Abstract
When applying GLI reference values to a local patient population, discrepancies may arise due to population or equipment characteristics. This is illustrated with a specific example and request to GLI for guidance. https://bit.ly/3jmATb9
To the Editor:
We have read with interest the official European Respiratory Society technical standards for Global Lung Function Initiative (GLI) reference lung volumes [1], and we welcome the user-friendly online computation module that facilitates translation of spirometry, transfer factor and lung volume variables into % predicted and z-scores [2]. The latter allows lung function researchers to stay up-to-date with the latest normative values of variables that are being measured in the routine lung function lab, while awaiting upgrades of their laboratory equipment to the latest reference values for carbon monoxide transfer factor (25 September 2020 update) or to the newly published lung volumes.
Now that GLI reference values are available for all major lung function measures, at least in Caucasians of European descent, we wonder how individual laboratories can deal with systematic differences that may arise, for instance for patients of different ethnicity or when using a specific lung function device. When we were previously confronted with a device-specific difference for which the manufacturer could not offer an explanation, we proposed a flexible z-score based criterion method, involving measurement of a limited number of normal subjects on the two devices, to align our local reference equations [3]. While it offered a solution to align the predicted values of the two devices, using either offsets or rescaling, it did not address variability that may also have been device-dependent. Possibly, a more sophisticated statistical approach could improve the method we proposed, to avoid individual laboratories having to measure a large sample of normal subjects [4]. This is what we hope the GLI team can also deliver, such that the applicability of GLI can be broadened. Specifically, if it were possible to measure a small number of local normal subjects on a local device on both sides of the age range of interest, would this not suffice to then align these data with the general trend lines revealed by GLI over that age range?
To illustrate the problem we are faced with when readily applying GLI, we computed % predicted and z-scores, using two previously published datasets obtained from normal subjects evenly distributed across the age range 20 to 80 years [5, 6]. In these studies, spirometry, transfer factor and lung volumes had been acquired to obtain a local normal database with the lung function equipment at the time of study. Equipment was from the same manufacturer (Vyaire, Mettawa, IL, USA): one device, currently in use, is a Masterscreen (MasterScreen PFT, SentrySuite 2.19) [6] and the previous device was Vmax (Vmax Encore 20c,22,22d) [5]. The results can be viewed in table 1, where the most consistent observation is that z-scores generally show an average value close to 0 with a standard deviation of 1 or less. The most obvious deviation from GLI can be seen for residual volume (RV)/total lung capacity (TLC), where one equipment obtains average z-scores close to zero whereas the other was in excess of 0.5, which is considered the threshold for a meaningful difference [1]. Unsurprisingly, the dataset showing z-scores close to zero had been obtained with the same device as 51% of the GLI lung volume data set [1]. The problem is that by applying GLI to our current device, normal subjects appear to have some degree of hyperinflation, feeding the concern that patients would be more readily crossing the upper limit of normal for RV/TLC.
With this very practical example of how individual laboratories can be confronted with challenges when adopting GLI, we believe it is now the time for the GLI experts to offer guidance in what to do when confronted with obvious equipment- or population-specific differences locally. Ideally, this could constitute an add-on feature on the current online computation module [2].
At its core, the Global Lung Function Initiative is a truly wonderful project in that all countries worldwide could contribute lung function data, and that data of acceptable quality would be processed in a consistent data analysis framework, with a sound statistical basis. At the same time, it also encouraged lung function specialists and clinicians alike to embrace z-scores. However, the success of GLI depends on data availability and quality, which has been a limiting factor when progressing from relatively simple spirometry [7] to more complex transfer factor [8] and lung volume measurement [1]. This explains why the normative values for the latter two are only available for Caucasians. With this correspondence we hope to encourage the GLI team to use all the data that they do have, to propose a sensible way to deal with equipment or population differences, while still enjoying the benefit of GLI equations and the general trends they reveal. Otherwise, other research laboratories will be faced with similar problems to ours, where we needed to undertake the extra steps described above, lest a large proportion of patients gets unduly diagnosed with hyperinflation, because of elevated z-scores for RV and RV/TLC in normal subjects.
Shareable PDF
Supplementary Material
This one-page PDF can be shared freely online.
Shareable PDF ERJ-01601-2021.Shareable
Footnotes
Author contributions: S. Verbanck, E. Vanderhelst and S. Hanon co-wrote the manuscript.
Conflict of interest: All authors have no conflict of interest to disclose.
Support statement: This project was supported by the Fund for Scientific Research-Flanders (FWO-Vlaanderen, Belgium). Funding information for this article has been deposited with the Crossref Funder Registry.
- Received June 6, 2021.
- Accepted June 22, 2021.
- Copyright ©The authors 2021. For reproduction rights and permissions contact permissions{at}ersnet.org