Introduction

Standard pulmonary function tests measure a range of indices that inform clinicians about different aspects of lung physiology [1]. Interpretation of pulmonary function tests requires an understanding of the range of values that can be expected in otherwise healthy individuals. The Global Lung Function Initiative (GLI) Network has established a protocol to collate previously collected healthy pulmonary function data and to derive a single standard for the interpretation of measurements against a healthy population [2]. All-age GLI reference equations are now available for spirometry, static lung volumes and gas transfer [3–5].

Multiple breath washout (MBW) is a lung function test based on efficiency of washout of an inert tracer gas from the lungs during tidal breathing. It assesses the unevenness of gas mixing within the lung, also known as ventilation inhomogeneity. This is expressed by a number of MBW variables, the most commonly reported of which is the lung clearance index (LCI). A substantial body of literature on LCI, predominantly derived over the past two decades, has consistently shown it to be more sensitive at detecting early obstructive lung disease than spirometry variables in children with cystic fibrosis (CF) [6–8]. MBW requires less active patient cooperation than spirometry and can be reliably measured in young children and infants, including unsedated infants [9–11]. LCI is now being routinely measured in some CF centres to assist clinical decision-making and has been identified in CF guidelines as a clinical outcome of interest [12]. Results obtained for LCI in other obstructive disease conditions in research studies suggest that it may also have future important applications in asthma [13], bronchiectasis [14, 15], primary ciliary dyskinesia [16, 17], early COPD [18], lung disease following haematopoietic stem cell transplantation [19], occupational lung disease [20, 21] and lung transplantation [22, 23].

The other index consistently generated from multiple breath washout is functional residual capacity (FRC), a measure of resting end-expiratory lung volume. This also provides a quality control check to assess the technical acceptability of the trials and overall test session [24]. Over time, the focus of MBW testing has evolved from FRC being the main index of interest, to one focused on LCI for research and clinical use. Though infrequently used on its own in clinical practice, FRC may provide complimentary information about gas trapping (due to peripheral airways disease) [25] and can also be used to assess response to therapies [26].

Technical standards for MBW equipment and how to perform MBW testing, along with frameworks for training operators, and applying quality control have been published [24, 27–30]. Commercially available devices have evolved significantly over the past decade and have incorporated several advances with improved accuracy and precision of measurements. Differing inert tracer gas options exist, with differing intrinsic properties, such as nitrogen (N2) and sulfur hexafluoride, which may impact on gas mixing and excretion [31, 32]. Importantly, the methodological approach chosen for indirect N2 concentration measurement has also been shown to directly influence results [33–35].

Differences in equipment, tracer gases and algorithms to calculate LCI have limited previous attempts to collate healthy reference equations for LCI and FRC derived from MBW [18, 36, 37]. Consequently, and in line with consensus recommendations [24], individual centres have developed their own normal ranges for LCI and FRC. This often restricts individual datasets to narrow age ranges, single devices and limited sample sizes [38–42]. Centre- and device-specific reference equations make it challenging to interpret measurements collected elsewhere, or to understand how much of the differences are due to sampling variability and technique compared to fundamental differences between sites, equipment and protocols. The lack of all-ages data also makes it difficult to accurately interpret how ventilation inhomogeneity changes with growth and ageing.

A single standard for the interpretation of MBW indices across all ages is an important resource for the use of MBW more widely in research and clinical practice. The aim of this study was to collate MBW data from healthy individuals and to derive GLI reference equations for LCI and FRC across all ages. In addition, we aimed to investigate how differences in equipment, tracer gas and equipment dead space volume impacted outcomes.

Methods

A European Respiratory Society (ERS)-approved task force was formed in May 2020 to develop all-age reference equations for LCI and FRC measured by MBW. The task force comprised researchers and healthcare professionals with expertise in developing international standards, lung physiology, pulmonary function testing and biostatistics.

Data sources

A pragmatic review of the literature was conducted to identify published studies that included measurement of LCI and FRC in healthy children and adults. Task force members searched for articles published between 2005 and 2023 in Embase, Ovid MEDLINE, Web of Science, PubMed and Scopus databases. Publications were then screened to remove duplicates, articles without MBW data, articles without healthy individuals and reviews/editorials without original data (details provided in the supplementary material). The authors of studies with ≥50 healthy participants were contacted and invited to share their data with the task force (this restriction was applied to ensure that data were collected only from centres with significant experience conducting LCI assessments). Invitations were also circulated through international and local respiratory societies (American Thoracic Society (ATS), Asia Pacific Society of Respirology, ERS, Latin American Thoracic Association, Pan African Thoracic Society, Thoracic Society of Australia and New Zealand) to solicit unpublished data.

Contributors were required to obtain approval from the local ethics committee to contribute data to the GLI Network. All data were pseudo-anonymised before submission and entered to a standard data template. Study data were collected and managed using Research Electronic Data Capture (REDCap) hosted at Dalhousie University (Halifax, NS, Canada) [43, 44].

Data were collated for LCI and FRC as well as demographic variables: sex, age, height and weight. The mean LCI and FRC from all acceptable trials were included in the analysis from test occasions with a minimum of two acceptable trials. If a centre contributed more than one data point per individual participant, one measurement only was randomly selected and included. Metadata included equipment type (and any modification of commercial equipment), tracer gas and equipment dead space volume. Due to the published impact of different software versions on MBW outcomes [33, 37], we only accepted data analysed using specific software versions (Eco Medics Exhalyzer D: Spiroware 3.3.1; ndd EasyOne Pro: Easy One Pro Lab/Easy One Connect V03.08.00.11 or higher until the date of data submission) [45, 46]. Sites were given detailed instructions and supported to reanalyse data from previously collected washout tests to maximise the inclusion of available data. Quality control assessments were performed by the individual sites according to international guidelines [24, 27, 28, 30, 47]. All submitted data were reviewed to identify missing data and implausible outliers; contributors were contacted directly to clarify discrepancies.

Healthy participants were defined as a person who was not a current smoker, had never smoked and was not obese. Participants aged <12 years with missing cigarette smoking status were assumed to be never-smokers. Incorporating any missing observations on smoking status in those aged >12 years, we created two definitions of healthy: a healthy participant was defined as a person with no reported currently smoking (not smoking now) and no reported smoking in the past (has not smoked >100 cigarettes). This definition includes individuals with a missing currently smoking or ever-smoked status (i.e. they were assumed to be healthy). A second definition “strictly healthy” was defined as an individual who was reported to be a never-smoker. This definition excluded participants who had missing data with respect to currently smoking or ever-smoked status (i.e. no assumptions were made on this subset). Body mass index (BMI) measurements were used to categorise subjects as overweight or obese. The Centers for Disease Control and Prevention growth charts were used for children aged 2–18 years, with overweight defined as BMI >85th percentile and obesity defined as BMI >95th percentile. For adults, we used measures of BMI >25 kg·m−2 to categorise an individual as overweight and those with a BMI >30 kg·m−2 to define obesity.

Statistical analysis

The generalised additive models of location shape and scale (GAMLSS) programme, previously used for other GLI projects, was used to develop the reference equations of LCI and FRC measured by multiple breath washout to 2.5% of starting inert tracer end-tidal concentration. Briefly, the GAMLSS technique allows the median value to be summarised as a function of multiple explanatory variables (e.g. height, age, sex), the spread of values around the median value to be constant or vary by a function of explanatory variables, and any departure from a normal distribution (skewness, kurtosis) to be transformed to normal using a Box–Cox transformation. Thus, the resulting model residuals are normally distributed. The BCCG (or Box–Cox Cole and Green family) distribution and a log transformation of the response variable was applied. All analyses were performed using the GAMLSS package in the statistical programme R (The R Project for Statistical Computing; www.r-project.org, version 4.3.1).

Age, sex, height, weight and total equipment dead space volume were treated as explanatory variables and were evaluated independently and in a multivariable model. Total equipment dead space constitutes both the pre-capillary dead space (volume between the airway opening and inert gas sampling point) and post-capillary dead space (volume between the gas sampling point and bias flow). These variables have previously been shown to influence LCI and/or FRC. We did not investigate ethnicity as a predictor as race and ethnicity are social constructs without a consistent definition globally, and recent statements endorsed by both the ATS and ERS have recommended against its continued use in reference equations [48, 49]. Variables that improved overall model fit were retained in the final model. The goodness of fit of each model was assessed by Schwarz Information Criteria (SBC), residual Q–Q plots and residual plots. Differences in LCI and FRC between equipment types and sites were investigated after the final models were derived. The upper and lower limits of normal for LCI and FRC from the final equations were compared to previously published equations. Sensitivity analyses were performed to determine the impact of the definitions of healthy and strictly healthy.

ResultsStudy population

Data from 3647 individuals were submitted by 23 studies from nine countries (figure 1). Since infant MBW is particularly challenging, and it became clear early in the analysis that approaches to infant MBW varied widely (e.g. positioning of infant, technological approaches, sedated versus unsedated), infants (defined as those aged <2 years) were excluded from analysis. After exclusions of ineligible studies, data from 1579 healthy individuals aged between 2 and 81 years from 17 studies and eight countries were available (table 1, figure 2). A large number of observations (n=474, 30%) were from children aged between 5 and 10 years (figure 2a).

FIGURE 1

Flow diagram of exclusions used to reach the final datasets.

TABLE 1

Summary of healthy participants by site

SiteParticipants nAge yearsFemaleOverweight#LCIFRC LCountryEquipmentTracer gasWas the equipment modified?Breathing pattern51405–5950.712.16.1 (5.8–6.4)1.9 (1.4–2.7)United KingdomInnocor closed circuitSF6 up to 0.4%YesTidal breathing10225–1650.04.66.4 (6.3–6.6)1.7 (1.1–2.2)United KingdomInnocor open circuitSF6 up to 0.4%YesTidal breathing12272–1155.60.06.6 (6.2–6.9)0.6 (0.5–0.8)United KingdomMass-spectrometerSF6 4%NATidal breathing137320–7750.732.95.9 (5.5–6.3)3.4 (2.8–4.1)United StatesEco Medics Exhalyzer DN2 (washout 100% oxygen)NoTidal breathing15753–6752.010.76.5 (6.2–6.8)2.0 (1.2–3.2)GermanyEco Medics Exhalyzer DN2 (washout 100% oxygen)NoTidal breathing197319–6560.332.96.3 (6.0–6.7)2.7 (2.2–3.2)AustraliaEco Medics Exhalyzer DN2 (washout 100% oxygen)NoFixed breathing protocol20545–5944.45.66.2 (5.9–6.5)2.7 (1.7–3.3)BelgiumEco Medics Exhalyzer DN2 (washout 100% oxygen)NoTidal breathing20415–3261.00.06.3 (5.9–6.7)1.7 (1.1–2.3)BelgiumNdd Easy One ProN2 (washout 100% oxygen)NoTidal breathing242755–2049.51.56.2 (5.9–6.5)1.0 (0.8–2.4)SwitzerlandEco Medics Exhalyzer DN2 (washout 100% oxygen)NoTidal breathing261533–6659.513.76.3 (6.0–6.6)2.0 (1.0–2.7)FranceEco Medics Exhalyzer DN2 (washout 100% oxygen)NoTidal breathing3323820–8150.028.66.4 (6.0–6.8)3.2 (2.7–3.8)BelgiumN2 analyzerN2 (washout 100% oxygen)NoFixed breathing protocol343721–7767.635.16.3 (6.0–7.2)3.2 (2.6–3.5)United KingdomEco Medics Exhalyzer DN2 (washout 100% oxygen)NoTidal breathing36446–4450.011.46.2 (6.0–6.6)1.6 (1.2–2.8)United KingdomInnocor open circuitSF6 up to 0.4%YesTidal breathing37956–5971.620.06.4 (6.2–6.7)2.8 (2.5–3)AustraliaEco Medics Exhalyzer DN2 (washout 100% oxygen)YesTidal breathing42345–6352.917.66.4 (6.0–6.9)2.0 (1.5–2.7)United KingdomInnocor open circuitSF6 up to 0.4%NoTidal breathing47415–7465.914.66.4 (6.2–6.9)3.0 (2.4–3.5)GermanyEco Medics Exhalyzer DN2 (washout 100% oxygen)NoTidal breathing52623–1158.10.06.4 (6.1–6.7)1.0 (0.8–1.2)AustraliaEco Medics Exhalyzer DN2 (washout 100% oxygen)NoTidal breathing59956–747.40.06.8 (6.5–7.2)0.6 (0.6–0.7)South AfricaEco Medics Exhalyzer DN2 (washout 100% oxygen)NoTidal breathing

FIGURE 2

a) Age distribution of individuals considered to be healthy; b) histogram showing lung clearance index (LCI) measurements of healthy individuals; c) histogram showing functional residual capacity (FRC) measurements of healthy individuals; d) scatter plot of LCI against age stratified by sex of healthy individuals; e) scatter plot of FRC against age stratified by sex of healthy individuals.

Reference equations

There was a nonlinear relationship between predicted LCI and age, and age also contributed to the between-participant variability (figure 3). In other words, LCI was on average higher in younger children and older adults, but relatively stable during childhood and early adulthood. Furthermore, there was more variability in LCI in younger children and older adults, resulting in wider limits of normal. FRC reference equations included age and height as predictors, as well as sex-specific equations. The coefficients for each prediction equation are presented in table 2. Look-up tables and a worked example are available in the supplementary material.

FIGURE 3

a) Predicted lung clearance index (LCI); b) graph in a) expanded to show only ages from 2 to 15 years; c) predicted functional residual capacity (FRC). Data are presented as the predicted values for age and 95% confidence limits. Predicted equations are overlaid on observed values.

TABLE 2

Coefficients and model fit for the Global Lung Function Initiative multiple breath washout equations

LCIM (median)S (variability around the median)L (skewness)FRC maleM (median)S (variability around the median)L (skewness)FRC femaleM (median)S (variability around the median)L (skewness)Limits of normal and comparison with existing reference values

The upper limit of normal for LCI (as both +1.64 and +1.96 z-scores) using the derived equations and previously published equations is presented in table 3. The upper and lower limits of normal for FRC (+1.64 and +1.96 z-scores) are presented in supplementary tables S1 and S2. The GLI equations were generally similar to previously published equations for FRC and LCI. The upper limits of normal for LCI from the derived GLI equations were higher in children aged <10 years compared to existing static reference values. The GLI upper limits of normal for FRC were lower in females, and higher in adult males compared to existing reference values. In the final models there remained small residual differences between sites and equipment types (figure 4).

TABLE 3

Upper limit of normal for lung clearance index (LCI) using the derived equations and previously published equations

Age yearsGLI MBWHorsley 2020 (Innocor closed circuit) ULN (1.64 z-scores) [38]Verbanck 2016 (N2 analyser) ULN (1.64 z-scores) [42]Kentgens 2022 (Exhalyzer D) ULN (1.96 z-scores) [41]PredictedULN (1.64 z-scores)ULN (1.96 z-scores)PredictedULN (1.64 z-scores)PredictedULN (1.64 z-scores)PredictedULN (1.96 z-scores)36.57.78.056.47.57.86.16.876.47.47.66.16.86.37.1106.37.27.36.16.86.37.1156.26.97.16.16.86.37.1206.16.97.06.1