The ICH publishes dozens of guidelines, and a new regulatory writer can spend a lot of energy feeling like they should have read all of them. You don't. There's a core set that most clinical and safety writers rely on week to week, though the exact set depends on what you write — requirements differ substantially among clinical, safety, CMC, and nonclinical writers. The rest range from "useful to have read once" to "someone on the CMC side owns this, not you." Knowing which is which is most of the battle.

This post covers the ones that matter most for a clinical or safety writer, with the sections you touch regularly and the one thing people most often get wrong about each. Then it runs quickly through the guidelines worth knowing exist so you recognize them when they come up. If you want the wider context of how these documents fit into drug development, the complete guide to regulatory writing is the place to start.

ICH E3 — the shape of a Clinical Study Report

E3, "Structure and Content of Clinical Study Reports," is the one you will live inside if you write CSRs. It lays out a 16-section skeleton: title page, synopsis, table of contents, abbreviations, ethics, investigators and administrative structure, introduction, objectives, investigational plan, study patients, efficacy evaluation, safety evaluation, discussion and overall conclusions, tables and figures, references, and appendices. Sections 9 through 13 — the plan, the patients, the efficacy and safety evaluations, and the discussion — are where you spend almost all of your time.

The gotcha: writers treat the 16-section list as a rigid template that must be reproduced exactly, empty headings and all. It isn't. In 2012 the ICH published an E3 Questions and Answers document precisely to clarify that E3 is guidance on content, not a mandatory format. You can reorganize, combine, or relocate material to suit the study, as long as the expected content is present and findable. The practical companion most experienced writers reach for is CORE Reference (Clarity and Openness in Reporting: E3-based), a jointly developed EMWA/AMWA resource that turns E3 plus later expectations into a working authoring reference. If you only internalize one guideline deeply, make it this one.

ICH E6(R3) — Good Clinical Practice, and why writers need it

E6 is Good Clinical Practice, and the R3 revision restructured it into a set of overarching Principles plus annexes. The Principles and Annex 1 (interventional trials) were adopted at ICH Step 4 on 6 January 2025; Annex 2, covering trials with decentralized, pragmatic, and/or real-world-data elements, reached Step 4 on 3 June 2026. The consolidated E6(R3) guideline, dated 16 June 2026, brings the Principles, Annex 1, and Annex 2 together — though regional implementation still varies. Writers sometimes assume GCP belongs to clinical operations and monitors, not to them. It doesn't stay in that lane. E6 defines the essential records a trial must generate and retain, and several of them — the protocol, the investigator's brochure, and the CSR itself — are documents you draft. It also sets expectations for what a protocol must contain and how changes to it are controlled.

The gotcha: confusing which documents are "essential" and who owns them, and assuming GCP is a QA concern rather than a drafting one. When a reviewer or auditor traces a claim in your CSR back through the protocol and its amendments, the trail E6 expects is one you helped build. Read it as a description of the paper reality your documents have to fit into.

ICH E9 and E9(R1) — describing the analysis correctly

E9, "Statistical Principles for Clinical Trials," is a statistician's document, but the 2019 addendum E9(R1) is one every writer needs to understand, because it changed how you are expected to describe what a trial actually measured. E9(R1) introduced the estimand framework: a precise description of the treatment effect built from five attributes — the treatment, the population, the endpoint (variable), the strategy for handling intercurrent events, and the population-level summary measure.

The gotcha: writing "the analysis was performed on the ITT population" as if that fully specifies the estimand. It doesn't. The population is one of five attributes, and the one writers most often skip is the intercurrent-event strategy — what happens to the analysis when a patient discontinues treatment, switches therapy, or dies. If you describe endpoints without saying how those events were handled, you have described an incomplete estimand — not a full one. (And an estimand is not the same thing as an estimator: the estimator is the statistical method used to estimate the estimand, not the target being described.) You don't have to derive any of this yourself; you have to render the statistician's estimand faithfully and keep it consistent everywhere it appears, from the synopsis to the efficacy section.

ICH E2F — the DSUR

E2F defines the Development Safety Update Report, the annual safety report for drugs still in development. If you write DSURs, this is your spec. It sets the reporting interval anchored to the Development International Birth Date, the content of the report, and the expectation of cumulative summary tabulations of serious adverse events across your investigational studies.

The gotcha: treating the DSUR like a periodic safety report for a marketed product. It isn't. Its scope is investigational, its reference for expectedness is the reference safety information in the investigator's brochure, and it carries both interval line listings of serious adverse reactions and cumulative summary tabulations of serious adverse events — not one in place of the other. Getting the reporting period and the data lock right, and reconciling the cumulative counts against the source safety data, is the actual work.

ICH E2C(R2) — the PBRER

E2C(R2) governs the Periodic Benefit-Risk Evaluation Report, which is what most people still call the PSUR. The name change is not cosmetic. The report is a benefit-risk evaluation for a marketed product, not just a safety data dump, and its interval is anchored to the International Birth Date.

The gotcha: calling it a PSUR and, more importantly, writing it like one — a catalog of adverse events with no integrated benefit-risk conclusion. E2C(R2) expects you to weigh the accumulated safety findings against the established efficacy and arrive at a reasoned position, with expectedness judged against the reference product information. The cumulative-table arithmetic and the listedness checks are the mechanical part; the benefit-risk narrative is the point.

ICH M4 — how the whole dossier is organized

M4 defines the Common Technical Document, the CTD, which is the organizing structure for a marketing application: Module 1 (regional administrative), Module 2 (summaries), Module 3 (quality), Module 4 (nonclinical), Module 5 (clinical). The M4 family splits into M4Q for quality, M4S for safety, and M4E for efficacy (Module 2.5 and 2.7). Nearly every document you write lands somewhere in this structure.

The gotcha: forgetting that the CTD is an organizational standard, not a content standard, and duplicating instead of summarizing — writing a Module 2 overview that simply repeats Module 5 rather than distilling it. If you're fuzzy on how the five modules relate, CTD explained: Modules 1–5 walks through it.

Worth knowing exists

You don't need to master these, but you should recognize them:

  • E2A–E2D — the clinical safety cluster: definitions and standards for expedited reporting (E2A), electronic transmission of individual case reports (E2B), and post-approval safety data management (E2D). This is where terms like "serious," "unexpected," and "expedited" are actually defined.
  • E8(R1) — General Considerations for Clinical Studies, revised in 2021 around quality-by-design. Useful background for anyone drafting protocols.
  • M11 — a harmonized clinical protocol template: a common structure and standardized content for study protocols so the same information lives in the same place across sponsors. The M11 guideline, template, and technical specification reached ICH Step 4 on 19 November 2025, though regional implementation still varies. As it is adopted it will shape how protocols are built.
  • The Q-series — quality guidelines covering stability (Q1), analytical validation (Q2), impurities (Q3), specifications (Q6A), and more. If you ever touch CMC / Module 3, these become relevant fast.

Why this list is good news for automation

Notice what these guidelines have in common. E3 gives a CSR its sections. E2C(R2) and E2F set the shape of the periodic reports. M4 organizes the dossier. E9(R1) defines which attributes an estimand description must carry. These documents are largely template-bound by design — much of the purpose of harmonization.

That structure is what makes a lot of regulatory drafting automatable. When a document's skeleton, expected content, and cross-references are specified in advance, software can hold those rules as constraints and render the draft against them, leaving the writer to do the interpretation the guidelines can't encode — which results are pivotal, whether the benefit-risk balance has shifted, how to read an ambiguous signal. That is the division of labor behind regulatory writing automation: the guideline supplies the form, the source documents supply the facts, and the writer supplies the judgment.