Forgotten Biology: The Missing Link Behind Trial Outcomes

Introduction

Clinical trials are often described as the science of uncertainty, and that uncertainty is expected. It is how evidence is built. However, not all uncertainty is scientific. In many cases, it arises when the biological context of an intervention is not carried through into the study population.

A product enters a trial with strong science. The mechanism is clear, the rationale is sound, and expectations are well-founded. Yet the outcome fails to support a meaningful claim. Decisions slow down, confidence weakens, and what should have been clarity begins to feel like doubt. What makes this situation particularly difficult is not just the result, but the contradiction within it. Because in many such trials, the data are not uniformly negative. A closer look often shows that some participants have responded meaningfully, in line with the expected mechanism. The product has worked, but not consistently across the population.

At this point, the question shifts.

It is no longer only about whether the product works, but whether the study preserved the biology well enough to show where it works.

Clinical studies begin with precision. Interventions are designed to act on specific biological pathways, and expected outcomes are grounded in that understanding. But as studies move from concept to execution, this precision is often diluted during population selection.

Participants are recruited based on diagnosis and eligibility criteria, an approach that is necessary and aligned with regulatory expectations. However, diagnosis is a broad clinical construct. It groups individuals with similar symptoms, not necessarily similar underlying biology. Each individual carries a unique biological context shaped by genetics, physiology, and lifestyle. Within a single study, participants may differ in what is driving their condition, how it is expressed, and how they are likely to respond. These differences are not peripheral; they directly influence outcomes.

As the study progresses, this diversity becomes visible in a predictable pattern. Some participants respond strongly, some show modest improvement, and some show little or no response. When these responses are combined, the overall effect becomes less distinct. As variability increases, the true biological effect (signal) gets diluted, making it harder to detect statistically. In practice, this means a study may appear non-significant overall, even when clearly responding subgroups exist within the population.

The industry is not unaware of this pattern, but it is often addressed too late. A large proportion of trials report subgroup differences, yet very few are designed to account for this variability at the stage where it matters most. As a result, variability is analysed after the study rather than shaped before it.

This creates a structural limitation. By the time variability is understood, it is already embedded in the dataset. The population is fixed, and the statistical outcome is effectively locked. The insight explains the result, but it does not change it.

The only stage where this can be meaningfully influenced is at recruitment. This is where the study population is defined, and where the clarity of the outcome is largely determined. Today, recruitment is primarily driven by eligibility with a simple question: Does the participant meet the criteria? If yes, enrol. This ensures compliance, but it does not ensure that the study population is aligned with the biology of the intervention. There is a meaningful difference between a participant who qualifies for a study and one who is likely to demonstrate a measurable response. That difference often determines whether the signal will be clearly seen or diluted.

This is where phenotype-based recruitment becomes relevant, not as a change to eligibility criteria, but as a refinement within them. Before formal enrollment, participants can be assessed through structured, low-risk pre-screening to better understand their functional state, lifestyle context, and relevance to the study endpoints.

This approach is already well demonstrated in clinical research, where selecting the right biology at entry has led to clearer outcomes.

In the KEYNOTE-024 trial, patients with non-small cell lung cancer were prospectively selected based on PD-L1 expression ≥50%, ensuring inclusion of tumors dependent on the PD-1/PD-L1 pathway. This biologically aligned recruitment reduced heterogeneity and enabled a clear separation in progression-free survival and overall survival between pembrolizumab and chemotherapy.[1]

Similarly, the FLAURA trial enrolled only patients with activating EGFR mutations (Ex19 deletion or L858R), which are direct drivers of tumor growth. By aligning the study population with the drug’s mechanism, the trial demonstrated a robust and consistent survival benefit with osimertinib compared to standard therapy, with reduced variability across participants. [2]

In the CLEOPATRA trial, patients were selected based on HER2 overexpression or amplification, a well-defined oncogenic driver. This ensured that dual HER2 blockade with pertuzumab and trastuzumab acted on a biologically relevant population, resulting in a strong and sustained improvement in survival outcomes. [3]

Beyond oncology, the CANTOS trial applied a similar principle using an inflammation-based phenotype. Participants were selected based on hsCRP ≥2 mg/L, identifying individuals with persistent systemic inflammation. This allowed the study to clearly demonstrate that targeting IL-1β with canakinumab reduces cardiovascular events—an effect that may have been diluted in a broader, unselected population. [4]

In more recent clinical research, this principle has become even more explicit. Trials in cardiometabolic and cardiovascular domains have increasingly relied on functional and risk-based phenotypes rather than broad disease labels.

For example, in the DAPA-HF (Dapagliflozin in Heart Failure) and EMPEROR (Empagliflozin in Heart Failure) trials, participants were selected based on reduced ejection fraction and elevated NT-proBNP levels, identifying patients with active cardiac stress biology. This enabled consistent demonstration of benefit with SGLT2 inhibitors across diverse populations. [5], [6]

Similarly, the SELECT trial (Semaglutide in Overweight/Obesity WITHOUT Diabetes- cardiovascular outcomes) focused on overweight and obese individuals with established cardiovascular disease but without diabetes, isolating a clearly defined cardiometabolic risk phenotype. This targeted selection allowed semaglutide to demonstrate a significant reduction in major cardiovascular events, which may have been less apparent in a more heterogeneous population. [7]

Regulatory thinking is also evolving in this direction. The U.S. FDA has emphasised enrichment strategies to improve the likelihood of detecting treatment effects by focusing on populations where an intervention is more likely to show benefit. In parallel, the updated ICH GCP E6(R3) (2025) framework places greater emphasis on how participants are selected, encouraging a more structured and risk-based approach to recruitment, including the thoughtful use of pre-screening processes, while maintaining ethical and patient-centric standards. [8], [9]

When this approach is applied, the study population becomes more aligned with the biological mechanism of the intervention. Variability does not disappear, but it becomes more structured. Participants become more comparable in ways that matter, and outcome measures respond more clearly. The product does not change. The study design does not change. What changes is the ability of the study to reflect the biology it was built to test.

Across multiple clinical evaluations, a consistent observation has emerged: study outcomes tend to be clearer, more stable, and easier to interpret when the characteristics of the study population are aligned with the underlying biological mechanisms being targeted. Conversely, a lack of such alignment can contribute to variability in outcomes, even when evaluating well-characterized interventions.

This has led to increasing interest in more structured participant selection strategies. One such approach involves incorporating phenotype-based pre-screening into recruitment workflows, with the aim of ensuring that relevant biological pathways are adequately represented within the study population. Rather than limiting enrollment, this strategy focuses on enhancing the interpretability and translational relevance of study findings by improving the biological fit between the intervention and the participants.

Seen through this lens, a non-significant result does not always indicate that a product is ineffective. It may simply mean that the study did not include the right population to demonstrate its effect clearly. Clinical studies begin with biology, but their success depends on whether that biology is preserved through execution. When it is not, the signal weakens. When it is, outcomes become clearer and more consistent.

In that sense, biology is not just part of science; it is the missing link that determines whether that science is ultimately seen.

References:

1. Martin Reck, et. al. Pembrolizumab versus Chemotherapy for PD-L1–Positive Non–Small-Cell Lung Cancer. November 10, 2016. N Engl J Med 2016;375:1823-1833, DOI: 10.1056/NEJMoa1606774, https://www.nejm.org/doi/full/10.1056/NEJMoa1606774
2.  Jean-Charles Soria, et al. Osimertinib in Untreated EGFR-Mutated Advanced Non–Small-Cell Lung Cancer. November 18, 2017. N Engl J Med 2018;378:113-125. DOI: 10.1056/NEJMoa1713137, VOL. 378 NO. 2. https://www.nejm.org/doi/full/10.1056/NEJMoa1713137
3. José Baselga, et. al. Pertuzumab plus Trastuzumab plus Docetaxel for Metastatic Breast Cancer. January 12, 2012. N Engl J Med 2012;366:109-119. DOI: 10.1056/NEJMoa1113216. VOL. 366 NO. 2. https://www.nejm.org/doi/full/10.1056/NEJMoa1113216
4. Paul Ridker, et. al. Antiinflammatory Therapy with Canakinumab for Atherosclerotic Disease. 21, 2017 N Engl J Med 2017;377:1119-1131. DOI: 10.1056/NEJMoa1707914, VOL. 377 NO. 12, https://www.nejm.org/doi/full/10.1056/NEJMoa1707914
5. John McMurray, et. al. Dapagliflozin in Patients with Heart Failure and Reduced Ejection Fraction. September 19, 2019. N Engl J Med 2019;381:1995-2008. DOI: 10.1056/NEJMoa1911303, VOL. 381 NO. 21. https://www.nejm.org/doi/full/10.1056/NEJMoa1911303
6. Milton Packer, et. al. Cardiovascular and Renal Outcomes with Empagliflozin in Heart Failure. August 28, 2020 N Engl J Med 2020;383:1413-1424. DOI: 10.1056/NEJMoa2022190. VOL. 383 NO. 15. https://www.nejm.org/doi/full/10.1056/NEJMoa2022190
7. A. Michael Lincoff, et. al. Semaglutide and Cardiovascular Outcomes in Obesity without Diabetes. November 11, 2023. N Engl J Med 2023;389:2221-2232. DOI: 10.1056/NEJMoa2307563. VOL. 389 NO. 24. https://www.nejm.org/doi/full/10.1056/NEJMoa2307563
8. FDA Guidance on Enrichment Strategies for Clinical Trials. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/enrichment-strategies-clinical-trials-support-approval-human-drugs-and-biological-products
9. FDA Guidance: E6(R3) Good Clinical Practice. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/e6r3-good-clinical-practice-gcp