(695) Correcting legacy multiple breath nitrogen washout data (MBWN­₂) is not equivalent to collecting new data directly using the corrected software

Recently Wyler et al 2022 [1], discovered a cross-talk error with commercial MBWN₂ software 3.1.6 which was in widespread use at the time. The error caused an absolute over-reading of N₂ of ~1%, i.e., at 2% N₂ read as 3%. This caused an extended tail to the washout and over-estimated lung clearance index (LCI_2.5) values. Subsequently, the same group, along with the software manufacturers released an updated and corrected software version, 3.3.1. Since there have been discussions within the field on how to correct legacy data, whether to migrate or completely “rerun" raw data A-files from 3.1.6 into 3.3.1. To our knowledge, no research has been published assessing whether either method is equivalent to directly collecting data in the new corrected 3.3.1 software.

Methods:

We prospectively recruited 19 participants, 10 adult healthy controls (HC: no current/ previous history of respiratory disease, or symptoms at the time of testing) and 9 people with cystic fibrosis (PwCF). MBWN₂ was performed using the Exhalyzer® D first on the 3.1.6 software (mandated by study protocol) and next, directly on 3.3.1. Participants completing ≥2 acceptable runs on each software version within an 80-minute period are included here. To assess for differences, we used a one-way ANOVA with repeated measures and post hoc Tukey’s test. Second, to assess the accuracy of migrating legacy data from 3.1.6 to create normative values, we migrated a large legacy internal dataset (N=203), collected from HC adults and children (6-45 years); to which we then applied a linear regression model. Statistical analysis was performed using Prism Version 9.1.

Results:

MBW data directly collected in 3.3.1 was significantly different from both migrated and rerun data. In the HCs, mean (SD) LCI_2.5 for the 3.3.1 collected data was 6.98±0.61 compared to 6.48(0.44) (P < 0.0001) for migrated and 6.52(0.39) (P < 0.001) for rerun data. The CF group had mean LCI_2.5 of 8.95(2.92) on 3.3.1 compared to 8.16(2.29) (P < 0.01) for migrated and 8.23(2.61) (P < 0.01) rerun data. The mean absolute difference between directly collected and migrated or rerun data was 0.51(0.28) [7.7(3.8)%] and 0.48(0.33) [7.2(5.0)%] respectively for HC group and 0.79(0.66) [8.4(5.0)%] and 0.72(0.45) [8.5(4.6)%] respectively for the CF group. Seven of the 19 participants (37%; 4 CF) had a relative difference >10% for both migrated and rerun data compared to 3.3.1 collected data. Second, the upper limit of normal (ULN, 97.5%) from our legacy internal dataset on 3.1.6 was 8.30. When this dataset was migrated into 3.3.1 the new ULN was 7.41. When we applied the linear regression model (Y= 1.283*X-1.329) to this dataset, the ULN became 8.10. All HC collected on 3.3.1 data set were ≤8.10.

Conclusions:

In summary, neither the rerunning of legacy raw data nor migration of data is equivalent to collecting data directly with the corrected 3.3.1 software. Our findings have implications for the Global Lung Initiative MBW normative values project. We urge the investigators to consider using solely data collected in the corrected software to generate this normative range. Further, caution must be used in clinical practice when comparing corrected legacy data vs 3.3.1 collected data for clinical interpretation. We recommend that a new baseline is collected directly on 3.3.1 before clinical interpretation and decisions are determined when comparing consecutive MBW tests.

 

Acknowledgements:

References: [1] Wyler F, et al. doi:10.1152/japplphysiol.00338.2021