A lean clinical investigation is one where every patient enrolled, every data point collected, and every site opened answers a specific scientific or regulatory question that cannot be answered more cheaply another way. For a MedTech startup, the path is: exhaust literature, equivalence, and harmonised standards first; define the single minimum scientific question the investigation must answer; size the study to the statistical minimum, not the bureaucratic maximum; run it at one or two strategically chosen clinical partner sites under a full ethics committee approval; notify the competent authority where the Regulation requires it; execute to the Good Clinical Practice standard EN ISO 14155:2020+A11:2024; and fold the results into the clinical evaluation per MDR Article 61. Done this way, a startup investigation can cost a fraction of a conventional industry trial without compromising one line of the Regulation.

By Tibor Zechmeister and Felix Lenhard. Last updated 10 April 2026.

TL;DR

  • A clinical investigation is the most capital-intensive single activity most MedTech startups ever run. The most important work happens before the first patient is enrolled.
  • The 600 to 3,000 bench and simulated-use tests a disciplined startup runs before the first patient is not overkill. It is the reason the investigation succeeds instead of failing mid-enrolment on a problem that should have been caught in the lab.
  • Literature review, equivalence under MDCG 2020-5, and the use of harmonised standards can remove entire investigation arms that a less disciplined sponsor would have run.
  • The minimum viable investigation answers one scientific question at one or two sites with the statistical minimum sample size. Every extra arm, site, endpoint, or patient must defend itself against the Regulation and the budget.
  • Good Clinical Practice compliance in the EU means EN ISO 14155:2020+A11:2024. This is not optional and not negotiable. The standard is the operational floor of MDR Articles 62 to 82 and Annex XV.
  • Testing the device on employees, friends, family, or friendly clinicians without full ethics committee approval is not a lean shortcut. It is illegal in every EU jurisdiction and career-ending for the founder and the device.

Why the lean investigation question matters for your startup

Felix has watched founders treat the clinical investigation as the moment the real work begins. That is the wrong frame. By the time the first patient sits down with your device, the investigation has either been set up to succeed or set up to fail, and almost none of what happens in the room between patient and device changes that.

The expensive failures Tibor has seen up close usually share the same story. A startup burns a seed round on a clinical investigation that should never have been run. The scientific question was too broad. The sample size was copied from a pharma template. The endpoints were bureaucratic, not clinical. The sites were chosen by whoever returned the email first. The ethics submission took nine months because the protocol kept changing. The data, when it finally came in, did not answer the question the Notified Body was going to ask.

The disciplined version of the same story is unrecognisable. The same startup spends months before the investigation running somewhere between 600 and 3,000 bench, simulated-use, and usability tests. They recruit one or two clinical partners on day one. Long before any sales conversation. And those partners become co-designers of the protocol, not just sites on a list. They write one scientific question on a whiteboard and let it kill every feature of the protocol that does not serve it. The investigation, when it finally runs, is small, fast, and produces clean data the Notified Body actually wants to see.

That is what lean looks like. The rest of this post is the procedure to get there.

The MDR text. What the Regulation actually says

Clinical investigations under MDR live in Articles 62 to 82 and Annex XV of Regulation (EU) 2017/745. The core article is 62.

"Clinical investigations shall be designed, authorised, conducted, recorded and reported in accordance with the provisions of this Article and Articles 63 to 80, acts adopted pursuant to Article 81, and Annex XV.". Regulation (EU) 2017/745, Article 62, paragraph 1.

Article 62 also sets out the general principles: investigations must be scientifically sound, the rights and safety of subjects must be protected, the risks must be justified by the expected benefits, and the investigation must be designed to produce reliable and robust data for the intended regulatory purpose. Article 63 governs informed consent. Article 70 governs the application procedure and competent authority notification. Article 72 sets out the sponsor's obligations for conducting the investigation. Article 80 requires the sponsor to record and report serious adverse events and device deficiencies that could have led to a serious adverse event, within defined timeframes.

Annex XV specifies, in detail, the content of the clinical investigation plan, the documentation of the device under investigation, the investigator's brochure, and the reporting requirements. And EN ISO 14155:2020+A11:2024. Clinical investigation of medical devices for human subjects. Good clinical practice. Is the harmonised standard that provides the operational framework for compliance with these articles.

None of this changes with the size of the sponsor. A two-person startup has the same obligations as a multinational running a Class III trial. What changes is the scope, not the discipline.

The lean investigation procedure, step by step

The procedure below runs in order. Do not start a later step while an earlier one is still unsettled. Running the steps out of order is the most expensive mistake a startup can make in this domain.

Step 1. Exhaust cheaper alternatives first

Before the word "investigation" is spoken, the clinical evaluation team asks whether an investigation is actually needed for the specific claims in the intended purpose. The MDR recognises three sources of clinical evidence: scientific literature, equivalence to a device already on the market, and new clinical investigation. A lean sponsor treats them in that order of priority.

Run a structured literature review against the specific clinical claims of the device. If the literature for the indication, user population, and mechanism of action supports the claims, a fraction of the intended investigation may disappear.

Evaluate equivalence per MDCG 2020-5. If a technical, biological, and clinical equivalent device is on the market with usable data, the equivalence route may remove entire investigation arms. The same pattern that saved the Graz-based company we described in post 065, where harmonised-standard coverage of established measurement methods cut 1 to 1.5 years and EUR 400,000 to 500,000 out of the plan.

Check whether harmonised standards already address the question. Electrical safety, biocompatibility, usability, and software lifecycle are all domains where recognised standards can carry presumption of conformity without a single patient being enrolled.

What remains after these three filters is the part of the evidence base that genuinely requires new clinical data. That, and only that, is the scope of your investigation. This step alone can change a planned EUR 2 million study into a EUR 200,000 one.

Step 2. Define the minimum scientific question

Write one sentence. It starts with "Does this device, in this patient population, under these conditions of use, achieve this specific clinical outcome compared to this comparator?" If the sentence has more than one primary outcome, cut until it has one. If the comparator is vague, make it specific. If the patient population is broad, narrow it to the exact group the intended purpose describes.

This is the hardest step for founders who love their device. Every feature wants its own endpoint. Every investor wants its own reassurance. Every clinical advisor wants their favourite measurement. A lean investigation resists all of that. The scientific question is what the Notified Body and the competent authority will assess. Everything else is cost.

A good test: if you can write down the one-sentence question, the primary endpoint, and the statistical analysis plan on a single page without hedging, the question is defined. If not, keep cutting.

Step 3. Size the study to the statistical minimum, not the bureaucratic maximum

Sample size has two drivers. One is statistical. The number of subjects required to detect the effect size you expect, at the power you need, with the alpha you can defend. The other is bureaucratic. The number of subjects that someone, somewhere, said "looks like enough for a Notified Body." The first is a calculation. The second is folklore.

A lean sponsor does the calculation. A competent biostatistician, given the expected effect size, the acceptable confidence, and the realistic dropout rate, produces a number. Take that number. Add the minimum justified buffer for dropouts. Stop.

Tibor has seen startups enrol twice the statistically required sample size because a consultant said "more is safer." More is not safer in a budget-constrained startup. More is the difference between finishing the investigation and running out of money before the last subject completes follow-up. If the number from the biostatistician is 42, the number in the protocol is 42 plus dropout buffer, not 80.

The Notified Body is not impressed by over-enrolled studies. They are impressed by well-designed ones that answer the question they are actually being asked.

Step 4. Choose one or two clinical partners strategically

This is where Felix's day-one rule earns its keep. Clinical partners are not vendors you hire when you are ready to enrol. They are co-founders of the clinical strategy, recruited at the start of the company, long before there is anything to sell. Their clinical judgement shapes the protocol. Their patient population defines the inclusion criteria. Their workflow constraints define what is operationally feasible. And, not incidentally, they become the first friendly customers once the device is on the market, because they have lived with it since the beginning.

A lean investigation runs at one or two sites, not ten. Multi-site investigations look impressive on a slide deck. They are operationally brutal for a small team. Every site adds a contract, a local ethics submission, a training process, a monitoring overhead, and a quality signal to watch. One excellent site with a principal investigator who understands the device and the intended purpose is worth more than five mediocre ones.

Strategic choice means: pick the sites whose patient population matches the intended purpose exactly, whose principal investigator has run investigations before, whose ethics committee is familiar with medical device studies (not just drug trials), and whose clinical workflow the device is designed to fit. Everything else is a distraction.

Step 5. Ethics committee approval path

Every clinical investigation under MDR requires ethics committee approval. The details of which committee, under which national procedure, and in which language depend on the member state or states where the investigation runs. This is one of the few parts of the MDR where national law still plays a significant role, and where a local partner or sponsor representative saves weeks of confusion.

The ethics submission is built from the clinical investigation plan required by Annex XV, the investigator's brochure, the informed consent documents required by MDR Article 63, the insurance documentation, the CVs of the principal investigators, and whatever national-specific forms the committee requires. Assembling this takes time. Assembling it badly takes much more time because the committee sends it back with questions.

A lean sponsor does two things. First, engage the ethics committee informally before submission. Most committees will meet with a sponsor to explain what they expect, and a 30-minute conversation prevents months of rework. Second, submit a clean, complete package the first time. Incomplete submissions are the single largest source of delay in clinical investigation projects Tibor has seen.

No ethics approval, no investigation. No exceptions. No testing in advance "just to check." This is the red line in Step 7 below, stated now because it matters at every step.

Step 6. Competent authority notification where required

MDR Article 70 requires the sponsor to submit an application to the competent authority of the member state or states in which the investigation is to be conducted, for certain categories of investigation, before starting. The specific procedure and the scope of authorities that must be notified or must authorise the investigation depend on the device class, whether the investigation is pre-market or post-market, and the purpose of the investigation (conformity assessment, further assessment of a CE-marked device, and so on).

The competent authority route runs in parallel with the ethics committee route. They are two separate approvals with two separate review processes. A lean sponsor submits them in parallel where the national procedure permits, not sequentially.

The competent authority review is substantive. They look at the device safety rationale, the risk management file per EN ISO 14971:2019+A11:2021, the pre-clinical data, the investigator's brochure, and the investigation plan. This is why Step 1 matters. Every pre-clinical test that was run before the investigation becomes the backbone of the safety argument that convinces the competent authority to authorise the study. The 600 to 3,000 tests are not over-engineering. They are the reason the competent authority says yes.

Step 7. GCP-compliant execution via EN ISO 14155:2020+A11:2024

Execution is the part that looks like "the clinical investigation" to outsiders. To the sponsor, it should be the most predictable part of the entire project because every difficult decision was made in Steps 1 to 6.

EN ISO 14155:2020+A11:2024 is the operational rulebook. It covers the responsibilities of the sponsor, the investigator, and the ethics committee; the content of the clinical investigation plan; the informed consent process; the data management and monitoring requirements; the handling of adverse events and device deficiencies; the reporting structure; and the archiving obligations. A lean sponsor follows it line by line. There is no leaner version. The standard is already the floor.

During execution, the sponsor must record and report serious adverse events and certain device deficiencies to the competent authority per MDR Article 80 and to the ethics committee per the national procedure. The timelines are short. The documentation must be real and current. Data is monitored against the plan. Deviations are documented and justified. The investigator, the monitor, and the sponsor each know their roles and stay in them.

AI can help with some of this work. Literature screening, adverse event coding, monitoring dashboards, document drafting. The limit, based on Tibor's experience with Flinn.ai and similar tools, is that the core design of a clinical investigation is not an AI-automation problem. The scientific question, the sample size, the endpoints, the risk-benefit analysis, the protocol deviations, the reporting decisions. These are human regulatory judgements. Use AI to speed up the parts that are mechanical. Do not use AI to replace the parts that require a human who knows the Regulation and the device.

Step 8. Reporting and integration into the clinical evaluation

When the last subject completes follow-up, the investigation is not finished. The clinical investigation report required by MDR Article 77 and the data integration into the clinical evaluation under MDR Article 61 are the steps that convert the study into regulatory evidence.

The report is structured per Annex XV and EN ISO 14155:2020+A11:2024. The results feed directly into the clinical evaluation, which lives in the technical documentation per Annex II. A lean sponsor plans this integration from the start. The endpoints of the study are designed to map one-to-one onto the GSPR claims in the clinical evaluation, so no post-hoc re-framing is needed.

If the results support the claims in the intended purpose, the clinical evaluation is complete for those claims. If the results do not support the claims, the intended purpose must be revised, and the process may need to go back to an earlier step. Cutting claims that the data does not support is not a failure. Shipping a device with claims the data does not support is a failure.

RED LINE. No testing on employees, family, or friendly clinicians without ethics approval

This sits separate from the procedure because it is not a step. It is a boundary.

Tibor has seen founders test prototype devices on themselves, on a co-founder, on a family member, on a friendly doctor's patients "informally," on hospital staff "just to see how it feels." Every one of these crosses the line between research and unauthorised clinical use. Every one of these is illegal in EU member states under national implementations of the Clinical Trials Regulation, the Medical Device Regulation, and general research ethics law. None of these are clever workarounds. All of these are career-ending for the founder, fatal for the device, and. Most importantly. Unsafe for the subjects.

The red line is absolute. If a device is being used on a human for any purpose that generates data about its performance, safety, or usability, that use is either covered by a properly approved clinical investigation with ethics committee approval and the required competent authority notification, or it is illegal. There is no third category. There is no "friends and family" exemption. There is no "early feedback" loophole. There is no "informal pilot" that the competent authority will later accept as legitimate evidence.

Tibor has watched multiple promising companies collapse because a founder thought this rule was negotiable for the right reason. It is not negotiable for any reason. If a lean investigation feels too expensive or too slow, cut scope from Steps 1 through 8. Do not cut the ethics approval. That is the one line that, once crossed, cannot be uncrossed.

The Subtract to Ship angle

The lean investigation is Subtract to Ship applied to the most capital-intensive activity in MedTech development. Every patient enrolled must answer a question the Regulation requires. Every endpoint must map to a GSPR. Every site must contribute evidence no other site can contribute. Every procedure in EN ISO 14155:2020+A11:2024 that applies to the study stays; every procedure that does not apply stays out.

The subtraction is not in the standard. The standard is the floor, not a menu. The subtraction is in the study design upstream of the standard. A well-designed investigation of 50 subjects at 2 sites with 1 primary endpoint costs a fraction of a poorly designed investigation of 200 subjects at 10 sites with 6 primary endpoints, and produces better regulatory evidence. The leaner study is not the worse study. It is the study that answers the question the Notified Body and the competent authority are actually asking.

If you cannot trace a feature of your investigation protocol to a specific question required by MDR Article 62, Annex XV, or the clinical evaluation plan, cut it. If you cannot trace a site, cut it. If you cannot trace a sample size beyond the statistical minimum plus dropout buffer, cut it. What remains is the investigation you can actually run on the runway you actually have.

Reality Check. Where do you stand?

  1. Have you exhausted the literature, equivalence, and harmonised standards routes before deciding that a new investigation is required?
  2. Can you write your scientific question, primary endpoint, and comparator on a single page without hedging?
  3. Do you have a biostatistician-calculated sample size, or a number that was suggested by a consultant or a grant template?
  4. Did you recruit your one or two clinical partners on day one of the company, or are you looking for sites now?
  5. Have you informally engaged the ethics committee before submission, or are you planning to submit cold?
  6. Do you have a current MDR Article 80 adverse event reporting process written down and rehearsed, before the first patient is enrolled?
  7. Is every endpoint in your protocol traceable to a specific claim in the intended purpose and a specific requirement in the clinical evaluation?
  8. Has any testing of the device on humans happened outside the boundaries of an ethics-approved investigation? If yes, stop reading and call a lawyer.

Frequently Asked Questions

How much does a lean clinical investigation actually cost for a MedTech startup? The cost depends on the device class, the indication, the sample size, and the number of sites, but a realistic range for a lean single-site or two-site investigation in Europe runs from a low-six-figure to a mid-six-figure euro budget, excluding internal team costs. Conventional industry trials for the same indication frequently run an order of magnitude higher because they are designed around contract research organisation cost structures, not startup constraints. The lean range assumes Steps 1 to 8 of this procedure are followed and that pre-clinical work was thorough enough to support the safety argument without additional investigation.

Can a startup run a clinical investigation without a contract research organisation? Yes, for small investigations, if the sponsor team has genuine competence in clinical operations, monitoring, data management, and regulatory reporting. The sponsor obligations under MDR Article 72 and EN ISO 14155:2020+A11:2024 do not require a contract research organisation. They require that the obligations are met. For a two-site investigation of 40 to 80 subjects, a disciplined in-house team with experienced advisors can run the study directly. For anything larger or more complex, outsourcing specific functions (monitoring, data management, pharmacovigilance) usually becomes more efficient than building them in-house.

Do I need the investigation to be finished before I submit the technical file? For first-time certification of a device that requires new clinical data per MDR Article 61, yes. The clinical evaluation in the technical file must reflect the results of the investigation, not a planned investigation. The clinical evaluation is not "to be completed later." The Notified Body reviews it as part of the conformity assessment. This is why the clinical evidence strategy is planned at the very start of development, not bolted on at the end.

What if my investigation produces negative results? Then the intended purpose, the claims, or the device must change, and the clinical evaluation must reflect what the data actually shows. Negative results are not a failure of the process. They are the process working. A founder who ships a device with claims the data does not support is a founder who has left the boundaries of the Regulation, regardless of how good the device is. The lean investigation is the honest investigation, and honesty sometimes means rewriting the intended purpose.

Can we use AI tools to design or run the investigation? AI tools can accelerate literature screening, document drafting, adverse event coding, monitoring dashboards, and data management. Tibor's direct experience with Flinn.ai and with the state of the tools in 2026 is that the design of the investigation itself. Scientific question, endpoint selection, sample size justification, risk-benefit analysis, ethics considerations. Is not a task to delegate to AI. The regulatory and clinical judgements require humans who know the device, the Regulation, and the patient population. Use AI for the mechanical parts and keep humans on the judgement parts.

Sources

  1. Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices, Article 62 (general requirements regarding clinical investigations conducted to demonstrate conformity of devices), Article 63 (informed consent), Article 70 (application for clinical investigations), Article 72 (conduct of a clinical investigation), Article 77 (information from the sponsor at the end, temporary halt or early termination of clinical investigations), Article 80 (recording and reporting of adverse events that occur during clinical investigations), Articles 62 to 82 (clinical investigations chapter), Annex XV (clinical investigations). Official Journal L 117, 5.5.2017.
  2. EN ISO 14155:2020+A11:2024. Clinical investigation of medical devices for human subjects. Good clinical practice.
  3. EN ISO 14971:2019+A11:2021. Medical devices. Application of risk management to medical devices.
  4. MDCG 2020-5. Clinical Evaluation. Equivalence: A guide for manufacturers and notified bodies, April 2020.

This post is part of the Clinical Evaluation & Clinical Investigations series in the Subtract to Ship: MDR blog. Authored by Felix Lenhard and Tibor Zechmeister. If you are sizing a clinical investigation against a runway that will not stretch to the conventional industry budget, Zechmeister Strategic Solutions works with founders on exactly this design decision. The one where the scientific question, the Regulation, and the bank balance have to agree on the same protocol.