LAUSR

Rationale Timely, accurate diagnosis of invasive aspergillosis (IA) is key to enable initiation of antifungal therapy in lung transplantation. Despite promising novel fungal biomarkers, the lack of a diagnostic gold-standard creates difficulty in determining utility. Objectives This study aimed to use latent class modelling of fungal diagnostics to classify lung transplant recipients (LTR) with IA in a large single centre. Methods Regression models were used to compare composite biomarker testing of bronchoalveolar lavage to clinical and EORTC-MSG guideline-based diagnosis of IA with mortality used as a surrogate primary outcome measure. Bootstrap analysis identified radiological features associated with IA. Bayesian latent class modelling was used to define IA. Measurements and Main Results A clinical diagnosis of fungal infection (P =<0.001) and composite biomarker positive Results (P =<0.001) had significantly increased 12 month mortality. There was poor correlation between clinical diagnosis, EORTC-based IA diagnosis and composite biomarker positivity. Tracheobronchitis was positively predictive of a clinical and composite biomarker positive diagnosis of IA (p=0.004;95% CI–1.79–21.28 and p=0.03;95% CI–0.85–15.62 respectively). Latent class modelling resulted in the formation of 3 groups: Class 1: likely fungal infection; Class 2: unlikely fungal infection; Class 3: unclassifiable. A. fumigatus PCR was positive in ∼90% of class 1 LTRs compared to only 1% in class 2. Analysis of mortality showed a trend towards significance comparing class 1 with class 2 (p=0.06;HR–4.7;95% CI(0.91–24)) (figure 1). Abstract S17 Figure 1 Conclusions This study demonstrates a latent class modelling approach for IA diagnosis in LTR with a combination of culture, composite biomarker testing, and radiology required for optimal IA diagnosis.