We get the same question from diagnostics teams almost every week: our assay works on the bench, so what actually breaks when we try to make 100,000 of them a month? The honest answer is that nothing about the chemistry breaks. What breaks is everything around it. The environment the device is built in, the processes you can no longer eyeball for pass or fail, the incoming membrane lot that shifted flow rate by 8 percent, and the paper trail that has to satisfy an FDA investigator who wasn't in the room when you designed the thing.
This is the arc that separates a working prototype from a commercial consumable, and it's the part the published literature mostly skips. There's plenty of good academic work on cleanroom-free prototyping of microfluidic devices, benchtop methods like xurography and screen-printing that get you a first article fast. There's plenty of FDA material defining what an IVD is and how it's classified. What sits in between, the operational work of turning a bench process into a validated, repeatable, auditable manufacturing line, doesn't come packaged anywhere. So let's package it.
The stakes are real and the market knows it. The global IVD contract manufacturing market was estimated at USD 21.1 billion in 2024 and is projected to reach roughly USD 37.45 billion by 2030, with reagents and consumables the largest product segment (Grand View Research, 2025). Companies are outsourcing consumable production because scaling it well is hard, specialized, and capital-heavy. Before your team blazes a new trail here, it's worth walking the tried-and-true path first.
The Framework You're Now Building Inside
Start with the ground rules, because they changed recently and a lot of scale-up plans still reference the old ones. As of February 2, 2026, the FDA Quality Management System Regulation (QMSR) took effect, amending 21 CFR Part 820 to incorporate ISO 13485:2016 by reference (FDA, QMSR final rule). Most of the legacy Quality System Regulation subpart prose that engineers memorized, the old design-control and process-control paragraph numbers, was removed. The requirements didn't disappear. They now flow through the ISO 13485:2016 clauses the regulation points to.
This matters for a scale-up because every artifact you build, the validation protocols, the acceptance records, the design-transfer package, now maps to a clause in ISO 13485 rather than a CFR subpart. When your contract manufacturer talks about process validation, they mean Clause 7.5.6. When they talk about your device file, they mean Clause 7.3.10. Getting the vocabulary right isn't pedantry. It's what keeps your submission and your quality agreement aligned with what an investigator actually asks for.
One note on scope. FDA classifies IVDs as reagents, instruments, and systems intended for use in the diagnosis of disease in specimens taken from the body, and routes them through premarket pathways by device class, up to premarket approval for Class III (FDA, Overview of IVD Regulation). The QMSR framework here is US-market. If you're also chasing EU IVDR, you carry separate QMS obligations even though they lean on the same ISO standard. Plan for both from the start rather than retrofitting.
Cleanroom Class: Match the Environment to the Failure Mode
The first scale-up decision that costs real money is where the product gets built. Cleanroom air cleanliness is classified under ISO 14644-1:2015 by the maximum permitted concentration of airborne particles per cubic metre at specified sizes, from ISO Class 1 (cleanest) through ISO Class 9, and classification requires a defined number of sampling locations and a defined occupancy state: as-built, at-rest, or operational (ISO 14644-1:2015). That last detail trips people up. A room that hits ISO Class 7 empty at rest is a different animal once you put ten operators, a laminating line, and in-process materials in it. Classify in the state you'll actually run in.
Here's the part the standard deliberately doesn't do: it won't tell you which class your product needs. ISO 14644-1 defines the measurement, not the requirement. The class is driven by your product's contamination sensitivity and your process risk assessment, and that assessment is where diagnostics gets interesting.
A large portion of consumables are more sensitive to particulate and humidity than to viable organisms. A stray fiber landing across a test line is a functional defect. Ambient humidity swings during conjugate drying change sensitivity lot to lot. For a lot of lateral flow work, a well-controlled ISO Class 8 environment with tight humidity control does more for yield than chasing a lower particle number. Microfluidic cartridges shift the calculus. Bonded channel features in the tens-of-microns range mean a single particle can occlude a channel or spoil a bond seam, which pushes critical assembly and bonding steps toward ISO Class 7 or cleaner.
Whatever class you land on, the QMSR framework obligates you to control it. ISO 13485:2016 Clause 6.4 requires establishing and maintaining the work environment needed to achieve product conformity, including documented requirements for cleanliness, personnel health and clothing where those factors could affect quality, and arrangements to control contaminated or potentially contaminated product (ISO 13485:2016 Clause 6.4). The clause states the obligation. It leaves the specific class and monitoring regime to your risk determination. That's freedom and responsibility in the same sentence: you get to choose, and you have to defend the choice with a documented rationale and environmental monitoring data that backs it.
One practical warning from moving programs off the bench. Cleanroom-free prototyping is genuinely useful early, and the fabrication literature is right that you can get far with plotter cutters and makerspace tools. But a geometry that works when a grad student cuts it by hand does not automatically survive a classified-environment production line running at volume. The transition from benchtop to controlled manufacturing is exactly where format assumptions get retested, and it's cheaper to retest them on purpose than to discover them in a failed PQ. We'll go deeper on this in a dedicated piece on cleanroom strategy for IVD manufacturing.
Process Validation: When You Can't Inspect Your Way to Confidence
Process validation is required under ISO 13485:2016 Clause 7.5.6 where the output of a process cannot be, or is not, verified by subsequent monitoring or measurement. You establish documented validation procedures including criteria for review and approval, equipment qualification, and revalidation (ISO 13485:2016 Clause 7.5.6). Read that trigger carefully, because it's the whole game in diagnostics.
Ask of every process step: can I fully verify the output of this step by measuring the finished device? For a lot of what makes a diagnostic consumable work, the answer is no. You cannot non-destructively confirm every critical technical requirement. You cannot inspect a heat-sealed barrier film for leak integrity on 100 percent of units without destroying them. When you can't verify downstream, you validate the process. That's the line, and it's where your validation budget goes.
The standard states the requirement and stops there. It does not prescribe IQ/OQ/PQ specifics for lateral flow or microfluidics, because those approaches come from industry practice and format-specific characterization, not from standard text. So here's how the practice tends to shape up for the two dominant formats.
For lateral flow, the high-leverage processes are reagent dispensing (line placement, volume, and concentration), drying and conditioning, lamination and registration, and slitting and cutting. Installation qualification confirms the dispensing and lamination equipment is installed and operating to spec. Operational qualification maps the process window: dispense volumes, line speeds, drying temperature and humidity, run at the edges of their ranges to find where the assay still performs. Performance qualification then proves the process holds at target across multiple lots, ideally with different operators and material lots deliberately included so you learn your real-world variability instead of your best-day variability. The assay chemistry is well characterized in the literature, membrane, conjugate pad, sample pad, absorbent pad, test and control lines (NCBI, PMC3312431). What's not in the literature is how you keep line placement and conjugate activity constant across a million units. That's the validation work.
Microfluidic cartridge assembly concentrates its risk in bonding and in feature fidelity. Bonding is the step where defects tend to cluster, in our experience, because bond strength depends on tightly coupled variables: surface prep, alignment, temperature, pressure, dwell time. Small drifts stack. Your OQ has to find the interactions, not just the single-variable limits, which is where design of experiments earns its keep. We'll treat this in depth in a dedicated piece on microfluidic cartridge assembly at scale, but the headline is that a bonding process you validate loosely is a bonding process that fails you at the variability margin, which is exactly where diagnostics fails.
Don't validate a process you're still redesigning. The most expensive validation is the one you run twice because the process wasn't frozen. Get the process locked, characterized, and stable, then validate it. Revalidation triggers, a new material supplier, an equipment change, a facility move, belong in the procedure from day one so a routine change six months post-launch doesn't quietly invalidate your PQ.
Lot Release: Deciding What "Good" Means Before You Ship It
Every lot you ship has to earn its release, and the QMSR framework is specific about the machinery. ISO 13485:2016 requires documented acceptance activities across receiving, in-process, and final acceptance, and product must not be released until planned acceptance activities are satisfactorily completed and release is authorized and recorded, including the identity of the person authorizing release (ISO 13485:2016 Clause 8.2.6 / 8.2.4). The framework is clear. What it deliberately doesn't do is tell you which tests belong on your lot-release panel. Test selection is product- and format-specific, and building that panel well is one of the higher-stakes decisions in a scale-up.
The trap teams fall into is treating lot release as a downstream inspection problem. It isn't. It's the mirror image of your process validation decision. Every function you couldn't verify at the unit level, and therefore validated the process for, still needs a lot-level statistical check to prove the validated process stayed in control for this specific lot. And every function you can measure directly needs an acceptance criterion tied to your design requirements, not to whatever looked reasonable during development.
A lot-release panel commonly reaches for functional assay performance against characterized reference panels (sensitivity and specificity at defined analyte concentrations), line intensity and background, flow characteristics, and stability-indicating checks tied to your shelf-life claim. For microfluidic cartridges, add leak and flow integrity, fill-volume accuracy, and any on-board reagent functionality. The specifics belong in a focused treatment, and we'll build one out on lot release testing for diagnostics. The principle that carries across both: your acceptance criteria are only as credible as the design requirements they trace to, which is why lot release cannot be designed in isolation from the DHF.
A word on sampling. Lot release is a statistical argument, not a vibe. The sampling plan and its acceptance number are part of the record, and "we test some units and they look fine" is not a plan an investigator accepts. Decide your sampling basis, tie it to risk, and document the rationale. If a test fails, the record says the test failed, and your nonconformance process takes it from there. Genericizing a failure into softer language in the record is how small problems become audit findings.
Supply Chain: Where Scale-Up Actually Fails
Assays don't usually fail at scale because the design was wrong. They fail because a raw material moved. Nitrocellulose membrane lots vary in flow rate and protein binding. Antibody lots vary in activity. A conjugate that behaved on one gold lot behaves differently on the next. At R&D volumes you absorb this by hand-selecting materials. At commercial volumes, hand-selection isn't a strategy, it's a liability.
The QMSR framework puts the obligation squarely on purchasing controls. ISO 13485:2016 Clause 7.4 requires documented purchasing procedures ensuring purchased product conforms to specified requirements, evaluation and selection of suppliers based on their ability to meet requirements, and records of supplier evaluation and any resulting actions, with control proportionate to the risk of the purchased product (ISO 13485:2016 Clause 7.4). "Proportionate to risk" is the operative phrase. Your critical antibody deserves a deeper qualification and tighter incoming acceptance than your outer carton.
The standard governs the controls. It does not hand you IVD-specific scale-up tactics, because those are industry practice, not standard text. So here are the ones that matter most for consumables.
Qualify materials by function, not just by certificate of analysis. A membrane that meets its spec sheet can still shift your assay if your assay is sensitive to a parameter the spec sheet doesn't control. Build incoming acceptance that tests the material the way your process actually stresses it. Where a material is both critical and single-sourced, treat dual-sourcing as a program, not a wish, because qualifying a second source is itself a validation exercise with lead time measured in months. And lock material specifications tightly enough that a supplier can't quietly reformulate under you, which happens more than anyone likes.
Lot reservation and forward-buying of critical biological materials deserve a mention. When a single antibody lot can carry months of production, the supply-chain decision and the lot-to-lot consistency decision become the same decision. Diagnostics lives or dies at the variability margin, and the raw material is where variability is born. Cost-down work belongs here too, but it belongs after stability, not before it. We'll take that up separately in a piece on cost-down strategies for lateral flow consumables. Chase COGS before your material chain is stable and you'll trade a few cents per unit for a yield problem that costs far more.
The Interface: How a CM's QMS Meets Your Design History File
Here's the question that decides whether a CM transfer goes smoothly or turns into a nine-month paperwork reconciliation: whose records live where, and how do they connect?
Start with the transfer obligation. ISO 13485:2016 Clause 7.3.8 requires that design and development outputs be transferred to production in a manner that ensures manufacturing capability meets design requirements, and Clause 7.3.10 requires the manufacturer to maintain a design and development file for each device type or family, containing or referencing the records generated across the design process (ISO 13485:2016 Clause 7.3.8 and 7.3.10). Clause 7.3.10 is the DHF concept FDA reviewers know by name. The DHF is typically among the first records an investigator asks to see, so its coherence matters well beyond the transfer event.
The part the standard doesn't spell out is the mechanics between two separate quality systems. You, the OEM, own the design and the DHF. Your contract manufacturer runs its own ISO 13485 QMS and generates its own records: validation protocols and reports, device history records for each lot, equipment qualification, environmental monitoring, nonconformance and CAPA. Those records don't automatically fold into your DHF. The bridge between the two systems is the quality agreement, and a vague one is where transfers go to die.
A quality agreement that actually works spells out which party owns each record, which CM-generated records are referenced into the OEM's design file versus retained at the CM and made available on request, how design changes get communicated and re-approved on both sides, how nonconformances and CAPAs are shared, and who has release authority and audit rights. When it's written well, your DHF references the CM's validation and device history records cleanly, the traceability holds under audit, and design transfer is a controlled handoff rather than a scramble. When it's vague, you find out during an inspection that a record you assumed was in your file lives in a system you can't produce on demand.
The practical move for a diagnostics team planning a transfer is to treat the quality agreement as a design input, not an afterthought signed at contract close. Map your DHF structure to the CM's record structure before the first validation lot runs. Decide together, in writing, which of the CM's records satisfy which of your design-transfer requirements. Do that early and the ISO 13485 QMS interface becomes an asset: independently maintained records that hold up in your submission and your audits without you rebuilding them. Do it late and you inherit a reconciliation project.
Sequencing the Whole Thing
If there's one theme across cleanroom, validation, lot release, supply chain, and the DHF interface, it's that they aren't a checklist run in parallel. They're a dependency chain. Your environmental class flows from your process risk assessment. Your validation scope flows from what you can't verify downstream. Your lot-release panel mirrors your validation decisions. Your supply-chain controls determine whether your validated process stays valid. And your DHF interface has to capture all of it in records that trace back to design requirements.
Get the sequence right and each stage feeds the next. Get it wrong, validate before the process is frozen, set acceptance criteria before the design requirements are locked, chase cost before the material chain is stable, and you pay for the same work twice. Diagnostics scale-up rewards teams that respect the order of operations.
One last note on the regulatory environment you're scaling into. FDA has been active on IVD validation expectations, including January 2025 draft guidance on validating certain IVDs for emerging pathogens under a Section 564 declared emergency (FDA, Federal Register 2025-01-07). That guidance is narrowly scoped to emergency use for novel pathogens and doesn't govern routine commercial scale-up. But it's a reminder that the agency's expectations for IVD validation keep sharpening, and a manufacturing program built on the current QMSR framework and current ISO editions is far easier to flex when they do.