Novel Clinical Tool to Estimate Risk of False-Negative KRAS Mutations in Circulating Tumor DNA Testing

Area of Science:

• Oncology
• Molecular Diagnostics
• Biostatistics

Background:

• Circulating tumor DNA (ctDNA) testing offers a noninvasive method for determining RAS mutation status in metastatic colorectal cancer.
• However, false-negative results in ctDNA testing can occur, potentially impacting critical treatment choices.
• Accurate assessment of RAS mutational status is crucial for guiding targeted therapy in colorectal cancer.

Purpose of the Study:

• To develop and validate a statistical model for assessing the probability of false-negative results in ctDNA-based RAS mutation detection.
• To provide clinicians with a tool to better interpret ctDNA results and mitigate risks associated with missed mutations.
• To enhance the reliability of noninvasive molecular diagnostics in oncology.

Main Methods:

• A Bayesian statistical model was developed using ctDNA data from two independent cohorts (n=172 and n=146).
• The model utilizes the frequencies of reference mutations (APC and TP53) to estimate the probability of KRAS false negatives.
• Model performance was evaluated through cross-cohort validation and simulations, using Guardant assays data.

Main Results:

• The model indicates that negative KRAS findings are reliable (posterior probability of false negative <5%) when the maximum frequency of APC or TP53 mutations is at least 8%.
• Validation studies confirmed the model's ability to accurately distinguish between false-negative and true-negative ctDNA results.
• Simulations demonstrated the practical utility of the proposed statistical approach in real-world clinical scenarios.

Conclusions:

• Clinicians can use this statistical tool to precisely quantify the likelihood of false-negative KRAS ctDNA results, particularly when reference mutation frequencies are below 8%.
• The methodology can be extended to predict false negatives for other genes and incorporate additional reference markers.
• An open-source R Shiny application is available to facilitate the widespread adoption and application of this tool.