Below is a clear, practical guide for how a project biostatistician conducts QC in a clinical trial. This focuses on statistical and scientific QC, not routine programming.

---

1. Prepare for QC

• Confirm final, approved documents:
• Protocol
• SAP
• TFL shells
• Understand:
• Primary/secondary objectives
• Estimands and analysis populations
• Blinding requirements

---

2. Verify Analysis Inputs

• Confirm correct:
• Dataset versions and timestamps
• Population flags (ITT, Safety, PP)
• Treatment variables and visit definitions
• Ensure traceability from datasets to outputs

---

3. Review Statistical Methods

• Verify analyses match the SAP:
• Models, tests, covariates
• Handling of missing data and intercurrent events
• Censoring rules
• Ensure no undocumented deviations or post-hoc changes

---

4. Review Results for Accuracy and Plausibility

• Check:
• Ns and denominators
• Direction and magnitude of effects
• Event rates and distributions
• Identify inconsistencies or unexpected results that require re-check

---

5. Cross-Check Consistency

• Confirm consistency across:
• Disposition, demographics, and analysis populations
• Related tables, figures, and listings
• Figures vs corresponding tables

---

6. Maintain Blinding

• Conduct QC at the appropriate blinding level
• Do not access unblinded comparative results unless explicitly authorized
• Coordinate with independent/unblinded statisticians when needed

---

7. Resolve and Document Issues

• Log issues with:
• Description
• Impact
• Resolution
• Confirm corrections are implemented and re-QC’d
• Document all decisions and rationale

---

8. Sign Off

• Confirm:
• All QC findings are resolved
• Outputs align with SAP and protocol
• Final versions are correct and labeled
• Provide documented statistical sign-off

---

What the Project Biostatistician Does Not Do

• Does not perform routine TFL programming
• Does not break the study blind
• Does not approve undocumented analysis changes

---

One-Sentence Summary

As project biostatistician, QC is conducted by verifying methodological correctness, reviewing results for plausibility and consistency, ensuring blinding compliance, resolving issues, and providing documented statistical sign-off.