Traditional systematic literature reviews in health technology assessments quickly become outdated, failing to incorporate new evidence as it emerges. This is not a minor inconvenience. It is a structural problem with significant commercial consequences. HEOR teams that submit an HTA dossier built on a static evidence base compiled six to twelve months before submission are submitting to a review body that may have received newer evidence they do not know about. NICE managed access agreement reviews require evidence updates at 12 to 36-month intervals. G-BA benefit reassessments create ongoing evidence obligations. JCA assessors reviewing a submission concurrent with EMA review expect the evidence to reflect the current landscape, not the landscape as it existed when the analysis was conducted.
The solution to the static evidence problem is not conducting more frequent SLRs. It is building a living evidence base pharma AI that updates automatically as new evidence emerges, maintaining current, sourced intelligence across clinical, regulatory, HTA, and competitive domains without requiring the HEOR team to initiate a new evidence project each time the landscape changes. This is what KnolAI delivers through the Knolens platform, and this post explains how to build this capability, what it requires technically and operationally, and how pharma teams are deploying it to maintain evidence currency across the full product lifecycle.
1. Why Static Evidence Bases Fail Pharma Teams
A systematic literature review is designed to answer a specific question at a specific point in time using a structured, reproducible methodology. The output is a static document: an evidence table, a forest plot, a PRISMA flow diagram, a narrative synthesis. From the day it is completed, it begins to depreciate in value as new publications, regulatory decisions, HTA outcomes, and competitive developments change the landscape it describes.
The depreciation rate depends on the therapeutic area. In fast-moving oncology indications, with multiple agents in late-phase development and frequent regulatory and HTA decisions, an evidence base may be materially outdated within three to six months of completion. In slower-moving indications, the evidence base may remain current for twelve to eighteen months before a new trial result or HTA decision changes the landscape. In every case, the HEOR team faces the same problem: when the next submission, the next managed access review, or the next payer meeting arrives, they must either repeat the evidence synthesis work from scratch or submit with evidence they know is potentially outdated.
ISPOR's Health Technology Assessment Council has formally recognised living HTA as the solution to this problem, describing it as a real-time, dynamic approach that uses explicit methods to determine the value of a health technology at different lifecycle points, from market access through continued evidence generation. Living systematic reviews, which combine contemporaneity and rigour to enhance data accuracy and utility for decision-making, are now widely accepted as an alternative to traditional single static reviews. The challenge that has prevented wider adoption is the operational burden: running a living review manually requires continuous analyst effort that most HEOR teams cannot sustain alongside their other workload.
2. What a Living Evidence Base Actually Is
A living evidence base pharma AI is a governed, continuously updated intelligence infrastructure that maintains current evidence across all relevant domains, automatically ingesting and validating new evidence as it is published, without requiring manual analyst intervention to initiate each update cycle.
The distinction from a traditional SLR is architectural. A traditional SLR is a project with a defined start date, a search date, and an end date. A living evidence base is an infrastructure with no end date. It runs continuously. When a new clinical trial publication is indexed in PubMed, the system identifies it as relevant to the defined evidence scope, screens it against the pre-defined PICOS criteria, extracts the relevant data points, validates them against the source, and adds the validated evidence to the knowledge layer. The HEOR team is alerted to the new addition. The evidence table is updated. No new SLR project is required.
This is not a theoretical capability. JMIR's 2026 review of AI tools for living evidence synthesis documented a rapidly expanding set of AI applications across abstract screening, data extraction, evidence table generation, and continuous monitoring, with AI improving both efficiency and accuracy across all phases of living evidence maintenance. The infrastructure to build and maintain a living evidence base now exists. The question for most HEOR teams is how to deploy it for their specific indication and evidence requirements without the substantial data engineering investment that building it from scratch would require.
3. The Five Components of a Living Evidence Base
Building a living evidence base pharma AI that genuinely updates automatically requires five operational components, each of which must be in place for the system to function without continuous manual intervention. [9]
Component 1, Pre-defined evidence scope with PICOS criteria: The living evidence base is defined by a pre-specified PICOS framework: population, intervention, comparator, outcome, and study design. This framework acts as the automated filter for all incoming evidence. New publications are screened against the PICOS criteria automatically. Only publications meeting the inclusion criteria enter the evidence base. The PICOS framework is documented and version-controlled so that scope changes are managed transparently.
Component 2, Continuous source monitoring: The system monitors all relevant evidence sources continuously, covering PubMed, Embase, ClinicalTrials.gov, CTIS, NICE website publications, G-BA assessment database, HAS decision database, FDA drug approval database, EMA EPAR updates, and conference abstract feeds from ASCO, ASH, AACR, and ESMO for oncology-focused evidence bases. Monitoring is automated: the system checks each source on a defined frequency, typically daily for high-priority sources and weekly for lower-frequency sources, without requiring an analyst to conduct the check. [4]
Component 3, Automated screening with dual-validation: New records identified by the monitoring layer are screened against the PICOS criteria automatically. Abstract screening applies the inclusion and exclusion criteria to each new record. Full-text screening applies more detailed criteria to records that pass abstract screening. Human review of borderline records is flagged automatically for HEOR team decision. The dual-validation process, applying AI screening with a simulated inter-rater reliability check, satisfies the PRISMA 2020 methodological requirements for evidence screening documentation. [4]
Component 4, Automated extraction and validation: Records passing full-text screening are processed for data extraction. Structured extraction pulls efficacy endpoints, safety data, patient population characteristics, comparator definitions, and study quality assessments from source documents with sentence-level source attribution. Extracted data points are validated against the source document before entering the living evidence base. Only validated, sourced data points update the evidence layer. [5]
Component 5, Stakeholder alert and review workflow: When new validated evidence enters the living evidence base, the relevant HEOR team members are alerted with a structured summary of what is new, how it compares to existing evidence in the base, and whether it affects any current conclusions or submissions. The review workflow enables human experts to assess the new evidence and approve its integration into any active submission documents before distribution. [7]
4. How KnolAI Builds and Maintains a Living Evidence Base
KnolAI, the research intelligence module of the Knolens platform, implements all five living evidence-based components as a pre-built, configurable system. Your HEOR team does not build the monitoring infrastructure, the screening algorithm, or the extraction pipeline. You configure them for your indication and evidence scope, and KnolAI runs them continuously. [8]
The PICOS framework for your indication is configured in the KnolAI protocol module during onboarding. KnolAI activates monitoring across all relevant source types for your indication immediately. New publications and regulatory decisions that match the scope are identified within hours of publication. Abstract screening runs within the same session against the pre-defined criteria. Full-text screening follows for records passing abstract screening, with flagged borderline records routed to the HEOR team for review.
Data extraction from included records is structured and sentence-level attributed: every efficacy figure, endpoint definition, patient population characteristic, and study quality assessment extracted from a new publication links to its specific source location in that publication. The extracted data is validated before being added to the Knolens knowledge graph as a new entity-relationship triple, carrying the source provenance of the publication from which it was extracted. [9]
When a new record enters the living evidence base, KnolAI generates a structured update alert. The alert identifies the new publication, summarises the new evidence it contains, compares the new data to existing evidence in the base, and flags any existing submission documents or evidence tables that may require updating based on the new evidence. The HEOR team reviews the alert, assesses the new evidence, and approves or declines its integration into active submission documents. The entire process from publication identification to team alert takes hours, not weeks.
5. Living Evidence Bases for NICE Managed Access Agreements
One of the most operationally significant applications of a living evidence base pharma AI is supporting NICE managed access agreements, where conditional reimbursement is tied to ongoing evidence generation requirements and periodic evidence reviews. NICE MAA conditions typically require evidence updates at 12 to 36-month intervals, with the updated evidence forming the basis for the reassessment that determines whether full reimbursement continues or is modified.[5]
Under a traditional static evidence model, each MAA review requires a new SLR project: a new search, a new screening exercise, a new data extraction, a new evidence synthesis, and a new evidence table. For a product with a 36-month MAA review cycle and three active indications, this means conducting up to three full SLR projects every 36 months on a rolling schedule, each requiring several months of analyst time. The total evidence maintenance cost across a product's lifecycle is substantial.
Under a KnolAI living evidence base model, the MAA review evidence package is a current evidence snapshot from a continuously maintained knowledge layer. When the MAA review date arrives, KnolAI generates the updated evidence summary and evidence tables from the living evidence base, already reflecting every new publication that has appeared since the last review. The evidence update is generated in days rather than months. The HEOR team's role shifts from conducting the update to reviewing and approving it.
6. Living Evidence for Competitive Intelligence: Monitoring While You Work
A living evidence base is not limited to published clinical literature. For competitive intelligence functions, the same living infrastructure monitors all relevant competitive signals: clinical trial registry updates, regulatory filing and approval events, HTA assessment decisions, conference abstract publications, and partner and licensing announcements. The competitive intelligence living evidence base operates on the same automated monitoring and extraction principles as the HEOR living evidence base, across different source types and with competitive-specific alert logic.
The value of a living competitive intelligence layer is most visible when a high-priority competitive event occurs: a competitor Phase III readout, a regulatory approval in your primary indication, or an HTA body decision that changes the evidence bar you will be assessed against. In a traditional CI function, this event appears in the CI team's awareness through scheduled source checks, conference attendance, or trade press monitoring, typically with days to weeks of latency between the event and the CI team's awareness of it.
In a KnolAI living CI layer, the competitor Phase III readout is identified within hours of appearing in a trial registry update or a press release indexed by the monitoring layer. The alert is generated with the specific efficacy data extracted from the source, a comparison to the existing competitive landscape in the knowledge base, and the strategic implications for the HEOR team's current submission evidence architecture. The CI team has actionable intelligence before most competitors have scheduled their internal analysis meeting.
7. The PRISMA 2020 Extension for Living Reviews: What Documentation Is Required
The PRISMA 2020 extension for living systematic reviews requires specific documentation elements that living evidence base implementations must satisfy to be used in HTA submissions and regulatory applications. The extension requires that the review protocol documents the trigger conditions for updates, the timing of updates, the screening methodology, the data extraction approach, and the mechanism for managing potentially out-of-date conclusions.
For a KnolAI living evidence base, this documentation is generated automatically. The PICOS framework that defines the monitoring scope is the pre-registered protocol. The screening decisions are logged with timestamps and inclusion or exclusion criteria for each record. The data extraction is documented at the sentence level. The trigger conditions for update alerts are defined in the configuration and documented in the audit trail. When an HTA body requests living review documentation, the KnolAI audit trail provides the complete methodology record from the first publication identified through every subsequent update cycle.
ISPOR's Vienna Principles for living evidence synthesis automation, covering automation across the spectrum of review tasks, continuous improvement, and integration with high-quality standards, are satisfied by the KnolAI architecture by design. The principles were developed to guide exactly the kind of automated living evidence synthesis that KnolAI delivers, and aligning with them ensures that KnolAI-generated living evidence is methodologically defensible in any HTA submission context.
8. Integrating the Living Evidence Base Into Your Submission Workflow
A living evidence base delivers its full value when it is integrated into the submission workflow, not when it is maintained as a separate monitoring function that must be manually reconciled with the submission at periodic intervals.
Within the Knolens platform, the living evidence base is the knowledge layer that KnolComposer draws from when generating dossier content. When a new publication enters the living evidence base and updates the clinical efficacy picture for the indication, the relevant sections of any in-progress submission documents are automatically flagged for review. The HEOR team reviews the new evidence, approves its integration, and the submission content is updated from the living knowledge layer without requiring a manual document revision exercise.
This integration eliminates the submission preparation problem that static evidence bases create: the SLR was completed six months ago and the submission is being finalised now, but three new publications have appeared in the interval that the team must manually reconcile with the existing evidence tables before submitting. With a living evidence base integrated into the submission workflow, the evidence tables are always current because they draw from a knowledge layer that is always current.
9. How Fast Can Your Team Deploy a Living Evidence Base with KnolAI?
Building a living evidence base pharma AI from scratch, with custom monitoring infrastructure, custom screening algorithms, and custom extraction pipelines, is a data engineering project requiring months of development and substantial ongoing maintenance. KnolAI eliminates this build requirement: the monitoring infrastructure, screening algorithms, and extraction pipelines are pre-built and configurable. Your team defines the evidence scope. KnolAI runs the living evidence programme.
Sprint 1, Weeks 1 to 2, Living evidence base active for your indication: The PICOS framework for your indication is defined and configured in the KnolAI protocol module. Source monitoring is activated across PubMed, Embase, regulatory databases, HTA decision databases, and competitive signal sources relevant to your therapeutic area. The first monitoring cycle runs immediately. Any publications and regulatory events from the past 30 days matching your PICOS scope are identified and presented for screening. The living evidence base is active from the first week.
Sprint 2, Weeks 3 to 4, Screening, extraction, and alert workflow configured: Abstract and full-text screening is calibrated against your PICOS criteria. The extraction template is configured for your indication-specific endpoints, population characteristics, and study quality domains. The stakeholder alert framework is configured to route new evidence alerts to the appropriate HEOR team members with the right context and review workflow. The first structured evidence update alert is delivered.
Sprint 3, Weeks 5 to 6, Submission integration and MAA review workflow live: The living evidence base is connected to your active submission documents. Evidence table update alerts are configured to flag affected dossier sections when new evidence enters the knowledge layer. For products with active NICE managed access agreements, the MAA review evidence generation workflow is configured: update cycles are scheduled, alert thresholds are set, and the evidence update package generation is automated for each review date.
Conclusion
A traditional systematic literature review is a project with an end date. A living evidence base pharma AI is an infrastructure with no end date. The shift from project to infrastructure is the shift that allows pharma HEOR teams to stop conducting evidence maintenance as a periodic emergency and start maintaining evidence currency as a continuous, automated background function.
At Pienomial, we built KnolAI's living evidence infrastructure because we believe that the most commercially important thing a HEOR team can do for a product's market access trajectory is to ensure that the evidence supporting every submission, every payer meeting, and every managed access review is the current best evidence, not the best evidence available six months ago. KnolAI makes this achievable without adding to the team's workload, because the living evidence system runs continuously whether or not the team is actively working on a submission.
CTA: See how KnolAI builds and maintains a living evidence base for your indication. Book a demo with the Pienomial team.
