Using artificial intelligence tools in answering important clinical questions: The KEYNOTE-183 multiple myeloma...

Area of Science:

• Clinical Oncology
• Artificial Intelligence in Medicine
• Hematology

Background:

• The KEYNOTE-183 phase III trial investigated pembrolizumab plus standard of care (SOC) for refractory/relapsed multiple myeloma (rrMM).
• The trial was halted due to increased deaths in the pembrolizumab arm.
• Traditional subgroup analyses may not fully explain these findings.

Purpose of the Study:

• To systematically identify prognostic and predictive risk factors for death using artificial intelligence (AI).
• To pinpoint subgroups contributing to overall or excess deaths in the KEYNOTE-183 study.
• To investigate the interaction between treatment arm and patient characteristics.

Main Methods:

• Utilized artificial intelligence tools for a systematic analysis of the KEYNOTE-183 dataset (data cutoff June 02, 2017).
• Identified prognostic factors for overall death and predictive factors for excess death in the pembrolizumab plus SOC arm.
• Conducted subgroup analyses based on identified factors, including ECOG performance status.

Main Results:

• Plasmacytoma was identified as a prognostic factor for death.
• ECOG performance status emerged as a predictive factor for death.
• A qualitative interaction between ECOG performance status and treatment arm was observed, with higher deaths in the pembrolizumab plus SOC arm for patients with ECOG performance status 1.

Conclusions:

• AI-driven analysis successfully identified key risk factors in the KEYNOTE-183 trial.
• ECOG performance status is a critical factor in predicting excess mortality associated with pembrolizumab plus SOC in rrMM.
• Findings highlight the importance of AI in understanding complex clinical trial data and informing patient risk stratification.