Cleaning Validation Data: Managing Limits, Swab Results, and Worst-Case Justifications Digitally

Why Cleaning Validation Is Now a Data Problem

Most pharma and biotech quality leaders inherited a cleaning validation program built around documents: a master plan, a limits memo, a set of protocols, and a stack of reports signed years ago. Each document was correct on the day it was approved. The trouble is that the facts inside those documents keep moving. New products come onto shared equipment. A toxicologist revises a permitted daily exposure. A batch size is scaled up. A line gets a new filling machine with a larger product contact surface. None of those events edits the limits memo on its own.

That gap is why cleaning validation data management has become a practical concern for leadership rather than a technical detail for the validation team. The regulators have not changed the science. What has changed is how clearly they now say that the program has to stay current and that its calculations and results have to meet data integrity expectations.

The Regulatory Picture as of October 2026

Several documents shape expectations today, and their status matters because it is easy to cite a draft as though it were in force.

• PIC/S PI 006-4, Recommendations on Qualification and Validation. It entered into force on 1 October 2026 and replaces PI 006-3 from September 2007, which covered the validation master plan, installation and operational qualification, non-sterile process validation, and cleaning validation.¹² Chapter 7 is a full, updated treatment of cleaning validation. Its cleaning chapter applies to finished products and offers supplementary, optional guidance for active pharmaceutical ingredients (APIs).¹
• EU GMP Annex 15, Qualification and Validation. The version in force came into operation on 1 October 2015. Section 10 covers cleaning validation and Section 11 covers change control.³ In February 2026, the European Medicines Agency and PIC/S opened a joint consultation on a concept paper to revise Annex 15, running from 9 February to 9 April 2026. The concept paper proposes extending the annex to active substance manufacturers and reflecting the revised ICH Q9(R1) guideline on quality risk management.⁴ A concept paper is a proposal to revise, not new text. The 2015 annex remains the version in force.
• EMA guideline on health-based exposure limits (HBELs). The guideline on setting health-based exposure limits for use in risk identification in shared facilities (EMA/CHMP/CVMP/SWP/169430/2012) was published in November 2014 and remains the current version on the EMA site.⁵ Its companion questions and answers document (EMA/CHMP/CVMP/SWP/246844/2018) is dated 19 April 2018.⁶
• WHO TRS 1033, Annex 2 (2021). A points-to-consider document on including HBELs in cleaning validation, with specific statements on calculations, data integrity, and process capability.⁷
• FDA Guide to Inspections, Validation of Cleaning Processes (July 1993). Still the FDA’s inspection guide on the topic. It states that the objective of the inspection is to ensure that the basis for any limits is scientifically justifiable.⁸
• Health Canada GUI-0028, Cleaning Validation Guide (June 2021). A detailed guide that addresses HBEL-based limits, new product introduction, and change control.¹¹

What Two 2026 Warning Letters Show

Two FDA warning letters from 2026 show how cleaning programs fail in practice, and both failures are about information that went stale or never existed.

In a January 30, 2026 letter to Cohance Lifesciences Limited, a finished pharmaceutical manufacturer, FDA described visible residues on nondedicated equipment that had been documented as clean after a product changeover. The firm’s own testing of those residues confirmed multiple APIs above its allowable limit. FDA also wrote: “Your 2018 cleaning validation study failed to use the identified worst-case drug product, making it inadequate to ensure cleaning effectiveness.” The next sentence is the one every quality leader should read twice: “You also did not re-evaluate your cleaning validation program when introducing new drug products on shared manufacturing equipment, and you lack routine cleaning verification.”⁹

In an April 7, 2026 letter to Chemco Corporation, a maker of topical over-the-counter drugs, FDA described white residue in a tank that the quality unit had verified as clean according to a logbook, and noted that the firm had not conducted cleaning validation studies for its non-dedicated equipment.¹⁰

The remediation requests in both letters are nearly identical. FDA asked each firm to identify and evaluate all worst-case drugs with higher toxicities, higher potencies, lower solubility in their cleaning solvents, and characteristics that make them difficult to clean, along with worst-case swabbing locations and maximum hold times before cleaning. FDA also asked each firm to describe the steps its change management system requires before a new product or new manufacturing equipment is introduced.⁹¹⁰ That request is a data management question. It asks whether the firm has a reliable way to know, at the moment of change, which limits, worst-case selections, and verification plans are affected.

The Data Objects Behind Every Cleaning Claim

When an inspector asks “how do you know this equipment is clean enough for the next product?”, the honest answer is a chain of linked records. If any link is missing, out of date, or held in a place nobody controls, the claim is weaker than it looks. We find it useful to name the objects explicitly, because naming them is the first step to managing them.

Equipment Master Data Is the Dependency Most Teams Miss

Underneath all four objects is equipment information: which items make up a process train, what materials they are made of, their product contact surface areas, and which parts are hardest to clean. WHO states that authorized equipment drawings should be current, accurate, and available, that surface area calculations should be documented and justified, and that the source data for those calculations should be available.⁷ PI 006-4 asks manufacturers to identify critical, hardest-to-clean areas using equipment drawings and practical evaluation of how the equipment is built.¹

In many sites, surface areas live in an engineering spreadsheet from the original qualification, and nobody updates them when a change part is added or a hose is lengthened. Every limit on that line then carries an error that no one can see. If you do only one thing after reading this article, find out where your shared surface areas are held, who owns them, and when they were last checked against the physical equipment.

Why the Links Matter More Than the Documents

Each object has an owner in a different function. Toxicology owns HBELs. Manufacturing science or technical services often owns the limit calculations. The quality control laboratory owns analytical methods, recovery studies, and results. Validation owns the protocols and worst-case rationale. Engineering owns equipment data. The failure pattern we see most often is not that any one of these is wrong. It is that one of them changes and the others are never told. A digital approach is worth the effort mainly because it can make those dependencies visible and enforce them.

Limits: From Toxicology Report to Acceptance Criterion

Annex 15 requires that limits for the carryover of product residues be based on a toxicological evaluation, that the justification be documented in a risk assessment with supporting references, and that acceptance criteria consider the cumulative effect of multiple items of equipment in the process train.³ PI 006-4 says the rationale for limits should be formally documented, based on therapeutic dose, potency, and toxicology, and be practical, achievable, and verifiable.¹ The FDA inspection guide uses similar language and notes that limits once mentioned by industry, such as 10 ppm and 1/1000 of the normal therapeutic dose, must still rest on a scientifically justifiable basis.⁸

The Calculation Chain

WHO TRS 1033 Annex 2 gives a clear worked method. The maximum safe carryover (MSC) of product A into product B is the PDE of product A multiplied by the batch size of product B, divided by the maximum daily dose of product B. The maximum safe surface residue (MSSR) is the MSC divided by the total shared product contact surface area. WHO recommends presenting these values in a table that lists each preceding product against each following product, and identifying the margin of safety.⁷ Those surface limits are then converted into a swab or rinse acceptance criterion using the swab area, the sample dilution volume, and the recovery factor.

Every term in that chain is a data point with its own source and its own reasons to change. The table below lists them, where they usually come from, and what tends to change them.

WHO is specific that these calculations and their data should be available and comply with data integrity principles, and that the calculation should include the PDE, maximum daily dose, batch size, total shared surface area, sample areas, sample dilution volumes, and recovery factors.⁷ That is close to a data model in plain words.

Calculated Limits Are Not Targets

A point that often gets lost when calculations move into a system: the HBEL-derived number is a ceiling for safety, not an operating target. PI 006-4 states: “Although this guideline may be used to justify cleaning limits, such evaluations are not intended to be used to set limits at the level of the calculated HBEL.”¹ The EMA questions and answers make the same point in Q6.⁶ A well-built tool should therefore hold at least two numbers for each product pair: the safety-based limit and the tighter limit the site uses in practice, with the rationale for the gap between them.

Legacy Limits, Alert Limits, and a Difference in Emphasis

Most sites moved from traditional limits to HBEL-based limits over the last decade, and many still carry both. The guidance documents treat the legacy numbers in slightly different ways, and your data model should be able to handle whichever position your quality system adopts.

• EMA Q&A 6 says that for existing products, historically used cleaning limits should be retained and can be considered alert limits, provided that, taking cleaning process capability into account, they give enough assurance that excursions above the HBEL will be prevented. Results above the alert limit should trigger an investigation, and repeated excursions are not acceptable where they show the process is not in control.⁶
• Health Canada GUI-0028 suggests considering alert limits where HBEL-derived limits are much higher than historic limits such as 1/1000 of a dose or 10 ppm.¹¹
• WHO TRS 1033 Annex 2 says a cleaning limit based on HBELs should be calculated and compared with the existing limit, that historic limits may be used when they are more stringent, and that alert and action limits should not be based on historic limits but on statistical analysis of existing data.⁷
• PI 006-4 says manufacturers should consider alert limits that can be based on historical limits and/or a statistical evaluation of past cleaning results.¹

The practical reading is that a legacy limit can play a role as an early warning, but the justification for any alert limit should be tied to how the cleaning process performs in routine use. That means the system needs the history of results, not just the current pass or fail.

The switch to HBELs also matters in the other direction. In a study of 140 drug substances, 9% (12 substances) had a PDE-based carryover limit that was stricter than the 1/1000 of minimum daily dose criterion, and the authors reported that antineoplastics consistently fell into that group.¹⁹ For those products, an old limit is not a conservative safety margin. It is too loose. A program that still calculates some limits the old way needs to know which products those are.

Where Spreadsheet Calculations Break

PI 006-4 adds several requirements that make simple spreadsheets fragile. Acceptance criteria should be set both for individual equipment and for the total cumulative carryover from multiple shared equipment items, which PI 006-4 calls the process train effect. The impact on unit doses should be considered, for example in compression or primary packaging. Calculations should not assume that a contaminant is spread evenly through a batch or across equipment, since residue on a surface might only reach the beginning of a batch.¹

Swab and Rinse Results: Making Sampling Data Traceable

A residue result is only meaningful if you can trace it back to where the sample was taken, how it was taken, how much of the residue that technique recovers from that surface, and what the method can detect. Annex 15 requires that recovery be shown to be possible from all product contact materials sampled in the equipment with all the sampling methods used, and that sampling materials and methods should not influence the result.³

Recovery Factors Are Data With a History

PI 006-4 says swab sampling is generally more effective than rinse sampling and should be used where feasible, with a combination of both when equipment access does not allow enough direct surface sampling. It notes that rinse samples average results across a large area, so localized high residues may not be found. The swab method should specify the sampling solvent, swab material, locations, and technique, and should demonstrate the percentage of residue recovered from every relevant material, such as plastic, silicone, and stainless steel. A correction factor for incomplete recovery may be needed.¹ PI 006-4 also expects the people who swab plant equipment to do it the same way it was done in the recovery studies.¹

Industry practice gives useful detail on what good recovery data looks like. One widely cited set of best practices recommends spiking at 125%, 100%, and 50% of the acceptable residue limit at a minimum, extending down to the limit of quantitation where practical, and running every level in triplicate. It recommends that recoveries at the three higher levels be at least 70% and agree within a relative standard deviation of 15%, and that average recoveries below 70% or above 105% prompt an investigation.¹⁸

Recovery data can also be reused across sites when it is managed as a dataset. In a study pooling 1,262 recovery factor values from 16 sites, covering 48 drug substances, formulations, or detergents on 29 materials of construction, the authors found that 97% of the remaining variability in recovery was within sites and only 3% across sites. They concluded that a stainless steel recovery study at one site could represent most materials used in drug product manufacturing across multiple sites.¹⁷ That kind of finding is only possible when recovery results are stored in a consistent structure with the material, substance, swab, solvent, and spike level recorded for each value.

The Fields Every Result Needs

PI 006-4 lists what a cleaning validation protocol should contain, including the rationale for sampling methods, clearly defined sampling locations and the reason for choosing them, recovery data where appropriate, analytical methods with their limits of detection and quantitation, the acceptance criteria and their rationale, and any other products, processes, and equipment covered under a bracketing approach.¹ Translating that list into fields gives a practical minimum for each sample result.

On results below the limit of quantitation, the FDA inspection guide makes a point worth building into the system: if contamination is not detected, that only means it is not present above the sensitivity of the method, not that there is no residue.⁸ Store the value as “below LOQ” with the LOQ recorded, not as a zero that later flatters a trend chart.

Non-Specific Methods and Visual Inspection Records

Where total organic carbon (TOC) or conductivity is used instead of a specific method, PI 006-4 says the result must be assumed to be entirely due to the target residue.¹ A tool that stores TOC results should apply that assumption automatically when comparing results with product limits, rather than leaving it to the reviewer.

Visual inspection is also data. PI 006-4 expects visual inspection to be performed and documented after all cleaning, with any failures recorded in the quality system and investigated, including failures found at the second-person check after manual cleaning.¹ Where visual inspection is used as a cleanliness criterion, the site should establish the threshold at which residues are readily visible under the actual lighting and viewing distances in production. PI 006-4 says products whose maximum allowable carryover is below the limit of visual detection should have cleaning verification after every product changeover.¹ The EMA Q&A adds that operators performing visual inspection need specific training, including periodic eyesight testing, and a practical competency assessment.⁶ Visible residue limits, operator qualifications, and inspection outcomes all belong in the same structured record as the analytical results.

Worst-Case Justifications That Survive a Product Change

Few sites validate cleaning for every product on every piece of equipment. Instead they group products and equipment and validate the worst case in each group. That approach is accepted, but only with a written rationale that stays true. Annex 15 requires a scientific rationale for the choice of worst-case product and an assessment of the impact of new products brought into the site, and it names solubility, cleanability, toxicity, and potency as possible criteria.³

What the Guidance Now Expects a Justification to Contain

PI 006-4 gives more detail than earlier texts. It says manufacturers should have a documented justification both for using a worst-case approach and for the specific worst-case product or products chosen. Factors can include HBEL values, cleaning difficulty, residue solubility in cleaning agents and water, the physical characteristics of the item being cleaned, and experience with similar products. It adds that it may be necessary to pick more than one worst case, and gives an example: “an insoluble product with a high HBEL value may be the most difficult product to clean but not necessarily the worst case compared to a partially soluble product with a low HBEL value.” PI 006-4 also says that all products being evaluated under the worst-case approach should be identified.¹

That last sentence matters for data management. A worst-case justification is not only a claim about one product. It is a claim about every product in the group. If the list of covered products is not maintained, the justification cannot be checked.

Practitioners also caution against leaning on a single factor. One ISPE Pharmaceutical Engineering article notes that worst-case selection based only on solubility data may prompt questions during an inspection, and recommends coupon testing to see how a residue behaves on a surface under realistic conditions, especially when new products are introduced to existing lines.¹⁵

Build the Rating Matrix as Data, Not as a Paragraph

The APIC guidance on cleaning validation in API plants describes a worst-case rating approach that translates well into a structured record. Substances are grouped by equipment class and cleaning procedure, each combination forms a group, and at least one worst case in each group is validated. Ratings for factors such as hardest to clean, solubility, and PDE are recorded in a matrix. If the lowest calculated limit in a group is reasonable to apply to all members, it can serve as a common limit; the substance with the lowest solubility in the cleaning solvent is tested; and any substance outside the bracket has to be validated on its own. The same guidance suggests a matrix of limits covering every possible previous and next product combination.¹²

When that matrix is a table in a validation report, it is correct only on the day the report is signed. When it is structured data, three things become possible. First, a new product can be scored against the same factors and placed into a group, or flagged as falling outside every group. Second, a change in any factor, such as a revised PDE, can be checked against the current worst case automatically. Third, an inspector’s question, such as “which products does this 2019 study cover?”, can be answered from a query rather than from memory.

Worst Case Beyond the Product

Worst case is not only about products. Annex 15 asks for an assessment of the variable factors that influence cleaning effectiveness, such as operators and the level of detail in procedures like rinsing times, and says that worst-case conditions should be the basis for validation studies.³ Dirty hold time, clean hold time, and campaign length are all worst-case parameters. PI 006-4 says campaign lengths should normally be defined by both time and number of batches.¹ PI 006-4 also asks manufacturers to consider that starting materials from different suppliers may behave differently during cleaning.¹ Each of these conditions should be recorded as a validated parameter with its supporting study, so that routine records can be checked against it.

Keeping the Program Consistent When Products Change

This is where most cleaning programs lose their footing. Change control was designed to route a change to reviewers. It was not designed to recalculate a limit matrix or re-run a worst-case rating. Unless someone knows to ask, a change can pass every signature and still leave the cleaning program out of date.

The guidance is clear that cleaning belongs inside change control. Annex 15 says quality risk management should be used to evaluate planned changes for their impact on validation, among other things.³ PI 006-4 says changes to validated cleaning procedures must be controlled, and that revalidation should be considered after equipment, process, or product changes, giving examples such as a change in detergent, a change in site water pressure, and repeat failures in cleaning verification.¹ Health Canada’s guide says a change control system should be in place to assess and document all changes that might affect the cleaning process, and that all new product introductions should be reviewed through quality risk management and change control to decide whether existing controls are enough.¹¹ WHO says the suitability of shared facilities should be evaluated both for products already made there and when new products are planned, for example through a change control procedure.⁷

Change Triggers That Should Prompt Recalculation

The practical step is to agree, in advance, which data changes must trigger a cleaning impact assessment, and to make those triggers automatic wherever the data lives in a system. A reasonable starting list:

• New product on shared equipment: new HBEL record, new rows and columns in the limit matrix, worst-case rating, possible new recovery study.
• Revised HBEL: every limit where the product is the previous product must be recalculated, and the worst-case ranking rechecked.
• New strength, indication, or dose: changes the maximum daily dose of the next product and therefore every limit where it follows another product.
• Batch size change: changes carryover limits for every product that precedes it.
• Equipment added, removed, or modified: changes shared surface area, may introduce a new material of construction that needs recovery data, and may add a new hardest-to-clean location.
• Detergent change or supplier formulation change: PI 006-4 expects the site to know the detergent’s composition and to be told of critical changes by the detergent supplier.¹
• Analytical method change or transfer: may change LOD, LOQ, and recovery.
• Repeat verification failures or visual inspection failures: a signal that the validated state may no longer hold.

A Sequence for New Product Introduction

Both 2026 warning letters asked the firm to describe the steps its change management system requires before a new product is introduced.⁹¹⁰ A defensible sequence, written so that each step produces a record the next step depends on, looks like this.

HBELs Are Not Permanent Numbers

It is tempting to treat a PDE as a fixed attribute of a molecule. The guidance does not. The EMA Q&A says HBELs should be established for all medicinal products and that the underlying toxicological or pharmacological data needs periodic reassessment throughout the product lifecycle.⁶ For investigational products, it says the basis for the HBEL should be regularly reviewed as new data appears.⁶ WHO says manufacturers should periodically consider new data on HBELs and update PDE reports where needed, and notes that when different people derive different HBELs for the same substance, the reason should be clarified and investigated.⁷ Health Canada recommends that HBELs be periodically reevaluated and that the impact on the cleaning validation program be assessed and documented.¹¹

For a biotech with a moving clinical portfolio, or a site that makes products for several sponsors, this means the HBEL register needs review dates, owners, and a link to every limit it feeds. The EMA Q&A also says contract givers should either provide a full HBEL assessment to their contract manufacturers or provide the data needed to produce one.⁶ Sponsors who outsource manufacturing should treat HBEL updates as information that has to reach their manufacturing partners, and should say so in the quality agreement.

Data Integrity, Records, and the Tools That Hold Them

Data integrity expectations apply to cleaning data in the same way they apply to batch and laboratory data. WHO says that data and results relating to HBELs, PDE reports, cleaning validation results, and calculations should comply with data integrity principles.⁷ PI 006-4 expects the validation master plan to define how data integrity is confirmed in qualification and validation reports, with that confirmation done independently by people who did not execute the work.¹

Records of Routine Cleaning

PI 006-4 lists what cleaning records should make readily available: the area or equipment cleaned, who cleaned it, when, the procedure used, the product previously processed, and any data or printouts from the cleaning. The record should be signed by the operator, anyone who supervised the cleaning, and the person who visually inspected the equipment afterward.¹ Both 2026 warning letters describe equipment recorded as clean that the investigator found visibly dirty.⁹¹⁰ A logbook entry is only as trustworthy as the process around it. Electronic cleaning logs that capture time, identity, and the visual inspection step as separate attributable entries make that kind of gap easier to detect during review.

For automated cleaning such as clean-in-place systems, PI 006-4 expects the cleaning sequence to be defined, including temperatures, concentrations, cleaning agents, number of applications, valve openings, spray rates, pressures, and volumes, with critical parameters monitored by sensors and alarms. It also lists “ensuring cleaning recipes and data are appropriately controlled, reviewed and protected from accidental/intentional modification or deletion” among the controls needed.¹ Health Canada likewise expects automated cleaning recipes to be protected against uncontrolled changes, and cleaning data, including alarms, to be controlled and reviewed.¹¹

Test-Until-Clean Shows Up in the Data

One of the oldest cleaning findings is also one of the easiest to detect digitally. Annex 15 states: “Repeated cleaning and retesting until acceptable residue results are obtained is not considered an acceptable approach.”³ PI 006-4 says: “Continued cleaning failures and/or testing until clean (i.e. continually cleaning and testing until acceptable results are achieved) are not acceptable.”¹ The FDA inspection guide made the same point in 1993, noting that constant retesting and resampling can show the cleaning process is not validated, because the retests document the presence of unacceptable residue.⁸

On paper, a failed first result followed by a re-clean and a passing result can look like two unrelated events. In a system where every sample is tied to a cleaning event and an equipment item, the pattern is visible in a simple query: how many cleaning events needed more than one cycle before release? Any failure during or after validation should be recorded and investigated, and PI 006-4 expects remediation if a process fails intermittently.¹

What a Calculation Tool Needs

Whether limits are calculated in a validated spreadsheet, a module of a laboratory or quality system, or a dedicated cleaning validation application, the same controls apply. The tool is a computerized system used for a GMP decision, so it falls under your computerized system validation approach (under EU GMP, Annex 11), scaled to its risk. At a minimum it should provide:

• Controlled master data for products, HBELs, equipment, surface areas, and recovery factors, each with an owner and version.
• An audit trail for every change to an input, a formula, or a limit, showing who changed it, when, and why.
• Limit versions with effective dates, so any historical result can be judged against the limit that applied at the time.
• Impact reporting: which limits, product pairs, and worst-case groups are affected when an input changes.
• A second-person review step for new or changed calculations, recorded in the system.

The choice of tool matters less than whether the links between the data objects are enforced. A dedicated application that is not connected to change control has the same weakness as a spreadsheet: it is correct until someone forgets to update it.

Trending, Process Capability, and Where to Start

PI 006-4 describes a cleaning validation program as three elements: development of the cleaning process with critical parameters identified, a validation study made up of a set number of verification assessments, and ongoing evaluation after the study to confirm the process remains in a state of control.¹ ASTM E3106, the standard guide for science-based and risk-based cleaning process development and validation, takes the same lifecycle view of development, qualification, and verification.¹³ The ISPE guide on the cleaning validation lifecycle, published in 2020, also covers acceptance criteria, visual inspection limits, residue limits, and change management within that lifecycle.¹⁴

Ongoing Verification Needs a Plan, Not Just Results

PI 006-4 says the frequency and scope of monitoring after validation should be set using quality risk management and may require analytical monitoring at each product change. It names areas of special concern: low-HBEL products, manual cleaning, products whose carryover limits are below the limit of visual detection, the amount of data available, and equipment or products with a history of failure or variable results. It adds that continued cleaning verification will normally be needed for low-HBEL products that are cleaned manually.¹ The APIC guidance describes the continued verification stage in terms of post-validation monitoring, change control, and periodic management review of deviations, changes, and trends.¹²

Use Statistics to Show Control

WHO recommends calculating process capability (Cpk) and using statistical process control to support cleaning verification results, with the capability presented against control limits and the margin of safety and discussed as part of continuous improvement.⁷ Because cleaning results have only an upper limit, one published study used the upper capability index, Cpu, which compares the spread of the data with the distance from the mean to the upper limit. The authors proposed that a desired Cpu, for example 1.33, could be used as the starting point for calculating a limit for all residues on all product contact surfaces, and they applied the idea to new product introduction.¹⁶

None of this works without clean historical data. Results stored as pass or fail, or as zero when below LOQ, cannot support a capability calculation. An ISPE article on maintaining cleaning programs recommends that cleaning data be statistically trended and reviewed by a statistician or cleaning subject matter expert, and that the quality unit conduct and document a formal review of the cleaning program at least annually, covering cleaning-related deviations, CAPAs (corrective and preventive actions), periodic monitoring, and change controls that affect cleaning validation.¹⁵

Where to Start

You do not need to replace every system to get most of the benefit. In our experience, the order of work matters more than the tool.

1. Inventory the data objects. For each shared train, list where the HBELs, limit calculations, recovery factors, surface areas, and worst-case justifications are held, who owns each, and when each was last confirmed.
2. Reconcile the product list. Compare the products now made on each train with the products named in the current worst-case justifications and limit matrices. Every gap is a finding waiting to happen.
3. Fix equipment master data. Confirm surface areas and materials of construction against current drawings and the physical equipment.
4. Write the change triggers into change control. Add explicit questions to the change form for new products, HBEL revisions, dose and batch size changes, equipment changes, detergent changes, and method changes, each naming the data owner who must respond.
5. Structure the results. Store results with location, product pair, recovery factor, LOQ, and limit version, so trending and capability are possible.
6. Then choose the tool. Decide on a spreadsheet, a module, or a dedicated application once you know the size of your matrix and how often it changes.

Conclusion

Cleaning validation has always rested on data: toxicology, doses, batch sizes, surface areas, recovery, and results. What has changed is that regulators now say plainly that the program has to stay current across the product lifecycle, and that the calculations behind it must meet data integrity expectations. PIC/S PI 006-4, in force since 1 October 2026, sets out ongoing verification, change control, and documentation expectations in more detail than before. The 2026 FDA warning letters show the practical failure: a worst-case study that was not re-evaluated when new products arrived on shared equipment, and equipment recorded as clean that was not. Cleaning validation data management, done well, is how a site can show at any moment that its limits, results, and worst-case claims still describe the products it makes today.

Sakara Digital works with pharma and biotech organizations on the data and systems side of quality and validation, including how records like HBELs, limit matrices, and worst-case justifications are structured, owned, and kept current through change control. If you are reviewing your cleaning validation program against PI 006-4 or planning to move cleaning data out of spreadsheets, we are happy to talk through where to start.

For Further Reading