Lost in Translation: Which biomedical fields struggle to translate?

At any life sciences conference, on any given day, you will find rooms full of genuinely good ideas. Researchers presenting years of careful work. Posters that describe real effects in real biological systems. Data that, taken at face value, suggests something useful might one day happen for patients. The atmosphere is energetic, the science is often impressive, and the implicit promise hanging over the whole event is that some of this will eventually matter in the clinic.

Most of it will not, and that gap between promise and outcome is one of the most fascinating and underappreciated realities in modern science. Not because the research is bad, but because the journey from a laboratory finding to an approved clinical product is one of the most fiercely attritional processes in any industry. Understanding where the losses happen, why they happen, and which research fields give scientists the best realistic chance of seeing their work reach a patient within a career is worth thinking about carefully, whether you are a researcher choosing a direction, a funder allocating resources, or simply someone curious about how medicine actually gets made.

The Scale of the Problem: A Numbers Game

The biomedical literature has grown at a rate that makes the scale of the upstream pipeline almost impossible to visualise. PubMed, a database of medical research literature, now contains over 33 million indexed papers, and more than one million new biomedical and life science articles are published every year. A 2019 analysis of PubMed records found that around 18% of research articles are ever cited by a clinical trial or clinical guideline, even when assessed over a fifteen-year window. For papers classified as basic science (molecular and cellular, with no direct human relevance), that figure drops to approximately 4 to 5% [8].

A separate study tracking over 2.4 million publications across 18 medical specialties found that only 0.4 to 2.4% of published studies are ever cited in a point-of-care clinical resource, and in 9 of the 18 specialties examined, fewer than 1 in 10 clinical trials were ever cited at all [8].

Put another way: of every 100 papers published in the biomedical literature, perhaps 4 to 18 will ever meaningfully influence a clinical trial. The rest will contribute to the accumulated body of knowledge, inform other researchers, and in many cases represent genuinely good science that never finds a pathway forward.

The figures in this table are drawn from multiple analyses of pharmaceutical development pipelines [4, 7]. They do not include the pre-Phase I attrition that reduces the pool of basic science findings to clinical candidates in the first place, which is where the largest absolute losses occur.

The Time Problem: Why Good Ideas Take Decades

Survival rate is only part of the story. Duration matters almost as much. A frequently cited analysis published in the Journal of the Royal Society of Medicine established that the average time from a research discovery to its adoption into clinical practice is approximately 17 years [1]. More recent analyses of FDA-approved drugs from 1999 to 2013 put the median lag from target discovery to regulatory approval at 22 years [3]. For some disease areas, including certain cancer biomarkers, the lag from the growth of supporting scientific literature to drug approval has been measured at over 30 years [2].

These timelines have several practical consequences. They mean that most of the researchers who generate a foundational discovery never see its clinical application within their active careers. They mean that the market landscape, competitive environment, and clinical need that motivated the original research may have changed fundamentally by the time a product approaches approval. And they mean that the compounding cost of capital over two or more decades adds enormously to the financial burden of translation, independently of the direct costs of development.

Those direct costs are themselves considerable. Estimates of the average cost to bring a new drug to market range from approximately $1 billion to over $2.5 billion depending on therapeutic area, methodology, and whether capital costs of failed programmes are included [5]. A 2024 Deloitte analysis found that major pharmaceutical companies spent an average of $2.23 billion per approved asset, with their total spending on subsequently terminated programmes amounting to $7.7 billion in that year alone [6]. Every approved product carries within its price the financial weight of the many candidates that did not make it.

The Patent Clock

There is a further consequence of these timelines that rarely appears in discussions of translation but shapes commercial decision-making profoundly: the collision between development duration and patent life. A standard patent has a term of 20 years from the filing date. Filing typically happens early, often around the time of the foundational discovery, to establish priority. By the time a drug has cleared preclinical development, Phase I, Phase II, and Phase III trials, and obtained regulatory approval, a decade or more of that patent term has already elapsed. For programmes at the longer end of the development timeline, the effective period of market exclusivity, the window during which the developer can recoup its investment before generic or biosimilar competition arrives, may be as short as five to eight years. This compression creates a structural incentive to pursue development only in areas where peak sales can be achieved quickly, and to abandon programmes in areas where the commercial window is too narrow to justify the remaining investment. It is one of the less visible reasons why some scientifically promising technologies, particularly those targeting chronic conditions or complex diseases with long trial requirements, never find a commercial developer willing to carry them through.

The First Valley: From Discovery to Development Candidate

The journey begins in a laboratory. A researcher identifies a mechanism, synthesises a material, or discovers that a particular biological entity has an effect of interest. The first challenge is one that is rarely acknowledged in academic settings: most good scientific observations are not the same thing as a product concept.

A hydrogel that supports chondrocyte survival in a well plate experiment demonstrates that a material is not acutely cytotoxic and can provide a substrate for cell attachment. It does not demonstrate that the material can be manufactured at scale, sterilised without losing its properties, stored with an acceptable shelf life, injected or implanted through a clinically realistic procedure, and retain its performance in the complex and mechanically demanding environment of a living joint. These are not small gaps. They are each a potential endpoint for the programme.

The structural problem at this stage is a mismatch between the incentives of academic science and the requirements of product development. Academic metrics reward novelty, publication, and citation. They do not reward negative results, which are often the most informative experiments for translational purposes. A researcher who rigorously establishes that a promising material fails at scale, degrades unpredictably under physiological conditions, or triggers an unexpected immune response has contributed enormously to the field, but this contribution is difficult to publish and impossible to fund through most academic grant mechanisms.

Universities and grant bodies have been aware of this problem long enough that several structural interventions now exist. NIH snapshot data tells its own story: only 3.9% of publications produced by basic research awards are translational by any reasonable measure. Applied research awards do roughly twice as well at 7.4%, and specifically targeted translational awards from the Clinical and Translational Science Award programme reach 13.4% [9]. The implication is that the mechanism of funding matters enormously, not just the quality of the science being funded.

Intellectual Property: The Invisible Barrier

The translation of a research finding into a commercial product almost always requires intellectual property protection, yet the management of IP in academic institutions is frequently a source of delay, confusion, and failed partnerships rather than a facilitator of translation.

University technology transfer offices exist to bridge this gap, but their effectiveness varies enormously. Patent applications filed too broadly provide weak protection; those filed too narrowly leave commercial freedom-to-operate open for competitors. Timing matters critically: a publication that pre-dates a patent filing can destroy novelty and render a patent invalid in most jurisdictions. Researchers under pressure to publish, which is to say virtually all academic researchers, can unknowingly close the IP window before it has been formally opened.

Biomaterials present particular IP challenges because the relationship between composition and function is often complex and not fully characterised at the time of filing. A scaffold that works by a combination of mechanical properties, surface chemistry, and degradation profile may be difficult to protect with a composition-of-matter claim if the active components interact in ways that were not anticipated when the patent was written.

The Preclinical Problem: When Animals Do Not Predict

For technologies that survive the IP and early feasibility stages, the next major attrition point is preclinical validation. Animal models are the dominant tool for this work, and their limitations are well-established. Translational success rates from animal studies to clinical outcomes vary across published studies from close to zero to around 70%, depending on the therapeutic area and the quality of the preclinical programme. The variation is itself informative: there is no reliable rule of thumb for how well a mouse, rat, rabbit, or pig model predicts human response.

For biomaterials specifically, an additional complication is that the immune response to an implanted material varies substantially between species. Macrophage polarisation, foreign body response, and fibrotic encapsulation behave differently in rodents and humans in ways that are not always predictable from the preclinical data. A scaffold that integrates cleanly in a rat subcutaneous model may provoke a chronic inflammatory response in a primate or a human, and this difference is often not discovered until a first-in-human study.

The failure of cartilage tissue engineering to translate from the laboratory to routine clinical use illustrates these dynamics with unusual clarity. Hundreds of scaffold formulations, growth factor combinations, and cell seeding strategies have produced impressive results in animal models over the past three decades. Very few have demonstrated durable efficacy in clinical trials, and of the cell-based articular cartilage products available globally, none are approved for severe osteoarthritis and none for rheumatoid arthritis [15]. The biological complexity of load-bearing cartilage in a human joint environment, and the challenges of achieving integration with host tissue in the presence of synovial inflammation, simply were not captured by the animal models in which the technologies were developed.

The Clinical Trial: Where Most of the Losses Happen

For the small fraction of technologies that enter clinical development, the clinical trial is where the majority of financial attrition occurs. The statistics are unambiguous: around 90% of all drugs that enter clinical trials do not receive approval. The most lethal stage, consistently, is the transition from Phase II to Phase III. Over the decade from 2014 to 2023, only 28% of programmes that entered Phase II successfully completed it and advanced [4]. The overall likelihood of approval for a drug entering Phase I has now fallen to approximately 6.7%, an all-time low [4].

The single most common reason for Phase II and Phase III failure is lack of efficacy [7]. This is, on the surface, a scientific failure, but it frequently has structural antecedents: the endpoint chosen did not capture the biological effect adequately; the patient population was too heterogeneous to detect a signal in the enrolled sample size; the dose was not sufficient because toxicology constraints had set a ceiling below the therapeutic threshold; or the preclinical model had been optimistic in a way that was not discovered until the programme was too far advanced to redesign.

A category that is sometimes underappreciated in discussions of translational failure is commercial discontinuation: programmes that are stopped not because they have failed scientifically but because the company has recalculated the market opportunity, a competitor has reached the market first, or the cost of completing the remaining trials exceeds what the payer environment will support in terms of eventual reimbursement. This is not a scientific problem, but it terminates scientifically sound programmes with considerable regularity.

How Many Projects Does It Take? A Comparison by Modality

The headline attrition figures above describe an average across all drug types and disease areas. The reality is considerably more varied. The rate at which a Phase I candidate reaches approval differs by a factor of nearly ten between the most and least favourable areas [7], and this variation has profound implications for how research investment is allocated and how likely a given technology is to survive.

The table and figure below show approximate Phase I to approval success rates, expressed as the number of Phase I projects required to produce one approved product, across the major modalities and disease areas for which published data exist.

Vaccines for infectious disease and ophthalmology drugs share the highest Phase I to approval rates, at approximately 33% each [7]. Vaccines benefit from endpoints that are objective and measurable, from animal models that are reasonably predictive for many pathogens, from manufacturing platforms refined over decades, and from regulatory environments actively incentivised to move effective products forward quickly. Ophthalmology drugs benefit from local delivery that reduces systemic safety risk, from objective endpoints such as visual acuity and optical coherence tomography imaging, and from the relative accessibility of the target tissue.

Both areas also benefit from something easy to underestimate: a well-established clinical infrastructure. Centres that regularly run vaccine or ophthalmology trials have the patient populations, investigator expertise, outcome measures, and regulatory relationships in place to conduct efficient trials. This infrastructure effect, which compounds over time as a field matures, is one of the structural reasons why some areas consistently outperform others on translation metrics [14].

Oncology sits in an interesting split. Conventional small molecule oncology is in the low zone, requiring up to 29 Phase I programmes per approval. But targeted and biomarker-driven oncology performs considerably better [7]. The oncology field has improved its translation rate substantially over the past decade almost entirely by becoming more selective about which patients drugs are tested in. A researcher working on immune checkpoint biology or a predictive biomarker is in a structurally different position to one working on broad cytotoxic mechanisms.

Research Impact by Field: Which Area Is Most Likely to Produce a Clinical Product?

The modality table above describes what happens once a technology has entered clinical trials. But the more personal question for most researchers is earlier and more fundamental: given that I work in this field, what is the realistic chance that any of my published work will contribute to an approved clinical product within my career?

This question requires combining two probabilities: the chance that a given paper in a given field enters the commercial pipeline at all (most simply do not), and the downstream clinical success rate once it does. The best available proxy for the first stage is patent citation: research that is cited in a patent has entered the IP pipeline and has at least some chance of reaching development. Patent citation rates vary dramatically by field, from above 12% in biotechnology, virology, and allergy/immunology, to below 2% in medicine and general surgery [10, 11].

Multiplying patent entry probability by downstream clinical success rates produces an order-of-magnitude estimate of the per-paper probability of contributing to an approved product. The figure below shows these estimates ranked from highest to lowest clinical impact.

Virology, biochemistry, and immunology sit firmly in the high-impact zone [10, 11]. These fields benefit from defined molecular targets, high patent citation rates, reasonably predictive preclinical models, and clear clinical endpoints. A researcher in one of these fields producing 100 to 150 papers over a career has a meaningful probability of contributing to at least one pipeline entry that reaches clinical trial, and a realistic if modest chance of contributing to something that eventually reaches approval.

Biomaterials and tissue engineering sit at the opposite end, with an estimated per-paper probability roughly 30 to 50 times lower than virology [10, 11]. This does not mean biomaterials research lacks clinical value: it underpins enormous advances in surgical practice, implants, diagnostics, and device engineering. But the pathway to an approved standalone clinical product is longer, depends on more factors outside the researcher's direct control, and is less likely to complete within a single career.

There is an important within-field nuance. Oncology splits sharply between conventional small molecule approaches (low impact zone) and targeted, biomarker-driven work (moderate zone) [7]. A researcher working on immune checkpoint biology or predictive biomarkers for patient selection is in a very different structural position to one working on broad cytotoxic mechanisms. Similarly, a biomaterials researcher working on ophthalmology-specific materials (corneal implants, vitreous substitutes) operates in a field with a 33% clinical approval rate rather than the 3 to 5% that applies to cartilage scaffolds and tissue engineering constructs [7]. The field average is real but it conceals substantial sub-field variation.

The Manufacturing Gap

A category of failure that receives less attention than clinical trial outcomes but is equally capable of terminating a programme is the manufacturing gap: the discovery, during scale-up, that a technology which works reliably at laboratory scale cannot be reproduced consistently at the volumes, quality standards, and cost of goods required for commercial supply.

For biomaterials and biological products this problem is particularly acute. A hydrogel synthesised by a skilled researcher in a well-equipped laboratory, where each batch is characterised by the same person using the same equipment and tacit knowledge, may behave very differently when produced under GMP conditions by operators following a written protocol. Batch-to-batch variability that was managed empirically in the research setting becomes a regulatory problem in a manufacturing context.

Cell-based therapies illustrate this problem at its most extreme. Ex vivo cell culture methods as practised in academic laboratories are largely manual, operator-dependent, and irreproducible at scale. The transition to a commercially viable cell therapy requires not only clinical evidence of efficacy but a manufacturing process that can produce cells of defined phenotype and function in quantities sufficient for clinical use, at a cost of goods that can be absorbed within a reimbursable price, with a shelf life compatible with the logistics of clinical delivery. Each of these is a separate engineering challenge, and the combination has been the graveyard of many scientifically well-founded programmes [12].

Reimbursement: The Third Valley

Even technologies that navigate scientific, regulatory, and manufacturing challenges successfully face a final filter that is rarely considered at the research stage: reimbursement. A product that is safe and effective but is not reimbursed by national health systems or insurance payers will not reach patients at scale, regardless of its clinical merits.

The gene therapy Glybera (alipogene tiparvovec), approved in Europe in 2012 for lipoprotein lipase deficiency, is perhaps the most instructive example. It was the first gene therapy approved in the Western world, represented genuine scientific achievement, and addressed a serious unmet need. It was withdrawn from the market in 2017, having treated only a single patient through routine clinical channels [13]. The price per treatment course, set to recover development costs over a very small eligible patient population, was simply beyond what European health technology assessment bodies were prepared to reimburse. The science worked. The economics did not.

This dynamic is becoming more prevalent as advanced therapies with high development costs and small patient populations enter the market. The structural mismatch between the cost of developing a one-time treatment and the willingness of payers to fund it at a price that reflects that cost is creating what has been described as an affordability and access valley of death that sits downstream of the regulatory valley and can be just as terminal for a programme [13].

What Gets Through and Why

Against this background of attrition, it is worth asking what characterises the technologies that do successfully translate. The answer is not simply that they are the best science. Several factors consistently appear in retrospective analyses of successful translation [14]. The technology addressed a need that was both clinically significant and not adequately served by existing approaches: it was meaningfully better in a way that a payer would recognise, not just incrementally better. It had a development champion who understood both the science and the commercial requirements and was willing to bridge the two, often over many years. It had a regulatory strategy considered from early in development rather than retrofitted at the end. And it had a manufacturing pathway that was, at least in principle, technically and economically feasible from the outset.

The role of the development champion is particularly underappreciated. The journey from laboratory to clinic requires a level of persistence, commercial awareness, and willingness to adapt that is qualitatively different from the skills that produce good academic science. Many technologies with genuine potential have not translated because no one with the right combination of skills and motivation stayed with them through the full journey.

The Structural Solutions Being Tried

The biomedical research community has been aware of the translation problem long enough that several structural interventions have been designed to address it. In the United States, the National Center for Advancing Translational Sciences (NCATS) was established specifically to fund work in the valley between basic discovery and clinical development. In the UK, NIHR Biomedical Research Centres and Innovate UK have similar mandates, funding translational feasibility work that falls between the remit of basic research councils and the risk appetite of industry [14].

Accelerators and incubators embedded within academic institutions now provide infrastructure, regulatory expertise, and access to seed capital for researchers with promising technologies. The Babraham Research Campus in Cambridge, where Cell Guidance Systems is based, is one example of this model: a research environment in which academia and industry coexist in proximity, with deliberate structures to facilitate the movement of ideas and people between them.

Adaptive clinical trial designs, biomarker-driven patient selection, and better use of surrogate endpoints have improved success rates in some areas [7]. None of these interventions has resolved the fundamental arithmetic of attrition. They have, at best, shifted the point at which failures are identified so that they consume less capital. The pipeline remains narrow.

What This Means for Researchers Choosing a Direction

If direct clinical translation within a career is a priority, field selection matters enormously, and the data above give a clearer picture of the odds than most researchers encounter when choosing a research direction. Virology, biochemistry, and immunology offer the highest per-paper probability of clinical contribution [10, 11]. Biomaterials and conventional oncology offer the lowest [7, 10]. The difference is not random: it reflects the structural properties of each field, the quality of preclinical models, the clarity of clinical endpoints, the maturity of manufacturing platforms, and the degree to which industry actively co-invests alongside academic grants.

For researchers already committed to a lower-impact-rate field, the data suggest where leverage can be found. Working on applications within biomaterials that target ophthalmology or cardiovascular repair, which have better-established clinical endpoints and higher device approval rates, improves the odds relative to soft tissue engineering or cartilage repair [7]. In oncology, designing research around patient selection and biomarker identification rather than mechanism alone substantially improves the downstream success rate [7]. In any field, forming genuine industry collaborations early rather than treating commercialisation as a retrospective exercise is one of the most reliably effective ways to improve the probability that good science becomes a product [14].

For researchers who believe their work has genuine translational potential, a few other principles emerge from the evidence. Engaging with the clinical question from the start, rather than as a retrospective justification for the research, improves the probability that the work will be designed in a way legible to the clinical community [9]. Thinking about IP before publishing is a practical necessity. And accepting that the journey, if it is undertaken at all, will take far longer and cost far more than the research phase that generated the original idea is essential preparation for what lies ahead [1, 2].

Conclusion: The Arithmetic and What Lies Behind It

The pipeline from scientific idea to clinical product is long, narrow, and expensive. The average time from discovery to clinical adoption is 17 years by conservative estimates and considerably longer by more rigorous ones [1, 2]. The probability that a drug entering Phase I will reach approval is approximately 7 to 14% [4, 7], and the probability that a basic science finding will ever directly influence a clinical intervention is lower still [8]. The cost of the successes includes the cost of all the failures that preceded them, now measured in billions of dollars per approved product [5, 6].

These numbers do not argue against the research enterprise. They argue for greater honesty about the difficulty of what translation requires, greater investment in the stages where the most avoidable losses occur, and greater awareness among researchers of how the structural properties of their chosen field shape the probability of impact. Researchers doing this work are doing something genuinely valuable, whether or not their individual projects ever reach patients. But the more clearly they understand the pipeline they are hoping to enter, and which parts of it they can influence, the better their chances of being among the small fraction whose work makes it all the way through.

References

[1] Morris ZS, Wooding S, Grant J. The answer is 17 years, what is the question: understanding time lags in translational research. J R Soc Med. 2011;104(12):510-520.

[2] Hanney SR, Castle-Clarke S, Grant J, et al. How long does biomedical research take? Health Res Policy Syst. 2015;13:1.

[3] Eder J, Sedrani R, Wiesmann C. The discovery of first-in-class drugs: origins and evolution. Nat Rev Drug Discov. 2014;13(8):577-587.

[4] Norstella. Why are clinical development success rates falling? 2025 analysis of Citeline data, 2014-2023. Available at: www.norstella.com.

[5] Wouters OJ, McKee M, Luyten J. Estimated R&D investment needed to bring a new medicine to market, 2009-2018. JAMA. 2020;323(9):844-853.

[6] Deloitte. Measuring the return from pharmaceutical innovation 2024. Deloitte Centre for Health Solutions, 2025.

[7] Hay M, Thomas DW, Craighead JL, Economides C, Rosenthal J. Clinical development success rates for investigational drugs. Nat Biotechnol. 2014;32(1):40-51.

[8] Dopp JM et al. Tracing the path of 37,050 studies into practice across 18 specialties. PLOS ONE. 2023.

[9] Li H et al. A snapshot of translational research funded by the NIH. PLOS ONE. 2018.

[10] Ke Q. Comparing scientific and technological impact of biomedical research. J Informetr. 2018.

[11] Ahmadpoor M, Jones BF. The dual frontier: patented inventions and prior scientific advance. Science. 2017;357(6351):583-587.

[12] NEWDIGS FoCUS. Clinical development success rates for durable cell and gene therapies. Tufts Medical Center, 2023.

[13] Hao D et al. Making advanced therapies affordable and accessible. PMC. 2025.

[14] Lim DJ et al. Traversing the valley of death. Nat Rev Bioeng. 2023;1:990-1003.

[15] Makris EA et al. Advances in cartilage tissue engineering innovation and translation. Nat Rev Rheumatol. 2024.

IMAGE Hydrogel CREDIT Targetting cancer, Bigstock

Learn more about powerful technologies that are enabling research: