Set the ground rule: count by where each unit ends up, not by line item
The usual way a submission list gets built is to write out every ordered test, assign one unit to each, and add up the total. That arithmetic tends to overshoot on active devices and undershoot on sterile ones, so it is wrong in both directions. The accurate method comes down to one question: after this unit finishes the current test, can it still go into the next one? If it can, share it. If it cannot, it has to be its own unit.
On that basis every ordered test falls into one of three buckets. Sort first, count second. Once the sorting is done, the quantity is largely settled.
| Disposition type | How to recognize it | Typical tests | Sampling rule |
|---|---|---|---|
| Consumed | The sample is destroyed, contaminated, or irreversibly changed once the test ends | Destructive mechanical methods, destructive seal integrity methods, extraction for biological evaluation | A dedicated unit per condition, per direction, per contacting material; never shared with anything else |
| State-changed | The sample looks intact but has been through processing, so it is no longer in as-manufactured condition | Sterilization, accelerated aging, transport and environmental preconditioning | Can only be passed downstream, never back upstream; get the order wrong and the whole group is repeated |
| Reusable | The test does not alter the sample, so it can move on to something else | Labeling and marking checks, construction and dimensional checks, most functional checks | Several tests may be chained on one unit, but the order has to be written on the request form |
The classification lands on the test request form: reusable items become one serial chain, state-changed items become one directional chain, and consumed items are listed individually. Once those three lines are drawn, the quantity is calculated rather than estimated.
The hard part is sequence, not headcount
The real difficulty in sample planning is not the total but the order. Most requests for additional samples are not caused by a shortage of units; they are caused by a reversed sequence that leaves the units already on site no longer representative.
The one-way chain runs like this. The finished product is built, goes into sterilization, goes into accelerated or real-time aging, then into sterile barrier performance and package integrity checks, and finally into the evaluations that presuppose finished-product condition. Every step down the chain narrows what the sample can still be used for, and no step can be undone.
Three positions on that chain account for most of the mistakes.
The first is scheduling evaluation work ahead of sterilization. Biological evaluation asks for samples in finished-product condition; under the ISO 10993 and GB/T 16886 framework the article being extracted should match the product as marketed, including sterilization method, packaging form and aging condition. Submit an unsterilized sub-assembly or a hand-built prototype and the testing will still run to completion and produce data, but the conclusion cannot be used in the submission dossier, so the round is wasted.
The second is scheduling packaging work ahead of aging while preparing only one set of samples. Under ISO 11607 the sterile barrier system is normally examined in the initial condition and again after aging, and those are two sets of samples rather than one set tested twice. Prepare one set and the aged condition has nothing left to test on.
The third is treating sterilization residual samples as ordinary stock. The residual work associated with ISO 11135 carries expectations about how the sample is stored after it leaves the sterilization batch and when it reaches the laboratory; representativeness decays if the sample sits around. These samples are consumed once tested and cannot return to any other test. For sample preparation notes tied to sterilization method and packaging condition, see sterilization and packaging validation.
Combining parallel tests on a shared unit
Once the chains are laid out, what remains can run in parallel. Parallel scheduling compresses the timeline, but sharing is only allowed when it does not change the starting condition of any test involved.
| Combination | Workable | Precondition | What it costs when it fails |
|---|---|---|---|
| Several non-destructive functional checks chained on one unit | Yes | No residue introduced between tests, assembly state unchanged | Low risk, and this is where most of the reduction comes from |
| Labeling and marking check chained ahead of everything else | Yes | Performed at the start of the chain | Run it after sterilization and the markings may already have been affected by processing |
| Destructive mechanical work sharing a unit with seal integrity | No | None | Whichever runs first destroys the unit, and the second has no carrier left |
| Biological evaluation sharing extraction samples with other consumed tests | No | None | Extraction conditions differ, and mixing them makes both data sets unusable |
| Electrical safety sharing one unit with EMC on an active device | Conditional | Irreversible safety items scheduled after EMC, or a separate complete unit provided | A unit damaged during safety work sends EMC back into the queue with a fresh submission |
The active device row deserves its own paragraph. Under IEC 60601-1 a subset of the work has irreversible effects on the unit, for instance the items involving enclosure strength, insulation and single fault condition simulation. After those, the unit is no longer in normal operating condition, while EMC testing requires the equipment under test to be in a declared normal operating mode that can be reproduced. The customary arrangement for active devices is therefore a separate unit for electrical safety and another for EMC, plus a decision on whether an extra engineering unit is worth having for pre-screening. Squeezing both onto one unit saves the cost of one unit; losing that bet costs the schedule of the entire project. Confirm the unit demand of the two test families separately before shipping instead of reporting one complete device and leaving it at that.
Biological evaluation gets counted on its own
Sample quantity for biological evaluation does not follow the number of ordered tests. It follows three variables: contact category, the number of distinct contacting materials, and extraction conditions. Change any one of them and another set of samples is required.
The common underestimate shows up on multi-material products. One device may have an enclosure that contacts skin, a connector that contacts mucosal tissue, and tubing that contacts body fluid. Those three cannot be merged into a single extraction, and they cannot be prorated against the assembled device either. Build a materials list cross-referenced against contact locations before placing the order and the count becomes reliable. For evaluation routes and sample preparation, start with biocompatibility evaluation.
The other underestimate comes from color and formulation differences. Whether products that share a structure but differ in masterbatch or formulation batch can be merged depends on whether the difference rationale holds up. If it does not, each needs its own set. That belongs to coverage rationale, which is a separate subject from the sampling logic here; the coverage material in the knowledge center deals with it, while this article stays on sampling.
Spares, retained samples and repeat-test units
Spares are not optional, they are part of the schedule. Three situations consume them directly: damage in transit, a mismatch between what arrives and what the request form describes, and an anomaly during testing that has to be reproduced.
- Transit damage. Sterile barrier products, glass and precision components take the worst of it, so the outer packaging has to survive ordinary handling.
- Description mismatch. If sample identification does not match the request form, the laboratory cannot start work, and the wait for replacements still eats schedule.
- Reproduction. When a borderline result appears, whether a same-batch unit is on hand to reproduce it immediately decides when the report can be issued.
Retained samples belong in the count as well. Not everything comes back after testing. Consumed units cannot be returned, and state-changed units, once returned, can no longer be treated as as-manufactured stock. Making this clear internally before the order goes out avoids awkward asset reconciliation later.
Failure modes and what each one costs
| What went wrong | Where it surfaces | Actual cost |
|---|---|---|
| One unit reported per ordered line item | Mid-test | Not enough units for destructive work, testing halts while replacements are shipped |
| Unsterilized sub-assembly used for finished-product-condition tests | When the report is used | Data cannot support the dossier, the whole round is repeated |
| Only an initial-condition set prepared for packaging work | After aging finishes | Nothing left to test in the aged condition, another aging cycle is needed and the timeline doubles |
| Electrical safety and EMC sharing one active device unit | After safety work ends | Unit damaged, EMC goes back into the queue |
| No spares held | Day of arrival | Any single discrepancy stops everything, with no fallback |
| Sequence never written on the request form | Before testing starts | The laboratory follows the customary order, and when that differs from expectations there is no way to trace responsibility |
The costs in that table share one property. Changed at the planning stage they cost nothing; changed once units are on site they cost a reschedule; changed after the report is issued they cost another full round. The whole value of sample planning is moving the decision forward into the planning stage.
Self-check before the samples ship
- Every ordered test has been tagged as consumed, state-changed or reusable.
- The order of the one-way chain is written on the request form and does not depend on a verbal agreement.
- For tests that require finished-product condition, the sterilization and packaging status of the samples has been confirmed.
- For tests that require both initial and aged conditions, two separate sets have been prepared.
- The material versus contact location cross-reference is complete, and multi-material products have not been counted as one assembled device.
- Separate units for electrical safety and EMC have been prepared for active devices, or the conditions for sharing one have been stated.
- Spares have been prepared against the vulnerable points rather than added as a rough proportion of the total.
- Sample identification matches the request form item by item, including how the model designation is written and the batch information.
- The physical submission format and packaging expectations have been checked against the sample submission requirements.
Work through that list and the quantity table can be finalized. Remaining deviations usually come from design changes to the product itself, which sit on a different control line.
Have someone lay the sample table out with you
If what you have is a multi-material, multi-configuration product with both a sterile aspect and an electrical one, the sample table rarely comes out right on the first pass. Send us the product structure, the materials list, the target tests and your planned milestones, and we will return a proposed quantity and sequence using the logic above, marking which tests need dedicated units, which can be chained, and where spares should be prepared in advance. SUNGO Lab (Shanghai Shage Medical Technology Co., Ltd.) is a medical device testing laboratory accredited by CNAS, CMA and IAS (USA), with laboratories in Shanghai and Hefei. Call +86 132 4819 8029 or request a quote, and settle at the planning stage what belongs at the planning stage.