Purified Water and Water for Injection systems are qualified through a three-phase protocol that is frequently treated as a documentation exercise. It is not. The phases are designed to establish, in sequence, that the system can produce compliant water, that it does so reliably under normal operation, and that it continues to do so across a full year of seasonal variation.
Why Three Phases
Each phase answers a different question, and the sampling intensity in the early phases is deliberately far higher than routine monitoring because the purpose is to characterise the system rather than to confirm it.
Phase 1 — Can it work?
Typically two to four weeks. Daily sampling from every point of use and from each stage of the generation train. Operating parameters are deliberately varied and the system is challenged. No water is released for manufacturing use during this phase.
The output is an understanding of how the system behaves — where counts run highest, how it responds to stagnation over a weekend, how quickly it recovers after sanitisation, and which points are the weakest. Facilities that treat Phase 1 as a pass/fail exercise miss its entire purpose.
Phase 2 — Does it work reliably?
Typically two to four weeks. Daily sampling continues, now under the standard operating procedures established from Phase 1 rather than under deliberate challenge. This confirms the system consistently produces water within specification when operated normally.
Water may generally be released for use during Phase 2, subject to the site's own risk decision.
Phase 3 — Does it keep working?
One full year. Sampling reduces to the routine schedule, with the objective of demonstrating performance across all seasons.
This is the phase most often shortened, and shortening it is the most consequential error in the whole protocol. In India, monsoon humidity, summer feed water temperature and winter demand patterns all affect microbial performance materially. A system qualified only across a cool dry period has not been qualified for the conditions it will actually face.
Sampling Points and Coverage
- Feed water, establishing the challenge the system is being asked to handle.
- After each unit operation — multimedia filtration, softening, activated carbon, RO, EDI, UV, final filtration — so a failing unit is attributable rather than inferred.
- Storage tank outlet, including the tank vent filter integrity.
- Loop supply and return, which bracket the distribution.
- Every single point of use. In Phase 1 and 2, all of them daily. This is what makes early-phase sampling volumes so high.
- Worst-case points identified from the design — furthest from the loop, lowest flow, most intermittent use.
Limits and Sample Volumes
| Parameter | Purified Water | Water for Injection |
|---|---|---|
| Action limit | 100 CFU/mL | 10 CFU per 100 mL |
| Typical alert limit | 50 CFU/mL | 5 CFU per 100 mL |
| Endotoxin | Not routinely required | Below 0.25 EU/mL |
| Total organic carbon | Below 500 ppb | Below 500 ppb |
| Conductivity | Stage-dependent per pharmacopoeia | Stage-dependent |
One detail is frequently wrong in practice. The sample volume must match the limit. Detecting 10 CFU per 100 mL requires filtering the full 100 mL through a membrane — a 1 mL plate count cannot resolve that limit at all. We regularly see WFI monitored by a method incapable of demonstrating the specification it is being compared against.
Note also that these are action limits, not specifications. Water exceeding them is not automatically unusable; it triggers an investigation and an assessment of product impact. Treating them as pass/fail specifications leads to unnecessary rejections and, worse, to pressure on the laboratory.
Deriving Alert Limits From Phase Data
The Phase 1 to 3 data set is exactly what alert limits should be derived from, and it is a wasted asset in most qualifications.
Rather than adopting the pharmacopoeial action limit and setting an alert at half of it, derive both from the observed distribution per sampling point, using the approach in microbiological trend analysis. A loop routinely running below 5 CFU/mL should carry an alert limit well below 50, or excursions will never be detected until they are large.
Cap everything at the pharmacopoeial action limit, but where your system is better than the compendium requires, your own limits are the useful ones.
Sanitisation and Its Verification
Sanitisation frequency should be derived from the observed rebound rate rather than fixed arbitrarily — sanitise, then sample at intervals across the following month and plot the recovery curve. The point at which counts approach the alert limit is where the next sanitisation belongs.
Methods in common use are hot water at 80 °C or above circulated for a validated period, ozone for ambient loops, and chemical sanitisation with peracetic acid or hydrogen peroxide. Whichever is used, effectiveness is judged by how long counts stay down, not by a single clean sample the day after. See biofilm in water systems.
Ongoing Monitoring After Qualification
Qualification establishes the baseline; routine monitoring detects drift.
- Point-of-use rotation, so every outlet is covered on a defined cycle rather than the same accessible taps being sampled indefinitely.
- Loop supply and return at higher frequency, since they reflect the whole system.
- Trend per point, never aggregated across the loop, because the average hides the one outlet that is deteriorating.
- Review TOC and conductivity alongside microbiology, since they frequently move first.
- Investigate rising trends within limits, which is where the early warning actually lives.
- Requalify after any modification — new outlet, changed sanitisation regime, replaced RO membranes, altered flow rates.
Common Qualification Failures
- Phase 3 truncated to a few months, so seasonal performance was never demonstrated.
- Sample volume inadequate for the limit, particularly for WFI.
- Alert limits set at half the action limit by default rather than derived from data.
- Not every point of use sampled in Phase 1 and 2, leaving parts of the loop uncharacterised.
- Sanitisation frequency fixed before the rebound rate was known.
- No dead leg survey during design review, so harbourages were built in.
- Requalification skipped after modification, because the change was made by engineering without quality involvement.
- Isolates not identified, so there is no baseline flora for the system against which a future problem can be compared.
The Indian Climate Dimension
Systems designed against temperate assumptions frequently underperform here. Ambient loops in central India run within the biofilm growth optimum for much of the year, and feed water quality varies sharply between the dry season and the monsoon.
For manufacturers around Indore, Pithampur, Dewas and Mandideep, we generally recommend that Phase 3 explicitly bracket the monsoon and the peak summer months, and that alert limits be reviewed seasonally rather than annually. A loop stable at 18 °C in January can behave quite differently at 35 °C in May, and a qualification that never observed the second condition has not tested the system.
Documentation the Qualification Must Produce
A qualification is judged on its documentation as much as on its data, and the packages we are asked to review are frequently strong on results and weak on rationale.
- Design qualification review, including the dead leg survey, material specifications, surface finish records and weld inspection documentation. A qualification that never examined the design cannot explain a design-caused problem later.
- Installation and operational qualification for each unit operation, with calibration records for every instrument the qualification relies on.
- The sampling plan with justification — why those points, why that frequency, why those volumes, and how worst-case locations were identified.
- Method documentation including the membrane filtration volume, media, incubation conditions and the demonstrated limit of detection, showing the method can resolve the limit being applied.
- All raw data across the three phases, not summary tables. Phase 1 data in particular has ongoing value for limit derivation and should not be reduced to a pass statement.
- Statistical derivation of alert and action limits from the phase data, with the method stated.
- Sanitisation rebound characterisation and the frequency derived from it.
- Isolate identifications from the phases, establishing the system's baseline flora for future comparison.
- A conclusion that states limitations — conditions not encountered during qualification, points that ran closest to limits, and what would trigger requalification.
That last element is the one that distinguishes a qualification written by someone who understands the system from one assembled to satisfy a checklist.
How DSTechnoverse Helps
We support water system qualification from the microbiological side and run the monitoring that follows it.
That covers Phase 1 to 3 protocol design and sampling execution, sample volume and method selection appropriate to the limit, alert and action limit derivation from your own phase data, dead leg and design review before qualification begins, rebound-rate characterisation to set sanitisation frequency, water system flora library development, per-point trend analysis, and requalification support after modification.
Our laboratory in Indore collects directly across Madhya Pradesh, which matters for water given the short hold time; clients elsewhere in India use validated cold-chain courier arrangements.
Frequently Asked Questions
Can Phase 3 be shortened?
Not defensibly. Its entire purpose is to demonstrate performance across a full annual cycle of feed water quality, ambient temperature and demand. A shortened Phase 3 has not tested the conditions the system will face.
Can water be used during qualification?
Not during Phase 1. During Phase 2 it generally can be, subject to a documented site risk decision. Phase 3 runs under normal production use.
What sample volume is needed for WFI?
The full 100 mL by membrane filtration, because the limit is expressed as 10 CFU per 100 mL. A 1 mL plate count cannot demonstrate that limit.
Are the pharmacopoeial values specifications or action limits?
Action limits. Exceeding one triggers investigation and product impact assessment rather than automatic rejection of the water.
How often should the system be sanitised?
At the frequency your rebound data supports. Sanitise, sample across the following month, and place the next sanitisation where counts approach the alert limit. A fixed monthly schedule chosen without that data is arbitrary.
How long does a full qualification take end to end?
Around fourteen months in practice. Phases 1 and 2 take four to eight weeks combined, Phase 3 runs a full twelve months, and protocol development plus final reporting add a further month or two either side. Systems are normally released for routine manufacturing use during Phase 3, so the twelve months is not lost production time — but the qualification is not complete, and the system is not fully characterised, until it finishes.
Can an existing system be retrospectively qualified?
Yes, and it is common for older installations, but it is more work than qualifying a new one rather than less. Retrospective qualification requires a design review to establish what was actually built including any undocumented modifications, a dead leg survey, and then the same Phase 1 to 3 sampling programme. The difference is that an operating system cannot easily be taken out of use for Phase 1, so the challenge testing has to be designed around production. Historical monitoring data supports the exercise but does not substitute for it, because it was collected under a sampling plan that was never designed to characterise the system.
Do we need to identify organisms recovered during qualification?
Yes, and it is routinely skipped. The phases generate the only comprehensive picture you will ever have of the system's baseline flora, and identifying those isolates gives you a reference set for every future investigation. When a product isolate appears two years later, being able to compare it against the organisms the system was carrying at qualification turns a months-long investigation into a days-long one. Archive the cultures as well as the identifications — see microbial identification.
What if the system fails during Phase 3?
Investigate, correct and assess how much of the phase remains valid. A single excursion with an assignable cause and effective correction does not usually restart the clock. A systemic problem — a design fault, a dead leg, an inadequate sanitisation regime — means the system being qualified is not the system that will operate afterwards, and Phase 3 should restart once the modification is complete and requalified.
What triggers requalification?
Any modification to generation, storage or distribution, including new points of use, changed sanitisation, replaced membranes and altered flow rates — plus a periodic review even where nothing has changed.
Water problems are usually system problems, not sampling problems. DSTechnoverse provides water and wastewater microbiology — potable, packaged, purified, WFI, process and effluent testing with biofilm and distribution system investigation. Based in Indore, Madhya Pradesh, serving clients across India. Talk to our microbiology team.