# Wearable Validation Study Plan & Statistical Analysis Plan (SAP): Template

*Palo Alto Health Labs, paloaltohealthlabs.com. Educational template only, not regulatory advice. Replace bracketed text. Structure follows DiMe's V3/V3+ framework ([V3](https://www.nature.com/articles/s41746-020-0260-4); [V3+](https://www.nature.com/articles/s41746-024-01322-2)).*

---
## Part A: Study Plan (Protocol)

### A1. Administrative
- Title, protocol ID, version and date, sponsor, principal investigator, sites
- Version history table

### A2. Background and rationale
- Clinical or research need; what the metric enables
- Prior evidence (verification, earlier validation, literature)

### A3. Intended use and context of use
- Measure or metric definition (concept of interest → measure → algorithm output)
- Target population, setting (clinic, home, free-living), and the decision it supports
- Claim tier: wellness / digital endpoint / SaMD or medical claim / payer evidence

### A4. Objectives and V3+ scope
| V3+ component | Covered in this study? | Objective |
|---|---|---|
| Verification | [Y/N; reference bench report] | |
| Analytical validation | [Y/N] | |
| Clinical validation | [Y/N] | |
| Usability validation | [Y/N] | |

### A5. Device and algorithm
- Hardware model, firmware, app, algorithm version (all **locked** for the study)
- Sensor placement, sampling rate, data export format (raw vs. summary)

### A6. Reference standard
- Reference device or method and its regulatory or validation status. Examples: ECG patch or Holter for HR/rhythm; PSG for sleep; clinical EEG for neuro signals; arterial SaO2 for SpO2
- Applicable standard (e.g., ANSI/CTA-2065-A, ISO 81060-2, FDA guidance)
- Time synchronization method and tolerance
- Reference-scoring procedure (e.g., blinded scorer(s), inter-rater agreement)

### A7. Study population
- Inclusion and exclusion criteria
- Diversity targets and how they're measured:
  - Skin tone: Monk Skin Tone **and** Individual Typology Angle (ITA) at the sensor site
  - Age bands; sex; BMI; relevant clinical conditions; wrist or anatomy factors
- Recruitment plan and screening log

### A8. Procedures
- Visit schedule; protocol segments (rest, scripted activities, free-living, sleep)
- Device fitting, wear instructions, adherence monitoring
- Usability tasks and data collection (if V3+ usability is in scope)

### A9. Endpoints
- Primary endpoint (e.g., agreement of HR with the ECG reference in 30-s epochs)
- Secondary endpoints (subgroups, activity conditions, detection accuracy)
- Prespecified acceptance criteria

### A10. Safety, ethics and regulatory
- IRB, informed consent or eConsent, data privacy (HIPAA/GDPR as applicable)
- Adverse events / device issues reporting
- Electronic records and signatures (21 CFR Part 11 where applicable)

### A11. Data management
- Data flow diagram; storage; de-identification; audit trail; data lock procedure

### A12. Publication and reporting plan

---
## Part B: Statistical Analysis Plan (SAP)

### B1. Analysis populations
- Enrolled; per-protocol; valid-epoch definitions (signal quality, wear time)

### B2. Data preprocessing
- Epoch length; alignment to the reference; artifact and signal-quality rules
- Handling of non-wear and missing data (prespecified, not post hoc)

### B3. Primary analysis: agreement (continuous measures)
- Bland-Altman: bias, 95% limits of agreement and their CIs, accounting for repeated measures per participant
- Error metrics: MAE, MAPE, RMSE; correlation (secondary only, not proof of agreement)
- Acceptance criteria: [thresholds and their justification]

### B4. Classification / detection analyses (if applicable)
- Sensitivity, specificity, PPV, NPV with 95% CIs; ROC curve and AUC
- Operating threshold: prespecified, or chosen on a development set and locked
- Unit of analysis (epoch, episode, participant), with clustering addressed

### B5. Subgroup analyses (prespecified)
- Skin tone (MST group / ITA category), age band, sex, BMI, activity condition, clinical status
- Interaction tests or stratified estimates; minimum n per subgroup

### B6. Sample size
- Precision-based: target CI width on the LoA or on sensitivity/specificity, overall and per key subgroup
- Assumptions, source of variance estimates, inflation for dropout and unusable data

### B7. Sensitivity analyses
- Alternative quality thresholds; per-participant averaged analysis; excluding the first/last epochs

### B8. Usability analysis (if applicable)
- Task success, use errors, close calls; qualitative root-cause analysis

### B9. Software, versioning and reproducibility
- Analysis code repository, package versions, locked analysis scripts before unblinding

### B10. Tables and figures shells
- Participant flow; demographics by subgroup; Bland-Altman plot; ROC curve; subgroup forest plot

---
*Need help tailoring this? Book a free 30-minute scoping call at [paloaltohealthlabs.com](https://paloaltohealthlabs.com/).*