A foundation to build on: what the ISPOR guidance means for systematic literature review teams

AI in SLRs

Cassandra Springate | 05/10/2026

The ISPOR Task Force’s new Good Practices Report is a welcome resource for systematic literature review (SLR) teams navigating GenAI adoption. Though it will inevitably date quickly, given how fast the AI landscape moves, its real value lies in giving the field a validated reference and baseline to build from.

For a field that’s been cautiously experimenting with GenAI, having a shared, evidence-based reference point for responsible use is genuinely valuable. The report is a welcome resource and marks a key milestone in the SLR community’s AI journey. What it is not, is a blanket green-light on using AI in SLRs. Alongside the recommendations, and examples of appropriate uses of AI, the authors acknowledge the limitations of making broad statements in a landscape where there are so many different models and versions of AI available [1].

Within its list of appropriate uses, the report confirms what disciplined teams are already finding: structured, high-volume work, such as title and abstract screening and data extraction, delivers a meaningful reduction in workload. While for the more interpretive, judgment-heavy parts of a review, such as risk-of-bias assessment and qualitative synthesis, the verification required can outweigh the efficiencies it offers.[1] That’s not to say it’s necessarily a flaw in the technology, just that it’s the nature of being accountable for our work. Every researcher needs to be able to defend the outputs of their SLR, and that means staying close to decisions and owning the insights and conclusions.

The report’s own admission that its assessments warrant periodic updating reflects how fast AI models are evolving, and the authors are transparent that publication bias will have skewed the evidence base. SLR teams should consider which models have been validated, under what conditions, and whether further validation is appropriate for the individual project [1]. Consistent with Cochrane’s RAISE guidance on responsible AI use in evidence synthesis [2] and the FDA’s draft guidance on AI in regulatory decision-making [3], the level of validation and oversight organisations apply should reflect the level of risk involved, rather than a fixed yes/no by activity. This principle is fundamental to continued exploration of the efficiencies and added value the AI can give us: The ISPOR report gives SLR teams a credible starting baseline, task by task – building on it, case by case, is the responsible next step.

That doesn’t mean the baseline is optional. It adds a line to our own risk assessment: a documented reason the deviation was justified, and an acknowledgment that anything built outside the recommended baseline carries a higher bar for external acceptance.

At HEOR we are using AI to strengthen our research capabilities and enhance the quality of our work. Specifically, we’re building bespoke, task-specific AI tooling suited to each project. In particular, we’ve found value in deploying GenAI alongside our existing quality framework as an additional layer of rigour, most notably during screening and data extraction. Used during scoping it can help us iterate toward a more efficient review design before the protocol is locked. Used in these ways, AI does not replace methodological judgement or accountability, it gives our researchers another tool to help set projects up for success.

References

[1] Fleurence RL, Qureshi R, Aggarwal R, Bian J, Dawoud D, Delnoij D, Faden RR, Klijn S, Sonabend R, Chhatwal J, GenAI for Systematic Literature Reviews ISPOR Task Force. Generative Artificial Intelligence for Systematic Literature Reviews: A Good Practices Report of an ISPOR Special Task Force. Value Health. 2026. doi:10.1016/j.jval.2026.08.007.

[2] Flemyng E, Noel-Storr A, Macura B, et al. Position statement on artificial intelligence (AI) use in evidence synthesis across Cochrane, the Campbell Collaboration, JBI and the Collaboration for Environmental Evidence 2025. Cochrane Database Syst Rev. 2025;(10). doi:10.1002/14651858.ED000178.

[3] US Food and Drug Administration. Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products – Guidance for Industry and Other Interested Parties. 2025. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-use-artificial-intelligence-support-regulatory-decision-making-drug-and-biological

 

Footnotes

HEOR uses approved enterprise AI tools to support its research while safeguarding client information through encryption, restricted access, and controlled data retention. All AI-assisted work is reviewed and verified to the same standard as any other deliverable, with named accountability and transparent disclosure where required.

More insights

Phil McEwan, CEO at HEOR, on describing the full value of a new technology in an HTA submission
The future of genomic medicine will be shaped not just by the data we generate, but by how well we help patients understand and benefit from it
The success of genomic screening will depend not only on what it can detect, but on whether acting on that information delivers meaningful value
The question is no longer whether genomics can improve healthcare. Increasingly, the challenge is whether healthcare systems are equipped to use genomic information effectively