Blog Series: Patient Stratification
Part 1: What is Informed Patient Stratification?
The challenge: small, heterogeneous datasets
Clinical trials today are increasingly conducted in smaller, more complex patient populations—especially in areas like oncology, rare disease, and in precision medicine settings.
These datasets are not just small—they are typically also highly heterogeneous.
Patients differ across biology, disease progression, prior treatments, and response patterns.
This variability introduces noise. And in small samples, even modest levels of noise can overwhelm the response signal.
The result:
- inconclusive results
- unstable effects
- missed insights
All of which can lead to promising therapies appearing to be ineffective and subpopulation insights being overlooked.
Patient stratification helps manage this variability before it undermines your study.
What is Patient Stratification?
Patient stratification segments participants based on characteristics expected to impact outcomes, such as:
- Demographics
- Biomarkers
- Disease severity
- Clinical history
Stratification influences both:
- Randomization (ensuring balance across arms), and
- Analysis (enabling meaningful subgroup interpretation)
In more advanced or hybrid designs, stratification also supports evaluating both enriched populations and the broader study population.
Stratification vs. Enrichment: What’s the Difference?
These terms are often used together, but they serve different purposes in trial design. Many trials use both.
- Stratification = balance and analysis
Patients are grouped based on key variables (e.g., biomarker status, disease severity) to ensure fair comparisons across treatment arms and enable meaningful subgroup insights.
- Enrichment = patient selection
Patients are included—or excluded—based on characteristics that increase the likelihood of detecting a treatment effect.
Enrichment increases the chance of detecting a signal.
Stratification ensures that signal is credible and interpretable.
In practice:
- Stratification helps you understand differences within your study population to ensure results are interpretable and regulatorily sound
- Enrichment helps you define who enters the study in the first place to increase the chance of success.
Used together, they strike a balance between maximizing signal and preserving generalizability.