point. A firm moving to recombinant reagents in a US-registered product is still making a method change that has to be justified and, where it affects a registered method, filed. The compendial status of <86> and the regulatory acceptance of a specific change in a specific application are different questions.

The divergence is the point

Read the two paragraphs above together and the practical conclusion is uncomfortable but important: as of 2026 the European and US frameworks have moved in opposite directions on the rabbit pyrogen test. Europe has removed it and pushed users toward a justified strategy built on BET and MAT. The US retains it, requires it by regulation for certain biological products, and names it as a fallback for precisely the cases where BET struggles.

For a product registered in both regions, that means the pyrogen strategy is one of the places where a single global approach may not be available, the registered method in each market governs, and a change in one market is a variation in that market rather than a laboratory improvement applied everywhere. Confirm the current requirement for each market and each product rather than generalising from either framework.

USP <151>, Pyrogen Test, remains an official USP chapter throughout.

The monocyte activation test in one paragraph

MAT uses human monocytes, or a monocytic cell line, as the detector. Pyrogenic material triggers the cells to release inflammatory mediators, typically measured as interleukin-6 or interleukin-1 beta by immunoassay, and the response is quantified against a reference. Because the readout is the human innate immune response rather than a crab-derived enzyme cascade, MAT responds to non-endotoxin pyrogens as well as endotoxin, which is its whole reason for existing. The cost is a more demanding assay: cell sourcing and qualification, donor variability where primary cells are used, and product interference that has to be characterised much as it is for BET.

Choosing the test

The last box is the one that catches people. A laboratory can validate a better method and still not be permitted to use it for release until the filing is updated in the relevant markets.

Sampling, sample handling, and timing

The number you report is only as good as the sample.

• Bioburden timing matters more than the count. Pre-sterilization bioburden must be sampled close to the sterilization step (after the last point where the load can change) and tested promptly, because holding a sample lets organisms grow or die and misrepresents the true load at sterilization. Define and validate the maximum hold time and storage condition.
• Endotoxin can grow in waiting samples. Gram-negative organisms in a held in-process sample can multiply and release endotoxin, biasing the result high; conversely, endotoxin can adsorb to container surfaces, biasing low. Test promptly or hold under validated conditions.
• Use endotoxin-free / depyrogenated labware for BET. Glassware is depyrogenated by dry heat (commonly 250 C for 30 minutes or a validated equivalent achieving a 3-log endotoxin reduction). A single contaminated pipette tip invalidates a run.
• Representative sampling. Sampling plans must justify number and location of units per the batch and the risk. For sterile filtration, bioburden of the bulk immediately before the final filter is the key control point.

Roles and responsibilities

Common mistakes and recurring inspection findings

These are the patterns that show up in citations against micro QC labs. None reference any specific company.

• No or inadequate method suitability. Reporting <61>/<62> or BET results on a product never shown to permit recovery / not to interfere. A passing result on an inhibitory product is meaningless. Inspectors check that suitability was done and re-done after formulation changes.
• Treating the <1111> count limit as the only question. Recovering an organism, calling it “within limits,” and never identifying it. Objectionable-organism assessment (211.113) is a separate obligation. Burkholderia cepacia complex findings in aqueous products are a recurring recall driver.
• Endotoxin limit / MVD errors. Using the wrong K (for example 5 EU/kg for an intrathecal product that requires 0.2), wrong body weight, or wrong dose basis. Diluting beyond MVD and reporting a “pass” that could not have detected the limit.
• Missing or failing positive product control / negative control invalidation ignored. Releasing on a run where the PPC was out of range, or where the negative control was positive (contaminated reagent water or labware). Both invalidate the run.
• Inadequate depyrogenation of labware or no validation of the depyrogenation cycle (no demonstrated 3-log endotoxin reduction).
• Sample-hold abuse. Bioburden or endotoxin samples held too long without a validated hold time, so the result does not represent the true in-process state.
• Invalidating OOS results without scientific justification. Re-testing a failing bioburden or endotoxin result into compliance without a sound lab-error investigation, or averaging away a failure. OOS handling must follow a defined process.
• Growth promotion of media not performed or not contemporaneous. Using media lots whose ability to grow the target panel was never confirmed.
• Data integrity gaps. Re-reading gel-clot tubes after the window, manual integration of BET curves without audit trail, deleting “atypical” runs. The same ALCOA+ expectations apply as anywhere in GxP.
• Confusing sterile vs nonsterile criteria. Applying <1111> count limits to a product that should be tested by <71> sterility, or vice versa.
• No LER assessment for an at-risk formulation. A surfactant plus chelator biologic formulation released on BET with no hold-time study, so the assay’s ability to detect endotoxin in that matrix was never established. A passing routine positive product control at the working dilution does not answer this question, because dilution is what breaks the masking.
• Applying <1111> limits to an in-process control. Using a finished nonsterile product acceptance criterion as the bioburden limit before sterile filtration, instead of the applicable in-process expectation and a sample size that can actually detect at that limit.
• Testing less than the specified sample volume. Assaying a small aliquot and scaling the result to report against a limit expressed per 100 mL, which does not have the sensitivity the limit assumes.
• Alert and action limits with no derivation. Round numbers with no link to process capability, or limits that were widened after an excursion with no recorded rationale independent of that excursion.
• Assuming a global pyrogen strategy. Treating a method change as a laboratory decision when the method is registered, or assuming that what is acceptable in one region is acceptable in another. The US and EU frameworks on the rabbit pyrogen test diverged in 2026.
• Media growth promotion not performed, not contemporaneous, or not covering the relevant panel, leaving every negative result produced on that lot unsupported.

Interview questions and strong answers

Q: What is the difference between bioburden and endotoxin, and can you have one without the other? A: Bioburden is a viable count of culturable organisms; endotoxin is LPS from Gram-negative cell walls and a chemical measure. Yes, you can have endotoxin with zero viable count, because killing the organism (or filtering it) leaves the heat-stable LPS behind. That is why sterilization and depyrogenation are different operations.

Q: Walk me through calculating an endotoxin limit. A: Limit = K / M. K is the threshold pyrogenic dose per kg per hour: 5 EU/kg for most parenterals, 0.2 EU/kg for intrathecal. M is the maximum human dose per kg per hour, using 70 kg standard body weight unless a defined population applies. Then express per mL using product strength. I would also state the LVP and WFI fixed limits, 0.5 EU/mL and 0.25 EU/mL.

Q: What is MVD and why does it matter? A: Maximum Valid Dilution, per USP <85>, is the most you can dilute a sample and still be able to read the endotoxin limit, computed as MVD = (endotoxin limit x product concentration) / lambda. It matters because you often must dilute to overcome interference, but diluting past MVD means a result at the limit would fall below the reagent’s sensitivity, so a “pass” would be meaningless.

Q: Which is your most important BET control and why? A: The positive product control. It proves the product is not masking endotoxin by inhibiting the reaction. A negative sample with a failing PPC is not a pass; it is an invalid test. Recovery must fall within 50 to 200 percent for photometric methods.

Q: Why is the gel-clot the referee method? A: USP <85> designates it as the deciding method when results from different methods conflict. It is the most direct readout of the cascade endpoint and least dependent on instrument calibration.

Q: What is method suitability for <61> and what panel do you use? A: It demonstrates the method recovers organisms from your specific product, within a factor of 2 of the inoculum control, using S. aureus, P. aeruginosa, B. subtilis, C. albicans, and A. brasiliensis at NMT 100 CFU. If recovery fails, the product is inhibitory and you change the method, often to membrane filtration with rinsing.

Q: An organism is recovered that meets the <1111> count limit. Are you done? A: No. I have to identify it and assess whether it is objectionable for the route under 211.113, even if it is below the count limit and not named in <62>. Burkholderia cepacia complex in aqueous non-sterile products is the classic objectionable organism missed this way.

Q: How would you defend choosing recombinant Factor C? A: It is endotoxin-specific (no beta-glucan false positives), reduces animal sourcing, and is recognized in USP <1085>, in USP <86>, and in Europe through Ph. Eur. 2.6.14 Method G, which replaces the standalone chapter 2.6.32 from 1 January 2027. I would still perform product-specific method suitability and route the change through change control with comparability data.

Q: What is Low Endotoxin Recovery and how would you test for it? A: LER is a formulation-driven loss of detectable endotoxin over time, characteristic of biologic formulations combining a surfactant such as polysorbate with a chelator such as citrate, phosphate, EDTA, or histidine. The formulation disperses the LPS aggregates the assay depends on, so endotoxin that is physically present stops being detectable. The critical subtlety is that a routine positive product control run at the working dilution can pass while the undiluted product is strongly masking, because dilution disrupts the masking. You test it with an undiluted spike-recovery study across at least four time points under process-relevant temperatures and hold times, and treat two consecutive points below 50 percent recovery as LER. The response is to fix the method and set the endotoxin specification as low as reasonably achievable, not to explain the result away.

Q: Where does the bioburden limit before sterile filtration come from, and what is it? A: Not from USP <1111>, which covers finished nonsterile products. For the sterile filtration control point the EMA sterilisation guideline, EMA/CHMP/CVMP/QWP/850374/2015, effective 1 October 2019, describes not more than 10 CFU per 100 mL TAMC as acceptable in most situations, tested on the bulk immediately before sterile filtration, on a 100 mL sample so the method can detect at the limit. The sample volume is part of the limit. I would also confirm what my own filing commits to, since that governs.

Q: Can we still use the rabbit pyrogen test? Can we stop using it? A: It depends on the region and the product, and the two major frameworks moved in opposite directions in 2026. Europe suppressed the rabbit test chapter, Ph. Eur. 2.6.8, with implementation from 1 January 2026, and replaced the prescriptive approach with general chapter 5.1.13 Pyrogenicity, which requires a justified strategy built on the bacterial endotoxins test or the monocyte activation test. The US retains it: 21 CFR 610.13(b) requires a rabbit pyrogen test for certain biological products, waivable under 610.9 on demonstration of an equivalent method, and FDA’s Pyrogen and Endotoxins Testing questions and answers, Edition 2, March 2026, points back to the rabbit test where non-endotoxin pyrogens may be present or where interference cannot be resolved within the MVD. So for a product registered in both regions I would expect to justify the strategy per market rather than assume one approach travels.

Q: What incubation conditions for TAMC and TYMC? A: TAMC on SCDA at 30 to 35 C for 3 to 5 days; TYMC on SDA at 20 to 25 C for 5 to 7 days.

Practical tips

• Calculate K, M, the endotoxin limit, and MVD once per product and lock them in the test method so analysts are not recomputing them per batch. Most release errors here are arithmetic.
• For interference, try dilution first; it is the cleanest fix and you have headroom up to MVD. Reach for divalent-cation buffers, pH adjustment, or endotoxin-specific reagents only when dilution alone will not get you there.
• Treat LAL reagent water as a critical reagent: one positive negative-control and the whole plate is suspect. Log and trend negative-control behavior.
• Keep depyrogenation validation current for any glassware used in BET; a lapsed cycle qualification is an easy finding.
• For bioburden, fix the sampling point and hold time in the procedure, because timing is the variable that most often makes a count unrepresentative.
• Build a microbial identification and objectionability decision tree so analysts know exactly when an isolate must be identified and escalated.
• Trend counts and endotoxin over time, not just pass/fail; a rising trend within limits is an early warning the contamination control strategy should catch before a failure.

Related reading

• Sterility Testing (USP <71>)
• Aseptic Processing and Media Fills
• Depyrogenation and Dry Heat Sterilization
• Environmental Monitoring Program
• Microbial ID and EM Excursions
• Water System Validation (USP <1231>)
• Compendial Method Verification
• OOS Investigation Process
• Annex 1 Contamination Control Strategy
• Container Closure Integrity Testing
• Sterilization Validation (Moist Heat)
• Visual Inspection of Injectables (USP <790>)