Developing a Risk-scoring Model for Ankylosing Spondylitis Based on a Combination of HLA-B27, Single-nucleotide Polymorphism, and Copy Number Variant Markers

Abstract

Objective. To develop a genotype-based ankylosing spondylitis (AS) risk prediction model that is more sensitive and specific than HLA-B27 typing.

Methods. To develop the AS genetic risk scoring (AS-GRS) model, 648 individuals (285 cases and 363 controls) were examined for 5 copy number variants (CNV), 7 single-nucleotide polymorphisms (SNP), and an HLA-B27 marker by TaqMan assays. The AS-GRS model was developed using logistic regression and validated with a larger independent set (576 cases and 680 controls).

Results. Through logistic regression, we built the AS-GRS model consisting of 5 genetic components: HLA-B27, 3 CNV (1q32.2, 13q13.1, and 16p13.3), and 1 SNP (rs10865331). All significant associations of genetic factors in the model were replicated in the independent validation set. The discriminative ability of the AS-GRS model measured by the area under the curve was excellent: 0.976 (95% CI 0.96–0.99) in the model construction set and 0.951 (95% CI 0.94–0.96) in the validation set. The AS-GRS model showed higher specificity and accuracy than the HLA-B27–only model when the sensitivity was set to over 94%. When we categorized the individuals into quartiles based on the AS-GRS scores, OR of the 4 groups (low, intermediate-1, intermediate-2, and high risk) showed an increasing trend with the AS-GRS scores (r² = 0.950) and the highest risk group showed a 494× higher risk of AS than the lowest risk group (95% CI 237.3–1029.1).

Conclusion. Our AS-GRS could be used to identify individuals at high risk for AS before major symptoms appear, which may improve the prognosis for them through early treatment.