Endotoxin testing in practice: limits, MVD and interference

How do you set the limit? Dose and body weight come first

The limit is what the body can tolerate, divided by the maximum dose a patient receives per kilogram per hour.

What the body can tolerate (K) depends on the route of administration. It is 5 endotoxin units (EU) per kilogram of body weight per hour for intravenous and similar injections, and 0.2 EU for intrathecal administration, where the drug goes in around the spinal cord. The larger the dose, the smaller the amount allowed per mL or per mg. The test methods in the US, European and Japanese pharmacopeias are harmonized, but the product limit itself falls outside the harmonization. Each company sets it product by product in its filing.

The body weight assumption is easy to overlook. For the average adult, 70 kg is commonly used in the US and 60 kg in Japan. The lighter the assumed weight, the larger the dose per kilogram, so the same product's limit comes out a little over 10% tighter under the Japanese assumption. For a pediatric product, calculate for the smallest patient in the target group receiving the largest dose. If you want to work out a limit from your own numbers, the endotoxin limit calculator does it.

Water used in manufacturing has a ceiling too. Water for injection is required to be below 0.25 EU/mL. How water systems are built is covered in pharmaceutical water systems (in Japanese).

Write the route of administration, the maximum dose and the body weight assumption into the rationale for the limit, because if the assumptions change, the limit changes.

How far can you dilute? To the maximum valid dilution, no further

Dilution is the standard way around interference, but it costs you detectability.

If you test the sample neat, the interference described below can stop you from measuring correctly. So you dilute it. You cannot dilute without limit, though. Dilute too far, and endotoxin at exactly the limit sinks below the lowest level the reagent can detect. A result of "not detected, so it passes" then means nothing. The ceiling on dilution is the maximum valid dilution (MVD, 最大有効希釈).

MVD is set by three things: the limit, the concentration of the sample and the sensitivity of the reagent. The more sensitive the reagent, the more room you have to dilute. When a strongly interfering sample leaves too little room to dilute, reviewing the reagent's sensitivity is one way out.

A test dilution has to meet two conditions at once: it must escape the interference, and it must stay inside the MVD.

How do you catch interference? Spike recovery, hold time and beta-glucan

Interference means a component of the sample weakens the reagent's reaction (inhibition) or strengthens it (enhancement), moving the result away from the true value.

Inhibition is the dangerous direction. Endotoxin that is present reads low, and a product that should fail could be shipped. Enhancement costs you a product that should have passed but does not threaten the patient directly. Common causes are a pH outside the range in which the reagent's enzymes work (roughly 6 to 8), too much or too little of divalent cations such as magnesium, chelators such as EDTA that take those cations away, and samples rich in protein or lipid.

You catch interference with spike recovery (添加回収), also called the positive product control (PPC): add a known amount of endotoxin to the sample and measure it. For quantitative methods read by turbidity or color, a recovery of 50 to 200% of the amount added means no interference. If the result falls outside that range, you fix it by diluting within the MVD, adjusting the pH or supplementing magnesium.

LER slips through spike recovery. In formulations that contain citrate or phosphate together with polysorbate, spiked endotoxin can be recovered right after it is added and then become unmeasurable as the sample sits. That is why a hold-time test is run: leave the spiked sample for as long as the product is actually stored, then check recovery. The details are in low endotoxin recovery (LER) (in Japanese).

There is a stumble in the opposite direction as well. LAL reagent, made from the blood cells of the horseshoe crab, has a pathway that responds to beta-glucans as well as to endotoxin. In a 2010 report on an intravenous immunoglobulin product, the high LAL result came from beta-glucan leaching out of cellulose filter pads, and the endotoxin had not increased. The clue is a dilution series whose results do not fall on a straight line. To confirm, use a buffer that blocks the glucan reaction, or a recombinant reagent that does not respond to beta-glucan.

Check spike recovery in every run, and from the formulation and the filter media the product has passed through, suspect LER and beta-glucan first.

Which reagent and method: gel clot, quantitative or recombinant?

Methods split by whether the answer is "over the limit or not" or a number.

The gel clot method checks whether endotoxin makes the reagent set into a gel. The equipment is simple and the method is rugged, and it suits a pass-or-fail call against the limit. Turbidimetric and chromogenic methods are quantitative: they follow the development of turbidity or color. They are easy to automate with a plate reader, and they are used to follow process trends.

The big change of the last few years is the recombinant reagent, which does not depend on horseshoe crab blood. One kind (rFC) makes only factor C, the entry point of the reaction, by recombinant technology. The other (rCR) makes every step of the reaction recombinant. The European Pharmacopoeia has listed an rFC method as a formal method since January 2021, and the USP made its recombinant reagent chapter <86> official in May 2025. FDA revised its testing Q&A in March 2026, extending wording that had been limited to LAL so that it covers recombinant reagents. The Japanese Pharmacopoeia (JP) also added, in its 18th revision, an alternative method using recombinant proteins to its general information chapters (参考情報).

What you confirm when you switch is whether your own product is measured the same way. In a Japanese study (AMED, fiscal 2015), three recombinant reagent products were used to measure 28 injectable products. The reagent with every step of the reaction made recombinant measured all of them within the MVD. Two factor-C-only products could not measure a heparin calcium product within the MVD. Even among recombinant reagents, then, the way interference appears differs by product. FDA likewise asks users of recombinant reagents to confirm that the reagent suits its purpose. Measure the same sample side by side with LAL, and compare whether spike recovery and interference behave the same. The differences between reagents are laid out in endotoxin testing and rFC.

When you change reagents, decide by comparing how interference behaves on your own product, not by the sensitivity printed in the catalog.

Checks to run on your test procedure or a contract lab's report

Open the test procedure or the contract lab's report and check the following. They tell you whether the numbers can be trusted.

1. Does the rationale for the limit state the route of administration, the maximum dose and the body weight assumption (adult 60 kg or 70 kg, or pediatric)?
2. Is the test dilution inside the MVD?
3. Does spike recovery (PPC) fall inside its range in every run, and is the procedure for an out-of-range result defined?
4. Does the formulation contain citrate or phosphate together with polysorbate? If so, was recovery checked after a hold time?
5. Does the process pass through cellulose filters or filter pads? Was the effect of beta-glucan checked?
6. Are glassware and containers depyrogenated?
7. If you are moving to a recombinant reagent, do you have data comparing it with LAL on your own product?

Reagent and kit candidates can be compared in the endotoxin test directory, and the instruments for quantitative methods in the plate reader directory. Depyrogenation of containers is covered in dry-heat depyrogenation and the FH value (in Japanese). Sterilizing filtration and sterilization remove or kill organisms but are not steps that remove endotoxin (see sterilizing versus bioburden-reduction filtration (in Japanese)). Removal in the process is covered in endotoxin control and removal (in Japanese), and monitoring of what is carried in in what bioburden is.