Review finds artificial intelligence use in low-resource cancer care remains at early stages
Across 18 studies in low- and middle-income countries, few systems had advanced workflow integration or evidence of readiness for safe scale-up.
High evidenceReviewInterpret with caution
Medical disclaimer: This article summarizes research findings and is for informational purposes only. It is not medical advice.
Editorial illustration — not from the study.
Researchers searched six databases for studies published from 2000 through 2025 and classified reported artificial intelligence systems according to how deeply they were integrated into radiation oncology workflows. The review included studies from low- and middle-income countries and assessed areas such as validation, deployment, workflow integration, and governance.
Half of the studies described artificial intelligence used for individual tasks, and about one-third described systems integrated into clinical workflows with clinician supervision. No study met the review's highest category for fully established operation. The findings describe the state of reported deployments, not proof that these systems improve patient outcomes or access to care.
What the review covered
The researchers conducted a systematic review using PRISMA 2020 methods. They searched six databases for relevant work published between January 2000 and December 2025, identifying 18 studies about artificial intelligence deployments or implementation readiness in radiation oncology in low- and middle-income countries. Studies were placed on a radiation-oncology-specific integration scale from Level 0, covering readiness or governance without clinical deployment, to Level 4, representing the most advanced operation. The review also examined workflow integration, validation, deployment, and governance.
What the evidence shows
Two of the 18 studies, or 11.1%, addressed readiness or governance without a clinical artificial intelligence deployment. Nine studies, or 50.0%, reported task-level applications, while six studies, or 33.3%, described integration into workflows under clinician supervision. One study, or 5.6%, reported coordination across multiple stages of the workflow. No study provided enough evidence to qualify for Level 4 operation. Validation was mainly retrospective, and the review found no reports of continuous performance monitoring. These findings suggest that reported deployments remained concentrated at early stages of integration.
Who this may apply to
These findings are most relevant to artificial intelligence projects in radiation oncology in low- and middle-income countries, particularly efforts to assess implementation readiness and scale-up. They may not describe systems in high-income countries or settings with different staffing, infrastructure, regulations, and governance. The review does not establish that any particular system is effective or safe for individual patients, and it does not provide evidence that the reported deployments improve outcomes.
The significance
Radiation oncology services in some low- and middle-income countries face limited specialist capacity and other resource constraints. Artificial intelligence may eventually be relevant to workflow support, but this review indicates that reported use has not yet demonstrated the infrastructure, validation, monitoring, and governance needed for safe and scalable implementation. The review evaluates deployment readiness rather than patient benefit, safety outcomes, or improved access to treatment.
Limitations & evidence assessment
The review included only 18 studies, so its conclusions depend on a relatively small evidence base. The abstract reports that validation was predominantly retrospective and that continuous performance monitoring was not reported; these limitations make it difficult to know how systems perform over time or under routine clinical conditions. No study reached the highest integration category, and the abstract does not provide enough detail to assess the quality, consistency, or possible publication bias of the individual studies. As a systematic review, the findings also reflect what was reported and may not capture every deployment in practice.
Why this evidence level: Systematic review aggregating primary studies with explicit methodology.
Evidence levels are editorial estimates derived from study metadata — they are not clinical appraisals.
// Source
npj Digital Medicine · 2026 · DOI: 10.1038/s41746-026-03391-x
This systematic review examined randomized human trials of concurrent chemotherapy and radiation for locally advanced head and neck, cervical, esophageal, and non-small-cell lung cancers. Across the included trials, combined treatment was reported to provide better overall survival than radiation alone, but it also caused more severe short-term side effects. The review did not provide enough information in its abstract to show how many trials or patients were included.
Researchers pooled results from seven human studies comparing multiple-staged with single-stage removal and reconstruction of congenital melanocytic nevi, sometimes called congenital moles. Staged surgery was associated with fewer major wound complications and recurrences, as well as slightly shorter and narrower scars, but the studies did not find a clear difference in cosmetic or overall patient satisfaction.
A cluster-randomized human study in 210 adults with recently detected stage 1 hypertension in Uzbekistan compared salt-sensitivity testing followed by assigned medication choices with conventional management. After 12 weeks, more participants in the testing-guided group reached the study's blood pressure target, although the study was small, short, and conducted in only three primary care centers.
