Central vs Local IRB: A Reliance-and-Accountability Decision, Not a Speed Contest

At a glance

• Central vs local IRB is fundamentally a question of who holds the IRB of record, not who reviews faster. The reliance agreement, not the IRB type, is the artifact that decides whether your sites are actually covered.
• For most federally funded multisite research in the US, the choice is already made: the 2018 revised Common Rule (45 CFR 46.114) requires reliance on a single IRB, with only narrow documented exceptions.
• Going central does not transfer local-context duties. Under 45 CFR 46.114(a) each participating institution remains responsible for safeguarding subjects, and 21 CFR 56.112 preserves institutional review on top of IRB approval.
• The investigator’s obligations under 21 CFR Part 312 do not change with the IRB model. The investigator still assures IRB review (312.66) and still reports SAEs and unanticipated problems (312.64, 312.66).
• The common failure mode is not a slow review: it is a reliance agreement that never specified who routes site-level SAEs and protocol deviations, so reportable events fall into a gap between site and central IRB.

The vendor-written comparisons frame this decision around speed and cost: central IRBs are faster, cheaper, and scale better, so go central. That framing is not wrong, but it answers the wrong question. The real decision is about accountability: which body becomes the IRB of record, what duties travel with that designation, and what duties stay nailed to the site no matter what. This article reframes central vs local as a reliance-and-accountability decision and walks through the paperwork that actually grants coverage.

The real question: who is the IRB of record (not who is faster)

A local IRB is the review board operated by, or designated for, a single institution that reviews the research conducted at that institution. A central IRB (often an independent or commercial board) reviews the protocol once and serves as the reviewing IRB of record for multiple sites that agree to rely on it. The difference that matters is not turnaround time. It is which body issues the approval your sites operate under and which body owns continuing review.

Both models share the same statutory job. ICH E6(R3) §1.2.1 states that the purpose of an IRB/IEC is to safeguard the rights, safety, and well-being of all trial participants, and that purpose does not change with the board’s address. Both a local and a central IRB must conduct continuing review at intervals appropriate to the degree of risk, as ICH E6(R3) §1.2.4 requires and 21 CFR 56.109(f) likewise mandates at least annually. Choosing central does not lower the bar; it relocates who clears it.

Local IRB vs central IRB, side by side

Dimension	Local IRB	Central IRB (reliance model)
IRB of record	The site’s own institutional board	One reviewing IRB for all relying sites
Local-context review	Held at the site	Still held at the site; not delegated (45 CFR 46.114(a))
Review timeline	Per-site review, often serial and slow	One review, typically faster to activate
Continuing review	Each site’s IRB (56.109(f), E6(R3) §1.2.4)	Central IRB of record, at least annually
Reportable-event routing	Site to its own IRB	Must be specified in the reliance agreement
Reliance paperwork	None between sites	Reliance agreement / IRB Authorization Agreement required
Cost / scaling	Multiplies with site count	Scales better across many sites
When required	Default historically	Required for many federally funded multisite studies (45 CFR 46.114(b))

Read the table by column for the model, but make the decision by row. The rows that bite are local-context review and reportable-event routing, because those are the ones teams assume the central IRB absorbed when it did not.

When the choice is made for you: the sIRB mandate

For federally funded cooperative research, central review is frequently not optional. Under the 2018 revised Common Rule, 45 CFR 46.114(b)(1) requires that any US institution engaged in cooperative research rely upon approval by a single IRB for the portion of the research conducted in the United States. The reviewing IRB is identified by the supporting Federal department or agency, or proposed by the lead institution subject to that agency’s acceptance. This is the single-IRB (sIRB) mandate, and where it applies, the central-vs-local debate is already resolved in favor of a single reviewing IRB.

The mandate is not absolute. 45 CFR 46.114(b)(2) carves out two exceptions: research for which more than single-IRB review is required by law (including tribal law), and research for which the supporting Federal agency determines and documents that a single IRB is not appropriate for the particular context. If you are relying on an exception, document it; the regulation makes the documentation the precondition, not an afterthought.

Two cautions on scope. First, the sIRB mandate is a Common Rule provision tied to federally conducted or supported cooperative research; an industry-sponsored trial regulated only under FDA’s 21 CFR Parts 56 and 312 is not swept in by 45 CFR 46.114, though sponsors routinely require central IRBs by contract for operational reasons. Second, FDA’s own IRB regulations (Part 56) and the Common Rule (Part 46) are separate frameworks that can both apply to the same study; where they do, satisfy both, because 21 CFR 56.103(c) is explicit that compliance with the IRB regulations does not render inapplicable other pertinent Federal, State, or local laws.

What a central IRB does NOT take off your plate

This is the load-bearing point. A reliance arrangement moves protocol-level ethical review to one board. It does not move the duties that are intrinsically local.

Stays with the site / institution	Owned by the central IRB of record
Institutional HRPP oversight and institutional review (21 CFR 56.112)	Protocol and consent-template approval
Safeguarding subjects at the site (45 CFR 46.114(a))	Continuing review of the protocol (56.109(f))
Local-context consent localization and site-specific consent elements	Determinations on unanticipated problems at study level
Site-level SAE and unanticipated-problem reporting (21 CFR 56.108(b))	Approval of protocol amendments
Investigator obligations under 21 CFR Part 312	Issuing the approval the sites rely on

Three of these deserve emphasis.

Institutional review survives reliance. 21 CFR 56.112 states that research approved by an IRB may still be subject to further appropriate review and approval, or disapproval, by officials of the institution; those officials cannot approve research the IRB has not approved, but they retain a veto and an oversight role. Relying on a central IRB does not dissolve your Human Research Protection Program.

Each institution still safeguards its own subjects. 45 CFR 46.114(a) is unambiguous: in cooperative research, each institution is responsible for safeguarding the rights and welfare of human subjects and for complying with the policy. The single-IRB requirement in 46.114(b) sits on top of that responsibility; it does not replace it.

Reportable-event handling is local and time-sensitive. 21 CFR 56.108(b) requires written procedures for prompt reporting to the IRB, to appropriate institutional officials, and to FDA of any unanticipated problems involving risks to subjects, any serious or continuing noncompliance, and any suspension or termination of IRB approval. When the IRB of record is a central board, the central board does not magically observe an event that happened at your site. Someone at the site has to route it, and the route has to be written down.

Consent localization also stays local. 21 CFR 56.109(b) lets the IRB require that information given to subjects as part of informed consent meets the standard in part 50, and 56.109(c) governs documentation; even under a central IRB, site-specific consent elements (local contacts, institutional language, applicable State requirements) are worked at the site within the central IRB’s approved template.

The reliance agreement: the artifact that actually grants coverage

The central IRB is not what covers your sites. The reliance agreement is. A reliance agreement (often executed as an IRB Authorization Agreement) is the document by which a participating institution cedes review to the reviewing IRB and the two parties allocate the duties that do not transfer. 45 CFR 46.114(c) recognizes such arrangements for cooperative research not subject to the single-IRB mandate, where institutions may enter a joint review arrangement, rely on the review of another IRB, or make similar arrangements to avoid duplication of effort. For studies under the mandate, the reliance agreement operationalizes the single-IRB designation.

Treat the agreement as the deliverable, not a formality. ICH E6(R3) §10.1 frames the principle precisely: the sponsor may transfer, and the investigator may delegate, tasks, duties, or functions, but they retain overall responsibility for their respective activities. A reliance agreement is exactly such a transfer of the review function, and §10.1 means the act of signing it does not offload accountability; it documents who does what and leaves overall responsibility where it started.

Reliance-agreement checklist. Before activation, confirm the agreement specifies:

• Who signs. The institutional signatory official (not the investigator alone) for each relying institution, plus the reviewing IRB.
• The reviewing IRB of record, named explicitly, and the scope of what it reviews.
• Local-context responsibilities retained by each site, including consent localization and applicable State or local requirements (consistent with 21 CFR 56.103(c)).
• Reportable-event routing: exactly who reports unanticipated problems, serious or continuing noncompliance, and suspensions to the central IRB, to institutional officials, and to FDA, on what timeline (mapping to 21 CFR 56.108(b)).
• SAE and protocol-deviation handling at the site, and the interface between site safety reporting and the central IRB’s determinations.
• HRPP and institutional review responsibilities that remain with each institution (21 CFR 56.112).
• Continuing-review responsibility and cadence, anchored to the central IRB (56.109(f)).
• Records, audit, and direct-access provisions.

The single most common gap is reportable-event routing. A study can run on a central IRB and still trip because no one wrote down who tells the central IRB about a site-level SAE or a protocol deviation. The board only acts on what reaches it, and 56.108(b) makes prompt reporting an affirmative written-procedure requirement, not a courtesy.

Investigator obligations are unchanged either way

Whatever you decide about the IRB, the investigator’s regulatory burden under 21 CFR Part 312 is constant. 21 CFR 312.60 makes the investigator responsible for ensuring the investigation is conducted per the signed investigator statement and applicable regulations, and for protecting the rights, safety, and welfare of subjects under the investigator’s care. That obligation does not shrink because review moved to a central board.

The IRB-facing duties are explicit. 21 CFR 312.66 requires the investigator to assure that an IRB complying with Part 56 will be responsible for initial and continuing review and approval, to promptly report to the IRB all changes in the research and all unanticipated problems involving risk to subjects or others, and to make no changes without IRB approval except to eliminate apparent immediate hazards. Note the phrasing: report to the IRB. When the IRB of record is central, the investigator’s promptly-report duty now points at the central board, which is precisely why the reliance agreement must specify the channel.

Safety reporting up the chain is also unchanged. 21 CFR 312.64 requires the investigator to immediately report to the sponsor any serious adverse event, with an assessment of whether there is a reasonable possibility the drug caused it. ICH E6(R3) §2.3.1 reinforces the underlying principle: the investigator may delegate trial-related activities but retains ultimate responsibility and must maintain appropriate oversight to protect participants and data reliability. The IRB model is orthogonal to all of this.

Where teams get it wrong

• Treating central as a delegation of accountability. It is a transfer of the review function only. Under ICH E6(R3) §10.1 the transferor retains overall responsibility, and under 45 CFR 46.114(a) each institution still safeguards its own subjects.
• Assuming the central IRB owns reportable events. It owns determinations, not detection. 21 CFR 56.108(b) puts prompt reporting on the site’s written procedures; if the reliance agreement is silent on routing, events stall.
• Letting the HRPP go quiet. 21 CFR 56.112 keeps institutional review alive; reliance does not retire your HRPP.
• Reading the sIRB mandate too broadly or too narrowly. 45 CFR 46.114(b) governs federally supported cooperative research and has documented exceptions in (b)(2); an FDA-only industry trial is not swept in by the Common Rule, though sponsors may still require central review by contract.
• Forgetting that two frameworks can stack. 21 CFR 56.103(c) confirms FDA’s IRB rules do not displace other Federal, State, or local requirements, so satisfy both Part 56 and the Common Rule where both apply.

A note on framing tension worth surfacing: the Common Rule’s 45 CFR 46.114(b) pushes hard toward one reviewing IRB for cooperative research, while both that same rule at 46.114(a) and FDA’s 21 CFR 56.112 insist that institution-level responsibility and review persist at every site. These do not contradict each other, but they pull in opposite operational directions: consolidate review, yet keep local accountability distributed. The reliance agreement is where you reconcile the pull, by writing down exactly which duties consolidated and which stayed local.

A decision checklist: choosing and operationalizing your IRB model

1. Determine whether the sIRB mandate applies (federally supported US cooperative research, 45 CFR 46.114(b)). If yes, plan for single-IRB review and document any (b)(2) exception.
2. Identify every framework in scope (Part 56, Common Rule, ICH E6(R3)) and plan to satisfy all of them; do not assume one displaces another (56.103(c)).
3. Name the IRB of record and confirm it can serve continuing review at the required cadence (56.109(f), E6(R3) §1.2.4).
4. Draft the reliance agreement against the checklist above, with reportable-event routing (56.108(b)) and retained institutional review (56.112) called out explicitly.
5. Confirm investigator obligations are wired correctly: IRB assurance and prompt reporting to the IRB of record (312.66), SAE reporting to the sponsor (312.64), overall responsibility retained (312.60, E6(R3) §2.3.1).
6. Localize consent within the central IRB’s approved template, keeping site-specific and State-specific elements at the site.

Remember the destination. This work supports compliance with the applicable IRB and investigator requirements; it does not certify that any single document or IRB model makes a study compliant. The sponsor and each institution stay responsible. Picking central over local can shorten activation, but only the reliance agreement, drafted against what does not delegate, determines whether your sites are actually covered.

For deeper treatment of the adjacent mechanics, see the sibling pieces on protocol-deviation handling, SAE and safety reporting, and informed-consent localization, and the broader IRB and ethics-review pillar.

Sources

• ICH E6(R3) Good Clinical Practice, version r3 (ICH) — https://www.ich.org/page/efficacy-guidelines
• 21 CFR Part 312 Investigational New Drug Application, version 2026-04 (FDA)
• 21 CFR Part 56 Institutional Review Boards, version 2024 (FDA) — https://www.govinfo.gov/content/pkg/CFR-2024-title21-vol1/pdf/CFR-2024-title21-vol1-part56.pdf
• 45 CFR 46 Protection of Human Subjects (Common Rule), version 2018 (HHS) — https://www.govinfo.gov/content/pkg/CFR-2024-title45-vol1/pdf/CFR-2024-title45-vol1-part46.pdf