# Long-term infection risks in haematological cancer survivors compared with individuals with no cancer history: protocol for a systematic review aided by artificial intelligence-based methods

## TL;DR

This study aims to use AI to better understand infection risks in long-term survivors of blood cancers compared to those without cancer history.

## Contribution

The novel use of AI tools like ASReview streamlines the systematic review process for assessing long-term infection risks in haematological cancer survivors.

## Key findings

- AI-based screening will be validated against manual review to ensure accuracy in identifying relevant studies.
- The review will summarize infection incidence and mortality by cancer type and time since diagnosis.
- Narrative synthesis and potential meta-analyses will provide insights into long-term infection risks in cancer survivors.

## Abstract

Infections are a major cause of morbidity and mortality among individuals with haematological cancers, but the duration of elevated risk in long-term survivors remains uncertain. Although previous attempts to summarise the existing literature on this topic would have been hampered by the sheer volume of studies on cancer and all-cause infections, emerging artificial intelligence tools now offer the ability to streamline the screening process, allowing for broader and more comprehensive reviews.

This protocol follows the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols guidelines. Eligible studies will include original observational data reporting long-term (≥1 year follow-up from diagnosis) infection-related outcomes in haematological cancer survivors compared with a general or cancer-free population. Screening will be supported by ASReview, an artificial intelligence-based tool for abstract prioritisation. An internal validation step will be conducted by comparing artificial intelligence-assisted screening results with manual review performed by two independent researchers on a subset of abstracts. The primary outcomes of infection incidence and infection mortality will be summarised by type of infection, type of haematological cancer and time since cancer diagnosis. Information on anti-cancer treatments received will also be described. Data synthesis will be mostly narrative due to the broad scope of the review, though meta-analyses will be performed in cases where studies are sufficiently homogenous. Risk of bias will be assessed using the Newcastle-Ottawa Scale.

Ethical approval is not applicable to this study. The results of the review will be disseminated to clinical audiences and submitted to a peer-reviewed journal.

CRD420251047091.