Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
csv
Languages:
English
Size:
< 1K
Tags:
clarus
clinical-trials
population-integrity
framework-transition
registry-audit
regulatory-review
License:
sample_id string | source_framework string | target_framework string | concept string | source_definition string | target_definition string | claim string | option_a string | option_b string | task string | correct_option string | required_note string | required_keywords string | axis string | difficulty int64 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
CPFT-0001 | Registry | Publication | Inclusion criteria | Age 18–75 | Age 18–65 | Claim same enrolled population | Invalid correspondence, age range narrowed | Valid correspondence, minor edit | Judge correspondence and write note | A | State 18–75 vs 18–65 and flag narrowed eligibility | 18–75|18–65|age|narrowed|eligibility | inclusion_criteria | 2 |
CPFT-0002 | Registry | SAP | Analysis set | ITT includes all randomized | mITT excludes no-dose patients | Treat ITT and mITT as identical | Invalid correspondence, analysis set changed | Valid correspondence, mITT fully mapped | Judge correspondence and write note | A | State ITT vs mITT and require explicit exclusion mapping | ITT|mITT|exclude|no dose|mapping | analysis_set | 3 |
CPFT-0003 | Protocol | Publication | Exclusion criteria | No prior anti-PD-1 | Prior anti-PD-1 allowed in analysis | Assume criteria unchanged | Invalid correspondence, exposure rule changed | Valid correspondence, same rule | Judge correspondence and write note | A | State prior anti-PD-1 change and flag population drift | prior anti-PD-1|allowed|exposure|drift|criteria | exclusion_criteria | 3 |
CPFT-0004 | Protocol | SAP | Stratification factor | Stratify by EGFR status | Stratify by EGFR and smoking | Assume same strata | Partial correspondence, strata expanded | Valid correspondence, unchanged | Judge correspondence and write note | A | State added smoking stratification and note comparability impact | EGFR|smoking|stratify|expanded|comparability | stratification | 2 |
CPFT-0005 | Registry | Publication | High-risk subgroup | High risk = prior MI | High risk = prior MI or diabetes | Treat subgroup as same | Partial correspondence, definition broadened | Valid correspondence, same subgroup | Judge correspondence and write note | A | State MI-only vs MI-or-diabetes and flag broadened subgroup | MI|diabetes|high risk|broadened|definition | subgroup_definition | 2 |
CPFT-0006 | Protocol | Publication | Missing data handling | All randomized included | Exclude patients missing baseline scan | Assume cohort unchanged | Invalid correspondence, exclusions added | Valid correspondence, same cohort | Judge correspondence and write note | A | State baseline scan exclusion and flag post-randomization selection | missing baseline|exclude|post-randomization|selection|cohort | missingness_selection | 3 |
CPFT-0007 | Registry | Publication | Enrollment geography | Global sites | EU sites only in analysis | Treat as same population | Invalid correspondence, geography restricted | Valid correspondence, same sites | Judge correspondence and write note | A | State global vs EU-only and flag restricted external validity | global|EU only|geography|restricted|external validity | population_scope | 3 |
CPFT-0008 | Protocol | SAP | Per-protocol set | Not defined | Per-protocol excludes major deviations | Report PP as primary without mapping | Invalid correspondence, primary set changed | Valid correspondence, PP secondary only | Judge correspondence and write note | A | State per-protocol introduced and require mapping from ITT | per-protocol|deviations|primary|mapping|ITT | analysis_set_shift | 3 |
CPFT-0009 | Registry | Publication | Run-in period | No run-in | 2-week placebo run-in excludes non-adherent | Treat as same enrolled group | Invalid correspondence, selection introduced | Valid correspondence, run-in documented and mapped | Judge correspondence and write note | A | State run-in selection and flag adherence-based exclusion | run-in|placebo|non-adherent|selection|exclude | selection_mechanism | 3 |
CPFT-0010 | Protocol | Publication | Substudy cohort | All participants eligible | Only biomarker-positive included | Treat substudy results as trial-wide | Invalid correspondence, cohort restricted | Valid correspondence, restricted cohort stated | Judge correspondence and write note | A | State biomarker-positive restriction and block generalization | biomarker-positive|restricted|cohort|generalize|trial-wide | generalization | 3 |
Clarus Population Framework Transition Integrity v0.1
What this dataset tests
- You track how the study population is defined as it moves across formal frameworks
- You detect when inclusion, exclusion, or analysis sets change without explicit mapping
- You flag population drift introduced between planning and reporting stages
Framework transitions covered
- Trial registry → protocol
- Protocol → statistical analysis plan
- SAP → publication
Scope
- One trial
- Multiple population definitions
- Multiple formal frameworks
Typical failure modes
- Registry eligibility differs from published cohort
- ITT redefined as modified ITT without mapping
- Post-randomization exclusions introduced
- Per-protocol sets promoted without justification
- Subgroup restrictions applied retroactively
What this dataset does not test
- Internal narrative rhetoric within a single section
- Endpoint definition changes
- Safety or adjudication rule drift
Task format
- Each row presents a population definition in a source framework and a target framework
- The model judges whether correspondence is valid
- The model must name the mechanism of drift when correspondence fails
Label meaning
- A indicates invalid or unsupported population correspondence
- B indicates valid correspondence with sufficient justification
Use cases
- Automated audit of trial registries versus publications
- Systematic review population vetting
- Regulatory submission consistency checks
Version note
- v0.1 establishes baseline population transition auditing
- Higher-resolution variants may add explicit mapping requirements
- Downloads last month
- 16