P32 Using genetic information to define idiopathic pulmonary fibrosis in UK biobank. (11th November 2022)

• Record Type: Journal Article
• Authors: Leavy, OC; Allen, RJ; Kraven, LM; Morgan, AD; Tobin, MD; Quint, JK; Jenkins, RG; Wain, LV
• Is Part Of: Thorax. Volume 77(2022)Supplement 1
• Page Start: A97
• Page End: A97
• Publication Date: 2022-11-11
• Subjects: Chest -- Diseases -- Periodicals; Thorax; Chest -- Diseases; Periodicals; Periodicals; 617.54
• DOI: 10.1136/thorax-2022-BTSabstracts.168
• Languages: English
• ISSNs: 0040-6376

Abstract

Abstract : Introduction: Idiopathic pulmonary fibrosis (IPF) is a complex, heterogeneous fibrotic lung disease with median survival of 3 years. IPF can be defined in population studies, such as UK Biobank, using electronic healthcare records (EHR). However, recent genetic studies of IPF using EHR have shown an attenuation of effect size for known genetic risk factors when compared with clinically-derived datasets, suggesting misclassification of cases. Method: Using various combinations of primary and secondary care EHRs and questionnaire data we defined IPF cases in UK Biobank and evaluated the definitions using association results for the largest genetic risk variant for IPF (rs35705950-T, MUC5B ). We further evaluated the impact of exclusions based on co-occurring codes for non-IPF pulmonary fibrosis and restricting codes according to changes in diagnostic practice. Results: Odds ratio estimates for rs35705950-T associations with IPF defined using EHR and questionnaire data in UK Biobank were significant and ranged from 2.06 to 3.09, which was lower than those reported using clinically-derived IPF datasets (OR: 4.99 to 5.06). We then evaluated the effect of excluding cases with co-occurring codes that might indicate misclassification, and excluding EHR codes that occurred before the most recent clinical guidelines for diagnosis of IPF. Code-based exclusions of cases and code occurrences gave slightly closer effect estimates to those previously reported, but sample sizes were substantially reduced. Conclusion: We show that when using UK Biobank to identify IPF, the effect size for the association between rs35705950-T and IPF risk is smaller than for clinically-derived IPF datasets. Further code-based exclusions did not lead to effect estimates closer to those expected. Though general population cohorts help to increase study sample size, future IPF research using general population datasets should take these limitations of EHR definitions of IPF into consideration.