Drug Discovery: Respect the Waves
The pharmaceutical industry operates with a fundamental misunderstanding of its own timelines. The standard narrative - 10 to 15 years from discovery to approval - captures only a portion of the true development arc. Measured from genuine scientific breakthrough to approved therapeutic, the journey more typically spans 25 to 30 years. This gap between expectation and reality has profound implications for how companies are built, financed, and managed.
A useful analogy comes from oceanography. When a tsunami forms, the ocean first recedes dramatically, exposing seafloor that was previously hidden. Then comes not one wave but a series of them, each with different characteristics. The first wave is rarely the largest. The entire process unfolds over hours, not minutes, catching many observers off guard.
Drug development follows a similar pattern. A breakthrough discovery creates energy, but that energy must travel through increasingly complex terrain before reaching patients. Progress comes in waves rather than steady streams. What appears to be failure often precedes the next advance. Understanding these dynamics might not accelerate development, but it may improve decision-making about where to deploy capital and effort.
(through mid-December)
Despite the long timelines and high attrition rates detailed in this paper, the system does produce results. The pattern holds: approximately 50 novel drugs reach patients each year. 2025 has continued this trajectory, with notable firsts including suzetrigine, the first non-opioid acute pain medication. The waves eventually reach shore, for those who understand how they travel.
Sources: FDA Novel Drug Approvals 2024; FDA Novel Drug Approvals 2025
The True Timeline
The conventional 10-15 year timeline appears in investor presentations, board decks, and regulatory discussions throughout the industry. It is also misleading. That figure typically measures from IND filing or clinical candidate nomination - the point at which a company formally begins development. It ignores the decades of foundational science that made development possible in the first place.
Research from Bentley University's Center for Integration of Science and Industry has measured a different metric: the elapsed time from foundational scientific discovery to approved therapeutic. The findings are sobering.1,2,3,4
When measured from initial scientific breakthrough rather than clinical candidate nomination, drug development averages approximately 30 years. The industry's standard 10-15 year figure captures only the final stage of a much longer process.
This distinction matters for risk assessment. A platform based on biology discovered in the past few years is not merely "early stage" - it has barely begun. Conversely, a company working on mechanisms that have been understood for two decades may be further along than conventional metrics suggest.
Three Technologies, Three Timelines
The extended timeline pattern holds across therapeutic modalities. Three of the most important platform technologies in modern medicine illustrate the dynamics:
Gene Therapy: 44 Years
The Cohen-Boyer recombinant DNA techniques that made gene therapy conceptually possible emerged in 1973. The first clinical trial occurred in 1990 - and it worked. A four-year-old girl with severe combined immunodeficiency showed improvement. The field appeared poised for rapid advancement.
Instead, it collapsed. Jesse Gelsinger died in 1999 from an immune response to the viral vector. Leukemia cases appeared in European trials when vectors activated oncogenes. The FDA did not approve a gene therapy in the United States until 2017. Forty-four years from the foundational science to a viable therapeutic platform. An entire generation of scientists spent careers on a technology that seemed perpetually a few years from clinical utility.5,6,7
The gene therapy experience illustrates a counterintuitive principle: a decade-long pause in a field's development is not necessarily wasted time. It may be essential learning.
Monoclonal Antibodies: 22 Years
César Milstein and Georges Köhler developed hybridoma technology in 1975 and received the Nobel Prize in 1984. The first therapeutic monoclonal antibody - muromonab-CD3 (OKT3) - reached the U.S. market in 1986. But this initial approval did not signal platform maturity. OKT3 was a fully murine antibody, and patients rapidly developed human anti-mouse antibodies (HAMA) that limited its utility. The immunogenicity problem would take another decade to solve.8,9,10,11
The commercially viable generation of monoclonal antibodies - chimeric and humanized constructs that avoided the HAMA response - did not emerge until 1994-1997 with the approvals of ReoPro, Rituxan, and Herceptin. From hybridoma discovery to a therapeutically robust platform: 22 years. Today, monoclonal antibodies generate approximately $250 billion in annual therapeutic revenue12 - although small molecules still dominate the pipeline by volume, representing roughly 60% of annual FDA approvals.13 The mAb platform's significance lies not in market share but in its demonstrated capacity for iterative improvement. Reaching that point required two decades of engineering.
CRISPR: 11 Years
CRISPR appears to be the exception. Gene-editing potential was demonstrated in 2012; regulatory approval for sickle cell disease came in late 2023. Eleven years is rapid by industry standards.14,15,16
Yet CRISPR's relative speed was enabled by decades of prior infrastructure. The gene therapy failures of the 1990s and 2000s generated the safety data, regulatory precedent, and clinical experience that CRISPR development required. Vector biology, delivery mechanisms, and manufacturing processes had already been worked out. Even with this foundation, the technology took over a decade and billions of dollars to reach approval.
The implication: truly novel mechanisms face the longest timelines because they lack validated models, established safety profiles, and regulatory precedent. Everything must be constructed from first principles.
The Valley of Death
Between basic research and viable clinical candidates lies the translational gap - often called the "valley of death." The terminology is not hyperbolic. Most promising compounds never emerge from this phase.17,18,19
Attrition at Scale
The numbers have remained stubbornly consistent for decades:
The overall likelihood of approval for drugs entering Phase I has actually declined - from approximately 10% a decade ago to 6.7% as of 2024. Despite advances in tools and understanding, success rates have moved in the wrong direction.2,20,21,22
Why Drugs Fail
The distribution of failure causes has remained relatively stable:
Lack of efficacy (40-50%): The compound does not produce the desired effect in humans despite promising preclinical results. This remains the dominant failure mode. Critically, research suggests that nearly 60% of clinical failures can be traced back to inadequate target identification and validation - the problem is often not the molecule itself but the underlying hypothesis that modulating a particular target will affect the disease.23 The disconnect between target engagement and clinical efficacy reflects the profound complexity of human disease, where single-target interventions often prove insufficient against multifactorial pathologies. Animal models are poor predictors of human response - over 92% of drugs that succeed in animal studies fail in human trials, a rate that has not materially improved in decades.24
Safety problems (30%): Unacceptable adverse effects emerge during clinical testing. Many compounds that engage their intended target also affect unintended pathways.
Poor drug properties (10-15%): Issues with absorption, distribution, metabolism, or excretion. A compound may work in controlled conditions but fail in the complexity of human physiology.17,25,26,27
The Economic Reality
Attrition at this scale carries substantial economic consequences:
The $2.23 billion figure incorporates the cost of failed programs - they are embedded in the cost structure rather than separate from it. This economic reality explains why drug development requires either substantial scale or exceptionally patient capital.
Therapeutic Area Variation
Not all therapeutic areas present equivalent challenges. The difference between the most and least difficult areas spans nearly an order of magnitude:
The most comprehensive study of clinical trial success rates - analyzing over 400,000 trial entries across 21,000 compounds from 2000-2015 - found success rates ranging from 33.4% for vaccines to 3.4% for oncology. While more recent data may show modest variation, the relative ranking of therapeutic areas has remained stable, and the tenfold difference between the most and least challenging areas has significant implications for portfolio construction and financing strategy.30
| Therapeutic Area | Success Rate | Median Time in Clinic | Key Factors |
|---|---|---|---|
| Vaccines | 33.4% | 5.9 years | Well-characterized biology, clear endpoints |
| Hematology | 26.1% | 6.2 years | Accessible tissue, measurable biomarkers |
| Infectious Disease | 19.1% | 6.5 years | Clear causation, definable endpoints |
| Cardiovascular | 8.7% | 7.0 years | Requires large outcomes trials |
| CNS/Neurology | 7.9% | 7.2 years | Blood-brain barrier, subjective endpoints |
| Oncology | 3.4% | 13.1 years | Tumor heterogeneity, resistance, evolution |
Success rates and durations from Wong et al., 2019; key factors reflect established characteristics in drug development literature.
Oncology presents a particularly challenging profile: the lowest success rate combined with the longest development timeline. At 13.1 years in clinical development versus 5.9-7.2 years for other areas, oncology compounds not only face greater risk but face it over a longer period - a compounding effect that significantly affects expected returns.
A 3.4% success rate over 13 years presents a fundamentally different risk profile than a 33% success rate over 6 years. These differences should inform capital allocation, milestone planning, and investor expectations.
Clinical Phase Dynamics
Once a compound enters clinical development, it faces a series of distinct hurdles. Each phase presents different probabilities of advancement:
Phase I: Initial Safety Assessment
Phase I primarily establishes the maximum tolerated dose (MTD) - not merely whether humans can tolerate a compound, but how much they can tolerate and what the dose-limiting toxicities are. This information is fundamental to designing subsequent efficacy trials. Approximately 47% of drugs advance from this stage on average. Notably, some analyses suggest Phase I success rates have declined substantially over the past 15 years, potentially reflecting more aggressive early termination of programs showing suboptimal biomarker signals.31
This decline is not necessarily negative. Companies are increasingly using biomarkers to identify problems earlier, concentrating failures in Phase I rather than Phase III. Expensive late-stage failures become cheaper early-stage failures - a shift in timing rather than overall success.
Phase II: The Critical Juncture
Phase II represents the primary point of attrition. Only 28% of compounds advance to Phase III. This stage provides the first substantial evidence of whether a drug will actually work in the target patient population.32,21
Phase II failures typically result from efficacy shortfalls: target engagement that does not translate to clinical benefit, biology that proves more complex than models suggested, or trial designs that were inadequate to detect real effects. Some analyses suggest that design issues contribute to as many as 90% of clinical failures.17,27,33
Phase III: High Stakes Confirmation
Phase III success rates are higher - approximately 55% - but the stakes are correspondingly elevated. By this point, typical investments range from 7-10 years and hundreds of millions of dollars. A Phase III failure often threatens company viability rather than merely program termination.32,21
The Iteration Pattern
Scientific breakthroughs rarely produce approved therapeutics on the first attempt. Instead, they typically spawn successive waves of increasingly refined approaches, each incorporating lessons from previous failures.
The tsunami metaphor proves most illuminating when examining the cyclical nature of pharmaceutical innovation - how initial breakthroughs spawn multiple successive waves of therapeutic development, each building upon and learning from previous surges. This pattern of iterative advancement, characterized by rapid progress followed by setbacks, retrenchment, and renewed advance, defines the actual experience of translational medicine far more accurately than linear development models suggest.
Monoclonal Antibodies: Four Generations
The evolution of antibody therapeutics provides a detailed case study in iterative development:
(1975-86)
(1986-94)
(1994-97)
(1997+)
Note: These metrics are independent measures, not complements. Immunogenicity risk measures the likelihood of anti-drug antibody responses; clinical success reflects overall approval rates. Factors beyond immunogenicity - including target selection, indication choice, and trial design - also influence success.
First generation (1975-1986): Fully murine antibodies. OKT3 was approved in 1986, but high immunogenicity (human anti-mouse antibody responses) severely limited clinical utility.8,10,11
Second generation (1986-1994): Chimeric antibodies combining mouse variable regions with human constant regions. Reduced immunogenicity but still problematic. ReoPro approval in 1994 marked the beginning of commercial viability.
Third generation (1994-1997): Humanized antibodies retaining only the complementarity-determining regions from mice. Rituxan (1997) and Herceptin (1998) became transformative cancer therapies.
Fourth generation (1997-2000s): Fully human antibodies produced via phage display and transgenic mice. This technology underlies today's blockbuster products.
Each generation built directly upon the limitations identified in its predecessor. The field did not abandon monoclonal antibodies after the first-generation immunogenicity problems - it engineered around them. In 2024, the FDA approved 13 monoclonal antibodies, the highest annual total since 2015. The platform that required 22 years from discovery to commercial maturity now produces dozens of approvals per decade.34
What appears as sudden success is typically the fourth or fifth iteration, constructed on lessons from earlier failures that most observers never witnessed.
Gene Therapy: The Decade-Long Pause
Gene therapy's trajectory was more dramatic. The initial wave in the 1990s ended with patient deaths and cancer cases attributed to viral vectors. Development effectively halted.6,4
Researchers did not abandon the approach. Instead, a decade was spent developing safer vectors, understanding failure mechanisms, and building the scientific foundation for a second attempt. When gene therapies began receiving approval in the 2010s and 2020s, they worked. The pause was not wasted time - it was prerequisite learning.
What Improves the Odds
Given prevailing success rates, identifying factors that materially improve outcomes takes on particular importance. The evidence points to several strategies with demonstrated impact.
Biomarker-Driven Development
The most validated method for improving clinical trial success is biomarker-driven patient selection. A study of 1,079 oncology compounds found a substantial effect:35
Drugs developed with biomarkers succeeded 24% of the time versus 6% without - a four-fold improvement. The larger MIT study confirmed this finding: biomarker-stratified trials demonstrate nearly double the success rate of unstratified approaches.30
Despite this evidence, biomarker-driven development remains underutilized. Many programs employ biomarkers opportunistically in early phases but fail to carry stratification strategies through to pivotal trials. The underutilization is partly structural: biomarker development requires upfront investment that delays IND filing, a trade-off many early-stage companies cannot afford despite the downstream benefits. This represents an underutilized opportunity for risk reduction.
Regulatory Pathway Optimization
Expedited regulatory pathways exist for drugs addressing serious conditions with unmet need. Their effect on development timelines is material:
In 2024, 66% of novel drug approvals utilized at least one expedited program. Breakthrough Therapy Designation, Fast Track, Accelerated Approval, and Priority Review are not merely administrative labels - they provide increased FDA interaction, faster feedback cycles, and clearer guidance on trial design.36,37,38,39
The Accelerated Approval pathway deserves particular attention. Drugs approved via this route based on surrogate endpoints must complete confirmatory trials to verify clinical benefit. The track record is encouraging but imperfect: approximately half of accelerated approvals convert to traditional approval within a median of 3.2 years.40 However, as of 2021, roughly 38% of accelerated approvals still had pending confirmatory trials, and some products have remained on the market for years without verified clinical benefit - so-called "dangling approvals."41 Recent legislative changes have given the FDA stronger tools to enforce post-marketing requirements and expedite withdrawals when confirmatory trials fail. Since 2011, 15 oncology products granted accelerated approval have been withdrawn, with 60% of those withdrawals occurring since 2020.42 The system is self-correcting, albeit slowly.
Platform Validation Effects
Once a platform technology produces its first approval, subsequent programs typically advance more rapidly. Regulatory pathways are established. Manufacturing processes are understood. Clinical endpoints are validated.
CRISPR's first approval in 2023 opened pathways for a pipeline of related therapeutics. A recent case demonstrated this effect: physicians developed a custom CRISPR therapy for an infant in approximately six months. That timeline was only possible because the underlying platform had already cleared regulatory and scientific hurdles.14,16
Persistent Challenges
Despite decades of investment in improved tools, faster sequencing, and computational approaches to discovery, overall success rates have not improved - and have arguably declined. Several structural challenges appear resistant to technological solutions.
The Exhaustion of Obvious Targets
One explanation for stubbornly stable success rates despite advancing scientific knowledge: the "low-hanging fruit" has been picked. The straightforward targets - enzymes with clear catalytic sites, receptors with well-defined ligands, pathogens with obvious vulnerabilities - were addressed first. What remains are the complex diseases: neurodegeneration involving multiple interacting pathways, cancers with redundant survival mechanisms, autoimmune conditions with unclear etiology. Each generation of drug discovery tackles harder problems than the last, which may explain why improved tools have not translated into improved success rates. The industry is running faster to stay in the same place.43
The Translation Problem
Preclinical models remain poor predictors of human outcomes. The statistics are sobering:
Over 92% of drugs that demonstrate efficacy in animal models fail in human trials. This rate has not materially changed in decades. The transition from preclinical models to human testing exposes a fundamental challenge: animal models consistently fail to predict human responses with sufficient accuracy. Rodent pharmacokinetics differ dramatically from human pharmacokinetics, tumor physiology in xenograft models bears limited resemblance to human disease complexity, and species-specific toxicities emerge without warning. These disconnects explain why approximately half of compounds fail during or immediately after Phase I trials despite having demonstrated safety and activity in animal studies.31
Increasing Complexity
Clinical trials have grown substantially more complex over time - more endpoints, more data collection requirements, more regulatory demands. Data collection in Phase III trials increased 283% over one decade. Greater complexity creates more opportunities for execution failures unrelated to underlying drug efficacy.46,47
The Innovation Penalty
The most novel therapeutic approaches - truly first-in-class mechanisms with potential for transformative impact - face the longest timelines and highest failure rates. They lack validated precedent; every element must be established from scratch. The economics of drug development effectively penalize genuine innovation, even though such programs represent the source of the most significant breakthroughs.17,48
The Translational Funding Gap
Academic institutions generate discoveries. Pharmaceutical companies conduct late-stage development. The territory between them - translational research and early clinical work - is chronically underfunded. Academic laboratories lack resources for clinical development. Venture capital increasingly favors later-stage opportunities with clearer return profiles. The valley of death persists in part because no single stakeholder owns it.18,19,49
Implications
The data presented in this paper suggest several considerations for those building, financing, or advising companies in the sector:
Timeline calibration matters. Thirty years from foundational discovery to approved drug is typical for novel mechanisms. Plans built on 10-year assumptions may encounter difficulties when entering year eight. Runway calculations should account for the full historical range of development timelines. The challenge, of course, is that few financing structures accommodate 30-year timelines. The companies that navigate these dynamics successfully typically do so through staged partnerships, platform licensing, or positioning for acquisition at validated inflection points rather than attempting to fund the entire journey independently.
Failure contains information. Gene therapy's decade-long pause was not wasted - it was prerequisite learning. Monoclonal antibodies required four generations of iteration before achieving commercial success. A failed first approach does not necessarily indicate flawed science; it may indicate an early position in a multi-generational development cycle.
Therapeutic area selection has outsized impact. A 3.4% success rate in oncology versus 33% in vaccines represents an order-of-magnitude difference in expected outcomes. This variation should inform portfolio construction, financing strategy, and milestone expectations.
Biomarker utilization represents available leverage. Four-fold improvements in success rates from proper patient stratification are documented in the literature. Yet most programs do not employ biomarkers consistently. This may represent the highest-impact modification available to many development programs.
Platform validation creates second-mover advantages. First approvals from novel platforms take decades. Subsequent programs move considerably faster. Being positioned as a fast-follower on a newly validated platform may represent a defensible strategic approach.
The next wave is already forming. Several technologies are positioned to follow the pattern of monoclonal antibodies - early promise, iterative refinement, eventual maturation. Bispecific antibodies and antibody-drug conjugates (ADCs) are extending the mAb platform; ADCs generated over $13 billion in 2024 sales across 21 approved products, with bispecific ADCs now entering late-stage trials.50 Non-viral delivery systems using lipid nanoparticles may solve the manufacturing and safety challenges that have constrained gene therapy - early data show in vivo CAR-T cell generation using targeted LNPs, which could transform cell therapy from a bespoke manufacturing process to an off-the-shelf injection.51,52 These technologies are at different points in their development arcs. Recognizing which wave is forming - and when it will arrive - may be as important as understanding past patterns.
Drug development will remain difficult. The biology is complex, the regulatory requirements demanding, and the capital requirements substantial. These constraints are not changing in the near term. However, understanding the actual dynamics - true timelines, the wave patterns of technological development, where programs fail and why - provides a foundation for navigating them more effectively. Approximately 50 novel drugs reach patients each year: 2024 delivered 50, and 2025 is on pace to match that figure. The pattern is clear: the waves do arrive, but only for those still standing when they reach shore.