cSDRG Explained: Writing the Clinical Study Data Reviewer's Guide for SDTM (with the PHUSE Template)

By Mohan Peddineni · · 1 min read

The Clinical Study Data Reviewer's Guide (cSDRG) is the SDTM half of the reviewer's guide pair. It sits next to your SDTM datasets and define.xml, and it answers the questions a reviewer has when the tabulation data does something they didn't expect. For example: a sponsor-defined domain, a SUPPQUAL that carries half the CRF, or a validation rule that still fires.

The analysis-side document, the ADRG, is covered in its own post. Here we stay on the SDTM side: what goes in each section of the PHUSE template, and how to write it so it saves the reviewer time.

Get the PHUSE cSDRG template

PHUSE publishes the cSDRG Package (version 1.3, November 2018). It contains the template, completion guidelines, completed examples and example traceability diagrams. It also includes an appendix for legacy data conversion, which you need when SDTM was converted from older non-CDISC data.

PHUSE cSDRG Package (template, completion guidelines, examples)

The examples are worth reading even if you have written a cSDRG before. They show the level of detail reviewers expect, which is usually less prose and more specifics than people write.

What goes in each section

1. Introduction

The purpose, acronyms, and the standards inventory: SDTMIG and SDTM versions, define.xml version, controlled terminology package date, MedDRA and WHODrug versions. As with the ADRG, take these from define.xml so they match exactly.

2. Protocol description

The protocol number and title, a short design summary, and a description of how the trial design datasets (TA, TE, TV, TI, TS) represent it. Note anything a reviewer could trip over, like an arm in TA that no subject was randomised to, or visits in TV that were removed by an amendment.

3. Subject data description

This is the core of the document.

  • Overview: anything that applies across domains. Examples are screen failures included in DM, how USUBJID is built, re-screened subjects, or data from an extension study.
  • Annotated CRF: a pointer to acrf.pdf, plus anything collected but not submitted ("Not Submitted" annotations) and why.
  • Domain table: every submitted domain with its observation class, key variables, and whether it has a SUPP-- dataset.
  • Domain-specific notes, only where needed: a custom domain and why no standard domain fit, how --STRESC was standardised in LB, why some EXSTDTC values are partial, or how central and local labs were combined.
DomainLabelClassKey variablesSUPP
DMDemographicsSpecial PurposeSTUDYID, USUBJIDYes
AEAdverse EventsEventsSTUDYID, USUBJID, AEDECOD, AESTDTCYes
EXExposureInterventionsSTUDYID, USUBJID, EXTRT, EXSTDTCNo
LBLaboratory Test ResultsFindingsSTUDYID, USUBJID, LBTESTCD, VISITNUM, LBDTCNo

4. Data conformance summary

Name the validator, version and rule set, and the date you ran it. Then add a table of remaining issues: rule ID, message, domain, record count, and a specific explanation. Group issues that share a cause rather than repeating the same sentence twenty times.

RuleDomainCountExplanation
(validator rule ID)LB14LBSTRESN is null where the central lab reported "<0.1". The original value is kept in LBORRES/LBSTRESC; no numeric conversion was agreed in the data transfer specification.

The example is illustrative. Always quote the exact rule ID and message from your own validation report.

5. Appendix

Use the legacy data conversion appendix when it applies, plus any extra traceability you need, such as a diagram showing how source data flowed into SDTM.

Writing tips from the programming side

  • Keep it in step with define.xml. Every SUPPQUAL, value-level metadata and codelist you describe should match the define exactly.
  • Explain decisions, not mechanics. The reviewer doesn't need your merge logic. They need to know why a mapping choice was made.
  • Be honest about data issues. Known issues like protocol deviations or a site with odd data belong here. A reviewer who finds them on their own will ask why you didn't mention them.
  • Start early. Write the domain notes as you map. By submission, nobody remembers why a variable sits in SUPPAE.
  • Remove template guidance text and empty headings before you create the PDF. If a section doesn't apply, say "Not applicable" in one line.

cSDRG or ADRG? Where each topic belongs

TopicDocument
Why a sponsor-defined domain was createdcSDRG
How screen failures are represented in DMcSDRG
How the safety population flag is derivedADRG
Why some EXSTDTC values are partialcSDRG
How partial dates are imputed for analysisADRG
How central and local labs were combined in LBcSDRG
Visit windowing for analysisADRG

The rule of thumb: if it describes what was collected and how it was tabulated, it goes in the cSDRG. If it describes an analysis decision, it goes in the ADRG.

Try it yourself: write the LB domain note

Your study has safety labs from a central laboratory, plus local lab results entered on the CRF for some unscheduled hematology tests. Both are in the LB domain. Write the short cSDRG note a reviewer needs.

Answer (one good version): "LB contains results from the central laboratory and from local site laboratories. Local results were collected on the CRF for unscheduled hematology tests not sent to the central lab. The source is identified by LBNAM. Original results and units are kept in LBORRES/LBORRESU, and all results were converted to SI units in LBSTRESC/LBSTRESN/LBSTRESU. Reference ranges come from the lab that performed each test (LBSTNRLO/LBSTNRHI), so ranges can differ between sites." It says where the data came from, how to tell the sources apart, and what a reviewer should expect when comparing values.

Key takeaways
  • The cSDRG explains your SDTM package. The ADRG explains your ADaM package. Both go in Module 5 next to the datasets and define.xml.
  • Start from the PHUSE cSDRG template and its completed examples.
  • Collection and tabulation topics go in the cSDRG. Analysis decisions go in the ADRG.
  • Put the detail where reviewers need it: the subject data description and the conformance explanations.

Related: ADRG explained · All eSubmission articles · SDTM articles

MP
Written by Mohan Peddineni

Manager, Statistical Programming. Writes about SAS, R, CDISC and regulatory submissions. More about me → · LinkedIn

↑