C72-34 Using Urinary Inflammatory Biomarkers to Develop Novel Phenotypes of Prematurity-Associated Lung Disease

Rationale Chronic respiratory disease affects many individuals born very preterm (≤32 weeks gestation). Prematurity-associated lung disease is complex and heterogenous, with individuals displaying phenotypic traits that mimic other respiratory conditions. A multidimensional model to determine individual phenotypes of lung diseases has been proposed as a first step in optimising the management of prematurity-associated lung disease. Topological data analysis is a framework of statistical methods, including cluster analysis, that can be used to identify unique groupings or “clusters” of individuals who share similarities, but are dissimilar to those in other clusters. This data-driven method can be used to identify phenotypes of disease. We aimed to identify novel phenotypes of prematurity-associated lung disease by applying Topological data analysis using the Mapper algorithm to urinary biomarkers. Methods Urine samples were collected from young-adults born very preterm (1997-2003) in Western Australia. Concentrations of urinary biomarkers (interleukin-8 & leukotriene-B4 [neutrophilic inflammation], cysteinyl leukotriene [eosinophilic inflammation], clara cell protein [low levels associated with epithelial injury] & 8-isoprostane [oxidative stress]) were measured via ELISA and normalised to specific gravity. Pulmonary function testing was completed as per ERS/ATS guidelines and perinatal history was collected at time of enrolment. The Mapper algorithm was applied to urinary biomarkers to develop clusters. Differences between groups was described using lung function data and perinatal characteristics. Results Of 127 very preterm individuals (median[IQR] age 19.3[18.4, 20.4] years), 56% (N = 72) provided urine samples, with 95% (N = 69/72) with complete biomarker data for analysis. The Mapper algorithm identified three clusters (1-3) from 67 individuals (Figure 1). Cluster 1 (N = 17) had the lowest levels of inflammatory markers. Cluster 2 (N = 25) had the highest levels of leukotriene-B4, cysteinyl leukotriene and 8-isoprostane. Cluster 3 (N = 25) had the lowest level of clara cell protein. Interleukin-8 levels did not differ between groups. Cluster 3 had significantly worse mean±SD oscillometry z-scores for resistance at 5Hz (Rrs5) (1.7±1.1) and area under the reactance curve (AX) (2.2±1.2) compared to cluster 1 (Rrs5 0.8±1.8; AX 1.4±1.3) but not cluster 2 (Rrs5 1.2±0.9; AX 2.0±1.0). Antenatal corticosteroid use was lowest in cluster 1 (65%) compared to cluster 2 (92%) and 3 (91%). No differences were seen in other perinatal characteristics or lung function measures between groups. Conclusions Increased Rrs5 and AX z-scores, with lower levels of clara cell protein may indicate a small airway disease phenotype in this population. This study highlights the importance of using multiple biomarkers for diagnostic testing.