The conclusion first: you are selecting for provability, not resemblance
Many manufacturers go wrong at the very first step of a wheelchair 510(k): open the public database, find the overseas product that looks closest to their own, note the number and start writing. What actually determines whether the submission clears first time is not how much the candidate resembles your product. It is three things — whether it is still legally marketed, whether its intended use covers the use you want to declare, and whether the technical differences between you and it can be closed with performance data you can actually produce. Resemblance is for people to look at. Provability is what the reviewer reads.
The cost of choosing the wrong predicate is not "revise and resubmit". At best you receive a request for additional information, the review clock stops, and you are asked for a further full round of bench testing. At worst you are found not substantially equivalent, all the preceding test cost and time is sunk, and a fresh submission has to be assembled. Bench test programmes for wheelchairs are not short to begin with, so one round of rework often costs an export season.
What a valid predicate has to satisfy, all at once
The substantial-equivalence determination follows a clear logical chain. Taken apart, a predicate has to clear only three gates.
Gate one: legally marketed. The candidate must be currently legally marketed in the United States and must not have been withdrawn for reasons of safety or effectiveness. This has to be checked against recall records and marketing status — not inferred from the fact that it is still on the manufacturer's website. A product recalled for a safety issue cannot serve as the comparison baseline no matter how well its specifications line up with yours.
Gate two: same intended use. Note that this is intended use, not product name. Two products can both be called "powered wheelchair" while one has indications for everyday indoor and outdoor mobility and the other for rehabilitation use with specific postural support. Those are not the same intended use.
Gate three: technological characteristics. Either the technological characteristics are the same as your product, or the differences do not raise new questions of safety or effectiveness and you can supply performance data demonstrating substantial equivalence nonetheless. The large majority of Chinese manufacturers are on the second route.
These three gates sit in parallel; they are not weighted factors that offset one another. Fail gate one and everything after it is void. Gate three determines workload: every additional difference that needs explaining adds a burden of proof, and there are only two forms that proof can take — existing test data with traceable methods, or new test data. Engineering analysis can help explain why a difference does not matter, but it rarely substitutes for data on its own.
Practical search order: classification first, product code second, individual records last
The correct search order runs counter to most people's instinct.
Step one is to confirm how your product is classified in the United States. Manual wheelchairs, powered wheelchairs and mobility scooters fall under different regulatory routes there, and some lower-risk mechanical products may be exempt and not require a 510(k) at all. Get this wrong and everything downstream is wasted. The specific classification, regulation number and product code must be taken from the current classification database record; any third-party restatement of a code has to be verified back against the source.
Step two is to use the product code to define the candidate pool. Once you have the code, search all historical records under it and sort newest first. Do not lock onto a well-known brand at the outset — there are usually many records under a single code, and among them are products with simpler construction, more complete public information, and far easier alignment than you would expect.
Step three, and only then, is to open records individually. Prioritise records that have a public 510(k) Summary with a detailed device description. Some records carry only a Statement and no Summary, with minimal public information; you cannot extract technical detail usable in a comparison table from those, and choosing one as your predicate leaves you with nothing to write from.
One further point worth noting: in review practice, older records are more likely to attract questions about whether their technology still represents the current state of safety and effectiveness. Formally compliant does not always mean easier to progress. Where candidates are otherwise comparable, prefer the more recent one.
What a public summary tells you — and what it never will
This step is badly underrated and easily misused. A well-drafted 510(k) Summary is normally organised into fixed blocks: applicant and device identification, device description, intended use and indications, an item-by-item comparison with the chosen predicate, non-clinical performance testing and the consensus standards cited, software or clinical statements where applicable, and finally the equivalence conclusion.
Three blocks let you work backwards to the test scope. The first is the non-clinical performance testing section, which states which categories of test were run. The second is the consensus standards list, which lays open the source of the test methods. The third is routinely overlooked: the rows in the comparison table that were singled out for explanation. Those rows are precisely where the review focused at the time, and they are very likely where you will have to concentrate your evidence if you carry comparable differences. Read all three together and you have been handed part of your own future test list in advance.
A de-identified example: in the summary for a folding powered wheelchair, the performance testing section covered strength and fatigue, stability, braking and speed, electromagnetic compatibility, and battery and charger safety, while the comparison table separately expanded on differences in drive layout and seat adjustment mechanism. If you are building a comparable product and happen to differ in those same two respects, you can reasonably predict that your test list will be no shorter than theirs, and that those two differences will need dedicated justification and data.
Be equally clear about what a summary never contains: raw test data and curves, sample counts and sampling method, specific acceptance limits and measured margins, any retest or redesign that followed a failure, material grades and suppliers, moulding process parameters, and the substance of any additional-information exchanges during review. In other words, a summary tells you which directions were tested. It does not tell you to what level, judged how. Copying a summary straight into a test protocol is a trap in the opposite direction.
Two further practical notes. First, a summary reflects the position of that submission at that time; do not assume it still holds today, and re-confirm every consensus standard it cites against the current effective edition. Second, the correct output from reading a summary is not a set of notes but a two-column table: left column "they did it, we have not"; right column "we have this function, they do not". The left column determines the test budget you must add; the right determines the extra justification you must write. That table should exist at project initiation, not when the quotation arrives.
For wheelchairs, the test directions read out of those three blocks usually fall into a few families: bench tests for strength and fatigue, stability-related tests, braking and speed tests (powered products), electrical and electromagnetic compatibility tests (powered products), battery and charger safety evaluation (powered products), and biological evaluation of body-contacting components. Internationally these map mainly onto the ISO 7176 series (wheelchairs) and the ISO 10993 series (biological evaluation). Most Chinese manufacturers already hold reports to the GB/T 18029 series, which corresponds to the ISO 7176 series in test method, but whether the standard number and edition cited and the accreditation scope of the issuing laboratory will be accepted has to be confirmed separately — equivalence cannot be assumed. Specific clauses, loads, cycles and acceptance limits are always governed by the current effective edition of the standard text; any figure written into a protocol from memory has to be checked back against the source.
Intended use: writing it broad digs your own hole
This is the most common self-inflicted wound we see. To give the product "more selling points", manufacturers write the indications very broadly — covering children, covering specific patient populations, covering demanding outdoor terrain, sometimes hinting at rehabilitative function. The results:
- Your intended use exceeds what the predicate covers, and the substantial-equivalence gate fails outright;
- Even if you force-fit a predicate with broader coverage, the technological characteristics you must align and the data you must add expand in step;
- Statements involving special populations can push the product into a higher level of scrutiny and trigger additional requirements.
The pragmatic approach is: narrow the intended use to the minimum viable set that matches what you actually sell, then go looking for a predicate. Being narrower than the predicate is safe; being broader means preparing separate evidence for the excess. This recommendation follows from the logic of the substantial-equivalence determination and from project experience — it is not a statistical finding, but it very rarely points the wrong way.
Which differences "raise new questions"
A difference is not in itself a problem; raising a new question of safety or effectiveness is. The working test is: does the difference change the nature of an existing risk, or only its degree?
Applied to wheelchairs, you can self-check along three lines. First, the energy path: do the drive, braking or battery change how energy is generated, transmitted and dissipated? Introducing regenerative braking, for instance, feeds kinetic energy back into the electrical system, coupling the braking and electrical lines so that previously independent test conclusions can no longer simply be added together. Second, failure modes: does the difference introduce a failure form that did not previously exist? Lift and standing mechanisms bring entrapment, unintended descent and abrupt centre-of-gravity shift — new risks that existing frame strength data cannot cover. Third, human–machine interaction: do the user's method of operation, the reachable controls, or the consequences of incorrect operation change? Remote and mobile-app control turns "unintended actuation" from a local error at arm's length into loss of control outside the user's line of sight.
If any of these three lines is touched, treat it as potentially raising a new question — assume data will be required and use analysis to remove items, rather than assuming none is required and arguing upward. On that measure, common differences fall broadly into two tiers:
- Differences that change degree only: a small change in frame tube wall thickness, a cushion fabric colour change, a minor armrest profile change. These are usually covered by existing strength and stability data plus a comparison statement.
- Differences that may raise new questions: change of drive arrangement (rear-wheel to mid-wheel drive materially changes the stability envelope and steering behaviour), change of battery chemistry, change of controller architecture, addition of standing or lift function, addition of wireless remote or app control.
The second tier has to be answered with measured data, and usually not one test but a set. This is exactly why the predicate has to be fixed at project initiation — it directly determines your test budget.
Predicate candidate screening table
The table below is the screening sheet we actually use at project kick-off. Run every candidate through it row by row; if any row flags, change candidate rather than hoping.
| Check | Question to answer | If it fails |
|---|---|---|
| Legal marketing status | Still legally marketed in the US? Withdrawn or recalled for safety or effectiveness reasons? | Eliminate outright, no workaround |
| Intended use | Do the indications fully cover the population, environment and functions you intend to declare? | Narrow your own intended use, or change candidate |
| Product form | Manual / powered / scooter, drive layout, presence of standing or lift functions — all consistent? | Crossing product form almost always requires a change |
| Key technological characteristics | How large are the differences in control system, battery chemistry, braking method, seat adjustment mechanism? | List every difference and assess whether data is needed |
| Contact materials | Are body-contacting components in the same material category? | Justify separately via the biological evaluation route |
| Completeness of public information | Is there a public summary? Is the device description sufficient to support a comparison table? | Insufficient information makes drafting impractical — change candidate |
| Record age | Is this a reasonably recent record? | Where candidates are close, prefer the newer |
Alignment priorities by product form
| Product form | Differences most often queried | Performance data usually required |
|---|---|---|
| Manual wheelchair | Frame construction, folding mechanism, load limit, wheel configuration | Static and impact strength, fatigue durability, stability |
| Powered wheelchair | Controller and drive layout, battery chemistry, braking logic | Stability, speed and braking performance, EMC, battery and charger safety |
| Mobility scooter | Steering mechanism, three- vs four-wheel differences, freewheel disengagement design | Stability and gradient performance, braking, electrical safety |
| Walkers and crutches | Not in the same comparison family as wheelchairs | Must find their own predicates; cannot be mixed in |
Four places projects come unstuck
One: using the other party's top-of-range product as the baseline. High-specification products often integrate functions you do not have — obstacle detection, multi-axis powered seating. Choose one and you then have to explain why you remain substantially equivalent despite lacking those functions, which increases rather than reduces your burden. A mid- or low-specification product closer to your own feature set is usually far less work.
Two: splitting two predicates. Taking the intended use from A and the technological characteristics from B is not encouraged within the substantial-equivalence framework and invites direct challenge. What is permitted is citing a reference device to support the scientific validity of a method or test protocol — but a reference device is not a predicate, and the role of each has to be stated explicitly in the submission rather than blurred.
Three: substituting an EU report for the US route. EN 12184 is the EU-side standard for powered wheelchairs and scooters; some of its tests share a common origin with the ISO 7176 series and the data has engineering reuse value. But the US route requires a declaration of conformity against the consensus standards it recognises, and the applicable standard, edition and laboratory competence all have to line up. Handing over a CE report as-is generally does not work. The sensible approach is to take the union of both sets of requirements at protocol design stage — one time on the rig, data for both sides.
Four: forcing a predicate that does not exist. If your product genuinely introduces a function with no precedent on the market, repeatedly assembling a predicate will only extend the review. In that situation, assess whether a different marketing route is appropriate rather than persisting inside the 510(k) framework.
What to prepare before testing and before submission
Once the predicate is fixed, the test protocol has a basis. Before testing starts, have these ready:
- The original public records for the candidate predicate (including the summary), and a first draft of your item-by-item comparison table — each difference row needs three columns filled: description of the difference, whether it raises a new question, and which type of data will be used to prove equivalence;
- Full technical specifications for your own product: overall dimensions and mass, load limit, drive and braking configuration, battery and charger specifications, control system description. Specifications must match the final production configuration, not a prototype with ad-hoc changes;
- Material list and supplier documentation for body-contacting components, stating material category, contact location and nature of contact, used to define the scope of biological evaluation;
- Existing domestic test reports (GB/T 18029 series and others), annotated item by item as to whether the test method shares a common origin and whether the issuing laboratory's accreditation scope covers that item — this determines what data can be reused and what must be repeated;
- Prototype configuration list and a statement of configuration differences. Where one platform is sold in several configurations, argue first which unit represents the worst case; otherwise you risk being asked to test every configuration, multiplying cost and schedule;
- The final text of the intended use and indications — once this changes mid-test, the protocol usually has to change with it.
Of these, the last is the most often neglected and the most expensive. Change the indications once and the comparison table has to be rewritten, possibly with the prototype configuration adjusted to match.
What SUNGO Mobility Testing Lab can do
SUNGO Mobility Testing Lab specialises in wheelchairs, mobility scooters, walkers and crutches, and is accredited by CNAS, CMA and IAS (USA), with laboratories in Shanghai and Hefei. To be clear: an accreditation mark demonstrates that the laboratory holds the corresponding technical competence within its accredited scope. It is not a commitment as to the outcome of market access in any target market; whether a submission is accepted and cleared depends on the overall quality of the dossier and the determination of the competent authority.
For US 510(k) projects we can, once the predicate is fixed, help build the test item list, organise bench and whole-product testing to the ISO 7176 series, and coordinate EN 12184 items where an EU submission runs in parallel — aiming for one set of samples and multiple report packages. Biological evaluation is scoped along the ISO 10993 route according to contact components. Specific clauses, loads and acceptance limits are always governed by the current effective edition of the standard text, and we go through them with you item by item at protocol stage.
Test priorities differ substantially by product form. Start with the scope of electric wheelchair testing, manual wheelchair testing and mobility scooter testing, or review the standards we cover on the standards and test basis page.
If you have a 510(k) project running without a predicate settled, or you hold domestic reports and want to assess how much can be reused, send us the product specifications: +86 132 4819 8029, or request a quote online. We will scope the test item list and schedule first, and discuss cost after that.