inst/testdata/QUICKSTART.md

In clinCompare: Dataset Comparison with 'CDISC' Validation for Clinical Trial Data

Quick Start Guide - SDTM Test Datasets

Overview

Two versions of realistic synthetic SDTM clinical trial datasets for Study CLIN-2025-042 with intentional differences to test comparison and validation tools.

Dataset Versions

• v1: Interim dataset (500 subjects, 33,495 records)
• v2: Final dataset (503 subjects, 33,568 records) - with corrections and new data

Files Available

Demographics (DM)

• dm_v1.csv - 500 subjects, 16 variables
• dm_v2.csv - 503 subjects, 17 variables (ETHNIC added)

Adverse Events (AE)

• ae_v1.csv - 1,495 adverse events
• ae_v2.csv - 1,545 adverse events (+50 new)

Laboratory (LB)

• lb_v1.csv - 16,000 lab test results
• lb_v2.csv - 16,000 lab test results (100 value corrections, 17 indicator corrections)

Vital Signs (VS)

• vs_v1.csv - 14,000 vital sign measurements
• vs_v2.csv - 14,020 vital sign measurements (+20 new, 50 value updates)

Exposure (EX)

• ex_v1.csv - 1,500 treatment exposures
• ex_v2.csv - 1,500 treatment exposures (11 dose corrections, 3 route corrections)

Key Characteristics

Study Design

• Study ID: CLIN-2025-042
• Subjects: 500 (+ 3 new in v2)
• Sites: 5 (SITE01-SITE05: 3 USA, 1 Canada, 1 GB)
• Arms: TRT01 (40%), TRT02 (40%), PBO (20%)

Demographics

• Age: 18-75 years
• Sex: 52% M, 48% F
• Race: 65% White, 15% Black, 15% Asian, 5% Other
• Ethnicity: 85% Not Hispanic, 15% Hispanic

Data Types

• All dates in ISO format (YYYY-MM-DD)
• Numeric values with 2 decimal places
• Standard SDTM variable naming conventions
• Realistic missing values (e.g., AEENDTC for ongoing events)

Common Use Cases

1. Test Data Comparison Tools

Compare v1 and v2 to identify: - New subjects (3 in DM) - New records (50 AE, 20 VS) - Updated values (100 LB, 50 VS, etc.) - Corrected values (10 RACE, 5 AGE, etc.) - New columns (ETHNIC in DM v2)

2. Data Validation

Check for: - SDTM compliance - Data type consistency - Valid value ranges - Referential integrity (USUBJID linkage) - Date sequence validity

3. ETL Testing

Process datasets for: - Data transformation pipelines - Subject key management - Visit scheduling - Laboratory normal ranges - Treatment assignment tracking

4. Reconciliation Tools

Practice reconciliation between: - Interim (v1) and Final (v2) datasets - Version tracking - Change detection - Audit trails

Column Quick Reference

All Files

• STUDYID (CLIN-2025-042)
• DOMAIN (DM, AE, LB, VS, EX)
• USUBJID (unique subject ID)

Domain-Specific Key Variables

DM: SUBJID, RFSTDTC, RFENDTC, SITEID, SEX, AGE, RACE, ETHNIC, ARMCD, ARM, COUNTRY v2 only

AE: AESEQ, AETERM, AEDECOD, AEBODSYS, AESEV, AESER, AEACN, AEREL, AEOUT, AESTDTC, AEENDTC

LB: LBSEQ, LBTESTCD, LBTEST, LBORRES, LBORRESU, LBSTRESN, LBSTRESU, LBNRIND, VISITNUM, VISIT, LBDTC

VS: VSSEQ, VSTESTCD, VSTEST, VSORRES, VSORRESU, VSSTRESN, VSSTRESU, VISITNUM, VISIT, VSDTC

EX: EXSEQ, EXTRT, EXDOSE, EXDOSU, EXDOSFRM, EXROUTE, EXSTDTC, EXENDTC, VISITNUM, VISIT, EPOCH

Intentional Differences (v1 → v2)

| Domain | Change Type | Count | Details | |--------|------------|-------|---------| | DM | New Column | 1 | ETHNIC added | | DM | New Subjects | 3 | SUBJID: NEW00, NEW01, NEW02 | | DM | Corrected RACE | 10 | Data corrections | | DM | Corrected AGE | 5 | Data corrections | | AE | New Records | 50 | New AE observations | | AE | Updated AESEV | 13 | Severity corrections | | AE | Corrected AEREL | 8 | Relationship corrections | | LB | Updated LBSTRESN | 100 | Result value updates | | LB | Corrected LBNRIND | 17 | Normal indicator fixes | | VS | New Records | 20 | New measurements | | VS | Updated VSSTRESN | 50 | Result value updates | | EX | Corrected EXDOSE | 11 | Dose adjustments | | EX | Corrected EXROUTE | 3 | Route corrections |

Sample Data

Load in Python

import pandas as pd

# Load demographics v1
dm = pd.read_csv('dm_v1.csv')
print(dm.head())
print(f"Subjects: {len(dm)}")
print(f"Sites: {dm['SITEID'].nunique()}")
print(f"Arms: {dm['ARMCD'].unique()}")

Load in R

library(readr)

# Load adverse events v1
ae <- read_csv('ae_v1.csv')
head(ae)
nrow(ae)  # Total AE records
unique(ae$AEBODSYS)  # Body systems

Load in SQL

-- Create table from v1 data
CREATE TABLE dm_v1 AS
SELECT * FROM (
  LOAD DATA FROM 'dm_v1.csv' FORMAT CSV
)

-- Compare v1 and v2
SELECT COUNT(*) FROM dm_v2 
WHERE SUBJID NOT IN (SELECT SUBJID FROM dm_v1)
-- Returns: 3 (new subjects)

Data Validation Checks

Expected Ranges

• Age: 18-75 years
• Lab Values: Physiologically realistic (e.g., ALT 7-56 U/L)
• Vital Signs: Normal ranges (e.g., SYSBP 90-140 mmHg)
• Doses: 0-100 mg as defined

Cardinality Checks

• DM: 1 row per subject
• AE: Multiple rows per subject (avg 3)
• LB: 500 × 8 tests × 4 visits = 16,000 rows
• VS: 500 × 7 params × 4 visits = 14,000 rows
• EX: 500 × 3 visits = 1,500 rows

Referential Integrity

• All USUBJID in AE, LB, VS, EX must exist in DM
• All VISITNUM should be 1-4 (with EX having 2-4)
• All SITEID should be SITE01-SITE05

Reproducibility

• Random seed: 42
• All values are deterministic
• Regenerating with same script produces identical data

Support Files

• README.txt - Detailed variable descriptions
• QUICKSTART.md - This guide
• GENERATION_SUMMARY.md - Complete technical documentation

Questions?

Refer to README.txt for detailed variable definitions and data specifications.

clinCompare documentation built on Feb. 19, 2026, 1:07 a.m.