Combining Gene Signatures Improves Prediction of Breast Cancer Survival
Author Information
Author(s): Zhao Xi, Rødland Einar Andreas, Sørlie Therese, Naume Bjørn, Langerød Anita, Frigessi Arnoldo, Kristensen Vessela N., Børresen-Dale Anne-Lise, Lingjærde Ole Christian
Primary Institution: Oslo University Hospital, Norway
Hypothesis
Can combining information from multiple gene signatures improve the prediction of breast cancer survival?
Conclusion
Combining the predictive strength of multiple gene signatures improves prediction of breast cancer survival.
Supporting Evidence
- Hierarchical clustering revealed two distinct risk groups with different survival probabilities.
- The combined-PI risk predictor was the most significant among all tested predictors.
- Using principal components analysis improved the prediction of survival outcomes.
- Patients with a positive prognostic index had a higher risk of relapse and death.
Takeaway
This study shows that using a mix of different gene tests can help doctors better predict how long breast cancer patients will live.
Methodology
The study used microarray data from two cohorts of breast cancer patients, applying a Cox proportional hazards model to analyze gene expression and predict survival outcomes.
Potential Biases
Potential bias due to the reliance on previously published gene signatures and the inherent variability in gene expression data.
Limitations
Some gene sets did not achieve model convergence, which may affect the robustness of the predictions.
Participant Demographics
The study included 123 patients in the training set and 81 patients in the test set, primarily early-stage breast cancer cases.
Statistical Information
P-Value
p=0.003 for recurrence; p=0.001 for breast cancer death
Confidence Interval
95% CI 1.37—5.80 for combined-PI risk predictor
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website