The trace

One value, followed through all five stops.

The study's primary result is the average drop in HbA1c on the 25 mg dose at Week 26: −1.14 percentage points. Below is where that number came from — top to bottom, each stop a real surface of the study served as-is, with the rule that made it, the document that authorised the rule, and the team responsible. In the study itself this is one click on the table cell.

Stop 1 · OutputsStatistics designs it programming produces it writing describes it

The number in the table.

Table 14.2.1 · finaltlf:as-locked/T-14-2-1/r14/c3

**Table 14.2.1** Least-squares mean change from baseline in HbA1c (%) by visit — Full Analysis Set
Week 26	Placebo (N=128)	10 mg (N=127)	25 mg (N=129)
n	117	116	118
Change, mean (SD)	−0.31 (0.912)	−1.01 (1.008)	−1.16 (1.000)
LS mean change (SE)	−0.32 (0.09)	−0.97 (0.09)	−1.14 (0.09)
Difference vs placebo (95% CI)		−0.65 (−0.90, −0.40)	−0.82 (−1.07, −0.57)
p-value		<0.0001	<0.0001
MMRM on change from baseline over the 118 contributing subjects (SAP §7.1), Full Analysis Set. Served from the study; every cell opens.

One number summarising 118 people. The writer will put it in a sentence. The statistician chose the model. Neither of them typed anything that went into it.

Stop 2 · AnalysisStatistics specifies programming builds

One of the 118 rows behind it — and the row that didn't count.

ADaM · ADLB · subject 0104-1003 · PARAMCD HBA1Cadam:as-locked/ADLB/NVT501301-0104-1003/44

ASEQ	AVISIT	ADT	ADY	BASE	AVAL	CHG	ANL01FL
14	Baseline	2023-12-25	−17	8.9	8.8	−0.1	N
6	Baseline	2024-01-11	1		8.9		Y
44	Week 26	2024-07-09	181	8.9	7.1	−1.8	Y

CHG = AVAL − BASE = 7.1 − 8.9 = −1.8. The baseline comes from the record flagged as baseline (SAP §4.1) — and two results fell in the baseline window, one at screening and one on Day 1. ANL01FL says which counts, by the rule in SAP §4.3: closest to the target day. The screening result is in the dataset and in no analysis — the layer a data manager rarely sees.

Stop 3 · TabulationProgramming maps what data management locked

The same result, in the standard's shape.

SDTM · LB · subject 0104-1003 · LBSEQ 44sdtm:as-locked/LB/NVT501301-0104-1003/44

USUBJID	LBTESTCD	LBORRES	LBORRESU	LBSTRESN	VISIT	LBDTC	LBDY
0104-1003	HBA1C	7.1	%	7.1	Week 26	2024-07-09	181

Nothing is invented here. LBDY = LBDTC − RFSTDTC + 1 = 181; LBSTRESN equals LBORRES because the laboratory already reports in the standard unit. Each column is a named rule — and each is a hop the study can show you.

Stop 4 · CaptureData management, the site, and a laboratory

A date a person typed. A result nobody at the site ever saw.

eCRF · Laboratory Collectioncrf:-/LC/LCDAT

Date sample collected

09JUL2024

Entered 2024-07-11 09:14 by the study coordinator, site 0104 · no queries on this subject

Central lab transfer · 686 recordsCLAB-20240715

SUBJECT    ANALYTE  RESULT  UNIT  COLLECTED
0104-1003  HBA1C    7.1     %     2024-07-09
… 685 more records, 49 subjects
transferred 2024-07-15

The HbA1c value was never typed into the CRF. It arrived in a file from the central laboratory six days later and was matched to the visit on subject and collection date. If the two disagree, that's a query — and the statistician will never know it happened unless they come here.

Stop 5 · DocumentsWritten before any data existed — every stop above cites one of these

The paragraph that put the field on the form.

Protocol §2.1 · SAP §4.3doc:-/protocol/2.1

Protocol 2.1 — Objectives and endpoints

To demonstrate the superiority of sematiglizin over placebo in glycaemic control after 26 weeks of treatment. Primary endpoint: change from baseline in HbA1c (%) at Week 26.

SAP 4.3 — Analysis visits and windows

Where more than one record for a parameter falls in the same window, the record selected for analysis is the one whose study day is closest to the target day. The selected record is flagged ANL01FL = 'Y'.

The endpoint decided the form needed a collection date. The SAP decided which of two baseline results would count, months before either existed. The people who built the form and the people who wrote the SAP mostly never met — this is where they do.

Trace this number in the study →

Every one of the study's 3.4 million lineage edges is a step like these.

In the studyTrace it yourself

The same cell, live — open the fan of 118 contributing rows and walk any of them down.

Back toWhy it matters

Four handoffs where everyone did their job right and the study still suffered.

PlansThe tour and this trace are free

The full study — every stage, handoff and role door — is a plan. See what's included.