How to Structure Clinical Evidence for SaMD That Influences Decisions

TLDR:

To structure clinical evidence for SaMD that influences decisions, demonstrate three key pillars—valid clinical association (linking SaMD output to the clinical condition), analytical validation (technical accuracy and reliability), and clinical validation (proven clinical benefit in the target population).

Use existing evidence or generate new data as needed, tailor evidence rigor to the SaMD’s risk level, and ensure your clinical evidence plan clearly defines endpoints, populations, and statistical methods. This comprehensive approach supports regulatory compliance and decision-making.

Why Clinical Evidence Matters for SaMD

FDA’s latest guidance signals a shift: for SaMD that influences clinical decisions, real-world performance data is becoming non-negotiable.

• Incomplete or misaligned evidence is the #1 driver of Additional Information (AI) requests.

• The 2025 draft prioritizes fit-for-purpose evidence tied to device function, patient impact, and risk class.

A well-structured clinical evidence plan is your submission's foundation—and credibility booster.

---

Tiers of Clinical Evidence

Tier	Purpose	Study Type	Key Deliverables
Analytical Validation	Confirm algorithm performance on known data	Bench testing; retrospective simulation	Accuracy metrics; confusion matrix
Retrospective Clinical Review	Assess performance on historical data	Chart review; registry analysis	Sensitivity/specificity tables; study summary
Prospective Observational Study	Demonstrate real-world use	Single-arm observational cohort	Clinical study report; endpoint analysis
Controlled Prospective Trial	Compare against standard of care	Randomized or matched cohort trial	Controlled trial report; statistical analysis

---

Designing Your Clinical Evidence Plan

Step 1: Define Intended Use and Target Population

• Clearly specify the SaMD’s clinical purpose (e.g., diagnosis, triage, monitoring, prediction).

• Define the clinical setting (e.g., ICU, outpatient, primary care).

• Detail inclusion and exclusion criteria for the intended users and patient population.

• This step ensures the evidence generated is relevant to the real-world use case.

Step 2: Establish Valid Clinical Association

• Demonstrate that the SaMD output is scientifically and clinically associated with the targeted clinical condition or physiological state.

• Use existing literature, clinical guidelines, or generate new evidence if novel.

• This forms the foundation for all further validation steps.

Step 3: Select Study Endpoints and Metrics

• Choose primary endpoints that reflect clinical relevance, such as diagnostic accuracy (sensitivity, specificity), time-to-alert, or prediction performance.

• Include secondary endpoints like impact on provider workflow, usability, or time saved.

• Define measurable, clinically meaningful metrics aligned with intended use.

Step 4: Plan Analytical Validation

• Design tests to verify the SaMD processes input data correctly and produces accurate, reliable, and precise outputs.

• Use bench testing, simulations, or retrospective data.

• Document accuracy metrics, confusion matrices, and error rates.

Step 5: Plan Clinical Validation

• Design studies to demonstrate clinical performance and impact in the target population and setting.

• Possible study designs include retrospective clinical reviews, prospective observational studies, or controlled trials.

• Collect data on clinical outcomes, usability, and safety.

Step 6: Define Statistical Methods and Sample Size

• Calculate sample sizes with adequate statistical power to detect meaningful effects.

• Plan confidence intervals and subgroup analyses (e.g., by age, sex, comorbidities).

• Ensure statistical rigor to support regulatory submissions and clinical claims.

Step 7: Identify Data Sources and Quality Control Measures

• Determine data sources such as electronic health records (EHRs), device logs, clinical registries, or structured case report forms (CRFs).

• Implement SOPs for data extraction, curation, and quality assurance.

• Use validated data collection tools with audit trails to ensure data integrity.

Step 8: Execute Evidence Generation and Documentation

• Conduct analytical and clinical validation activities per plan.

• Analyze data using appropriate statistical methods.

• Document results thoroughly in clinical evaluation reports following regulatory standards.

Step 9: Engage with Regulatory Authorities and Plan Lifecycle Management

• Seek early and ongoing feedback from regulators to ensure alignment.

• Plan for continuous clinical evaluation, including post-market surveillance and real-world data collection.

• Update clinical evidence as the SaMD evolves or new clinical insights emerge.

---

FAQ

What level of evidence is enough for a 510(k)?

For moderate-risk SaMD (IMDRF Cat II–III), retrospective + prospective observational evidence with analytical validation is typically sufficient.

Do I always need a controlled trial?

No. Controlled trials are ideal for high-risk SaMD (Cat IV) but not always required if strong real-world data exists.

Can real-world evidence replace prospective studies?

Not entirely. Real-world data can supplement but not replace planned studies unless justified via Pre-Sub.