Oropharyngeal cancer management in the era of artificial intelligence for optimal decision-making.

OBJECTIVE

To assess the accuracy of AI systems in informing clinical decision-making processes for the management of oropharyngeal malignancies.

METHODS

Structured inquiries derived from established clinical recommendations for oropharyngeal cancer management were formulated. These inquiries were systematically processed through multiple Large Language Model (LLM) platforms. The Claude platform was specifically employed to identify discrepancies between established guideline recommendations and LLM-generated responses. Subsequently, a panel of three subject matter experts conducted an assessment of these identified differences to evaluate their potential clinical significance.

RESULTS

Analysis of the 28 generated responses revealed that 21 cases (75%) showed congruence between LLM outputs and established guideline recommendations. In three instances, the guidelines provided significantly more comprehensive content than LLM responses. One response from the LLM contained additional generated relevant information, while three responses were contradictory to the guidelines.

CONCLUSION

The LLM demonstrated 75% concordance with guideline recommendations and should be considered a complementary tool for established clinical guidelines.

LEVEL OF EVIDENCE: Level 3.