Every clinical study report rests on a package of tables, figures and listings, usually called the TFLs. They are produced by the statistical programming team from the analysis datasets, and the CSR text refers to them hundreds of times. When the numbering is clean, a reviewer can move from a sentence in section 11 to the table behind it in seconds. When it is not, every cross-reference becomes a question.
This guide explains what ICH E3 does and does not say about TFL numbering, the convention most sponsors have settled on, and how to keep the text and the TFLs in step.
What ICH E3 actually specifies
E3 does not prescribe a numbering scheme for individual tables. What it does is define two places in the report where TFLs live, and the headings inside them.
Section 14, "Tables, figures and graphs referred to but not included in the text." E3 says important demographic, efficacy and safety data should be summarised in the text, "however, if these become obtrusive because of size or number they should be presented here, cross-referenced to the text." It then sets out the headings:
- 14.1 Demographic data
- 14.2 Efficacy data
- 14.3 Safety data
- 14.3.1 Displays of adverse events
- 14.3.2 Listings of deaths, other serious and significant adverse events
- 14.3.3 Narratives of deaths, other serious and certain other significant adverse events
- 14.3.4 Abnormal laboratory value listing (each patient)
Section 16.2, "Patient data listings." These are the subject-level listings:
- 16.2.1 Discontinued patients
- 16.2.2 Protocol deviations
- 16.2.3 Patients excluded from the efficacy analysis
- 16.2.4 Demographic data
- 16.2.5 Compliance and/or drug concentration data
- 16.2.6 Individual efficacy response data
- 16.2.7 Adverse event listings (each patient)
- 16.2.8 Listing of individual laboratory measurements by patient, when required by regulatory authorities
Section 16.1 holds study information (protocol, sample CRF, statistical documentation and so on), 16.3 holds case report forms, and 16.4 the individual patient data listings that some authorities require.
The convention most sponsors use
Because E3 gives the section headings but not table numbers, the industry has converged on extending E3's numbers with further levels. A typical scheme:
| Output | Example number | Lives in |
|---|---|---|
| Summary table, demographics | Table 14.1.1 | Section 14.1 |
| Summary table, primary efficacy | Table 14.2.1.1 | Section 14.2 |
| Figure, Kaplan-Meier of primary endpoint | Figure 14.2.1.1 | Section 14.2 |
| Summary table, TEAEs by SOC and PT | Table 14.3.1.2 | Section 14.3.1 |
| Listing, deaths | Listing 14.3.2.1 | Section 14.3.2 |
| Listing, adverse events by patient | Listing 16.2.7.1 | Section 16.2.7 |
The first three levels come straight from E3. The fourth level and beyond are the sponsor's own, usually defined in the statistical analysis plan's table shells or a separate TFL specification. That is a convention, not a requirement, and you will see variations: some sponsors restart numbering per section, some use a single running number with the E3 section as a prefix, and some put efficacy figures in their own sequence.
The E3 Questions and Answers (R1) support that flexibility. E3 "is a guidance, not a set of rigid requirements or a template," and the Q&A quotes E3's own introduction: the sequence and grouping of topics may change if alternatives are more logical, and "the numbering should then be adapted accordingly."
In-text tables versus section 14 tables
Most CSRs have two kinds of tables:
- In-text tables sit in sections 10 to 12 of the report body. They are short, summarise the main result, and are numbered by their position in the text (for example, "Table 11-3").
- Section 14 tables are the full outputs from the TFL package, numbered in the 14.x scheme.
Every in-text table should say which section 14 table it comes from. That way a reviewer can always find the full version, and the QC team can check that the numbers match.
The E3 Q&A gives one concrete example of where things go. The example adverse event table printed in E3 section 12.2.2, with subject identifiers, "is not meant to be presented in section 12.2.2 of the report, but in section 14.3.1." The text should carry a summary table of common adverse events, comparing treatment groups, without subject identifiers or verbatim terms.
Where the CSR sits in the submission
In a CTD submission the whole CSR, including its section 14 and 16 outputs, goes into Module 5, under the study report sections set out in ICH M4E(R2). The numbering inside the CSR is independent of the CTD numbering, which is one reason it pays to keep the E3 structure recognisable: reviewers in every region know where to look.
Keeping text and TFLs in step
TFL numbering problems almost always come from change: a table added late, a shell renumbered after a SAP amendment, or a rerun that shifts outputs. A few habits prevent most of them.
- Fix the TFL list before programming starts. Agree the shells and their numbers in the SAP or TFL specification, and treat renumbering after that point as a change that needs sign-off.
- Never reuse a retired number. If Table 14.2.3 is dropped, leave the gap. Reusing it means every old cross-reference silently points to the wrong table.
- Cross-reference by number and title. "Table 14.3.1.2 (TEAEs by SOC and PT, safety population)" survives renumbering better than a bare number, because the mismatch is obvious.
- Use the same population labels everywhere. If the table says "Safety Analysis Set", the text should not say "safety population" in one place and "treated set" in another.
- Reconcile at the end. Before sign-off, check that every TFL cited in the text exists in the final package with the same number and title, and that every number quoted in the text matches its source table.
Common reviewer findings
- Text cites a table that no longer exists, or exists under a different number.
- In-text tables with values that differ from the section 14 table after a rerun.
- Listings for deaths and SAEs that do not reconcile with the narratives in 14.3.3.
- Sections of 16.2 left out without explanation, even where E3 says the content is relevant.
- Mixed numbering styles in one report, often where two programming vendors contributed.
Where automation helps
The cross-reference check in step 5 is mechanical and slow by hand: hundreds of citations, each to be matched against a table number, title and value. Asthra's CSR writing tool drafts the report text from your TFL package and cites each number to the output it came from, so mismatches show up as flags rather than reviewer questions. Decisions about structure and presentation stay with the writer and statistician.
FAQ
Does ICH E3 require tables to be numbered 14.x.x? E3 defines the section headings (14.1, 14.2, 14.3 and their subsections, and 16.2.1 to 16.2.8) but not individual table numbers. Extending E3's numbers with further levels is an industry convention, set in the sponsor's SAP or TFL specification.
Can I change the E3 structure? Yes. The E3 Q&A (R1) confirms E3 is guidance, not a template, and that numbering should be adapted if the structure is changed. If you leave out a section that is relevant to the study, say so and give the reason.
Where do patient narratives go? In section 12.3.2 of the text if there are few, or in section 14.3.3 if there are many. See our guide on writing SAE and death narratives for a CSR.
Sources
- ICH E3: Structure and Content of Clinical Study Reports, sections 12.2.2, 14 and 16
- ICH E3 Questions and Answers (R1), FDA guidance page
- ICH M4E(R2): The CTD, Efficacy