Search in the Abstract Database

Search Abstracts 2016

* = Presenting author

P733 Genetic risk scores for risk and therapy response prediction in Slovenian Crohn’s disease patients

K. Zupančič*1, K. Skok1, K. Repnik1, P. Skok2, U. Potočnik1

1Medical faculty, University of Maribor, Centre for human genetics and pharmacogenomics, Maribor, Slovenia, 2University medical Centre, Department of gastroenterology, Maribor, Slovenia

Background

Genome-wide association studies have provided a comprehensive catalogue of more than 140 risk loci associated with Crohn´s disease (CD). Attempts have been made to develop accurate genetic risk profiles based on a large number of patients enrolled from several ethnically and genetically heterogeneous cohorts. The aim of our study was to develop CD risk model based on a homogeneous Slovenian cohort. In addition, we also examined the possibility if common variants could sufficiently predict response to CD treatment.

Methods

In our study, we included 160 CD patients (90 refractory [CDr] and 70 non-refractory CD patients) and 209 healthy individuals from Slovenian population. The association study was performed for 74 single nucleotide polymorphisms (SNPs), obtained from Immunochip. We constructed genetic risk scores (GRS) as a sum of risk alleles using weighted additive risk model. Discriminatory accuracy was measured by an area under ROC curve (AUC).

Results

The highest accuracy of AUC 0.78 (0.79 and 0.77 for CDr and non-CDr, respectively) was achieved with GRS combining 33 SNPs with optimal sensitivity and specificity of 75.0% and 72.7%, respectively. For risk evaluation, we classified individuals in different risk groups according to their GRS and corresponding likelihood ratios of a test (LR > 5 and LR < 0.2 for high- and low-risk group, respectively). Individuals with the highest risk (GRS > 5.54) showed significantly increased odds of developing CD (OR 26.65; 95% CI 11.25–63.15, p < 0.0001) compared with the individuals with the lowest risk (GRS < 4.57). However, the proportion of CDr patients did not differ significantly between the high- and low-risk groups (OR 1.09, 95% CI 0.31–3.84, p = 0.82). To discriminate between CDr and non-CDr, we used a different combination of SNPs and achieved the highest accuracy of AUC 0.73 in GRS comprising of 30 SNPs with optimal sensitivity and specificity of 75.6 and 64.3%, respectively. In the high-risk group (GRS > 3.55), there was a significantly greater proportion of CDr patients in comparison with the low-risk group (GRS < 2.54) with OR of 14 00 (95% CI 3.53–39.92, p = 0,001).

Conclusion

Our results suggest that genetic risk model of common variants shows good discriminatory accuracy for stratifying individuals at a higher CD risk and might be useful for screening purposes in targeted population (positive family history and prolonged gastrointestinal disturbances). Further, common risk variants may also serve as a promising therapy response predictors to distinguish between different CD sub-phenotypes. In the future, a larger Slovenian CD cohort will be needed to verify the results and to further optimise population-specific risk prediction.