For four decades, pharmacovigilance has run on a heartbeat. Cases arrive, get coded and narrated, accumulate in a database, and then, on a calendar cadence, a team pulls the period's data, computes disproportionality, evaluates signals, and writes a periodic report that is a snapshot of a window that has already closed. The PSUR is fundamentally a retrospective document: by the time it is read, the period it describes is months gone. This batch rhythm made sense when case data arrived on paper and computation was expensive, but it is now an artifact of constraints that no longer exist. The E2B(R3) standard, FDA-mandated for IND safety reports from 1 April 2026 and moving to postmarketing transmission via the ESG NextGen gateway from 1 October 2026, gives the case stream a structured, machine-readable spine, and on that spine an always-on signal layer can evaluate safety continuously rather than in quarterly snapshots. This lesson traces the shift from batch ICSR flow to continuous case-stream signal evaluation, what it does to the QPPV role, and the genuinely open conversation about whether the periodic report itself should change once the signal picture never stops updating. The capability is durable. The governance is still being written, and a Level 5 leader helps write it.

The Batch Paradigm and Its Hidden Cost

The batch paradigm in pharmacovigilance has a hidden cost that the calendar cadence obscures: the latency between a signal becoming detectable and a signal being evaluated. In the periodic model, a disproportionality pattern that crosses a threshold in week two of a reporting period is not formally evaluated until the period closes and the report is assembled, which can be months later, and a real safety signal that was statistically visible early sits unexamined while cases continue to accrue under it. The periodic report is not the place safety is monitored; it is the place safety monitoring is summarized and documented after the fact. The industry has always known this, which is why expedited 15-day reporting exists as a fast lane around the batch rhythm for the most serious cases, but the structural latency for the broad signal-detection function remained, because computing and evaluating disproportionality across the whole database was an expensive batch operation tied to the report cycle.

The deeper cost is that the batch rhythm shapes how the entire function thinks about its work. When the periodic report is the deliverable, the PV team's calendar orients around report cycles, the database is queried in periodic pulls, and the signal-evaluation muscle activates on a schedule rather than continuously. This was a reasonable adaptation to the constraints of paper intake and expensive computation, but it bakes in a reactive posture that the modern data environment no longer requires. A Level 5 leader has to see the batch paradigm clearly as an adaptation to obsolete constraints, not as an intrinsic feature of good pharmacovigilance, because only then does the continuous alternative become visible as a genuine choice rather than a vendor pitch. The question the leader must pose to the function is simple and disorienting: if the case data arrived continuously and computation were nearly free, would you still wait for the period to close before evaluating a signal? The honest answer is no, and that answer is what the E2B(R3) spine and the continuous signal layer make actionable.

The E2B(R3) Spine Makes the Stream Analyzable

The reason continuous pharmacovigilance is now a real architecture rather than an aspiration is the structured case stream that E2B(R3) provides. E2B(R3) is the ICH standard for the electronic transmission of individual case safety reports, and it structures each case into a rich, machine-readable schema covering patient, reaction, drug, and assessment data far more granularly than its predecessor. FDA mandated E2B(R3) for IND safety report transmission from 1 April 2026, and postmarketing ICSR transmission moves to the ESG NextGen gateway from 1 October 2026, which means the case stream feeding the safety database becomes a structured, continuous feed rather than a series of batched, partially structured submissions. When every incoming case carries the same structured fields, the database is continuously analyzable in a way that a heterogeneous batch of differently formatted submissions never was, and the disproportionality computation that used to be a periodic batch job can run continuously against the live stream.

On this structured spine, the AI systems already in 2026 production carry the operational load that makes continuous evaluation feasible at volume. ArisGlobal LifeSphere NavaX and Oracle Argus AI handle large shares of case intake, MedDRA coding, narrative drafting, and the disproportionality computation of PRR, ROR, and EBGM against the accumulating database, which means the mechanical work of turning a case stream into a continuously updated signal picture is automatable. The structured E2B(R3) feed and the AI processing layer together convert pharmacovigilance from a function that periodically reconstructs a signal picture into one that continuously maintains it, and this is the durable core of the continuous-PV thesis. The leader should treat the structured case stream as the foundation, exactly as the structured content base is the foundation for AI-native submissions, because in both cases the structure is what makes the AI layer safe and the continuous operation possible. Without the E2B(R3) spine, continuous PV is a faster way to process unstructured chaos; with it, it is a continuously maintained, analyzable, governed signal picture.

Continuous Signal Evaluation: Triage at Machine Speed

What continuous pharmacovigilance actually does is move signal triage from a periodic event to a continuous process, and the distinction between triage and evaluation is the crux of doing this responsibly. The continuous signal layer ingests the structured case stream, recomputes disproportionality as cases arrive, and surfaces potential signals at machine speed, prioritizing them by strength, seriousness, and novelty so that the human attention is directed to what matters as soon as it becomes visible rather than at the next report cycle. This is triage: the machine determines what deserves a human look and ranks it, which is exactly the kind of high-volume pattern work AI does well and humans do slowly. The value is that the latency between a signal becoming detectable and a human evaluating it collapses from months to near-real-time, which is a genuine patient-safety improvement, not merely an efficiency gain, because a safety signal evaluated sooner is a risk mitigated sooner.

The bounded reality is that triage is not evaluation, and the line between them is where the human judgment and the regulatory accountability live. The continuous layer can surface and rank a candidate signal, but the determination of whether it is a real signal, whether the causal association is plausible, whether the reaction is listed or unlisted, and what risk action follows, remains a human medical judgment that the WHO-UMC and Naranjo frameworks structure but do not automate. The danger of the continuous layer is the seduction of letting the machine's prioritization become the evaluation, where a low-ranked signal is implicitly dismissed without a human ever applying medical judgment to it, which is how a continuous system could miss what a periodic review would have caught. The leader's governance task is to ensure the continuous layer is explicitly a triage-and-prioritization tool whose outputs enter human evaluation, never a system whose ranking substitutes for the medical assessment, because the moment the ranking is treated as the conclusion, the function has automated away the judgment the regulation requires it to perform.

The QPPV Under Continuous AI Triage

The qualified person responsible for pharmacovigilance is the named individual personally accountable for the safety profile of the product and the operation of the PV system, and continuous AI triage changes the QPPV's work without diluting the accountability in the slightest. In the batch world, the QPPV oversaw a periodic process; in the continuous world, the QPPV governs a continuously running triage model and owns the signals it surfaces, which is a shift from supervising a cadence to governing a system. The QPPV does not read every narrative in a continuous-PV operation any more than they read every case in a high-volume batch operation, but they remain accountable for the model that decides what reaches human evaluation, which means the QPPV's oversight moves upstream to the design, validation, and monitoring of the triage layer itself. The accountability is unchanged; the locus of the QPPV's attention shifts from the individual case to the system that triages the cases.

This is a more demanding role, not a lighter one, and the leader must resource it accordingly. Governing a triage model means the QPPV must understand the model's behavior well enough to be accountable for what it surfaces and, more dangerously, for what it does not surface, because a signal the model never elevates is a signal the QPPV is still accountable for missing. This requires the QPPV to own the model's validation, its performance monitoring, and its failure modes, which is why the horizon lesson named the AI-aware QPPV as an emerging role: the QPPV who can govern a triage model and still own a signal is a different professional from the QPPV who supervised a periodic database review. The standing principle the leader must establish is that the continuous triage layer is a tool the QPPV governs and remains accountable for, never an autonomous decision-maker that absorbs the accountability, because the regulation places the safety profile on a named human and no amount of triage automation transfers that. The QPPV's signature on the safety position is the load-bearing control, exactly as the named author's signature is on the submission.

The Periodic-Report Reshaping Conversation

Here is where the lesson reaches a genuinely open question that a Level 5 leader should engage rather than wait out: if the signal picture is continuously maintained, what is the periodic report for? The PSUR and PBRER under ICH E2C(R2) were designed as periodic reconstructions of a signal picture, but if that picture is continuously maintained and continuously evaluated, the periodic report's role shifts from reconstruction to summary, from building the safety picture at the report cycle to documenting and communicating a picture that already exists and is current. This is not a small change in document production; it is a question about the purpose and cadence of a core regulatory deliverable, and the honest position is that it is unsettled. The capability to maintain the picture continuously is durable and arriving; whether the regulatory framework should change the periodic-report cadence in response is a policy question the agencies are only beginning to consider, and the leader should hold this distinction precisely rather than overclaiming that the PSUR is obsolete, which it is not.

The measured projection is that the periodic report increasingly becomes a structured summary of a continuously maintained signal picture rather than a from-scratch quarterly reconstruction, which improves its currency and reduces its production burden without yet changing its mandated cadence. Whether the cadence itself should change, whether a continuously maintained and continuously evaluated signal picture could support a different reporting rhythm or a more event-driven model, is exactly the kind of question that gets decided through regulator engagement, and a Level 5 leader's role is to bring evidence to that conversation rather than wait for a rule to land. The leader who is running a mature continuous-PV operation has the operational evidence the agencies need to evaluate whether the cadence should evolve, and bringing that evidence to the FDA and EMA workplans through the public-consultation channels is precisely the industry-leadership posture this level is about. The capability is durable, the governance is still being written, and the leaders who built the capability are the ones positioned to help write the governance, which is the difference between being subject to the rule and helping shape it.

Building the Continuous-PV Operation Responsibly

Translating the thesis into a defensible operation requires the same sequence discipline the AI-native submission required, because the failure modes are parallel. The foundation is the structured E2B(R3) case stream, without which the continuous layer is processing chaos rather than a governed feed, so the leader's first move is to ensure the organization is fully transitioned to E2B(R3) intake and the postmarketing ESG NextGen pathway on the mandated timelines. On that foundation, the AI processing layer for intake, coding, narrative, and disproportionality is the operational engine, and it must be validated under the organization's quality system with the same rigor as any GxP system, treating the triage model as a tool with documented intended use, performance criteria, and monitoring. Only with the structured feed and the validated processing layer in place does the continuous signal evaluation become a defensible operation rather than a fast and ungoverned one, which is the same structure-before-capability discipline that governs every frontier in this chapter.

The governance gates that make continuous PV defensible are specific and the leader must build them in as non-bypassable controls. The triage-not-evaluation gate ensures every surfaced signal enters human medical evaluation and that the model's ranking never silently dismisses a signal without a human judgment, which is the control against automating away the medical assessment. The QPPV-accountability gate ensures the QPPV governs the triage model's validation and monitoring and owns both what it surfaces and what it misses, which is the control against accountability leaking to the vendor. The audit gate ensures the continuous operation produces a complete, ALCOA+-compliant trail of which cases were triaged, how the model ranked them, what humans evaluated, and what was concluded, which is the control that makes the operation inspectable. Built this way, continuous pharmacovigilance is a genuine patient-safety advance that collapses signal latency while keeping medical judgment and regulatory accountability firmly human, and the leader who builds it in this order has both a better safety operation and a seat at the table where the periodic-report future is decided.

Key Takeaways

• The batch PV paradigm is an adaptation to obsolete constraints, not an intrinsic feature of good pharmacovigilance. Its hidden cost is the latency between a signal becoming detectable early in a period and being evaluated only after the period closes and the report is assembled. The periodic report is where safety monitoring is summarized after the fact, not where it happens, and the honest answer to whether you would still wait for the period to close if data arrived continuously is no.
• The E2B(R3) structured case stream is the spine that makes continuous evaluation a real architecture. Mandated for IND safety reports from 1 April 2026 and postmarketing via ESG NextGen from 1 October 2026, it turns the case feed into a structured, continuously analyzable stream on which ArisGlobal LifeSphere NavaX and Oracle Argus AI run intake, coding, narrative, and disproportionality. Without the structured spine, continuous PV is faster chaos; with it, it is a continuously maintained, governed signal picture.
• Continuous PV moves signal triage to machine speed, but triage is not evaluation, and the line between them is where accountability lives. The layer surfaces and ranks candidate signals so latency collapses from months to near-real-time, a genuine patient-safety gain, but whether a candidate is a real signal, whether causation is plausible, and whether the reaction is listed or unlisted remain human medical judgments. The danger is letting the model's ranking become the evaluation and silently dismissing a low-ranked signal.
• The QPPV's accountability is unchanged; the locus of attention shifts from the individual case to the triage system. The QPPV governs the triage model's validation, performance monitoring, and failure modes, and remains accountable for both what the model surfaces and what it never elevates. This is a more demanding role, the AI-aware QPPV, and the QPPV's signature on the safety position is the load-bearing control that no triage automation can absorb.
• Whether the periodic-report cadence should change is genuinely open, and the leader brings evidence to that conversation rather than waiting for a rule. The measured projection is that the PSUR and PBRER become structured summaries of a continuously maintained picture rather than from-scratch quarterly reconstructions, improving currency without yet changing the mandated cadence. Build the operation on the structured spine with triage-not-evaluation, QPPV-accountability, and audit gates, and you have both a better safety operation and a seat where the periodic-report future is decided.