When and how to update a systematic literature review for HTA submissions

Why currency is an HTA question, not only a methods question

Systematic reviews go out of date faster than most evidence plans assume. In a survival analysis of 100 quantitative systematic reviews, Shojania and colleagues found a median of 5.5 years before a signal for updating appeared, but a signal occurred within two years for 23% of reviews and had already occurred at the time of publication for 7%.^[1] In fast-moving therapy areas such as oncology, where comparators, data cuts and lines of therapy change within a single launch cycle, the effective shelf life of a review built for HTA is often much shorter.

HTA bodies do not treat this as a matter of academic tidiness. NICE expects evidence on outcomes to come from a systematic review that identifies, appraises and synthesises the relevant studies.^[2] Under the EU HTA Regulation, the rule is now explicit and time-bound. Annex I of Implementing Regulation (EU) 2024/1381 states that the cut-off date for the searches shall be a maximum of three months before submission of the dossier, and Article 12 gives the health technology developer 100 days from notification of the first request to submit it (60 days in certain accelerated or variation cases).^[3] The HTA Coordination Group has confirmed that any data informing the assessment of relative effectiveness and safety should be based on information retrieval with that cut-off.^[4]

A review that is methodologically excellent but six months old is therefore not compliant for a joint clinical assessment, and may draw questions from a NICE external assessment group. Updating is part of the submission, not an optional refinement.

A decision framework: the BMJ consensus and Cochrane guidance

The most widely used framework for deciding whether to update is the consensus checklist published in the BMJ by Garner and colleagues on behalf of the Panel for Updating Guidance for Systematic Reviews.^[5] It works in sequence:

1. Assess currency: Does the review still address a question that matters for practice, policy or research? Were its methods valid? If the question is current but the methods are weak, the panel suggests a new protocol may be more appropriate than an update.
2. Identify new methods, studies and information: New studies, longer follow-up, retractions, and newer methods such as updated risk-of-bias tools or GRADE.
3. Assess the effect of updating: Would the new studies or methods change the findings or their credibility?
4. Record the decision: Update, postpone or archive, with the rationale stated.

The panel defines an update as a new edition of a published review that includes new data, new methods or new analyses, asks a similar question, runs a new search and revises the findings as needed.^[5] The Cochrane Handbook adopts the same definition and is candid that there is no consensus on when to update. It recommends periodic assessment, notes that the original search may be re-run with date limits but that an information specialist should be consulted, and advises returning to the earliest records if the eligibility criteria or PICO have changed substantially.^[6]

Translating the framework into HTA terms

The consensus framework was written for reviews whose audience is clinicians and guideline developers. For a review that feeds a submission, each step maps onto a narrower question.

Our reading

For a review that will be read by NICE, NCPE or JCA assessors, the third consensus question is usually answered by the submission date itself. The practical decision is less about whether to update and more about the scope, timing and depth of the update.

Surveillance: knowing when the evidence has moved

Good updating depends on good surveillance between updates. The Cochrane Handbook lists surveillance searches, expert contact and quantitative or qualitative assessments as legitimate ways to decide that an update is due.^[6] In our experience, a surveillance plan for an HTA review usually has five components:

• Registry monitoring for the product and every scoped comparator, covering new trials, status changes and posted results.
• Data-cut tracking for pivotal trials, including planned interim and final analyses for overall survival.
• Regulatory and HTA outputs, such as assessment reports and published appraisals for comparators, which often contain data not yet in journals.
• Congress monitoring for the main meetings in the therapy area.
• Scheduled database alerts built from the review strategies, screened at a fixed cadence.

The JCA framework makes surveillance part of the dossier. The guidance on filling in the dossier template asks the developer to report whether and when new data relevant to the assessment scope might become available during the assessment period.^[7] The HTA Coordination Group has also stated that results for all scoped outcomes must be reported for the most recent prespecified data cut, and that for overall survival the last available data cut must be provided where it differs.^[4] A team without a data-cut calendar will find it harder to meet either expectation.

Timing the final update: a worked example for JCA

The three-month rule interacts with the 100-day submission window in a way that is easy to underestimate. Consider a standard JCA procedure.

The consequence is that, for a standard 100-day procedure, a final update after the scope arrives is effectively unavoidable. That update must also cover any PICO the company did not anticipate, which may require new search blocks rather than a simple date-limited re-run.^{[3, 6]} In shorter 60-day procedures the arithmetic differs, but the principle holds: the final search should be planned backwards from the submission date, with time left to screen, extract, re-run the evidence synthesis and update the dossier text.

In practice we plan at least two updates for a JCA: one shortly before the expected scope, so that the bulk of screening is done in advance, and a final targeted update after the scope arrives. The second is small if the first was well designed.

How to run and report the update

Searching and screening

Where the PICO is unchanged, the original strategies can be re-run with date limits. Where the population, comparators or outcomes have changed, Cochrane advises that the search may need to go back to the earliest records.^[6] Garner and colleagues recommend refining the original strategy using an analysis of the previous search, by database, terms and language.^[5] In our practice, date limits are set with a deliberate overlap with the previous search window, to allow for indexing delays, and new records are de-duplicated against the full set of records screened previously, not only against the included studies.

The JCA dossier guidance names the minimum sources. Bibliographic searches should at least cover MEDLINE and the Cochrane Central Register of Controlled Trials; registry searches should at least cover ClinicalTrials.gov, the Clinical Trials Information System, the EU Clinical Trials Register and the EMA clinical data platform; HTA reports and the EMA submission file are also listed sources.^[7] An update designed for NICE alone may not meet these requirements, so the sources should be checked whenever a review is repurposed for a JCA.

Reporting

PRISMA 2020 states that it can be used for original, updated or continually updated reviews, and its flow diagram template accommodates updates.^[8] PRISMA-S, the 16-item extension for reporting literature searches, includes an item on how searches were updated (for example, re-running searches or email alerts) and an item requiring the date of the last search for each strategy.^[9] The Cochrane Handbook describes three ways to present an updated search, integrated, incremental or replacement, and asks that the flow diagram reflect the chosen approach.^[6]

For a JCA, the dossier guidance requires that every search strategy be fully documented in an appendix, that the latest date of every search be stated, that a PRISMA flow chart be provided and that all search results be supplied as RIS files.^[7] An update that was run correctly but documented loosely creates avoidable assessor questions.

Auditing a review built by someone else

Many updates start from a review written by another team, often years earlier and for a different market. Before updating it, we audit it. An audit is not a criticism of the original authors; reviews are built for a purpose and the purpose may have changed. A structured audit typically asks:

• Can the searches be reproduced? Are the full strategies, dates and interfaces documented, and do re-runs return comparable numbers?
• Do the sources meet the target agency's requirements? For a JCA, this means the mandatory databases and registries above.
• Do the eligibility criteria cover the current PICOs? A review limited to one comparator class cannot support a JCA scope with several.
• Are exclusions traceable? Is there a list of excluded full texts with reasons?
• Are the methods current? The consensus checklist explicitly treats new risk-of-bias tools and GRADE as reasons to update.^[5]
• Are included studies adequately documented?

The last point matters more than it once did. JCA assessors verify completeness with their own searches. In the published JCA report on tovorafenib, the assessors ran their own searches, identified no additional studies, but described deficiencies in information retrieval in an appendix.^[10] The same report excluded a comparator study for which results were available only from conference abstracts, stating that studies with only limited summary data should be excluded because their methods and results cannot be assessed.^[10] A review that leans on abstract-only evidence for key comparators should be audited with that standard in mind.

What an update does to the network meta-analysis

The review is rarely the end product. Its outputs feed indirect comparisons, and an update can change them in four ways:

1. Network geometry: A new trial may add a comparator, close a loop or connect a previously disconnected treatment. NICE expects the network to include, ideally, all technologies named in the scope.^[2]
2. Effect estimates: A later data cut can change a hazard ratio, and with it the proportional hazards assessment and the choice between constant and time-varying effect models.
3. Consistency and heterogeneity: New direct evidence may reveal inconsistency that was invisible before. NICE asks for heterogeneity and inconsistency to be reported and, where found, explained.^[2]
4. Effect modifiers: Where population-adjusted methods are used, the HTA Coordination Group expects covariates to be identified through a systematic review, and expects any existing review of prognostic factors and effect modifiers to be sufficiently timely.^[4]

The practical lesson is to design the update and the synthesis together. Data extraction templates should already match the NMA input format, and every analysis dataset should be traceable to a dated search. When the search is refreshed, the team should be able to say within days whether the network or the relative effects have changed, and by how much.

A useful discipline

Keep a single evidence log that links each included study, data cut and extracted value to the search in which it was found. It turns the assessor question "is this the latest evidence?" into a lookup rather than a reconstruction.

Living reviews for multi-market launches

Cochrane describes a living systematic review as one that is continually updated, with new or newly identified evidence incorporated as soon as it becomes available. It suggests living reviews where the question is a priority for decision makers, certainty in the existing evidence is low or very low, and new research is likely; surveillance is frequent, for example monthly, and new information is incorporated in a timely manner, for example within a month of identification.^[11] The Handbook also notes that considerable resources are needed to support the process.^[6]

For a product launching across many markets over several years, these conditions are often met. A global review refreshed for each submission tends to produce slightly different search dates, eligibility decisions and extraction conventions each time. A living core review, with market-specific PICO filters applied on top, can be more consistent and, in our experience, often more efficient over the launch sequence.

These cadences are our suggestion, not an agency requirement. One statistical caution applies: repeated meta-analysis raises the risk of spurious findings, and Cochrane discourages formal sequential meta-analysis methods for updated meta-analyses in most circumstances.^[11] A living review should therefore report updated estimates with their uncertainty rather than re-test significance at each refresh.

AI-assisted screening, with human oversight

Updates are an obvious use case for machine-learning screening: the eligibility decisions from the original review provide training data, and the volume of new records is predictable. Agencies have now set out how they expect such tools to be used.

NICE's position statement on the use of AI in evidence generation, published on 15 August 2024, states that AI methods should augment, not replace, human involvement. It notes that AI may support primary and full-text screening, describes AI-based data extraction as less well established, and expects submitting organisations to declare and justify any use, with reference to checklists such as PALISADE.^[12] For the EU, the HTA Coordination Group adopted general principles on the use of AI in preparing the JCA dossier on 15 July 2026. They state that a human must remain ultimately responsible for quality and accuracy, that no step should be fully automated, that AI use must be documented in the dossier, including for information retrieval and screening, and that the tool name, version, date, developer and purpose should be reported, with prompts recorded and available on request.^[13]

In our practice, this translates into a simple rule: AI may prioritise or act as a second screener, but every exclusion that could affect the evidence base is confirmed by a human reviewer, and the AI step is described in the methods with enough detail to be reproduced or checked.

What this means for evidence planning

• Plan updates backwards from submission dates: For a JCA, assume a final update after the assessment scope arrives, within the three-month cut-off.
• Build surveillance in from the start: Registry alerts and a data-cut calendar make updates smaller and support the dossier requirement to report expected new data.
• Design for the strictest audience: A review that meets JCA source and documentation requirements will usually satisfy national agencies too; the reverse is not always true.
• Audit inherited reviews before updating them: Reproducibility, source coverage and PICO coverage come first.
• Link the review to the synthesis: Every NMA input should trace back to a dated search.
• Consider a living core review for multi-market launches, with documented market-specific filters.
• Use AI transparently: Declare it, keep humans accountable for decisions, and record versions and prompts.

Evidaxia designs and runs systematic literature reviews and updates for HTA, including audits and updates of reviews built by other teams, living reviews for multi-market launches and JCA-ready information retrieval. See SLR updates, systematic literature reviews and EU JCA dossier support.

References

1. Shojania KG, Sampson M, Ansari MT, Ji J, Doucette S, Moher D. How quickly do systematic reviews go out of date? A survival analysis. Ann Intern Med. 2007;147(4):224-33. Available from: https://pubmed.ncbi.nlm.nih.gov/17638714/
2. National Institute for Health and Care Excellence. NICE health technology evaluations: the manual (PMG36), chapter 3: Evidence [Internet]. London: National Institute for Health and Care Excellence; [date unknown] [cited 2026 Oct]. Available from: https://www.nice.org.uk/process/pmg36/chapter/evidence
3. European Commission. Commission Implementing Regulation (EU) 2024/1381 of 23 May 2024 laying down procedural rules for the interaction during, exchange of information on, and participation in the preparation and update of joint clinical assessments of medicinal products for human use at Union level, as well as templates for those joint clinical assessments. Official Journal of the European Union [Internet]. Brussels: European Commission; 2024 [cited 2026 Oct 30]. Available from: https://eur-lex.europa.eu/eli/reg_impl/2024/1381/oj/eng
4. Member State Coordination Group on HTA. Questions & Answers on general methodological and procedural issues for joint clinical assessments [Internet]. Brussels: European Commission; 2026 May 15 [cited 2026 Oct 30]. Available from: https://health.ec.europa.eu/document/download/b89dc548-10d7-4ef7-9c3b-deea90b639ca_en?filename=hta_20260518_QA_methodological_procedural_questions_HTACG_en.pdf
5. Garner P, Hopewell S, Chandler J, et al. When and how to update systematic reviews: consensus and checklist. BMJ. 2016;354:i3507. Available from: https://www.bmj.com/content/354/bmj.i3507
6. Cumpston M, Flemyng E. Chapter IV: Updating a review. In: Higgins JPT, Thomas J, Chandler J, et al, editors. Cochrane Handbook for Systematic Reviews of Interventions, version 6.5 [Internet]. London: Cochrane; 2024 [cited 2026 Oct 30]. Available from: https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-iv
7. Member State Coordination Group on HTA. Guidance on filling in the joint clinical assessment (JCA) dossier template: medicinal products [Internet]. Brussels: European Commission; 2024 Nov 28 [cited 2026 Oct 30]. Available from: https://health.ec.europa.eu/publications/guidance-filling-joint-clinical-assessment-jca-dossier-template-medicinal-products_en
8. Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. Available from: https://www.bmj.com/content/372/bmj.n71
9. Rethlefsen ML, Kirtley S, Waffenschmidt S, et al. PRISMA-S: an extension to the PRISMA Statement for Reporting Literature Searches in Systematic Reviews. Syst Rev. 2021;10:39. Available from: https://systematicreviewsjournal.biomedcentral.com/articles/10.1186/s13643-020-01542-z
10. Member State Coordination Group on HTA. Joint Clinical Assessment Report: Tovorafenib, version 1.0 [Internet]. Brussels: European Commission; 2026 [cited 2026 Oct 30]. Available from: https://health.ec.europa.eu/publications/joint-clinical-assessment-report-tovorafenib-ojemda_en
11. Thomas J, Askie LM, Berlin JA, et al. Chapter 22: Prospective approaches to accumulating evidence. In: Higgins JPT, Thomas J, Chandler J, et al, editors. Cochrane Handbook for Systematic Reviews of Interventions, version 6.5 [Internet]. London: Cochrane; 2024 [cited 2026 Oct 30]. Available from: https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-22
12. National Institute for Health and Care Excellence. Use of AI in evidence generation: NICE position statement, version 1.0 [Internet]. London: National Institute for Health and Care Excellence; 2024 Aug 15 [cited 2026 Oct 30]. Available from: https://www.nice.org.uk/position-statements/use-of-AI-in-evidence-generation-NICE-position-statement
13. Member State Coordination Group on HTA. General principles on the use of artificial intelligence in the preparation of the JCA dossier [Internet]. Brussels: European Commission; 2026 Jul 15 [cited 2026 Oct 30]. Available from: https://health.ec.europa.eu/publications/general-principles-use-artificial-intelligence-preparation-jca-dossier_en

Sources accessed October 2026. This article is analysis for information and is not legal, regulatory or tax advice.