This post was motivated by my experiences advising startups, participating in acceleration programs, and working with academic research groups. Over the years, I’ve seen the challenges of translating biomarker discoveries into impactful diagnostic products, and I hope to share some lessons learned to help others navigate this complex but rewarding journey.
Note on AI usage: I used Perplexity.AI to obtain most of the data cited, and ProwritingAid and Perplexity.AI to polish the text. Otherwise the content is mine.
Biomarkers and IVDs
In vitro diagnostics (IVDs) are essential tools in modern healthcare. They support or influence up to 70% of medical decisions, but they account for less than 2% of total healthcare expenditure1. IVDs play a critical role informing modern medical decisions, from the humblest blood sugar measurement to the most advanced multi-omics test. IVDs are also the foundation of the emerging paradigms such as Precision Medicine, Functional Medicine, Longevity Medicine, etc. as they enable personalised and cost-efficient prevention and treatment strategies. Because of this, the IVD market has experienced and is predicted to continue experiencing significant growth, with projections estimating its value to increase from $90.4 billion currently to $146.5 billion by 2035, driven by advancements in molecular diagnostics, next-generation sequencing (NGS), and artificial intelligence2.
IVDs rely on biomarkers—molecular signatures that arise from research into disease mechanisms or translational efforts aimed at identifying new drug targets or stratification tools. Biomarkers are perhaps the most common biomedical research outputs, now increasingly driven by developments and investments in high throughput data generation in genomics, proteomics, and metabolomics. A quick search of Pubmed shows that in 2024, 82,610 papers were published mentioning biomarkers, raising from 66,936 ten years before and 28,526 in 2004. For an indication on the significance of this type of research, the global biomarker research services market is expected to grow from $10.98 billion in 2025 to $310.66 billion by 2032, a compound annual growth rate (CAGR) of 61%3!
Developing IVDs is a high risk adventure
Taking biomarkers to market as IVD products typically takes 3 to 7 years and costs between $20 million and $100 million4, depending on the complexity of the diagnostic and regulatory pathway (e.g., 510(k) vs. PMA in the U.S., IVDR in the EU). Even longer timelines and higher costs are required for an IVD to become profitable, estimates suggesting it can take up to 10 years and additional investments in sales, marketing, and evidence generation to achieve market adoption5.
Given these high costs, long timelines, and low success rates—less than 2% of biomarkers identified in research ever make it to clinical use6—IVDs are often perceived as a risky investment. This perception is reflected in venture capital trends: diagnostics companies receive only a fraction of healthcare venture capital funding. For instance, molecular diagnostics represented just 3% of healthcare VC investments in 2004, and while this has improved slightly over the years, diagnostics still trail behind therapeutics in attracting funding7.
Why I’m Sharing These Insights and Who This Is For
I would like to see more biomarker research translating into products that can improve healthcare, but this is far more challenging than most scientists think. I have walked this path in my career, from academic research to entrepreneurship, giving me a unique perspective on this process:
- As an academic researcher, I took part in biomarker discovery projects—publishing papers, identifying novel biomarkers, and contributing to the understanding of disease mechanisms, but most importantly, exploring multiple paths for biomarker discovery.
- Later, I founded and ran a startup offering oncogenomics testing services based on liquid biopsies and tumour sequencing, giving me an understanding on the side of the clinical laboratory as well as of the medical community interaction with Dx innovation.
- Now, as the founder and CEO of Ophiomics, I am developing a portfolio of IVD-based products for liver diseases, which took me on a journey to product development in the Dx space (in vitro diagnostics medical devices to use the formal designation). While Ophiomics’ products incorporate algorithmic components (which I’ll discuss in another post), this post focuses solely on some lessons I learnt from the IVD development process.
So I’m writing this post for scientists contemplating a leap into entrepreneurship—or tech transfer offices and investors supporting such ventures—hoping that my experiences provide useful insights into what it takes to turn biomarker discoveries into impactful diagnostic products.
Three Major Uncertainties in IVD Development
Since investment will inevitably be required, and raising investment and driving adoption is all about risk-reduction, I am structuring this post around the concept of risks and strategies for risk reduction.
I believe that there are three major classes of uncertainty in an IVD development journey:
1- Technical Uncertainty – can we distinguish clinical meaningful subsets of patients?
2 – Clinical uncertainty – does it change and improve patient management?
3- Market uncertainty – will the clinical deciders, laboratory users, payers, health technology assessment organizations, patients, etc accept and adopt the product?
So risk mitigation will involve reducing and ideally eliminating these uncertainties. I share below some lessons learnt while building Ophiomics.
1. Technical Uncertainty: From Discovery to Validation
I’ll start with the first uncertainty that needs to be addressed, that we can identify or distinguish meaningful groups and distinguish them with some sort of assay performed on a biological sample. For scientists like my co-founder Joana Cardoso and myself, this is the thrilling phase—exploring biomarkers in biological samples and leveraging public datasets. We’ve been doing this for a while relying heavily on bioinformatics, but at Ophiomics, we’ve streamlined this process to make it as efficient and predictable as a discovery journey can be. Soon we learnt that data and samples were key and invested in building our own internal data and sample warehouse. In the past few years though, we came to realise that while single biomarkers or single-class biomarkers are still feasible the future lies in integrating clinical variables with multi-omics data (e.g., genomics, proteomics) and AI-driven interpretation. Our approach now centers on building our own liver-centric “data lake” to bring together and organise complex datasets (LiverSphere). But while exciting, this phase is probably the least relevant of the entire journey. Amongst scientist there is a perception that discovery is king but in fact, we just have to read the literature to find the vast majority of brilliant discoveries that went nowhere. Many, solving unmet medical needs that only exist in the imagination of scientist that never bother to study the relevant patient journey. There is also a perception among scientists that there is a single, magical set of biomarkers that alone can solve the clinical problem. However, what the literature tell us is that completely unrelated sets of biomarkers can generate equally effective products for the same clinical decision—breast cancer diagnostics exemplify this with multiple gene expression signatures offering comparable utility for the same clinical decision. So in fact what counts is execution, and the most important part of this initial step is proving that the test works reliably.
This phase is less glamorous but critical: demonstrating that the assay performs consistently across reagents, operators, and real-world conditions (e.g., sample storage, transport). During HepatoPredict’s development, we faced costly iterations due to unstable reagents and cross-reactivity. Moving from a business model that assumed a central laboratory to a kit-based assay made it even more challenging, with Ophiomics having to become a certified manufacturer of medical devices, and raising the technical validation demands associated with market authorisations (regulatory hurdles, discussed below). But it also motivated us to understand our assays better and develop them into in robust, re-usable platforms (HYSEN Liquid for ctDNA analysis from liquid biopsies and NOLO Express for analyses of FFPE tumour tissues: check Ophiomics Technology) if curious). This enables us to have now reduced validation timelines and costs, allowing us to develop a portfolio of new products focusing on clinical validations rather than technical troubleshooting.
Lessons Learned:
- Validate the Medical Need Early: Partner with clinicians during discovery to ground your research in real-world problems.
- Execution Beats Discovery: Focus on delivering actionable results rather than chasing perfection.
- Drive Actionable Outcomes: Stratifying patients into groups with identical clinical decisions is futile.
- Design for Regional Business Models: Build assay flexibility early to accommodate regional preferences (e.g., centralized lab models in North America/GCC vs. kit-based models in Europe).
- Align Technology with Use Cases: Choose technologies compatible with clinical workflows and cost considerations (e.g., rapid turnaround for emergency settings).
- Logistics Costs Bite Hard: Optimize packaging and supply chain logistics during development—shipping costs can exceed product costs if not accounted for early.
2. Clinical Uncertainty
Clinical validation is what matters most to clinicians: Does the test provide real clinical benefit? There’s no shortcut here—clinical performance must be demonstrated through clinical validation studies, and outcomes cannot be inferred. This principle applies broadly to medical technologies but became starkly clear to us during HepatoPredict’s development. For example, while tumour biology (assessed by HepatoPredict) should remain consistent across surgical approaches, validating it in living donor liver transplantation requires entirely new studies despite prior validation in deceased donor cohorts. Clinical studies demand navigating ethical approvals (which can take years, even for retrospective analyses), negotiating with stakeholders (who often lack your urgency), and managing spiralling costs – in fact, this is where most of the money is going to disappear. This makes this stage perhaps the one most important to get right from the onset, and my opinion is that understanding the nature of the evidence that is required is critical. However, unlike drug development, which follows strict phased trials (Phase I–IV), biomarker validation lacks a universally standardized framework. I suggest a simple, general scheme, but make sure to investigate how these apply to the clinical condition you are studying :
Level III – Evidence Type: retrospective (Analysis of archived samples (e.g., biobanks). Fast, cheap, but low impact.)
Level II – Evidence Type: Prospective-Retrospective studies (Pre-planned analysis of samples from completed trials. Moderate rigour.)
Level I – Evidence Type: Prospective (Dedicated trials with predefined endpoints. High rigour but time-intensive.)
Biomarkers research often begin at Level III (retrospective) but must advance to Level II/I for clinical adoption. In oncology, Level I evidence typically requires 5+ years of follow-up—and adding transplantation extends timelines further due to patient recruitment challenges. Clinicians Ignore Low-Level Evidence. A biomarker validated only retrospectively (Level III) will gather dust. Clinicians and guidelines (e.g., NCCN, ESMO) demand at least Level II evidence for adoption.
So, the burden falls on the team to identify early, clinically relevant outcomes that can drive real-world evidence generation. Without these interim milestones, few investors are likely to stay the course. For HepatoPredict it was relatively simple to start obtaining Level III evidence from multiple centres, but the small number of liver transplantations performed worldwide (less than 40K!) means that prospective evidence is not easy to come by. We are fortunate to have been invited to join a large European consortium that aims to reassess organ (liver) allocation approaches, to propose new European Guidelines, the LEOPARD Consortium. LEOPARD enables us to test HepatoPredict in a large prospectively collect cohort, supported by an EU grant, but it will be years before we see the final result. In addition, we initiated a prospective clinical trial with the same surgical team involved in the development of HepatoPredict, which will also take many years to complete. Note that for a start up, time is your worst enemy, and running a trial for more than 7 years the worst nightmare of entrepreneurs and investors alike. So we had to complement these studies by identifying other uses for HepatoPredict where relevant clinical validation could be obtained faster so that we can approach clinical use and initial market adoption earlier. So as part of our R&D approach, in developing our pipeline we are prioritising level II and I evidence as early as possible and for HepatoDetect (liquid biopsy for early detection of liver cancer), we skipped Level III entirely, embedding prospectively collected samples from the start.
Lessons learned :
7. Time and Cost Will Exceed Expectations. Really! I mean it!!
8. Prioritise Higher Level of Evidence Early (Level II/I). Don’t over-invest in retrospective validation (Level IIII).
3. Market Uncertainty
Market uncertainty is the final hurdle for new IVD products. You may develop the best diagnostic test in the world, but when you bring it to market, it might not gain traction – it was all for nothing. Pricing could be misaligned with market expectations, the format might be inadequate for clinical workflows, regulatory costs and time may exceed your runway, or reimbursement challenges could stall adoption. By engaging potential users early for HepatoPredict, we’ve gained invaluable insights into objections and hesitations that guided our validation strategies, opened up doors to new collaborative studies and cohorts, and helped us learn what business model made sense for several territories, in that way guiding the development in terms of kit format and start us preparing for technology transfer possibilities. I believe that prioritising early market feedback to guide product development and validation strategies is strategically sound, but it will be even more powerful if you take is as a true “commercial” effort: everybody will tell you that your assay is a good idea and the collaborative study you propose and pay for is wonderful, until you ask them to actually use your product in patient management, or even better, to pay for it – only then will you truly know if you are on a path to market adoption, and what you need to know to get there. But note, if you start your “commercial efforts” early to gain information, make sure to manage expectations in your investors and in your team – you are not yet at the point of making any money!
To reduce time to market and refine our products, we established an ISO 13485-certified pilot manufacturing facility, which allows us to prototype quickly and efficiently. This certification ensured compliance with stringent quality management standards, facilitating global early market entry even using a RUO (Research Use Only) product for that specific market.
Since our products all have an algorithmic component based on Machine Learning, we needed to have a means to give users access to it globally, independently of clinical and computational environment. We opted for a cloud-deployment of our internally-developed, user-directed MyOphiomics. It became a versatile digital platform that serves to manage user access, post-market surveillance, algorithm control and protection(!) and managing distributors and commercial relationships. It took time to develop but now deploying a new algorithm in a country-specific cloud is (mostly) trivial.
Regulatory approval is often the “boogeyman” for those unfamiliar with the process. However, in my experience, it becomes manageable if you work with experienced regulatory experts (internal or subcontracted) and dedicate sufficient time and budget to navigate complex frameworks like IVDR in Europe. We have been lucky, despite early non-productive local relationships, but this is where operating out of a country in which there is no critical mass of companies developing related products, in a continent heavy in regulations, makes it harder to think globally from the onset. We are now working with a mix of local and international advisors to understand global regulatory frameworks and taylor our go-to-market and regulatory strategy accordingly.
Finally, reimbursement remains one of the most challenging aspects of market adoption. Without clear health economic data demonstrating cost-effectiveness, even validated tests can struggle to gain traction. For HepatoPredict, we are working on a comprehensive health economics study using cohorts from the LEOPARD Consortium. Independent validation cohorts must be sufficiently large to support reimbursement applications—a process that demands careful planning and additional time if cohorts are unavailable at earlier stages.
Lessons Learned:
- Engage the Market Early, i.e. interact with potential users and clients as soon as possible to gather feedback that shapes product development.
- Invest in manufacture, your own or subcontracted. ISO certifications (e.g., ISO 13485) are not just formalities; they are often unofficial requirements for regulatory compliance.
- Seek Real-world evidence generation as early as possible, i.e. people actually using your product in a clinical workflow, even as a RUO, as it can provide early insights while preparing for full regulatory clearance and even support it.
- Engage international regulatory advisors early.
- Plan for Reimbursement Early. Health economic studies require independent validation cohorts and robust data collection—start planning these from day one to avoid delays later.
Final Thoughts
If you’re working on biomarker-based diagnostics, considering entering this field as an entrepreneur or investor, or supporting such ventures as part of a tech transfer office, incubator, or acceleration program—and would like feedback on your strategy—I’d be happy to connect! Feel free to reach out via LinkedIn or through this contact form.
Data Sources
1 MedTech Europe: Value of Diagnostic Information; PubMed: The Value of In Vitro Diagnostic Testing in Medical Practice
2 MarketsandMarkets: IVD Market Report; Coherent Market Insights: Biomarker Research Services Market
3 PMC: Why Your New Cancer Biomarker May Never Work; Academic Entrepreneurship: Developing In Vitro Diagnostics for Commercialization.
4 Premier Research: Navigating IVDR for Diagnostics Development.
5 BioCentury: Healthcare VC Trends.
6 Greenlight Guru: IVD Devices Guide.
7 PMC: A Business Model for Diagnostic Startups.; NIST: Companion Diagnostics for Personalized Medicine.