Predicting Treatment Outcomes for Kidney Cancer Using Multiple Data Types
Author Information
Author(s): Chen Meixu, Wang Kai, Kapur Payal, Brugarolas James, Hannan Raquibul, Wang Jing
Primary Institution: University of Texas Southwestern Medical Center
Hypothesis
Can a multi-modal ensemble model improve the prediction of treatment outcomes in clear cell renal cell carcinoma (ccRCC) patients?
Conclusion
The multi-modal ensemble model (MMEM) outperformed single-modality models in predicting overall survival and disease-free survival for ccRCC patients.
Supporting Evidence
- The MMEM achieved C-indices of 0.820 and 0.833 for overall survival and disease-free survival, respectively.
- AUROC values for binary predictions were 0.831 and 0.862 for patient death and cancer recurrence at 3 years.
- The clinical feature-based model had the highest weight in the ensemble model.
- Using multiple data types improved prediction accuracy compared to single-modality models.
Takeaway
Researchers created a smart model that uses different types of patient data to better predict how well kidney cancer treatments will work.
Methodology
The study used clinical data, multi-omics data, and histopathology images to develop a multi-modal ensemble model for predicting patient outcomes.
Potential Biases
Potential bias due to reliance on a single dataset and the subjective nature of histopathological analysis.
Limitations
The model needs external validation and could benefit from incorporating additional data modalities.
Participant Demographics
{"gender":{"male":146,"female":80},"age":{"median":60,"range":"26-86"},"race":{"white":186,"african_american":38,"other_and_unknown":2},"ethnicity":{"hispanic_or_latino":8,"not_hispanic_or_latino":181,"unknown":37}}
Statistical Information
P-Value
p<0.05
Confidence Interval
{"os":"0.820 (0.721, 0.908)","dfs":"0.833 (0.773, 0.891)"}
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website