Consider a scenario that plays out routinely in 2026: the same oncology product is assessed by ICER in the United States, NICE in England, and the G-BA in Germany, all within the same quarter. Three different methodologies, three different comparator definitions, and three different evidentiary standards produce three different conclusions on value. HEOR teams attempting to satisfy all three bodies from a single shared evidence base, without producing inconsistencies that a diligent reviewer can identify and challenge, face a structural problem that no amount of manual coordination fully solves.

Pienomial's Knolens platform was built for exactly this challenge. As an enterprise intelligence platform purpose-designed for life sciences, Knolens enables HEOR teams to build a single governed evidence layer that can be contextualised for each HTA body without losing factual consistency or source traceability. This post explains what ICER, NICE, and G-BA each require, where the evidence architecture gaps typically appear, and how AI closes them.[8]

1. ICER, NICE, and G-BA: What Each Body Actually Requires

ICER (United States): The Institute for Clinical and Economic Review does not operate a mandatory submission process. It publishes independent evidence-based assessments proactively. The primary metric is the incremental cost-effectiveness ratio expressed as cost per QALY gained and cost per evLY gained, with value-based price benchmarks calculated at $100,000 and $150,000 per QALY thresholds.[3] ICER assessments increasingly carry weight with US commercial payers, pharmacy benefit managers, and state Medicaid programmes, making proactive engagement and evidence monitoring a strategic necessity for products entering the US market.

NICE (England): The National Institute for Health and Care Excellence uses QALY-based cost-effectiveness analysis as its primary decision framework. The standard ICER approval norm is between £20,000 and £30,000 per QALY gained, with evidence-based flexibility for end-of-life conditions, highly specialised technologies, and innovative products. [4] NICE requires probabilistic sensitivity analysis to characterise decision uncertainty, prefers EQ-5D utility measurement, and increasingly requests real-world evidence for products where trial follow-up is limited or the patient population is narrow.

G-BA (Germany): The Gemeinsamer Bundesausschuss uses the AMNOG added benefit assessment process. The framework evaluates a product's additional benefit compared to an Appropriate Comparative Therapy, known as the zweckmäßige Vergleichstherapie or ZVT, which is determined by the G-BA itself before trial initiation.[1] Assessment considers only patient-relevant endpoints, specifically mortality, morbidity, quality of life, and adverse events. The QALY framework is not used. The dossier must be submitted on day one of EU market launch, making evidence readiness at the time of EMA approval a non-negotiable operational requirement.[2]

ISPOR's 2026 to 2027 Top 10 HEOR Trends report places HTA cross-country collaboration at number eight in the global priorities list, directly reflecting the growing complexity of simultaneous multi-body submission management.[7]

2. The Evidence Architecture Problem: Three Bodies, Three Realities

The deepest challenge in multi-HTA submission is not workload, it is structural evidence conflict. Each body defines the appropriate comparator independently, and those definitions frequently diverge.

In a typical oncology example, G-BA may define the ZVT for a second-line NSCLC product as docetaxel monotherapy, reflecting the German guideline standard of care. NICE may position the same product against an atezolizumab-based combination that is available in England. ICER may evaluate the product against the full range of available second-line options in the US market. If the Phase III trial was designed against docetaxel alone, NICE and ICER will both require indirect treatment comparisons against their preferred alternatives, comparisons that may not be feasible given the available trial network.

Endpoint conflicts compound the problem. G-BA rejects surrogate endpoints such as PFS unless they are validated as surrogates by IQWiG methodology. NICE accepts PFS with appropriate survival extrapolation modelling and characterisation of uncertainty. ICER evaluates both OS and PFS within an integrated cost-effectiveness framework. The same clinical data therefore requires three different evidential framings, and those framings must be internally consistent, because a reviewer with access to more than one submission will identify discrepancies immediately.

The pharma competitive intelligence strategy implication is significant. Understanding how each body has assessed analogous products, what evidence they accepted and what they challenged, is essential input to designing an evidence package that works for all three. That intelligence must be structured, attributed, and current.

3. How AI Builds a Unified Evidence Layer for Multi-HTA

The solution to the multi-HTA evidence architecture problem is not three separate dossier teams building from independent evidence bases. It is a single governed knowledge layer from which body-specific content is generated with full consistency and source attribution.

Knolens operates as an enterprise knowledge & AI memory platform that ingests all relevant clinical data, published evidence, HTA precedent decisions, and real-world evidence into a structured, continuously updated knowledge base. Evidence modules are tagged by HTA body applicability. A PFS hazard ratio result from a pivotal trial is stored once, with its source. For NICE it is retrieved and framed within a cost-effectiveness modelling context. For G-BA it is retrieved and assessed against IQWiG surrogate validation criteria. For ICER it is retrieved and incorporated as a clinical effectiveness input.

The consistency mechanism is architectural. Because all three submissions draw from the same evidence modules, factual claims are identical across submissions. Contextualisation occurs at the output layer. HTA body-specific language, methodology framing, and uncertainty characterisation are applied by KnolComposer without altering the underlying factual evidence. A reviewer comparing submissions across bodies will find the same clinical numbers, attributed to the same sources, presented in the methodological language each body expects.

4. AI-Powered Indirect Treatment Comparisons for Multi-HTA

Indirect treatment comparisons are required when head-to-head evidence against all relevant comparators is unavailable. That requirement applies in the majority of multi-HTA submissions, because G-BA, NICE, and ICER each independently define the appropriate comparator and those definitions rarely align perfectly with the trial's active comparator.

Methodology requirements differ meaningfully across bodies. NICE follows NICE DSU Technical Support Document guidance, preferring Bayesian network meta-analysis with probabilistic sensitivity analysis to characterise uncertainty. G-BA accepts adjusted indirect comparisons where NMA is not feasible, with strict heterogeneity documentation and relative treatment effects expressed as hazard ratios with 95% confidence intervals. ICER accepts NMA with sensitivity analyses and an explicit discussion of network limitations.

KnolAI structures the multi-HTA ITC by identifying all relevant trials in the treatment network from the knowledge layer, mapping common comparators, quantifying heterogeneity signals, and assessing the feasibility of the specific adjusted comparison required by each body. The output is a single structured ITC dataset with traceable source attribution for every data point. KnolComposer then generates the ITC methodology section in each body's required format, drawing from the same dataset, with appropriate statistical presentation and language for NICE, G-BA, and ICER respectively.

5. KnolComposer: Generating Body-Specific Dossier Sections from One Evidence Layer

The KnolComposer authoring workflow converts the shared evidence layer into body-specific dossier content through a structured five-step process.

Step 1, Evidence module activation: The HEOR team defines the submission scope including compound, indication, target HTA body, and submission timeline. KnolAI retrieves the relevant evidence modules from the knowledge layer, covering clinical trial results, RWE summaries, ITC outputs, and precedent HTA decisions for analogous products.

Step 2, Body-specific contextualisation: For each evidence module, KnolComposer applies the methodology requirements of the target body. For NICE, this means cost-effectiveness framing, QALY context, and probabilistic uncertainty characterisation. For G-BA, it means patient-relevant outcome focus, ZVT comparator framing, and the exclusion of QALY language. For ICER, it means US cost model inputs and budget impact context.

Step 3, Dossier section generation: Structured text is generated for each section with automatic source attribution. Every clinical claim in the generated text links to its evidence module, which links to its source publication.

Step 4, Cross-submission consistency check: KnolComposer compares clinical claims across all three dossier versions and flags any inconsistency in endpoint definition, comparator framing, or outcome direction before submission.

Step 5, Reviewer query simulation: KnolPersona simulates the most likely reviewer questions from each body and generates response frameworks grounded in the shared evidence layer, so the team is prepared for clarification requests before they arrive.

6. Audit Readiness: What HTA Bodies Now Require from AI-Assisted Submissions

Growing scrutiny of AI-assisted evidence generation is reshaping what audit readiness means in HTA submissions. NICE published a formal position statement in 2024 requiring that when AI is used in evidence generation, reporting must be transparent and must use relevant checklists including PALISADE to justify AI use and TRIPOD+AI to explain AI model development. Submitting organisations must explicitly declare AI use, explain the method choice, and document human oversight at each evidence generation stage. The submitting organisation remains accountable for all submitted content.

G-BA's transparency requirements for computational methods used in submissions are increasing, with particular scrutiny of automated evidence synthesis and indirect comparison methodology. ICER's evidence standards increasingly mirror PRISMA-level methodology documentation for systematic reviews and evidence syntheses.

What audit readiness requires in practice, for AI-assisted multi-HTA submissions: every search string used in literature retrieval is documented and reproducible; every screening decision is logged with its classification rationale; every data extraction is linked to a specific location in a specific source document; and every AI-generated text is identified as AI-assisted and linked to its evidence source. Knolens generates a complete audit trail for all KnolAI and KnolComposer actions, satisfying the NICE, G-BA, and ICER documentation requirements within a single governed workflow.

7. Building the Cost-Effectiveness Evidence Base for NICE and ICER

Health economic models require evidence inputs that are traceable, methodologically appropriate, and current. The most common model input failures that drive NICE and ICER rejections or unfavourable uncertainty assessments include transition probabilities from outdated or methodologically weak studies, utility values mapped from non-EQ-5D instruments without validated mapping functions, and comparator treatment effects from indirect comparisons without appropriate uncertainty characterisation.

KnolAI searches the evidence base for the highest-quality available data for each model input category, documents the search methodology, and flags inputs where evidence quality is limited, enabling the modelling team to characterise uncertainty appropriately rather than discover data limitations during the review. The output is a structured model input pack with source attribution for every parameter, formatted for NICE technical team review and ICER evidence report inclusion. [3]

The life sciences AI platform advantage is that the model input pack is a living document. As new clinical publications, registry data, or RWE analyses become available, KnolAI identifies updates relevant to existing model parameters and alerts the HEOR team, ensuring that the cost-effectiveness model reflects the current best evidence at the time of each submission rather than the evidence available at the time of initial model construction.

8. How Quickly Can Your Team Go Live with Knolens?

The G-BA AMNOG dossier is due on day one of EU market launch, which means evidence must be ready at EMA approval, not after it. [1] The good news: with Knolens already built and pre-configured, your team does not need months of infrastructure setup before seeing value. Most HEOR teams are running their first live multi-HTA evidence use case within six weeks of onboarding.

Knolens ships with pre-built evidence templates for NICE, G-BA, and ICER, pre-loaded HTA precedent data, and ready-to-run ITC and evidence synthesis pipelines. There is no bespoke development phase. The platform is configured to your indication and submission priorities, not constructed from scratch.

Sprint 1, Weeks 1 to 2, First evidence output live: Knolens is connected to your indication scope. Pre-built multi-HTA templates are activated. KnolAI runs the first comparative evidence analysis across NICE, G-BA, and ICER requirements simultaneously. Your team receives the initial cross-body evidence gap report. No IT build required.

Sprint 2, Weeks 3 to 4, Data source expansion: Additional sources relevant to your indication, such as specific national HTA precedent databases, RWE registries, or internal clinical data, are added to the knowledge layer. Evidence modules deepen automatically as new sources are ingested.

Sprint 3, Weeks 5 to 6, Workflow configuration: Body-specific output formats, alert thresholds, and comparator monitoring scope are adjusted to your team's submission workflow. KnolComposer is configured to generate NICE, G-BA, and ICER dossier sections in parallel from the same evidence layer.

From Sprint 3 onward, Knolens operates as a continuous multi-HTA intelligence layer. New HTA decisions, competitor trial registrations, and RWE publications are ingested automatically. ICER programme schedules are monitored continuously without analyst intervention. The resource model shifts from three parallel dossier teams building independently to a core evidence team reviewing AI-generated, pre-attributed content. 

9. Managing Simultaneous Reviewer Queries Across NICE, G-BA, and ICER

HTA bodies issue clarification requests during assessment, and for products under simultaneous multi-body review, these queries frequently arrive in the same week, each requiring a different contextualisation of the same underlying evidence. NICE issues a formal list of clarification questions approximately four weeks after submission. G-BA requests hearing attendance within twelve weeks. ICER may issue public comment periods at any stage of its assessment process.

The simultaneous query challenge is where evidence architecture quality becomes most visible. A NICE query about survival extrapolation methodology, a G-BA question about ZVT comparator relevance, and an ICER question about budget impact model assumptions may all arrive within days of each other. Each requires a response grounded in the same evidence base but framed in the methodology language of the requesting body.

Knolens enables rapid simultaneous query response: the relevant evidence module is retrieved from the knowledge layer, KnolComposer generates the response in the requesting body's required format with full source attribution, and the cross-submission consistency check ensures the response is factually consistent with the evidence presented to all other bodies. For HEOR teams managing simultaneous submissions, this capability compresses the response cycle from weeks to days and eliminates the cross-body inconsistency risk that arises when query responses are drafted under time pressure by separate teams.[8]

Final Thoughts

Multi-HTA evidence generation is not primarily a resource problem. It is an evidence architecture problem. ICER, NICE, and G-BA each assess the same clinical evidence through different methodological frameworks, and the teams that achieve the best outcomes across all three are those that treat the submissions as a coordinated programme drawing from a single governed evidence base, not three separate projects each building independently.

Pienomial's Knolens platform, as the enterprise intelligence platform and market access analytics platform purpose-built for this challenge, enables HEOR teams to build that shared architecture, generate body-specific dossier content with full traceability, and satisfy the growing audit requirements of NICE, G-BA, and ICER within a single governed workflow. The teams deploying this infrastructure before their next submission cycle will achieve faster, more consistent, and more defensible market access outcomes.CTA: Book a Multi-HTA Evidence Architecture Demo with Pienomial.