Scientific literature reviews and the potential of AI for evidence synthesis

The most comprehensive type of review, systematic literature reviews (SLRs), are valued for their rigorous and reproducible approach to evaluating key research questions. The transparent methodology limits bias, and SLRs are considered the gold standard for evidence-based medicine.¹ Other types of scientific literature reviews, such as comprehensive targeted reviews, employ less stringent methods but still provide extensive information on topics like treatment and disease landscapes. Both SLRs and comprehensive targeted reviews guide strategy and support regulatory submissions, making them trusted sources for drug development.

However, SLRs and comprehensive targeted reviews are labor- and time-extensive processes, often taking months or even years. These lengthy timelines and the rapidly growing number of published articles and available journals mean new evidence can quickly supersede older scientific literature reviews.² This can hinder decision-making processes, particularly in fast-paced fields like healthcare and drug development, where timely and accurate information is crucial.

In response, there has been growing interest in tools and methodologies that can support greater efficiencies, such as artificial intelligence (AI), that could speed up the time involved in assessing and updating scientific literature reviews.³

It’s vital, however, that such tools are validated and that their use is guided by expert knowledge of when and where it is most appropriate, as well as clear understanding of how it will be applied, particularly in contexts of use where scientific rigor of method and results are paramount, such as SLRs.

Staying ahead of the evidence

The most rigorous and extensive scientific literature reviews are SLRs carried out as part of health technology assessment (HTA) submissions, which adhere to strict guidelines from HTA agencies. These SLRs should be completed by experienced researchers. Due to the limitations of AI-assisted reviews, including the chance of missing relevant research, AI is not advised for SLRs informing HTA decisions.

However, the guidance in this space is evolving. The UK’s National Institute for Health and Care Excellence (NICE) has published a position statement on the use of AI in evidence generation which emphasizes early engagement with NICE and dialogue with NICE technical teams if using AI methods.⁴ Eventually, using AI as a second reviewer may be appropriate if the AI tools have been appropriately validated.

SLRs are also carried out for internal strategy and for publication. These types of SLRs are also crucial for identifying evidence gaps and can inform integrated evidence plans (IEPs). Scientific rigor is still critical for these SLRs. Therefore, it is advised to consider AI as a second reviewer for the first phase of literature screening.

Having one human reviewer and one AI reviewer with conflicts resolved by a third human reviewer provides a built-in quality check. Additionally, any AI-assisted methods should be transparently outlined in SLRs reports or publications, which is aligned with the 2020 PRISMA guidelines on the need for a “transparent, complete, and accurate account of why (a) review was done , what the authors did (such as how studies were identified and selected), and what they found (such as characteristics of contributing studies and results of meta-analyses).”⁵ If these types of SLRs can be conducted more quickly and at lower cost with the assistance of AI, SLRs can be used earlier in the drug development process and can be updated more frequently. This can better inform IEPs and lead to more effective evidence generation planning.

oncology researchers viewing results on a tablet

Another type of scientific literature review that is particularly helpful in the HEOR space is the comprehensive targeted review. Comprehensive targeted reviews often cover disease and treatment landscapes and inform early drug development. AI-assisted methods are particularly useful for broad topics where there is a large amount of published literature. Using AI to complete the first phase of literature screening is suggested as an option for time- and cost-savings. Although comprehensive targeted reviews have less stringent methodology requirements than SLRs, human-in-the-loop processes are still important, and implementing quality checks of a pre-determined percentage of references reviewed by AI is also a suggested option.

Finally, updating scientific literature reviews – be they SLRs or comprehensive targeted reviews – is another type of evidence synthesis where AI use can bring benefits. Existing scientific literature reviews can be used as training data, and AI tools can be used to assess whether there is sufficient new, relevant published literature to perform a review update. Monitoring the recently published literature will ensure that review updates will be conducted in a timely and appropriate manner.

Cautious adoption of AI in evidence generation

Finding ways to meet the requirements of the HTA authorities while reducing some of the time-intensive processes involved is a big priority for companies, since every day that a company’s product is not reimbursed is lost revenue. NICE guidance does go some way to support companies in the judicious use of AI to enhance decision-making in evidence generation. However, NICE has cautioned about the transparency and trustworthiness of AI.⁴

In the right context and properly applied, AI tools can empower companies by making it easier and more cost efficient to carry out SLRs more frequently and consequently help them make better-informed decisions.

PRISMA provides guidance on the use of automation in its 2020 expanded checklist. This includes reporting how automation tools were integrated within the overall study selection process as well as the application of machine learning in the screening process and what validation was carried out to understand the risk of missed studies or incorrect classifications.⁶

Surveys of Healthcare decision-makers show there is growing comfort with AI tools for evidence aggregation and evidence summarization. However, there remain legitimate concerns about security and privacy when using AI tools, as well as possible bias inadvertently introduced by reviewers. Furthermore, it is important to emphasize that AI requires considerable input from researchers and medical experts and that it is vital to ensure accuracy when researching topics that could impact people’s health and healthcare.

Some of these concerns can be addressed by utilizing literature review workflows where validated AI tools are guided by experienced researchers. Human-in-the-loop quality checks also provide a way to mitigate risk and maintain research integrity. There is scope for AI to mature as a key tool for literature reviews on condition there is a solid basis of training and validation in its development.

Appropriate and trusted AI-enabled tools that are geared toward the different parts of the evidence generation process can go a long way to removing some of the burden, giving companies the insights they need to meet their market access objectives.

About the authors

Kimberly Ruiz is Senior Director, Evidence Generation and Value Communication at Cencora. She leads teams of researchers in the conduct of systematic and targeted literature reviews, and various types of medical communications work, including scientific publications and dossier development. She has more than 20 years of experience as a researcher in the consulting, corporate, and non-profit sectors of the healthcare industry.

Malia Gill is Manager, Evidence Generation and Value Communications at Cencora. She conducts systematic and targeted literature reviews to provide a comprehensive understanding of trends in the literature base. Her work supports HTA submissions, meta-analyses, economic and epidemiological models, and scientific publications.

Kimberly Ruiz

Senior Director, Evidence Generation and Value Communication, Cencora

Malia Gill

Manager, Evidence Generation and Value Communications, Cencora

This article summarizes Cencora’s understanding of the topic based on publicly available information at the time of writing (see listed sources) and the authors’ expertise in this area. Any recommendations provided in the article may not be applicable to all situations and do not constitute legal advice. Readers should not rely on the article in making decisions related to the topics discussed.

Report

Anticipating the top market access trends globally for 2025

The 2025 healthcare landscape is poised for significant change. Cencora’s report, “Anticipating the top market access trends globally for 2025,” highlights key trends that will impact patient access to innovative therapies.

Key insights include navigating global policy changes, understanding the evolving dynamics between manufacturers and payers, addressing drug pricing pressures, leveraging digital communication for stakeholder engagement, and enhancing equity in US healthcare

Read the report

Connect with our team

Our team of leading value experts is dedicated to transforming evidence, policy insights, and market intelligence into effective global market access strategies. Let us help you navigate today’s complex healthcare landscape with confidence. Reach out to discover how we can support your goals.

Contact us

References:

^1. OCEBM Levels of Evidence, Centre for Evidence-Based Medicine. https://www.cebm.ox.ac.uk/resources/levels-of-evidence/ocebm-levels-of-evidence
^2. Literature search approaches in an era of increasing publication volume, ISPOR poster. https://www.ispor.org/docs/default-source/euro2024/isporeurope24cadarettesa57poster146827-pdf.pdf?sfvrsn=39a72ef6_0
^3. How much can we save by applying artificial intelligence in evidence synthesis? Results from a pragmatic review to quantify workload efficiencies and cost savings, Front Pharmacol. 2025 Jan. https://pmc.ncbi.nlm.nih.gov/articles/PMC11826052/
^4.Use of AI in evidence generation: NICE position statement. https://www.nice.org.uk/about/what-we-do/our-research-work/use-of-ai-in-evidence-generation--nice-position-statement
^5. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews, BMJ, 2021. https://www.bmj.com/content/372/bmj.n71
^6. PRISMA 2020 expanded checklist. https://www.bmj.com/content/bmj/suppl/2021/03/29/bmj.n71.DC1/pagm061899.w2.pdf

Related resources

Guide

HEOR Fundamental Education Series

Learn more

Article

Insights from ISPOR 2024: Looking for improvements in tools used in systematic review assessments

Learn more

Young, Black, female employee looks at a clipboard

Guide

Checklist: Why you need a learning solutions partner with market access and reimbursement expertise

Learn more

Cencora.com is providing automated translations to assist in reading the website in languages other than English. For these translations, reasonable efforts have been made to provide an accurate translation, however, no automated translation is perfect nor is it intended to replace human translators. These translations are provided as a service to users of Cencora.com and are provided "as is." No warranty of any kind, either expressed or implied, is made as to the accuracy, reliability, or correctness of any of these translations made from English into any other language. Some content (such as images, videos, Flash, etc.) may not be accurately translated due to the limitations of the translation software.

Any discrepancies or differences created in translating this content from English into another language are not binding and have no legal effect for compliance, enforcement, or any other purpose. If any errors are identified, please contact us. If any questions arise related to the accuracy of the information contained in these translations, please refer to the English version of the page.