King's College London

Research portal

Prediction of ankylosing spondylitis in the HUNT study by a genetic risk score combining 110 single-nucleotide polymorphisms of genome-wide significance

Research output: Contribution to journalArticlepeer-review

Sina Rostami, Mari Hoff, Matthew A. Brown, Kristian Hveem, Oddgeir L. Holmen, Lars G. Fritsche, Vibeke Videm

Original languageEnglish
Pages (from-to)204-210
Number of pages7
JournalJournal of Rheumatology
Issue number2
Early online date1 Aug 2019
Accepted/In press19 Mar 2019
E-pub ahead of print1 Aug 2019
Published1 Feb 2020


King's Authors


Objective. The genetic component of ankylosing spondylitis (AS) development is ~90%. Of the known heritability, ~20% is explained by HLA-B27, and 113 identified AS-associated single-nucleotide polymorphisms (SNP) account for ~7.4%. The objectives were to construct a weighted genetic risk score (wGRS) using currently known genome-wide susceptibility SNP, and to evaluate its predictive ability for AS in the Norwegian population-based Nord-Trøndelag Health Study (HUNT).

Methods. AS cases (n = 164) and controls (n = 49,032) were from the second (1995-1997) and third (2006-2008) waves of the HUNT study, to which the entire adult population of the northern region of Trøndelag was invited. A wGRS based on 110 SNP weighted by published OR for AS was constructed, representing each person's carriage of all risk variants. Logistic regression models including the wGRS alone or in combination with HLA-B27 carrier state and other adjustment variables (sex, age, smoking, body mass index, and hypertension) were developed. Discrimination among models was compared using area under the curve (AUC).

Results. The wGRS was associated with AS (OR 1.7, 95% CI 1.4-2.1), but showed low discrimination (AUC 0.62, 95% CI 0.58-0.67). HLA-B27 was significantly associated with AS (OR 50, 95% CI 32-81), showing high discrimination (AUC 0.88, 95% CI 0.85-0.90). Combining the wGRS and HLA-B27 improved prediction (AUC 0.90, 95% CI 0.87-0.92; p < 0.001 vs wGRS alone, p < 0.01 vs HLA-B27 alone). Further inclusion of adjustment variables gave a small improvement (AUC 0.91, 95% CI 0.89-0.94; p = 0.03).

Conclusion. Prediction in a population-based setting based on all currently known AS susceptibility SNP was better than HLA-B27 carrier state alone, although the improvement was small and of uncertain clinical value.

Download statistics

No data available

View graph of relations

© 2020 King's College London | Strand | London WC2R 2LS | England | United Kingdom | Tel +44 (0)20 7836 5454