Reference safety information: how the RSI decides what counts as expected

Reference safety information, or RSI, is the list of expected serious adverse reactions for an investigational medicinal product, and a sponsor compares every suspected serious adverse reaction against it to decide whether that reaction is expected. The CTFG Q&A on Reference Safety Information (November 2017) states its primary purpose: to serve as the basis for expectedness assessments of suspected serious adverse reactions by the sponsor, for expedited reporting of suspected unexpected serious adverse reactions and for annual safety reporting. One list therefore decides two things: which cases leave the building as SUSARs, and which reactions count as expected in the development safety update report.

What reference safety information contains

The CTFG describes the RSI as a list of expected serious adverse reactions classified using Preferred Terms according to MedDRA. Three rules shape what gets on it.

Only serious reactions qualify. The CTFG says the RSI section of an investigator's brochure should contain only expected serious adverse reactions to the investigational product, and that a broader description of the safety profile belongs elsewhere in the brochure. That broader description is where non-serious adverse reactions sit, along with suspected serious adverse reactions that have occurred only once, and fatal and life-threatening reactions that are considered unexpected.

Causality has to be established first. A reaction earns its place when, after a thorough assessment by the sponsor, reasonable evidence of a causal relationship between the event and the product exists. The CTFG notes that a suspected serious adverse reaction occurring more than once is not on its own sufficient to make it expected.

Fatal and life-threatening reactions are treated apart. MHRA's inspectorate is direct about it in part III of its RSI series: "fatal and life-threatening (LT) SARs should not be considered expected (unless explicitly stated in the RSI and approved by the NCA)."

Where the RSI sits, and how it is laid out

If the sponsor prepares an investigator's brochure for the products in a trial, the RSI belongs there, in a separated specific section of its own. For an authorised product, an approved summary of product characteristics may serve as the RSI. MHRA adds a limit on that shortcut: copying section 4.8 of the summary of product characteristics straight into the brochure is not acceptable, although the summary itself may be used as RSI where there is a good rationale for using it to support the conduct of the trial.

The CTFG asks for a specific shape. The RSI is presented as a table, listing reactions by system organ class using Preferred Terms from the latest MedDRA version, followed by the frequency, which is calculated on an aggregated level from previously observed suspected serious adverse reactions to the product.

Frequencies follow the categories used for the summary of product characteristics, Very Common, Common, Uncommon and so on, with the CTFG pointing to ICH E2C(R2) for the detail. Where too few subjects have been exposed for those categories to mean anything, which is normal early in development, the sponsor gives the number of observed suspected serious adverse reactions together with the number of patients exposed. MHRA rules out one answer entirely: a frequency of "not known" is unacceptable, because it makes it impossible to assess whether new reports show an increased occurrence.

How the RSI turns a listed reaction into an unexpected one

Expectedness is decided at Preferred Term level, and a term that looks listed can still produce a SUSAR. The CTFG answers the question plainly: "A provision of severity grades using CTCAE grading system in the RSI is not required. However, in accordance with ICH E2A guidance, reports which add significant information on specificity or severity of a known, already documented SAR represent unexpected events."

MHRA states the consequence as a rule for the assessor: "Any SAR which adds additional information about the specificity and/or severity of an expected reaction must be considered unexpected."

The CTFG's own worked examples show how little distance it takes.

Listed in the RSI	Reported in the case	Unexpected because of
Acute renal failure	Interstitial nephritis	Specificity
Hepatitis	Fulminant hepatitis	Severity
Cerebral vascular accident	Cerebral thromboembolism	Specificity
Exfoliative dermatitis	Stevens-Johnson syndrome	Severity and specificity
Transient increase in liver function tests	Increased liver function tests persisting for several months	Severity

Source: CTFG Q&A on Reference Safety Information, table 2.0.

Read that table from the site's side and every decision turns on the words in the case: hepatitis against fulminant hepatitis, a transient rise against one that persisted for months. A case recording "liver enzymes up, drug related" gives the assessor nothing to compare, so the comparison waits on a query.

Which RSI version applies when a reaction occurs

The onset date selects the version, and nothing later moves it. The CTFG's answer to its eighteenth question: "The RSI in place at the time of occurrence of the suspected SAR should be used to assess expectedness for follow up reports to Eudravigilance (EV). SUSARs should not be downgraded in EV on the basis that the RSI was updated after the occurrence of the event."

MHRA puts the same rule in one sentence: "The version of the RSI at the moment of occurrence of the SUSAR applies." A case reported late, or one whose follow-up arrives months afterwards, is still assessed against the RSI that was in force on the day the reaction began.

What an approved RSI update changes mid-trial

An updated RSI starts working only once it is approved everywhere the trial is running. The CTFG: the newly updated RSI can be used for expectedness assessments for expedited reporting only after approval of the substantial amendments in all of the member states where trials are ongoing, and any additional SUSARs that occur before the new RSI is approved are reported as SUSARs in the usual expedited manner.

The annual report runs on a different selection again. For a development safety update report, whose format comes from ICH E2F, the CTFG asks for the version of the RSI most recently approved in all member states, and its worked example keeps the older brochure as the RSI for that reporting period while both the old and the new brochure are submitted with the report. The reporting period is assessed on one baseline from end to end.

So a single trial can be running three selections at once: the in-force RSI for cases arriving today, an earlier RSI for a case whose onset predates the update, and the period's baseline RSI for the annual report. MHRA's advice follows from that, which is clear change control with documented tracking of when each approval happened and when it was implemented across every affected trial.

How US regulation asks the same question

The term RSI belongs to the EU and UK framework. US regulation reaches the same decision through the brochure itself. 21 CFR 312.32(a) defines it: "An adverse event or suspected adverse reaction is considered 'unexpected' if it is not listed in the investigator brochure or is not listed at the specificity or severity that has been observed."

The two deciding words are the same two words, specificity and severity. What differs is the packaging: an EU or UK sponsor compares against an approved RSI section with approved frequencies, while a US sponsor compares against the brochure as a whole. Both comparisons start the clocks in 21 CFR 312.32 for IND safety reports.

What the expectedness check asks of the clinical record

For a sponsor safety team, this is label work. The expectedness decision made during the trial determines which reactions reach the agency as SUSARs and which are carried as expected in the annual report, and that record is what the warnings, the monitoring schedule and the dosing language are eventually written from. A reaction recorded precisely enough to compare against a Preferred Term helps elucidate the safety profile while the trial is running, when it still shapes the label.

Two things decide whether a case can be assessed on arrival: a term specific enough to sit beside the RSI entry, and attribution scored against every drug in the regimen with its reasoning available to read. Attribution is the unstable half: a recurring adverse event's attribution changes between visits for the same patient in 31 to 36% of cases (Hillman et al., JCO 2010). The coding half runs on MedDRA terms and the consistency of that coding, which is also what makes the annual tabulation readable.

How Burna AI records the event the check runs on

Burna AI works on the site side of this path, where the term and its evidence are produced.

• Every adverse event in a note is graded against the CTCAE version the protocol names, v5.0 or v6.0, with the MedDRA Lowest Level Term assigned, so a graded event arrives carrying a candidate code the safety team confirms in the MedDRA version the report declares. Look a term up at ctcae.ai.
• Each event is scored for attribution against every drug in the regimen using the WHO-UMC and Kramer algorithms, with the reasoning shown.
• Every grade traces to the line in the medical record that produced it, and a grade whose citation set is incomplete is flagged before it is signed.
• A grade becomes a record when a qualified investigator signs it with a 21 CFR Part 11 electronic signature, and the signature carries what was accepted, what was overridden and the citation set as signed.
• The platform produces FDA MedWatch, IRB and serious adverse event reports from the graded record. The dated capability record lists what is available today.

The RSI, the expectedness assessment and the SUSAR decision stay with the sponsor's safety team, where the CTFG and 21 CFR 312.32 both put them. What changes is the quality of the term those decisions run on. AI suggests, clinicians decide. Human-in-the-loop, always.

Common questions about reference safety information

Is the RSI the same as the investigator's brochure?

The RSI is one section inside it. The CTFG asks for the RSI to sit in a separated specific section of the brochure, holding only the expected serious adverse reactions, while the rest of the brochure carries the wider safety profile including non-serious reactions and reactions seen only once.

Does a CTCAE grade decide whether a reaction is expected?

A grade measures severity, and the CTFG says severity grades using the CTCAE system are not required in the RSI. A grade still matters to the decision, because a report that adds significant information on the severity of an already documented reaction is unexpected. The comparison itself runs on MedDRA Preferred Terms.

Who performs the expectedness assessment?

The sponsor. MHRA describes the assessment as serving the sponsor's reporting obligations and the regulator's evaluation of risk and benefit rather than being investigator-driven. The investigator's judgment of seriousness and causality feeds it, and the reporting determination belongs to the sponsor.

Can an updated RSI be applied to cases that have already happened?

No. The CTFG is explicit that a SUSAR should not be downgraded on the basis that the RSI was updated after the event occurred, and that the new RSI applies to expectedness assessments only after the substantial amendment is approved in every member state where the trial is running.

Do fatal reactions ever appear in the RSI as expected?

MHRA's position is that fatal and life-threatening serious adverse reactions should not be considered expected unless the RSI explicitly says so and the national competent authority has approved it.

---

The AI platform for accelerated clinical research, from first-in-human trials through postmarket surveillance. Burna AI performs automated adverse event grading and attribution in real time, providing clear rationale and audit-ready evidence, including generation of IRB, sponsor and FDA reports. Burna AI delivers CTCAE v5 and v6 grades and MedDRA LLT codes on every encounter; enabling standardization, cross-site consistency and accuracy of clinical research data. AI suggests, clinicians decide. Human-in-the-loop, always.

If your safety team is waiting on queries before it can assess expectedness, we can show you what arrives when the term and its citation come from the note itself: book a 15-minute conversation.