Clinical trial documents: anonymising reports before publication

When a pharmaceutical company applies to sell a medicine in the EU, regulators later publish the clinical study reports behind that decision. Before that happens, someone has to strip out the patient data. EMA and the EU's Clinical Trials Information System, CTIS, each set rules for how.

This guide is for people who prepare or review those reports: sponsors, CROs, medical writers. It explains what gets published, what must be anonymised, and why patient narratives are the hardest part. It is not legal advice; a specific case should go to your regulatory affairs team or to EMA directly.

What do EMA and CTIS actually publish?

EMA adopted Policy 0070 on 2 October 2014, and it took effect on 1 January 2015, according to EMA's own implementation guidance. Since 2016, EMA has published clinical study reports, protocols and statistical documents for centrally authorised medicines once a marketing decision is made. EMA paused this publication for medicines other than COVID-19 treatments. It resumed for medicines with a CHMP opinion from September 2023 onwards, and the first packages appeared in January 2024.

CTIS covers something wider: every interventional clinical trial run in the EU or EEA, not just approved medicines. It has been live since 31 January 2022. Structured fields, such as the trial title and endpoints, are public from the day a country decides on the trial application. Documents like the protocol follow, redacted where needed.

What is the difference between anonymisation and redaction?

EMA's guidance treats the two as different tools. Redaction means blacking out a value completely, such as a patient's name or signature. It is simple, but it removes information for good. Anonymisation instead transforms the data: it can shift a date, or turn an exact age into a range, while keeping the report useful for research.

Every applicant must fill in an Anonymisation Report Form for each data package. EMA reviews it to check the applicant followed its guidance and applied its chosen method consistently throughout the report. The form is not optional, and an incomplete package is sent back for the applicant to fix.

Why are patient narratives and listings the hardest part?

A structured table is easy to anonymise: mask one column, done. A narrative paragraph describing one patient's case is not. EMA's guidance flags rare diseases as a particular risk: in the EU, a disease affecting no more than 5 people in 10,000 counts as rare, and a trial for one may have very few participants to hide among.

What anonymisation techniques does EMA recommend?

EMA's guidance names three main techniques. Masking, also called redaction, blocks out a value entirely and works best for direct identifiers such as a name or an address. It is the simplest option, but on its own it can leave a report harder to read.

• Randomisation: change a value's truth without deleting it. The usual form is date offsetting. Give each patient one random shift, applied to every date they have, so the order of their visits stays intact.
• Generalisation: widen the value, for example replacing an exact age of 56 with a 50-to-60 range.

EMA prefers offsetting and generalisation over plain masking, because they keep more of a report's scientific value. A date of birth, though, should still be redacted down to the year, EMA's guidance says, to protect age without losing it entirely.

Are investigators' names published too?

Patients are anonymised, but the people who ran the trial mostly are not. EMA's CTIS guidance lists names that stay in the published version: the principal investigator, the sponsor's and coordinating investigator's signatories, and whoever signed off the trial site as suitable. Their private phone numbers and emails are still redacted; only a professional contact stays public.

What should you check before you submit?

Health Canada runs a similar programme, called Public Release of Clinical Information. It aims to publish anonymised clinical data within about 120 days of starting the process, and asks for its own anonymisation report. Its guidance sets a re-identification risk threshold close to EMA's, and, with certification, lets a sponsor reuse documents EMA has already redacted.

• Confirm which document type you have: report, protocol, or listing, since the rules differ.
• Separate direct identifiers from indirect ones before choosing a technique.
• Check the rare-disease and small-population risk, even if no single value looks sensitive.
• Fill in the Anonymisation Report Form; do not leave it for the end.
• Keep a private, dated master copy: what you publish should never be your only record.

For a single report shared outside EMA's formal process, ONYRI Sanitize can help: it runs in your browser and masks names, dates, ID numbers and ICD-10 diagnosis codes. Type a subject ID pattern into Search the document, and it masks every occurrence. It does not generalise values, offset dates, write an anonymisation report, or handle a full submission package the way Policy 0070 requires.

None of this replaces reading EMA's own guidance for your specific report. The rules keep moving: CTIS tightened its transparency rules again in 2024, and EMA's external guidance was last updated in May 2025. Check the current version before you submit.