Device FIH safety reporting in Brazil is not drug pharmacovigilance with the logo swapped. The clocks are different, the form is different, and “SUSAR” is a useful overlapping idea — not the name of the ANVISA device channel.

If your FIH data are meant to support a later FDA investigational or marketing file, you still report to ANVISA under device rules. You then keep a narrative and causality trail that an IDE medical officer can read without asking why Brazil used a drug module.

The statute you actually report under

Current device clinical-investigation procedure is RDC 837/2023, announced by ANVISA in its December 2023 update. It replaced RDC 548/2021, which had already revoked RDC 10/2015. Chapter VII is the safety-monitoring chapter. Do not write SOPs against RDC 10/2015 clocks, and do not paste RDC 945/2024 drug-trial SUSAR text into a device FIH plan.

Law 14.874/2024 still applies to the ethics layer. Article 26 requires the sponsor to notify investigators, the institution, ethics bodies, and the sanitary authority of findings that may adversely affect participant safety. Article 55 makes serious adverse events reportable to the CEP that approved the research, and, for clinical trials intended to support sanitary registration, also to the sanitary authority. The law does not replace RDC 837/2023’s device form or day counts.

NotivisaEC is the device channel; VigiMed is not

ANVISA’s manual for adverse-event notification and safety monitoring in clinical trials involving investigational medical devices is the operational text. It tells sponsors to notify unexpected serious adverse events to ANVISA on the electronic NotivisaEC form:

https://pesquisa.anvisa.gov.br/index.php/847543?lang=pt-BR

The manual is explicit that those notifications go exclusively through that form, and that login is not required to notify. It also states that a SUSAR (suspected unexpected serious adverse reaction) sits inside the criteria for a notifiable serious event — but the RDC criteria are not limited to the drug-style SUSAR definition. If your safety database only flags “SUSAR = yes,” you will miss device cases that are unexpected, serious, and causally possible even when a medical monitor refuses the drug label.

VigiMed-Pesquisa Clínica is ANVISA’s channel for SUSARs in drug and biologic trials, under the drug clinical-trial framework (currently discussed by the Agency against RDC 945/2024). ANVISA’s own clinical-trial notification page describes VigiMed in that medicine context and requires a study-specific VigiMed registration, including the DEEC expediente. That is not the device DICD file. Do not open a VigiMed-Pesquisa Clínica account as a substitute for NotivisaEC on an implant FIH.

Post-market technovigilance (commercial devices after registro or notification) is a different duty and a different system generation. Keep investigational NotivisaEC cases out of the commercial technovigilance queue until the unit is actually on the market.

What is notifiable, in device language

RDC 837/2023 defines an adverse event as any unfavorable medical occurrence in a participant that is not necessarily causal. A serious adverse event includes death; a life-threatening event; persistent or significant disability; hospitalization or prolongation of hospitalization; congenital anomaly; suspected transmission of an infectious agent via a medical device; or a clinically significant event.

An event is unexpected when it is not described as an adverse reaction in the investigator brochure or in the instructions for use / operator manual of the investigational device. That is why the brochure is a reporting instrument, not a marketing appendix.

The sponsor must notify ANVISA of unexpected serious events that occurred in Brazil and whose causality versus the investigational product is possible, probable, or definite. Expected serious events, unrelated events, and non-serious events still belong in the sponsor’s file and in the annual tabulated report. They are not the 7- and 15-day electronic cases.

Causality is not a hallway opinion. The device manual points sponsors to the WHO-UMC causality scale (possible / probable / certain, plus the lower categories) when classifying every event, including non-serious ones. Train Brazilian investigators on that scale before site initiation. A PI who only knows “related / not related” will force you to re-code under the clock.

The clocks — investigator, sponsor, ANVISA

RDC 837/2023 is blunt on time. Count from knowledge, not from “when safety closed the query.”

• Investigator → sponsor: serious adverse events or death within 24 hours of the investigator becoming aware.
• Sponsor → ANVISA, fatal or life-threatening, unexpected, causality possible/probable/definite, Brazil: document and notify on the electronic form within 7 calendar days of sponsor awareness. Complementary follow-up goes on the same form within 8 calendar days after that initial notification.
• Sponsor → ANVISA, all other unexpected serious events meeting the same causality and territory tests: 15 calendar days from sponsor awareness.

Immediate protective measures are not optional. On a notifiable serious event, the form expects the measures already taken, the plan if the same event recurs, the care location, and identifiers that can trace the event and the participant. Notification is due even if you are in the middle of a brochure update, a protocol amendment, an annual report, or an early termination.

Temporary safety suspensions of the investigation must be notified to ANVISA within 7 calendar days of the suspension, with reasons, scope, treatment interruption, risk-minimization measures, and recruitment status. Restart waits for ANVISA approval published in the Official Gazette. That is a calendar item, not a medical-monitor footnote.

DSUR / ICH E2F is not a substitute for the device annual report

ANVISA publishes the ICH E2F Development Safety Update Report in its medicines clinical-trial library. E2F is a drug-development periodic-safety standard. It is a sensible internal template if the same legal entity also runs an IND. It is not the device annual report that RDC 837/2023 requires.

For a DICD, the sponsor files an annual follow-up report as a secondary petition to the DICD, covering Brazilian centers only, in tabulated form: title, reference number, recruitment status, amendment history, enrollment by site, deviations and violations by site, and a description of all adverse events in the period, with the case-report-form participant codes. The anniversary is the date of the Brazil start-of-investigation notification. The petition is due within 60 calendar days of that anniversary. Missing the report can cancel the investigation.

The final report is a different secondary petition, due within 12 months of the investigation’s end date, with demographics, statistical analysis, all events with causality, endpoint results, and any design changes. If you already produce a DSUR for a companion drug or combination product, map the Brazilian device tables into an annex. Do not file the DSUR and assume ANVISA has been served.

Pivotal high-risk (Class III/IV) designs are expected to constitute a data-monitoring committee. Recommendations go to ANVISA as a DICD addendum within 30 calendar days of the sponsor receiving them. An FIH that is not labeled “pivotal” can still justify a DMC on risk. Write that justification in the plan before first patient, including why you did not constitute one.

What must stay aligned if the FIH will later support FDA

FDA device investigations do not use the IND SUSAR clock. They use unanticipated adverse device effect (UADE) rules under 21 CFR 812.150. An investigator reports a UADE to the sponsor and reviewing IRB as soon as possible, and no later than 10 working days after first learning of it. The sponsor evaluates immediately and reports the evaluation to FDA, all reviewing IRBs, and participating investigators within 10 working days of first notice. If the UADE presents an unreasonable risk, termination clocks in 21 CFR 812.46 are 5 working days from that determination and 15 working days from first notice. FDA’s IDE FAQs repeat the same 10-working-day sponsor evaluation report.

Those clocks will not match Brazil’s 24-hour / 7-calendar / 15-calendar structure. Alignment is not “pick the longer one and hope.” Alignment is a single source narrative:

• One event identifier from the Brazilian CRF code through NotivisaEC and, if an IDE exists or will exist, through the UADE file.
• One expectedness baseline — the same brochure version cited in the DICD and in the IDE investigational plan. If Brazil updates the brochure after an unexpected event (RDC 837/2023 requires investigators to be informed and the brochure to be updated), send that version into the FDA file in the same week.
• One causality sentence that a device reviewer recognizes (device, procedure, disease, concomitant product). Do not let the Brazilian form say “possible” and the IDE memo say “unrelated” without a dated reconciliation.
• Device identifiers — lot, serial, software version, accessory — on every case. Import traceability under the DICD DOU number is the same data FDA will ask for when the unit later supports a 510(k), De Novo, or PMA.
• Quality-system evidence that the investigational unit was built under a system you can defend as you move toward the QMSR. A NotivisaEC case that cannot find the device history record is already a future FDA inspection finding.

If there is no U.S. IDE yet, still write Brazilian cases as if 812.150 will apply later. Reconstructing UADEs from a drug-safety database two years on is how FIH datasets lose credibility.

Where programs actually break

The bottleneck is rarely the web form. It is the weekend between the PI’s 24-hour email and a U.S. medical monitor who is still arguing expectedness against an old brochure. Build a Brazil on-call roster that can file NotivisaEC without waiting for California business hours. Give the Brazilian authorized representative and the ORPC (if one is the DICD face) read-only access to the safety tracker. Import holds, ethics queries, and ANVISA inspections all ask for the same case list.

Second break: treating CEP notification and ANVISA notification as the same send. Law 14.874/2024 sends serious events to the CEP. RDC 837/2023 sends a narrower unexpected-and-related set to ANVISA on NotivisaEC. Your SOP should have two lines, two clocks, two receipts.

Third break: annual-report amnesia. The 60-day secondary petition is how ANVISA sees aggregate harm. If your data-management lock cannot produce site-level AE tables from Brazilian CRF codes, you will miss a petition that can cancel the DICD — after first patients are already in.

A practical next step

Before site initiation, run one tabletop: a fatal unexpected device-related event at 18:00 Brasília on a Friday. Walk the 24-hour investigator notice, the 7-calendar-day NotivisaEC filing, the 8-day follow-up, the CEP notice under Law 14.874/2024, the brochure update, and — if an IDE is live or planned — the 10-working-day UADE evaluation to FDA. If any of those hands is “we’ll figure it out,” rewrite the safety SOP and name the NotivisaEC filer. First-patient week is the wrong time to discover that VigiMed was the only account anyone opened.

Most first-in-human (FIH) delays I see in Brazil do not start on the ANVISA clock. They start earlier, when regulatory affairs treats class as a registration afterthought instead of the decision that decides whether you even file a clinical investigation dossier, what the import petition can carry, and what the investigator brochure must treat as expected.

This is not another walkthrough of Brazil’s 90-day ANVISA review window, and it is not a recap of Law 14.874/2024. Those calendars matter, but they sit on top of a classification call. If that call is wrong, the clock you are watching is the wrong clock.

RDC 751/2022 is the class rule, not a slogan

Collegiate Board Resolution RDC No. 751 of 15 September 2022 is ANVISA’s current framework for medical-device risk classification, labeling and instructions for use, and the notification versus marketing-authorization (registro) regimes. Article 5 places devices in four intrinsic-risk classes:

• Class I — low risk
• Class II — medium risk
• Class III — high risk
• Class IV — maximum risk

Articles 6 and 7 then split the premarket path: Classes I and II are subject to notification; Classes III and IV are subject to marketing authorization. ANVISA’s English overview of the same split is on the Agency’s medical devices page.

Classification is not a marketing preference. It is a rules exercise against the manufacturer’s intended purpose. Annex I of RDC 751/2022 sets the classification rules (duration of use, invasiveness, active energy, implants, special rules). The most stringent applicable rule wins. Software as a medical device is classified on its own intended purpose when it is standalone. In vitro diagnostic devices sit on a separate IVD classification track; do not force an IVD into the same clinical-investigation box as an implant.

For an FIH program, lock three statements in the same document: intended purpose in Portuguese that will appear on the DICD and later on the registro file; the Annex I rule (or rules) you applied; and the resulting Class I–IV. If clinical, quality, and the Brazilian authorized representative cannot repeat those three lines without hedging, you are not ready to talk about first-patient dates.

RDC 10/2015 is not the current FIH filing rule

Sponsors still arrive with “RDC 10/2015” in the regulatory plan. That resolution existed. It is not the current device-investigation statute.

RDC No. 548 of 30 August 2021 revoked RDC No. 10 of 20 February 2015 (Article 84 of RDC 548/2021). RDC No. 837 of 13 December 2023 then replaced that 2021 text. ANVISA announced the update and the alignment with international practice in its 18 December 2023 notice (in force early January 2024). Article 80 of RDC 837/2023 revokes RDC 548/2021.

Current device clinical-investigation procedure is RDC 837/2023. Use RDC 10/2015 only as history. Do not cite it as the filing basis for a 2026 FIH.

How class drives whether you file a DICD

RDC 837/2023 is explicit about scope. A Dossiê de Investigação Clínica de Dispositivo Médico (DICD) is required for clinical investigations involving Class III or Class IV devices that are not yet registered in Brazil. A registered device studied for an indication different from the indication in the sanitary registration also goes in as a DICD.

The same resolution carves out investigations that do not go to ANVISA as a DICD, including:

• clinical performance studies of IVDs;
• usability/human-factors-only studies (unless the clinical investigation also includes those endpoints among others);
• investigations of devices already registered in Brazil for the same approved indications;
• clinical investigations of Class I or Class II devices.

Class I and II investigations are still expected to follow good clinical practice. RDC 837/2023 points to the Document of the Americas and ISO 14155 as the GCP standard. They still need ethics approval. They do not get a free pass on import controls or on manufacturing quality of investigational units. What they do not get is a DICD as the ANVISA on-ramp.

That is the classification bottleneck. A U.S. significant-risk implant that an RA lead quietly “simplifies” to Class II to avoid a dossier is not a timeline win. It is a first-patient hold when import, ethics, or a later registro reviewer asks why the intended purpose looks like Rule 8 (long-term surgically invasive / implant) and the file says Class II. The opposite error is also expensive: forcing a true Class II tool through a DICD because someone copied a Class III playbook wastes petition fees and a review cycle you did not owe.

What the DICD actually has to carry

When class puts you in DICD scope, RDC 837/2023 consolidates the old two-dossier habit into one submission. The petition typically includes the DICD form, the sanitary-surveillance fee (GRU) or exemption, the investigator brochure, a safety-experience summary (prior human use and any foreign post-market experience), the investigational-device dossier, the clinical investigation plan under GCP, and proof of registration in a WHO ICTRP or ICMJE-recognized trial registry (or that proof at start-date notification if it is not ready at first protocolization).

ANVISA’s petition checklist for DICD subject 80104 is published in the Agency’s SAT instruction list. Use that list, not a recycled IND table of contents.

Two operational rules from the same resolution save weeks if you lock them before translators start:

• The party who files the DICD owns every later secondary petition. Do not let a consultant file “just this once.”
• An organização representativa de pesquisa clínica (ORPC) may file only when the sponsor has no headquarters or branch in Brazil. If you already have a Brazilian legal presence, that presence is the face of the file.

RDC 837/2023 gives ANVISA 90 calendar days (dias corridos) to issue a manifestation on the DICD, with a tacit-start path after ethics approval if the Agency is silent. Law 14.874/2024 Article 58 speaks of 90 business days (dias úteis) for primary sanitary analysis of clinical-trial petitions intended to support registration. Those are not the same unit of time. Treat the mismatch as a planning flag; do not invent a hierarchy in a slide. The clock posts already cover activation math. Classification decides whether that math applies to you at all.

Import, IOR/AR, and QMSR sit on the same class decision

Investigational units do not enter Brazil on a commercial registro. RDC 837/2023 Chapter IX puts exclusive clinical-use import under sanitary inspection and SISCOMEX release. The importer must cite the Diário Oficial da União resolution number for the DICD grant. Outer packaging must carry that DOU number, storage conditions, and a lot, identification, or serial code that can trace the unit. Quantities are limited to what the investigation needs. Commercialization of those units is prohibited.

If someone other than the DICD holder imports, both parties sign a delegation-of-import-responsibility document. That is your import authorized representative problem, not a freight problem. The same legal person who will later hold notification or registro as authorized representative should see the investigational import list now. A first-patient kit that is not on the DICD product list will sit on the dock.

On quality: RDC 837/2023 requires investigational devices, and any placebo or sham, to be manufactured to good manufacturing practice and labeled so they are identifiable as investigational. ANVISA may inspect the manufacturing site against the technical, production, and quality-control story in the DICD. If the same design will later support an FDA marketing file, build those units under the quality system you intend to defend — including the United States Quality Management System Regulation transition — not under a “prototype exception” that you cannot reconstruct.

What the RA lead must lock before first patient

Write these eight items as a one-page gate. Do not open the site until each line has an owner and a document ID.

1. Intended purpose. The Portuguese indication that will classify the device under RDC 751/2022 Annex I and that will appear in the investigator brochure. Changing “long-term implant” to “intraoperative aid” after ethics approval is a new class, not a wording tweak.
2. Class and rule. Class I–IV plus the governing Annex I rule. Record why neighboring rules were rejected.
3. DICD yes/no. Apply RDC 837/2023 Articles 2 and 4. Class I/II, same-indication post-market, IVD performance, and usability-only studies are out of DICD scope. Class III/IV unregistered, and new indications on a registered device, are in.
4. Filing face. Sponsor legal entity in Brazil versus ORPC. Name the person who will own secondary petitions: start/end dates, amendments, annual report.
5. Import list. Every investigational SKU, spare, and accessory that will clear SISCOMEX, mapped to the DICD form. No “we’ll add the introducer later.”
6. Expectedness language. The brochure and operator manual define “unexpected” for later safety notification. If the FIH protocol lists risks the brochure omits, you have built a 7- and 15-day reporting trap.
7. Ethics path. CEP submission under Law 14.874/2024 in parallel with, not after, the sanitary file when a DICD is in scope. Single-CEP review for multicenter protocols is a legal design, not a courtesy.
8. Start-date discipline. RDC 837/2023 requires start- and end-date forms as secondary petitions within 30 calendar days of each Brazil date. First patient is a regulatory event, not only a clinical one.

Amendments after that gate are expensive. Substantial changes to the brochure or device dossier that can affect quality or safety wait for ANVISA manifestation (again, 90 calendar days in RDC 837/2023). Substantial protocol amendments that affect participant safety or scientific value wait the same way, except amendments that remove an immediate risk, which you implement and notify. Classification mistakes tend to surface as those amendments.

A practical next step

This week, run a 90-minute classification huddle with RA, clinical, quality, and the Brazilian representative. Put the intended-purpose sentence, the Annex I rule, the Class I–IV result, and a yes/no DICD decision on one page. Attach the RDC 751/2022 English text and the RDC 837/2023 scope articles. If those four people cannot sign the page, do not book first-patient week. The calendar you save is not ANVISA’s — it is the month you would have spent unscrewing a class you never locked.