CORRESPONDING AUTHOR
Professor Paul D Robinson
The Children’s Hospital at Westmead The Sydney Children’s Hospital Network Respiratory Medicine Department Cnr Hawkesbury Road and Hainsworth Street Locked Bag 4001, Westmead 2145, NSW Australia
t: (02) 9845 3395/97 | f: (02) 9845 3396 | e: dr.pdrobinson@gmail.com
To the Editor,
The recent identification of a sensor crosstalk (XT) error in the Exhalyser D MBW device1 led to the release of new software (3.3.1). Understanding impacts of transition between MBWN2 software updates, and how to best handle historical data, is critical2. Should users simply migrate legacy data, by uploading into the new software version to enable XT correction (XTC), or perform a more time-consuming reload of raw data A‐files to apply all advances between that and historical software versions (3.1.6): XTC, O2 drift correction and novel gas signal flow synchronisation methodology (dynamic delay correction, DDC)? What are the differences and which approach most accurately reflects MBWN2 Lung Clearance Index (LCI) values collected within 3.3.1?
This issue has been investigated to varying degrees across three studies published in 2022. Two studies have drawn differing conclusions comparing agreement between migration and reloading. Oestreich et al3 (n=44, healthy and Cystic Fibrosis, CF) reported a mean (95% CI) LCI difference of −0.16 (−0.27 to −0.04) turnovers (TO) or −1.3%, with an observed magnitude-dependent bias, which they felt supported a recommendation to reload data (and not migrate). Jensen et al.4, in a reply to that letter, “respectfully disagreed”. In their independent analysis of a separate, similarly sized cohort, in which they attempted to isolate the impact of XTC on 3.1.6 data, mean (95% CI) difference between migration and reload was −0.10 (−0.26, 0.06), or -1%, was not statistically significant, without a magnitude-dependent bias when correctly expressed as relative difference.
In the next study to be published, by Short et al, this discussion was extended by introducing direct comparison to 3.3.1 collected data. Subjects (n=19, healthy and CF) performed ≥2 technically acceptable trials on both 3.1.6 and 3.3.1 in a fixed order (mandated by study protocol). They confirmed the non-significant difference between migrated and reloaded data reported by Jensen, but also showed a large difference between collected 3.3.1 data and 3.1.6-collected data ‘corrected’ via either migration or reloading into 3.3.1: mean absolute difference 0.5-0.8, relative difference 7.2-8.5%. The authors felt these differences raised concerns about ‘corrected’ 3.1.6 data being used by the current GLI normative values project to derive reference equations applicable for 3.3.1, and proposed that, ideally, only data collected in 3.3.1 should be included. This conclusion would have significant implications on the available data to derive these equations.
Here we present results from an independent Australian dataset that the 3 CORCs have scrutinised and agree offers methodological improvements to help maximise the utility of legacy data. In this study, performed under local ethics approval (2021/ETH11854), recruited participants performed ≥2 technically acceptable trials in both 3.1.6 and 3.3.1, in a randomised order, on a single Exhalyser D device at a single visit. 3.1.6 data were both migrated and reloaded in 3.3.1 to generate ‘corrected’ LCI values to compare to data collected directly on version 3.3.1. Aspects of processing within 3.3.1 were identical for reloaded and 3.3.1 collected data, as done by Short et al, but with the addition of consistent O2 and CO2 sensor offset values. The latter was not performed in the study by Short et al where instead, software and equipment specific offsets were used to produce effective synchronisation (assessed visually) during the washout phase.