Influence of ventilation inhomogeneity on diffusing capacity of carbon monoxide in smokers without COPD

To the Editor:

Early detection of subclinical lung function impairment may enable a window of opportunity to slow down the progression of developing COPD [1, 2]. Single-breath carbon monoxide uptake in the lungs (D_LCO) can be used as a screening test for mild lung function impairment in smokers [3]. Yet despite being readily used in common practice, the physiology is complex and depends on gas ventilation in the airways, diffusion through the alveolar membrane and the volume of haemoglobin in the capillaries supplying ventilated alveoli [4]. As a result, mild changes in the peripheral airways often remain undetected and subsequent false normal D_LCO values limit the clinical utility of the test [5]. Increased ventilation inhomogeneity (VI), arising from uneven convective and diffusive gas transport, occurs early and may influence the D_LCO measurement method in smokers. The nitrogen multiple-breath washout (N₂MBW) method can reliably measure VI and is known to sensitively detect small airways disease in smoking adults with well-preserved forced expiratory volume in 1 s (FEV₁). Previous studies suggest that D_LCO correlates with VI; however, this association may have been mediated by airflow limitation from obstructed airways [6, 7]. Thus, until now the association of lung clearance index (LCI) with D_LCO remains unclear. The objective of this study was to investigate the influence of VI on D_LCO in vivo without inherent or induced airflow limitation. We hypothesised that D_LCO is associated with VI in smokers with preserved spirometry.

We performed a cross-sectional analysis in adult smokers enrolled from two prospective studies: a randomised controlled trial for tobacco cessation (ESTXENDS, www.clinicaltrials.gov identifier NCT03938298) and the Swiss Idiopathic Interstitial Pneumonia cohort (Swiss-IIP) [8]. For both cohorts, participants were enrolled consecutively in the pulmonary outpatient clinic and by advertisement between October 27, 2016 and November 30, 2019. Inclusion criteria were: age ≥18 years and self-reported smoking of ≥5 cigarettes per day for at least 12 months. Individuals with chronic lung disease, any inhaler medication or abnormal spirometry defined as FEV₁/forced vital capacity (FVC) <0.70 were excluded [9]. The study was approved by the local ethics committee (KEK BE 246/15, Basec PB 2016-01524; KEK BE 2017-02332) and written informed consent was obtained from all participants. The study setting was a pulmonary outpatient clinic, University Hospital, Bern, Switzerland.

Lung function testing was performed in accordance with current guidelines in the following order: N₂MBW (Exhalyzer D, Eco Medics AG, Duernten, Switzerland, Spiroware 3.1) during tidal breathing, D_LCO and spirometry (Jaeger MasterScreen™; CareFusion, Hochberg, Germany) [10, 11]. N₂MBW indices included LCI, a marker of global VI, S_cond, a marker of convection-dependent VI, and S_acin, a marker of diffusion-convection-dependent VI [12]. Breath-by-breath quality control was applied. Mean values of at least two acceptable MBW trials were reported. Additionally, demographics (age, sex, body mass index (BMI)) and smoking exposure in pack-years (PY) (packs of cigarettes per day×years of smoking) were assessed. To determine abnormal lung function, previously published upper and lower limits of normal (ULN) were applied (mean+1.96×sd) [12, 13].

Multivariable regression modelling was performed using D_LCO (%predicted) as the primary outcome. We included predictor and confounder variables selected a priori (LCI, S_cond, S_acin, FEV₁, sex, age, BMI, PY). Variables with p>0.2 were excluded stepwise in likelihood-ratio tests. Model coefficients and their precision were reported. Validity of regression assumptions were evaluated. For sensitivity testing, outliers identified using a leverage-versus-residual-squared plot were excluded and the analysis was repeated using only individuals with BMI≤35 kg·m⁻² to assess possible dependence of LCI and S_acin on BMI. A p-value <0.05 was considered statistically significant and analyses were performed using Stata 14.2 (StataCorp LP, College Station, TX, USA).

In total, 65 smokers were assessed for eligibility (n=36 ESTXENDS, n=29 Swiss-IIP). Reasons for exclusion were airflow limitation (FEV₁/FVC<0.70; n=6), failed quality control (N₂MBW, n=2; spirometry, n=8), and incomplete assessments (n=6). We analysed data from 42 individuals (45.2% females). Mean (sd) age and BMI were 39.3 (12.6) years and 26.4 (5.2) kg·m⁻². Median (interquartile range) smoke exposure was 18.5 (1.5–58.5) PY. FEV₁ and D_LCO were 96.5 (77.0–114.0) and 91.0 (60.0–128.0) % predicted. LCI, S_cond and S_acin were 7.6 (6.2–11.3), 0.025 (0.002–0.047) and 0.081 (0.046–0.256) units. In 14 (33.3%) individuals, LCI was above the ULN (8.25), whereas D_LCO was below 80% predicted in four (9.5%) (figure 1). Multivariable regression modelling showed that LCI and S_acin partially predicted D_LCO: 12.7% (adjusted regression R²) of the variance in D_LCO was explained by LCI and S_acin, p=0.010. Predicted D_LCO=135.5–(6.1×LCI)+(63.4×S_acin). Per one unit increase in LCI, D_LCO decreased by −6.1% predicted. Sensitivity testing excluding outliers (predicted D_LCO=135.1–(6.1×LCI)+(63.7×S_acin); adjusted R²: 0.142, p=0.024) confirmed the primary analysis.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

Association of carbon monoxide diffusion capacity and ventilation inhomogeneity. Diffusing capacity of carbon monoxide (D_LCO) in % predicted is plotted versus a) lung clearance index (LCI) and b) S_acin normalised for tidal volume (V_T) as recommended (S_acin·V_T). Dashed lines display upper and lower limits of normal. Individuals are displayed as circles with a colour scale indicating smoking exposure (tertiles): open circles, 1–10 pack-years; light filled circles, 11–29 pack-years; dark filled circles, 30–60 pack-years.

In an additional sensitivity analysis in individuals with BMI ≤35 kg·m⁻² we found no influence of BMI on the association between LCI, S_acin, and D_LCO (data not shown).

In this study, results revealed that D_LCO may be influenced by VI in smokers with normal spirometry. LCI and S_acin, quantifying global and diffusion-convection-dependent VI, partly explained the variance in D_LCO, and furthermore were inversely related to D_LCO. These findings suggest that LCI and S_acin may be refined biomarkers in this population to quantify small airway dysfunction if airflow limitation is absent.

Given the relatively low R² value, the relationship of LCI and S_acin with D_LCO appears to be complex. Ventilation inhomogeneity may lead to CO maldistribution, affecting the D_LCO estimate. Several technical and physiological considerations should therefore be taken into account. More specifically, whilst both D_LCO and N₂MBW methods capture similar physiological aspects, they differ in measurement principles. CO and N₂ have almost identical molar masses (28.01 g·mol⁻¹) and susceptibility to VI should be comparable. D_LCO requires a maximal inspiration effort, whereas MBW is performed during relaxed tidal breathing. CO rapidly diffuses through the alveolar membrane during D_LCO measurement. Hardly any of the lungs’ N₂ fraction passes the membrane during N₂MBW [11]. To isolate possible artefacts from VI on D_LCO, an airway model would be required.

Our findings are supported by previous studies suggesting that small airways dysfunction in current ex-smokers with COPD [14]. Importantly, LCI was more sensitive than D_LCO in capturing lung function abnormalities. Age and smoking history are known factors of lung function decline; however in our study, the association of VI with D_LCO was independent of BMI, age and smoking exposure. Diffusion-convection-dependent VI, global VI and D_LCO could be influenced by a common smoking-induced structural airway pathology. D_LCO measurement is considered sensitive to emphysema, which usually requires computed tomography scans for a definite diagnosis [5]. We hypothesise that VI may positively confound the association between D_LCO and structural airway pathology. In addition, D_LCO can also be confounded by other smoking-related changes, such as pulmonary vascular changes or increased CO-haemoglobin [15]. The utility of D_LCO to screen for early abnormalities in the lung function may be suboptimal and should be prospectively compared to MBW.

This study comes with a number of limitations which should be mentioned, such as the cross-sectional design, normative reference equations derived from diverse populations, and the lack of lung imaging to assess specificity of lung function abnormalities. Further evaluation of MBW breathing protocols requires additional studies.

In conclusion, our study suggests that the D_LCO measurement is influenced by VI. Ventilation inhomogeneity assessed by MBW may therefore become a refined biomarker in smokers with preserved spirometry to evaluate pre-COPD.