Water for injection (WFI): grades, specs and loop design

What separates WFI from purified water? One biological limit

Chemically, nothing. Both are held to conductivity of 1.3 µS/cm or less (stage 1, at 25°C) and TOC of 500 ppb or less. What WFI adds is a bacterial endotoxin limit of 0.25 EU/mL or less.

Injectables go straight into the bloodstream. Endotoxin comes from the cell wall of Gram-negative bacteria, and even a trace of it causes fever. So WFI gets one limit that purified water does not have, and a tighter microbial level to match: 10 CFU/100 mL against 100 CFU/mL for purified water (both are action levels, explained below).

Four water types are in play, and each has a job. Potable water (常水) is the feed: municipal-quality water that is not used in a drug directly and has no pharmacopeial specification, so no endotoxin limit either. Purified water (精製水, PW) goes into oral dosage forms, the dissolving of raw materials and the first rinse of containers. WFI goes into injectable manufacturing, the final rinse of sterile containers and, importantly, the feed for pure steam.

Pure steam (純蒸気) is easy to overlook. It is steam, not liquid, and it is blown into pipework and vessels for sterilization in place (SIP). Its quality is judged on the same ground as WFI: the condensate must meet the WFI specification, including endotoxin of 0.25 EU/mL or less. Saturated steam has no microbial limit of its own. That is as far as the pharmacopeia goes. Separately, in the steam sterilizer field, EN 285 asks for a dryness fraction of 0.95 or more, non-condensable gases of 3.5 mL or less per 100 mL of condensate, and superheat of 25 K (about 25°C) or less when the steam is expanded to atmospheric pressure.

WFI is not purified water made cleaner: the chemistry is the same, and the extra step is a deeper cut on the biological side, microbes and endotoxin.

Distillation is no longer the only way to make WFI

For a long time WFI meant distillation. Membranes are now an option, with conditions.

A still boils the water and takes off only the steam. Non-volatile impurities, microbes and endotoxin stay behind in the boiler. Because heat does the work, distillation is the easier method to back up with evidence of biological cleanliness. Stills come as multiple-effect (MED) or vapor compression (VC) designs. Organo's DWV series is a multiple-effect design that draws stable output when primary steam pressure varies and starts up quickly from standstill to draw-off. MECO has long built both vapor compression and membrane systems. In Japan, Kurita Water Industries and Nomura Micro Science supply pharmaceutical water equipment.

The change came in 2017. The European Pharmacopoeia monograph for WFI (Ph. Eur. 0169) was revised and took effect on April 1, 2017. It allows WFI from a membrane method built around reverse osmosis (RO) if the producer can show quality equivalent to distillation. Before the revision, only distillation was allowed. "Equivalent" means more than meeting the numbers: EDQM explains that it also looks at how dependable the production method is. A switch to membranes is notified to the competent authority before it is put into practice. The revision brought Europe in line with USP and the JP, which already accepted RO or ultrafiltration (UF). In the JP, WFI has long been framed as water that can be produced by ultrafiltration as well as by distillation.

A membrane train is basically one or two stages of RO followed by electrodeionization (EDI) or UF. BWT's OSMOTRON and Stilmas lines cover both membrane and distillation systems. What matters most is pretreatment, more than the RO membrane itself. Unless hardness, chlorine and organics are removed first, the RO membrane fouls quickly. Endotoxin is held back by the final-stage UF. With no heat, a membrane train saves energy, but it takes on two jobs in exchange: continuous monitoring and biofilm control. So the question for a supplier is not "distillation or membrane?" Ask how far the pretreatment is built out, and what backs up the endotoxin removal.

Membranes also opened a second option, cold WFI, made and distributed at ambient temperature. Conventional WFI is stored and circulated hot to suppress microbes, and keeping it hot costs energy, wastewater handling and maintenance. Ambient operation removes that load. But it also removes heat as an ally, so sanitization is set up separately, with ozone for example. An industry rule of thumb is to store unsanitized cold WFI for 24 hours at most, and to drain it or start sanitization before that limit.

Neither method is right in general: usage volume, energy and how far pretreatment is built out decide which one fits.

Conductivity, TOC and endotoxin decide release; the microbial figure is an action level

The WFI specification has four items, and they split into two kinds.

Conductivity and TOC measure chemical cleanliness, and both can be measured online. Conductivity is judged on the raw reading, without temperature compensation. USP <645> sets stage 1 at 1.3 µS/cm or less at 25°C. If a sample fails, testing moves to stage 2 (2.1 µS/cm or less at 25°C) and then to stage 3, which depends on pH. TOC under USP <643> is 0.50 mg/L (500 ppb) or less. The other two items measure biological cleanliness: endotoxin under USP <85> is 0.25 EU/mL or less, and microbes carry a reading trap.

The microbial figure of 10 CFU/100 mL is not a pass/fail line. It is the recommended action level in USP <1231>, a trigger to start an investigation. For purified water the guide figure is 100 CFU/mL. FDA has indicated that none of the limits for pharmaceutical water is a pass/fail line; they are action levels, and exceeding one means an investigation, not instant rejection. Microbes also need days of incubation, and by the time a count comes back the water has already been used. So microbes are managed by trend, not by release testing. The alert level sits below the action level and is set from your own operating history.

The endotoxin limit can be traced back to dose. Start from the premise that for parenteral routes other than intrathecal, the threshold dose K is 5 EU/kg per hour. Take an infusion of 1 L (1,000 mL) in one hour to a 70 kg patient. The whole product may then carry 5 × 70 = 350 EU, which is 0.35 EU/mL or less. WFI's 0.25 EU/mL sits inside that infusion-based benchmark. The 0.35 figure is a worked example under those assumptions only. A real product's limit is calculated per product from its dose and K, so keep the worked example and an individual product's specification apart.

Treat conductivity, TOC and endotoxin as release limits and microbial counts as a trend, and quote the pharmacopeia, the water type and the temperature with any value you cite.

The pharmacopeias agree on WFI and differ on purified water

For WFI, USP, Ph. Eur. and the JP line up closely: conductivity 1.3 µS/cm or less at 25°C, TOC 500 ppb or less, and endotoxin 0.25 EU/mL.

The differences are in purified water. USP applies the same 1.3 µS/cm at 25°C to purified water as to WFI. Ph. Eur. is looser for bulk purified water (Aqua purificata, online conductivity), allowing 5.1 µS/cm or less at 25°C. The JP, for water in containers, checks conductivity and TOC and also keeps a control based on potassium permanganate-reducing substances (過マンガン酸カリウム還元性物質), because cases of TOC rising during storage in containers were seen.

Assume "purified water is 1.3 everywhere" and you will misread the European purified water specification. This is also the reason to name the pharmacopeia and the temperature whenever a value goes into a specification or a contract.

When you pull a limit from a document, check which pharmacopeia, which water type and which temperature it belongs to.

Water quality is lost in the loop, not in the still

Most water trouble starts in piping, not in the generating equipment.

Water runs in one line from feed to use point: potable water, pretreatment, primary purified water (RO plus EDI), WFI generation (distillation or UF), the storage tank, the circulation loop (supply and return) and then each use point. You can make very good water and ruin it while it is stored and distributed. Even in nutrient-poor water, microbes attach to the inner wall of the pipe and build a film called biofilm. The basic defense is to keep the water moving.

The hot loop is the classic answer. WFI loops are widely designed to circulate above 65°C, and many run near 80°C. At that temperature microbes do not grow. BWT's LOOPO is one example of a product line aimed specifically at building hot loops. The weak point is the branch to a use point. Between the main line and the valve, water stagnates while the valve is closed. This stagnant section is called a dead leg.

A rule of thumb limits the branch length to a multiple of the pipe diameter (L/D). It originates in FDA's 1972 draft guideline for large-volume parenterals (21 CFR 212.49), which set 6 times for a hot circulating system. Today's ASME BPE calls 2 or less an achievable target with current technology. It is a target, not an absolute requirement, and ASME BPE itself notes that some systems cannot reach it. In water systems, 3 or less is what teams can realistically achieve, and some biotech systems tighten to 1.5. Note that 6 and 2 are measured from different reference points: the old 6 from the center of the main pipe, the recent values from the pipe wall.

Ambient loops go with cold WFI. Hot circulation consumes energy, so ambient operation sanitized with ozone is also used. Ozone loses effectiveness as the temperature rises, and given its half-life the practical range extends to about 35°C. As a guide, use above 50 ppb for short contact (1 to 2 hours); for contact longer than 6 hours, 20 ppb or more may be enough. After sanitization, UV breaks down the residual ozone before the water is used. An ambient system turns stagnation in a dead leg straight into a breeding ground, so it takes more operating discipline than a hot one.

Keeping the water moving matters more than making it clean, and the decisions that count are hot circulation versus ambient ozone and how short the dead legs are.

Steps for specifying a water system

These are the items to settle in order, because each one forces rework if it is decided late.

1. Decide the use first. Oral dosage forms and dissolution take purified water. Injectables, the final rinse and the feed for pure steam take WFI.
2. Confirm whether an endotoxin limit applies. For WFI, 0.25 EU/mL or less is added, and the weight of piping and monitoring changes from here.
3. Choose the production method: distillation or membranes, and whether to make cold WFI. Decide by usage volume, energy and how far pretreatment is built out. Ask each supplier how pretreatment is built and what backs up endotoxin removal.
4. Design the distribution loop: a hot loop above 65°C, or an ambient loop with ozone. Keep dead legs to L/D 2 or less as the target.
5. Set the monitoring and sanitization cycles. Follow conductivity and TOC continuously, follow microbes and endotoxin by trend, and sanitize before counts rise. Set the alert level below the action level, from your own history.
6. Back it all with equipment qualification (IQ/OQ/PQ). The PQ of a water system is built in three phases and takes about a year, so that seasonal variation is covered.

These related guides go deeper, one for each part. For the biological indicators, see bioburden versus endotoxin (in Japanese) and endotoxin (in Japanese). For removal, see endotoxin control and removal (in Japanese). Dead legs are covered in biofilm and dead legs (in Japanese), the evidence in equipment qualification (IQ/OQ/PQ) (in Japanese), and the overall position in contamination control strategy (CCS) (in Japanese). Design and operation of the distribution side continue in pharmaceutical water system design and operation (in Japanese).