Prototyping a diagnostic kit: lab validation steps before you seek a licensee
Diagnostic kit validation before licensing: analytical and clinical tests, stability and lot checks, CDSCO test licence forms, and a worked sensitivity example.
Diagnostic kit validation before licensing means showing, with data, that a prototype detects what it claims, across lots and over its shelf life. Cover analytical sensitivity, specificity, precision and stability, then clinical sensitivity and specificity against a reference method. In India, making kits for testing needs a CDSCO test licence, applied for on Form MD-12 and issued on Form MD-13.
A diagnostic assay that works on ten samples in the inventor's hands is a result. A kit that works on hundreds of samples, in other people's hands, from three different lots, after six months on a shelf, is a product a company might license. The gap between the two is validation, and it is where academic diagnostic projects can stall.
This guide sets out the lab validation steps to complete before you approach a licensee, the regulatory forms that apply, and a worked example of the core calculations.
What does a licensee want to see?
A company scout reads a kit's data looking for risk. Each missing piece is a cost they will have to bear.
| Question the licensee asks | Evidence that answers it |
|---|---|
| What is the lowest level it detects? | Analytical sensitivity (limit of detection) study |
| Does it react to the wrong things? | Cross-reactivity and interference studies |
| Does it give the same answer every time? | Repeatability and reproducibility across runs, days and operators |
| Can it be made consistently? | At least two or three lots made and compared |
| How long does it last? | Real-time and accelerated stability data at the intended storage temperature |
| Does it work on patients? | Clinical sensitivity and specificity against a reference method |
| Where do reagents come from? | Source and supply of antibodies, antigens, enzymes or primers |
Our post on diagnostic kit technologies to license shows the same checklist from the buyer's side.
Step 1: analytical performance
Analytical work tells you what the kit can do under controlled conditions.
| Study | What you do | What you report |
|---|---|---|
| Limit of detection | Test serial dilutions of the target near the expected limit, many replicates at each | The lowest level detected reliably |
| Cross-reactivity | Test related organisms or analytes that could give false positives | Which, if any, react |
| Interference | Spike samples with common interferents for the sample type | Effect on results |
| Precision | Repeat tests within a run, across days and across operators | Agreement between results |
Write the protocol and acceptance criteria before you start, and keep them signed and dated in the validation file. A study designed after seeing the data persuades nobody.
Step 2: lots and stability
A single lot proves the protocol once. Make at least two or three lots, ideally with different reagent batches, and test them side by side on the same panel. The CDSCO overview cited in our TRL for medical devices post notes that test batches should generate quality, stability and validation data, with at least three batches.
For stability, run accelerated studies at raised temperature for an early estimate, and start real-time studies at the intended storage temperature at the same time. Real-time data is what supports the shelf life on the label. If the kit includes a freeze-dried component, test it after the drying step, in final packaging.
Validating a diagnostic kit prototype
Step 3: clinical performance
Test the kit on real patient samples against an accepted reference method, with the operator blind to the reference result. Use samples that represent the people the kit will be used on, including weak positives and samples from people with related conditions.
Research use of patient samples needs institutional ethics committee approval. Where samples or materials go to a collaborator, put a material transfer agreement in place first.
How should you document it?
A licensee's technical team will read your validation file before your pitch deck. Keep it in one place, in a fixed order.
| Document | Contents |
|---|---|
| Intended use statement | What the kit detects, in which sample, for whom, by whom |
| Protocols | Each study's design and acceptance criteria, signed and dated before testing |
| Raw data | Instrument outputs and lab notebook pages |
| Reports | One per study, with results against criteria |
| Lot records | Reagent sources and batch numbers for each lot |
| Stability records | Time points, storage conditions, results |
| Approvals | Ethics committee approval, MTAs, test licence if held |
Write the intended use statement first. It decides the sample type, the reference method, the clinical panel and, under the Medical Devices Rules, the risk class. Changing it later means repeating studies, so agree it with your clinical collaborators and any target licensee before the first protocol is signed.
What does the regulator require?
Under the Medical Devices Rules, 2017, in vitro diagnostics are classed by risk from Class A (low) to Class D (high). The forms, as listed on CDSCO's IVD page:
| Step | Application | Licence or permission |
|---|---|---|
| Test licence, to make kits for testing and evaluation | MD-12 | MD-13 |
| Clinical performance evaluation of a new IVD | MD-24 | MD-25 |
| Manufacturing licence, Class A or B (state authority) | MD-3 | MD-5 |
| Manufacturing licence, Class C or D (CDSCO) | MD-7 | MD-9 |
A research lab validating a prototype for licensing usually needs the test licence if it makes kits for testing and evaluation. The manufacturing licence is the licensee's job.
Inventor's lab or licensee's plant
The more of the left column you finish, the stronger your licence terms.
A worked example
Illustration with invented results. A team tests its rapid kit on 300 samples against a reference laboratory method.
| Reference positive | Reference negative | Total | |
|---|---|---|---|
| Kit positive | 92 | 6 | 98 |
| Kit negative | 8 | 194 | 202 |
| Total | 100 | 200 | 300 |
| Measure | Calculation | Result |
|---|---|---|
| Clinical sensitivity | 92 ÷ 100 | 92% |
| Clinical specificity | 194 ÷ 200 | 97% |
| Positive predictive value in this panel | 92 ÷ 98 | 93.9% |
| Negative predictive value in this panel | 194 ÷ 202 | 96.0% |
Predictive values depend on how common the condition is in the panel, so report them with that caveat. A licensee will also ask for the confidence interval around each figure; with 100 positives, the uncertainty around 92% is wide, which is a reason to test more samples before negotiating.
Budget, also invented: analytical studies ₹3,00,000; three lots ₹2,50,000; stability over 12 months ₹1,50,000; clinical panel and independent testing ₹4,00,000. Total ₹11,00,000.
Common mistakes
- Validating a single lot.
- Testing only strong positives.
- No written protocol or acceptance criteria before testing.
- Relying on accelerated stability alone for the label claim.
- Using patient samples without ethics approval or an MTA.
My view: lot-to-lot data is the piece academic teams often skip, and it is the piece that tells a licensee whether the kit can be manufactured. I am less certain how many clinical samples a particular licensee will expect, since it depends on the analyte and the risk class; ask a target licensee early what panel size would satisfy them.
If your team needs help with prototyping or lab validation work, Zernyx Labs does prototyping and lab R&D for biotech, pharma and cosmetics.
Common questions
What should a diagnostic kit prototype show before a company will license it?
Data on the lowest level it detects, cross-reactivity and interference, repeatability across runs and operators, at least two or three lots made and compared, stability at the intended storage temperature, and performance on real clinical samples against a reference method.
How are clinical sensitivity and specificity calculated?
Sensitivity is true positives divided by all samples the reference method calls positive. Specificity is true negatives divided by all samples the reference calls negative. If a kit detects 92 of 100 reference-positive samples, its sensitivity is 92%.
Do I need a licence to make diagnostic kits for testing in India?
Yes. Under the Medical Devices Rules, 2017, making an in vitro diagnostic for testing and evaluation needs a test licence from CDSCO, applied for on Form MD-12 and issued on Form MD-13. Clinical performance evaluation of a new IVD needs permission on Form MD-25, applied for on Form MD-24.
Should validation be done in-house or independently?
Both, ideally. Do the analytical work in-house to learn the kit's limits, then have performance checked by an independent laboratory or clinical site. Licensees and regulators are likely to give more weight to independent data.