By Andreas Dimakakos, Director of Clinical Data; Kate Smietana, Director, Knowledge and Thought Leadership; Panos Karelis, VP of Commercial; and Dimitris Skaltsas, Co-founder & CEO
In biopharmaceutical R&D, where fewer than 1 in 10 development programs ultimately reach approval, risk assessment remains central to portfolio strategy and investment decisions. For more than two decades, benchmark studies1,2,3,4,5 as well as consortia gathering data from select companies have shaped how the industry thinks about development risk.
Those benchmarks remain valuable as they allow us to understand the historical success rate across the industry. However, in modern drug development, sponsors and clinicians operate at a level these benchmarks were never designed to measure, and the increasingly dynamic nature of clinical development makes those static and high-level benchmarks insufficient to enable good decisions. Today's assets are no longer single development stories. A single molecule may support multiple biomarker strategies, treatment lines, combination regimens, disease stages, and expansion pathways simultaneously.
When we analyzed more than 11,000 oncology development programs using Intelligencia AI's program-centric framework, we found that some of the industry's most trusted benchmarks are increasingly looking at development risk from 30,000 feet—just as many critical decisions are being made on the ground.
Three blind spots emerged repeatedly:
Blind Spot #1: One Number, Many Different Realities
Traditional benchmarks often treat a therapeutic area or a broadly-defined indication as a single market. But beneath the average, multiple development environments can exist with dramatically different probabilities of success. Looking only at the headline number can be like evaluating one neighborhood based on an entire city.
Blind Spot #2: Not All Risk Looks the Same
Two programs can ultimately fail at similar rates while creating completely different business outcomes. One may fail quickly after Phase I. Another may fail after years of pivotal development and hundreds of millions of dollars in investment. Traditional TA-focused benchmarks are often insufficient to represent how the risk profile across phases is affected by factors like technology platforms.
Blind Spot #3: The Biggest Risk May No Longer Be Biology
In many mature therapeutic areas, biological success is only part of the challenge. The real hurdle is demonstrating meaningful differentiation against increasingly effective standards of care. Programs are no longer competing only against disease—they are competing against a constantly advancing standard of care.
Traditional benchmark studies remain the gold standard for understanding high-level development risk. Our analysis suggests they may increasingly represent the beginning of the benchmarking conversation rather than its end.
The Benchmark Illusion
The challenge with most benchmark studies is not that they are wrong, but that they often stop precisely where strategic decisions begin. Let’s consider making an acquisition decision on an asset in a breast cancer space - a decision which requires an immediate investment of millions or even billions of dollars (vide recent acquisition of a breast cancer drug candidate for up to $3Bn in total with $2Bn upfront6 - which is just one of many high-value bets in the oncology space) and the upside is largely driven by risk and could be written off entirely if the acquired program fails to make it to approval. While thorough and granular enough risk assessment does not guarantee success, it can help identify and prioritize the best opportunities and inform valuation adjustments for the riskier choices.
At the overall disease level, breast cancer appears relatively attractive. In our dataset, Phase III-to-Registration success approached 46%, while Approval Given Phase III reached approximately 40%. For portfolio planners, investors, and business development teams, those numbers describe a mature therapeutic area with relatively favorable development dynamics.
But breast cancer is not a single development environment.
HER2-low programs achieved Approval Given Phase III rates approaching 67%. HER2-positive programs reached approximately 39%. Triple-negative breast cancer (TNBC) achieved only 12%.
The difference between 67% and 12% is not statistical noise. It is the difference between two fundamentally different investment theses hiding inside the same disease.
Drug-level benchmarks answer the question: "How risky is cancer drug development?" or - if more detailed - “How risky is drug development in breast cancer?”
Increasingly, executives need to answer a different question: "Which breast cancer strategy is most likely to succeed?"
The first question can be answered at the drug level. The second requires a program-level view.
Why It Matters That Technologies Have Differential Risk Distribution
Traditional benchmarks are excellent at measuring how much risk exists. They are less effective at showing where that risk actually resides.
In our analysis, novel modalities demonstrated only 17% Phase I-to-II transition success, compared with approximately 27% for established approaches. Viewed through a conventional lens, the conclusion appears straightforward: novel modalities are riskier.
But the story changes dramatically later in development. Once technical and biological feasibility was established, novel modalities achieved approximately 57% Phase III transition success, compared with only 34% for established modalities. Registration-stage success approached 98% (Figure 2, Panel A).
Novel modalities are not necessarily riskier. They simply concentrate risk earlier. That distinction carries major implications for capital allocation and portfolio construction.
TA-level benchmarks measure the amount of risk. Program-level benchmarks reveal the shape of risk.
As development costs continue to rise, understanding where risk resides may become as important as understanding how much of it exists.
The Biggest Risk May No Longer Be Biology
Historically, development success was viewed primarily through a scientific lens. The central question was simple: does the mechanism of action work on the disease? Increasingly, another question matters just as much: Can the program differentiate?
Across multiple oncology indications, we observed a consistent pattern: as therapeutic areas matured and standards of care improved, late-stage development became progressively more difficult. Comparator arms strengthened, regulatory expectations increased, and new entrants were pushed into narrower patient populations.
The result is a paradox that traditional benchmarks often struggle to capture: the largest markets are not always the most attractive development environments.
Breast cancer, lung cancer, and colorectal cancer collectively account for more than 1,500 oncology development programs in our dataset. Yet intense competition often makes incremental improvements increasingly difficult to translate into approval and commercial success.
The same dynamic appears when examining development strategy. Biomarker-driven programs consistently outperformed non-enriched populations during late-stage development, achieving approximately 41% Phase III-to-Registration success compared with 38% for non-biomarker approaches, while registration-stage success exceeded 91% versus 86% (Figure 2, Panel B).
Biology remains essential. But increasingly, success depends on the interaction between biology, patient selection, competitive positioning, regulatory strategy, and development execution. In many mature therapeutic areas, development failure is becoming as much a strategic problem as a scientific one.
Where Success Actually Lives
Future competitive advantage in drug development may come from measuring risk differently, where it actually lives.
Understanding the average probability of success is no longer enough. Strategic decisions increasingly require understanding where that probability changes and what drives it.
Organizations that continue to treat a TA- or asset-level success rate as an adequate representation of development risk may increasingly miss the differences that determine outcomes: the gap between a 67% and a 12% approval probability, the distinction between early-stage and late-stage risk concentration, or the reality that competitive density can transform an attractive market into a challenging one.
Drug-level benchmarks remain indispensable. But the decisions that create—or destroy—value often occur within the blind spots they cannot fully illuminate. Because increasingly, the most important question is not whether an asset can succeed. It is whether you can see the risks, opportunities, and strategic trade-offs hiding beneath the average before your competitors do.
Sources:
1. DiMasi et al. (2013) Clin Pharmacol Ther. https://pubmed.ncbi.nlm.nih.gov/23739536/
2. Hay et al. (2014) Nat Biotechnol. https://pubmed.ncbi.nlm.nih.gov/24406927/ (2021) BIO. https://www.bio.org/clinical-development-success-rates-and-contributing-factors-2011-2020
3. Smietana et al. (2016) Nat Rev Drug Discov. https://pubmed.ncbi.nlm.nih.gov/27199245/
4. Wong et al. (2019) Biostatistics. https://pubmed.ncbi.nlm.nih.gov/29394327/
5. Zhou et al. (2025) Nat Commun. https://pubmed.ncbi.nlm.nih.gov/41162353/
6. https://www.fiercebiotech.com/biotech/novartis-pays-synnovation-2b-breast-cancer-program-rivals-circle