A139
Predictive power of the LLP risk model with SEZ6I SNP
Olaide Y. Raji1, Stephen W. Duffy2, John K. Field1
1Roy Castle Lung Cancer Research Programme, University of Liverpool Cancer Research Centre, Liverpool, UK, 2Cancer Research UK Centre for Epidemiology, Mathematics and Statistics Wolfson Institute of Preventive Medicine, London, UK
Background
Risk prediction models are essential tools for identifying individuals at high risks of developing a disease. Within the Liverpool Lung Project (LLP), we have previously developed a risk prediction model for lung cancer using five important epidemiological risk factors. A recent DNA pooling from three independent studies indicated a role of the SEZ6L Met430IIe polymorphic variant found on 22q12.2 region in increasing lung cancer risk. This work extends the LLP risk model by including the SEZ6L SNP and estimates the degree of improvement obtainable.
Method
Data from high-throughput DNA allele frequency genotyping using a dense genome-wide single nucleotide polymorphism (SNP) marker available on a subset of patients participating in the Liverpool Lung Project (LLP) are combined with the previously identified risk factors. Multivariable conditional logistic regression model that includes the five epidemiological predictors and the gene marker is currently being developed. The model fit will be assessed using the area under the receiver-operating characteristic curves (AUC) from the bootstrap validation process.
Results
Results from the model fitting will be presented summarising the measure of fit and odds ratios. The improvement over the original model will be quantified and assessed for statistical significance. The absolute risks of lung cancer within a 5-year period for hypothetical individuals with different profiles will be presented.
Conclusion
The method represents an experimental approach to incorporate genetic biomarkers in risk model for predicting lung cancer occurrence in high risk population. However, this single SNP is expected to have minimal impact on the model predictive power.
