We get the same question from diagnostics teams every few weeks. Some version of: we can build a hundred cartridges a week in R&D, how do we get to a hundred thousand a month without the assay falling apart? It is a fair question, and it is rarely the one that keeps the team up at night once they dig in. The harder problem is not throughput. It is holding assay performance steady while everything around it changes: the deposition equipment, the reagent lots, the operators, the room, and the quality system that now has to prove every unit is fit to release.
Diagnostic consumables are unforgiving that way. A lateral flow strip or a microfluidic cartridge is a small stack of tightly coupled variables. Membrane, conjugate, striping position, drying profile, housing tolerance, bond integrity. Shift one and the control line moves, the flow front stalls, or the cartridge leaks. R&D can hand-tune around variability. A commercial line cannot, because the whole point of commercial production is that unit 1,000,000 behaves like unit 1.
This article walks the full lifecycle a contract manufacturer sees: the prototype-to-production handoff, the cleanroom decision, the process validations that actually matter for consumables, lot release, and the scale-up traps that show up specifically in lateral flow and microfluidics. It also maps that work onto the quality framework that governs it in the US as of 2026, which changed in a way that matters more than most teams realize.
The regulatory floor moved on February 2, 2026
Start here, because the ground under IVD manufacturing shifted this year. Effective February 2, 2026, the FDA's Quality Management System Regulation (QMSR) amended 21 CFR Part 820 to incorporate ISO 13485:2016 by reference, and the agency retired the Quality System Inspection Technique (QSIT) it had used to structure inspections for two decades (FDA QMSR Final Rule).
What that means in practice: the operative language for finished-device acceptance, process validation, and contamination control no longer lives in the old Part 820 subpart prose. It lives in ISO 13485:2016 clauses. If you or your CM are still citing the retired 820.30 design controls or 820.70 environmental-control text as current requirements, the file is out of date. The requirements did not disappear. They flow through named ISO clauses now, and an investigator is going to expect your quality system, and your CM's, to speak that language.
One caveat worth stating plainly, because diagnostics teams sell into more than one market. The QMSR governs US FDA jurisdiction. It does not by itself govern EU IVDR conformity. If your commercial plan includes Europe, IVDR runs in parallel and your CM's ISO 13485 system is the shared backbone, not a substitute. Keep the two straight in your documentation from the start.
The prototype-to-production handoff is where timelines live or die
Most of the delay in getting a consumable to market is not manufacturing. It is the handoff into manufacturing. The engineering literature has mapped this chain well. A peer-reviewed Biomicrofluidics perspective walks the microfluidic cartridge path from concept and lab prototyping through commercial scale-up, using commercialized molecular diagnostics platforms as reference points (Biomicrofluidics 16(2):021301). What that body of work consistently leaves open is the quality-system layer a CM has to bolt onto the engineering. That gap is exactly where projects stall.
Under ISO 13485:2016, design and development transfer is not a formality. Clause 7.3 requires documented procedures that verify design outputs are suitable for manufacturing before they become production specifications, and that production is actually capable of meeting product requirements (ISO 13485:2016). Read that as: you cannot freeze a spec on a bench-built unit and assume the line will hit it. You have to prove the line hits it.
For a diagnostic consumable, three transfer failures show up over and over. Reagent lot variability that the bench never saw because R&D used one heroic lot of antibody. Material substitutions made for supply or cost reasons that quietly change surface energy or bond chemistry. And tooling scale differences, where a prototype fixture behaves nothing like a production mold or a reel-fed striping head. None of these are exotic. All of them will move an assay result if they go uncontrolled. The transfer package has to name them, bound them, and test them before the spec locks.
Honestly, the single best thing a diagnostics team can do to protect its timeline is bring the CM in before design freeze, not after. A specification written with the production process in view is a specification the line can actually hold.
Choosing the cleanroom: risk first, class second
Cleanroom class is the question teams ask first and should decide almost last, because the class follows from the contamination risk, not the other way around. ISO 14644-1:2015 classifies cleanroom air cleanliness by airborne particle concentration on a scale from ISO Class 1 to ISO Class 9, and ISO Class 7 and ISO Class 8 are the classes most commonly specified for medical device manufacturing (ISO 14644-1:2015).
That "most common" is a usage pattern, not a rule you can inherit. The class an IVD consumable actually needs depends on what the product cannot tolerate. Protein and reagent handling, membrane striping, and conjugate drying each carry their own contamination and cross-contamination concerns that a generic ISO 8 assumption may not address. Particulate on a nitrocellulose membrane can nucleate a false line. Aerosolized reagent from one striping station can cross into another. The class has to be justified by the product's risk profile, and that justification has to be written down.
The QMS obligation behind the cleanroom sits in ISO 13485:2016 Clause 6.4. It requires the organization to document the work-environment requirements needed to achieve product conformity, and where personnel contact with product could adversely affect quality, to establish requirements for health, cleanliness, and clothing (ISO 13485:2016). Clause 6.4 sets the contamination-control duty. It does not name a cleanroom class. You select the class under the ISO 14644 series, justify it against the product, and then Clause 6.4 is where you prove the environment is controlled and monitored to that decision.
One more thing that trips teams up. A cleanroom class is a specification for a state, and it is not a single state. Classification is verified at defined occupancy conditions, and a room that passes at rest can drift under a full crew running reel-fed equipment. Environmental monitoring is not a certificate you hang on the wall once. It is an ongoing data stream that has to survive an audit, and it is a topic that earns its own deep-dive in this hub.
Process validation is the difference between an assay and a product
Here is the clause diagnostics teams underweight. ISO 13485:2016 Clause 7.5.6 requires validation of any production process whose output cannot be, or is not, verified by later monitoring or measurement, including defined criteria for review and approval of the process, equipment qualification, and provisions for revalidation (ISO 13485:2016).
Read that against a lateral flow line and the special processes jump out. You cannot fully verify a conjugate drying step by inspecting the dried pad, because the thing you care about, the released and functional conjugate at time of use, is not something you can see downstream. Same for membrane striping, where line position and dispense volume drive assay performance you will only fully confirm at readout. Same for cartridge bonding, where a weak or partial bond may pass visual inspection and fail under flow. These are the processes that demand IQ, OQ, and PQ, because you are validating the process to guarantee the output, not inspecting the output to catch the process.
The clause tells you that validation is required. It does not, and cannot, hand you the acceptance ranges. Nobody publishes the OQ window for your specific conjugate on your specific membrane. Those ranges come out of process characterization, ideally with a designed experiment behind them, and they become the criteria your CM holds the line to lot after lot. A CM that validates conjugate drying to a characterized humidity and temperature window is protecting your control-line intensity three months into commercial production, not just at first article.
This is also where the engineering literature and the quality system have to meet. A peer-reviewed study demonstrated high-throughput roll-to-roll manufacturing, using UV nanoimprint lithography and extrusion coating as scalable alternatives to injection molding for foil-based microfluidic lateral flow chips, exceeding 200 chips per minute (PMC12024825). That is a real throughput story, and it is specific: foil-based microfluidic chips on a particular fabrication route. It does not automatically transfer to nitrocellulose strip assays or to injection-molded cartridges, and the study addresses fabrication throughput, not lot release or reagent QC. The lesson is not "go roll-to-roll." The lesson is that a faster process is only a better process once it is validated and once its output is provably in spec. Speed without Clause 7.5.6 behind it is just faster scrap.
Lot release: the panel the standard makes you invent
Product release is where the quality system stops being paperwork and starts being a gate. ISO 13485:2016 Clause 8.2.6 requires the organization to define and implement receiving, in-process, and final acceptance activities, and it does not allow product release until the planned arrangements are satisfactorily completed, with records identifying the person who authorized the release (ISO 13485:2016).
That is the frame. The clause deliberately does not tell you what to test, because the release panel for a diagnostic consumable is yours to design. For a lateral flow product that panel typically reaches past dimensional and visual checks into functional performance: reagent and conjugate QC, control-line and test-line intensity against a defined standard, and flow-front timing that confirms the assay wicks and develops the way the design intends. For a microfluidic cartridge it extends to bond and leak integrity and fluidic performance under the actual run conditions. The specifics belong to the product. The discipline of defining them, justifying them, and gating release on them belongs to the standard.
Lot release for diagnostics carries a second burden that general device manufacturing does not feel as sharply: lot-to-lot consistency is the product. A diagnostic that drifts at the variability margin fails in the field even when every individual lot passed a loose spec. The release panel has to be tight enough to catch drift before it ships, which means it has to be built on characterized limits, not convenient ones. Lot release testing for diagnostics is deep enough that it gets its own spoke in this hub, because the panel design is where a lot of commercial quality problems are quietly won or lost.
Scale-up traps that are specific to diagnostics
The pitfalls that bite IVD consumables are rarely the ones a generic manufacturing playbook warns about. A few that show up again and again.
Reagent lot variability outruns process control. R&D optimizes on a favorite antibody lot. Commercial buys antibody the way commercial buys everything, in lots, from suppliers, with real lot-to-lot spread. If the assay was tuned so tight that only the hero lot works, the line will chase reagents forever. The fix is incoming reagent qualification and an assay design with enough margin to absorb normal biological variability, decided during transfer, not discovered during launch.
Drying and environment stop being invisible. At bench scale, a conjugate pad dries in ambient lab air and nobody logs it. At line scale, humidity swings across a shift change the release rate of your conjugate, and suddenly control-line intensity has a day-of-week pattern. This is why conjugate drying is a validated special process under Clause 7.5.6, not an unmonitored step.
The housing and the assay were designed by different people. A lateral flow strip that reads perfectly on an open bench can behave differently once it is snapped into an injection-molded housing that compresses the membrane stack or changes the wicking path. Cartridge bonding introduces the same class of problem in microfluidics, where a tolerance stack-up that was invisible in a machined prototype becomes a leak path in a molded part. Design transfer under Clause 7.3 is supposed to catch this, which is another reason the housing and the chemistry need to be validated together, not in sequence.
Automated inspection has to be qualified, not assumed. Moving from a technician eyeballing strips to an automated vision system is a genuine scale-up win, but the inspection method is itself a process that needs qualification. An unqualified vision check that passes marginal control lines is worse than no automation, because it launders bad units as good ones with a data trail that looks clean.
Cost-out gets attempted at the worst time. The pressure to reduce cost per unit is real and it usually arrives right at commercial launch, exactly when the process is least stable. Material substitutions and process changes made for cost in that window are change controls that touch assay performance, and they belong under the same validation and transfer discipline as the original design. Cost-down for lateral flow consumables is worth doing deliberately, and it is its own topic in this hub, because doing it fast and doing it right are not the same project.
Bridging the study phase into commercial production
Many diagnostics consumables are built at meaningful volume before commercial clearance, to support clinical and analytical studies. That phase has its own regulatory logic worth planning around. FDA guidance explains that certain investigational IVD studies are exempt from most Investigational Device Exemption requirements under 21 CFR 812.2(c)(3) when specified conditions are met, such as noninvasive sampling and a non-significant-risk determination with the required consent and labeling conditions (FDA IVD Device Studies FAQ).
That exemption governs the study, not the factory. It says nothing about cleanroom class, lot release, or scale-up. The trap teams fall into is treating study-phase builds as throwaway and then rebuilding the process from scratch for commercial. The better move is to make study-phase builds a genuine dress rehearsal for commercial: same CM, same environment decisions, same evolving controls, so the process that generated your clinical data is recognizably the process that ships. Before a team goes off blazing a new trail for the commercial line, the tried-and-true path is to make the study line and the commercial line the same line, matured.
What to look for in a CM's quality system
When a diagnostics team evaluates a contract manufacturer for consumables, the quality system is not a checkbox behind the capability. It is the capability. A few concrete things to press on.
Ask how the CM handles the QMSR transition in its own documentation. The right answer references ISO 13485:2016 clauses, not retired Part 820 subpart language, and can explain that the QMSR incorporates ISO 13485:2016 by reference as of February 2, 2026 (FDA QMSR Final Rule). A CM still describing its system in old QSR terms is telling you something.
Ask how it decides cleanroom class for an IVD consumable, and listen for a risk justification tied to the specific product rather than a default ISO 8 answer. Ask which of your consumable's steps it treats as validated special processes under Clause 7.5.6, and whether it characterizes those processes with designed experiments before setting acceptance ranges. Ask how it builds a lot release panel under Clause 8.2.6 for a functional assay, and who authorizes release. And ask how it manages design transfer under Clause 7.3, specifically how it handles reagent lot variability and housing-to-assay interaction, because those are the two failure modes that quietly wreck consumable launches.
The reason to keep assembly, cleanroom environment, and functional testing under one roof is not tidiness. It is that every handoff between a fabrication step, an assembly step, and a test step is a place for variability to enter and for traceability to fray. Diagnostics fail at the variability margin, so the fewer seams in the process, the better the odds that unit 1,000,000 behaves like unit 1.
That is the whole job of a diagnostics consumable CM, stated plainly. Take an assay that works on a bench and prove, unit after unit and lot after lot, that it still works at commercial volume, inside a quality system an FDA investigator will recognize on sight.