Leveraging artificial intelligence for decision-making in pediatric progressive and refractory solid tumors.
AI interpretation is pending for this paper.
Open original publication →What the AI sees
Not AI summarized yet.
Research significance
Pending deeper interpretation.
Source abstract
Pediatric patients with progressive and refractory solid tumors face a challenging prognosis. Despite advancements in treatments like immunotherapy and targeted therapy, survival rates remain low for certain tumor types. Decision-making in these complex cases often necessitates a multidisciplinary approach, integrating risk-based management, precision medicine, and access to clinical trials. Artificial intelligence (AI) technologies, particularly large language models (LLMs), hold promise for improving clinical reasoning and decision support in pediatric oncology. This study evaluated the decision-making capabilities of five AI tools-ChatGPT, Gemini, Claude, Perplexity, and OpenEvidence-in six hypothetical cases of refractory or progressive pediatric solid tumors. Each AI tool was presented with two sequential queries: a request to generate potential treatment options and then a request to identify and justify the most appropriate option from its initial list. The AI tools generated a total of 124 treatment recommendations, with an average of 24.8 per tool. Clinical trial enrollment was the most frequently selected "best option," accounting for 55.2% of cases. Other notable recommendations included targeted therapy (17.2%), surgery (10.3%), chemotherapy (10.3%), best supportive care (10.3%), and immunotherapy (3.4%). Notably, the AI tools exhibited distinct tendencies in their decision-making approaches, with some favoring aggressive interventions and others emphasizing supportive or palliative care. AI tools demonstrate potential for assisting with complex treatment decisions in pediatric oncology, particularly by identifying clinical trial options. However, the observed variability in recommendations underscores the need for careful human oversight to ensure that AI-generated suggestions align with clinical evidence, patient and family preferences, and the overall goals of care. Future research should explore how AI tools can be further refined to incorporate nuanced patient-specific information and address the emotional and psychological impact of AI-assisted decision-making.