Validating Risk Prediction Models for Multiple Primaries and Competing Cancer Outcomes in Families With Li-Fraumeni...

Area of Science:

• Genetics and Genomics
• Oncology
• Clinical Decision Support

Background:

• Disseminating risk prediction models into clinical practice faces barriers.
• Research cohorts are meticulously collected, unlike real-world clinical data.
• Demonstrating model utility with incomplete clinical data may overcome these barriers.

Purpose of the Study:

• To evaluate the performance of risk prediction models using a clinical counseling-based (CCB) cohort.
• To assess if models developed for Li-Fraumeni syndrome (LFS) risk prediction perform well with real-world clinical data.
• To compare model performance against standard clinical criteria.

Main Methods:

• A CCB cohort of 3,297 individuals across 124 families was used.
• Genetic counselors collected family history data for LFS risk assessment.
• The LFSPRO software suite was applied to predict TP53 mutations and cancer onset, with performance measured by AUC and O/E ratio.

Main Results:

• For predicting deleterious TP53 mutations, an AUC of 0.78 and an O/E ratio of 1.66 were achieved.
• The LFSPRO.MPC model predicted second cancer onset with an AUC of 0.70.
• The LFSPRO.CS model achieved AUCs between 0.70-0.83 for predicting first primary cancers like sarcoma and breast cancer.

Conclusions:

• Previously validated risk prediction models demonstrate good performance in a CCB cohort.
• These models outperformed standard clinical criteria for LFS risk assessment.
• The findings suggest that genetic counselors can utilize these mathematical models for improved risk counseling.