Background:Improved understanding of the complex genetic architecture of psoriatic arthritis (PsA) along with the reduction in the cost of genetic testing provide an opportunity to assess the application of genetic testing for PsA diagnosis in clinical setting.Objectives:The study aimed to assess the performance of the multi-SNP genetic test in predicting a clinical diagnosis of PsA by a rheumatologist among psoriasis patients with musculoskeletal symptoms.Methods:328 patients with psoriasis and musculoskeletal symptoms that were referred to a rapid access clinic for suspected PsA were enrolled. Patients with a prior diagnosis of PsA were excluded. A rheumatologist evaluated all patients and classified them as “PsA” or “not PsA”. All patients who were classified as not PsA at baseline were reassessed 1 year later to determine whether they have developed PsA. We tested 2 outcomes: 1) diagnosis of PsA at baseline; and 2) diagnosis of PsA at baseline or at 1 year. A custom multi-SNP genetic assay was genotyped on a MassARRAY system (Agena Biosciences). The custom PsA weighted genetic panel included 42 variants in or near 20 genes based on genome-wide significance in PsA studies. We tested the ability of each genetic maker individually to predict PsA using logistic regression models adjusted for age and sex. Machine-learning methods including logistic regression, naïve bayes and random forest were used to identify the optimal prediction model. Age and sex were included in the prediction models.Results:78 patients were classified as PsA (PsA-baseline) and the remaining 250 patients as not PsA. After 1 year, 17 additional patients developed PsA resulting in 95 patients with PsA at 1 year (PsA-1 year). The association between the tested SNPs and PsA is shown in Table 1. 5 SNPs located at LCE3A (rs10888503), TNIP1 (rs146571698) and IL-23R (rs4655683, rs2201841 and rs12044149) were associated with PsA-baseline or PsA-1 year (all p<0.05). Of the three machine-learning methods used, logistic regression was found to have the best prediction properties (highest AUC). Overall, the performance of prediction models to classify patients as PsA was modest (see Table 2). The AUC, sensitivity and specificity of the models to predict PsAat baselinewere: 0.62, 0.15, 0.97, respectively andat 1 year: 0.62, 0.15, 0.94.Table 1.The association between (selected top markers)ChromSNPGenePsA diagnosis at baseline (N=78)PsA diagnosis at 1 year (N=95)OR95% CIP valueOR95% CIP value1rs10888503LCE3A1.781.21, 2.620.0031.521.06, 2.170.026rs12191877HLA-C0.520.31, 0.890.0170.630.39, 1.010.0576rs2894207HLA-C*60.570.35, 0.930.0250.620.39, 0.990.0456rs13214872HLA-C0.570.34, 0.960.0350.650.41, 1.050.0786rs12189871HLA-C0.550.30, 1.010.0550.680.40, 1.170.1696rs4406273HLA-C0.570.32, 1.020.0570.670.40, 1.120.1225rs146571698TNIP12.030.97, 4.240.0612.071.02, 4.180.0441rs4655683IL-23R1.420.97, 2.080.0691.521.06, 2.170.0221rs2201841IL-23R0.720.50, 1.050.0870.680.48, 0.960.0541rs12044149IL-23R1.340.88,2.060.1671.601.07, 2.390.021Table 2.Performance of the genetic assay in predicting PsA2A. Prediction of PsA at baseline (N=78)MethodTPTNFPFNAccuracySensitivitySpecificityAUCLogistic Regression122428660.770.150.970.62Naïve Bayes2720347510.700.350.810.61Random Forest1423713640.770.180.950.562B. Prediction of PsA at 1 year (N=95)MethodTPTNFPFNAccuracySensitivitySpecificityAUCLogistic Regression1521815800.710.150.940.62Naïve Bayes2120330740.680.220.870.62Random Forest1620429790.670.170.880.55Conclusion:Despite the association of several genetic markers with PsA, genetic testing has marginal effect on predicting a diagnosis of PsA among patients with psoriasis and musculoskeletal symptoms.Disclosure of Interests:Lihi Eder Grant/research support from: Abbvie, Lily, Janssen, Amgen, Novartis, Consultant of: Janssen, Speakers bureau: Abbvie, Lily, Janssen, Amgen, Novartis, Quan Li: None declared, Dana Jerome: None declared, Chandra Farrer: None declared, Tanya Burry: None declared, Proton Rahman Grant/research support from: Janssen and Novartis, Consultant of: Abbott, AbbVie, Amgen, BMS, Celgene, Lilly, Janssen, Novartis, and Pfizer., Speakers bureau: Abbott, AbbVie, Amgen, BMS, Celgene, Lilly, Janssen, Novartis, Pfizer