Background: Commonly used FeNO cut-offs (ATS 2011) do not consider individual differences in expected normal FeNO. Also, dual cut-offs for rule-in and rule-out may be preferable for a biomarker.

Methods: Data from three independent datasets (US NHANES 2007-2012 and European ADONIX and WSAS datasets) were used to derive de novo reference equations to predict FeNO, using the lambda-mu-sigma (LMS) method, in non-smoking Caucasians (6-79 years) without self-reported respiratory disease and atopy/hay fever. Individualized FeNO cut-offs (%predicted) were tested against methacholine hyperresponsiveness combined with atopy, as a surrogate for type-2 airway inflammation, in a separate asthma cohort (MIDAS, n=531), and were compared to the general cut-offs 20 and 25 ppb (< and ? 12 years, respectively). All FeNO measurements were performed with NIOX instruments.

Results: Sex-specific reference equations, including height and age, were derived from 6,092 subjects (52% NHANES, 39% ADONIX and 9% WSAS). The accuracy (AUC [95%CI]) of FeNO %predicted (0.732 [0.683-0.781]) was significantly improved compared to measured absolute FeNO (0.707 [0.664-0.747]) (p=0.001; DeLong?s test). Compared to ATS cut-offs, a FeNO of 202 %predicted for rule-in had slightly higher specificity (88.9% vs. 83.3%) and similar sensitivity (48.5% vs. 50.6%); while 100% predicted for rule-out had significantly higher sensitivity (78.4% vs. 46.1%, p<0.01), while maintaining a fair specificity (49.5%).

Conclusion: The individualized LMS-based FeNO %predicted cut-offs improved the clinical interpretation of FeNO measurements, particularly for ruling out type-2 inflammation-related airway diseases, and can be a useful tool in asthma management.