Skip to main content

Appendix A — SAP template: diagnostic clinical performance study

Blank structure with prompts. Fill each section; delete prompts.

Structure borrowed from ICH E9 (by analogy) and adapted to IVDR Annex XIII Part A 2.3.2 and ISO 20916:2019. See sap_ptau217_mock.qmd for a filled example.

Field Entry
Document ID / version / date
Protocol / CPSP reference
Author, independent reviewer, approver
Applicable SOPs
Version history and lock date (before unblinding of reference-standard data)

1. Objectives and hypotheses

  • Intended use and intended-use population (as in labelling).
  • Primary objective; each primary endpoint with a performance goal and a pre-specified success rule.
  • Secondary and exploratory objectives.

2. Design, reference standard, cut-offs

  • Design (prospective, banked, case-control; sites; settings).
  • Reference standard: definition, readers/adjudication, blinding both ways, timing window, handling of unreadable results.
  • Cut-offs: how and where derived, lock date and file, rule for indeterminate results, convention at the boundary (≤ vs <).

3. Analysis populations

  • Enrolled, evaluable (primary), per-protocol. State the diagnostic analogue of ITT: who is in, who is out, why.

4. Endpoints and statistical methods

  • Each endpoint: definition, numerator/denominator, CI method (Wilson, exact, other), software function.
  • Treatment of indeterminates: separate category, counted as non-agreement, excluded in a sensitivity analysis.
  • Prevalence dependence of predictive values.
  • Agreement metrics when the comparator is not a reference standard: PPA/NPA/OPA (FDA 2007).

5. Sample size

  • Assumed expected performance, goals, alpha, power, category frequencies, loss to non-evaluable; method and cross-check.

6. Missing, invalid and unreadable results

  • Repeat rules, exclusion rules, flow diagram (STARD), worst-case and imputation sensitivity analyses.

7. Subgroups

  • List (age, sex, kidney function, cognitive status, setting, other), minimum n for reporting, descriptive vs inferential.

8. Multiplicity, interim analyses, sensitivity analyses

  • Co-primary rule, hierarchy, adjustments; interim analyses (if any) and who sees what; sensitivity analyses list.

9. Data standards

  • Datasets (SDTM/ADaM-style by analogy; CDISC device implementation guides exist), derivations, controlled terminology.

10. Software, validation and QC

  • R version, renv.lock, validated functions and tests; double programming of primary endpoints; QC checklist.

11. Tables, figures, listings (shells)

  • Disposition, baseline, 3×2 table, primary, secondary, subgroups, sensitivity, listings of discordant cases.

12. Deviations and amendments

  • How deviations are recorded; amendment procedure after lock.

References