SAT0710 Risk association for ankylosing spondylitis using a genetic risk score combining 110 snps of genome-wide significance in the population-based hunt study

Background The genetic component of AS development is estimated to ~90%. About 20% of the known heritability for AS is attributed to HLA-B27 and about 7.4% to 113 SNPs found in genome-wide association studies,1 with a further ~60% of heritability determined by as yet unmapped variants. Objectives To evaluate whether a weighted genetic risk score (wGRS) developed based on the currently identified SNPs is predictive of AS in the Norwegian population-based Nord-Trondelag Health Study (HUNT). Methods HUNT invited the entire adult population of Nord-Trondelag county. We used data from HUNT2 (1995–97) and HUNT3 (2006–8). AS cases were diagnosed from hospital records using the Modified New York Criteria. Participants with other inflammatory arthritis were excluded, leaving 181 AS cases and 55 586 controls. Genotyping was performed with Illumina HumanCoreExome arrays. Imputation was performed with Minimac3 based on European ancestry data. Imputed or genotyped data were available for 110 (97%) of the 113 SNPs. We used rs4349859 to indicate HLA-B27 carrier state (positive/negative). The wGRS was calculated by addition of risk alleles and weighting by the published OR from,1 representing each person’s carriage of all risk variants. Data were analysed using logistic regression. Four models were compared using AUC analysis: 1) the wGRS only; 2) HLA-B27 only; 3) the wGRS and HLA-B27; 3) the wGRS, HLA-B27, age, gender, BMI, hypertension, and smoking (never/previous/present). HUNT had ethical approval and participants gave informed consent. Results At baseline, mean age for cases was 43 years, 61% were men and 87% were HLA-B27 positive. The corresponding figures for controls were 46 years, 47% men and 12.7% HLA-B27 positive. The mean wGRS was 14.37 (range: 10.93 to 17.41). The wGRS alone was associated with AS (OR=1.7 for one unit increase, p Conclusions The wGSR was associated with AS, but had low predictive ability in a population-based setting. HLA-B27 was a much better predictor. Addition of clinical variables only slightly improved prediction, in accordance with the high genetic component in AS pathophysiology. Discovery and inclusion of more genetic risk variants, epigenetic factors, other demographic factors, and interaction terms, in addition to more efficient statistical approaches such as genome-wide risk score development, could improve prediction. Study limitations are false positive- or –negative AS diagnoses and potential selection bias of participants in HUNT. Reference [1] Ellinghaus D, et al. Analysis of five chronic inflammatory diseases identifies 27 new associations and highlights disease-specific patterns at shared loci. Nature Genetics2016;48(5):510–18. Disclosure of Interest None declared