Low-dose CT is an appealing approach to quantify simultaneously pulmonary and airway
abnormalities. Our study shows that the use of low-dose CT combined with modern post
processing software tools provides reliable information on airway wall thickness and
the extent of emphysema in a heavy smoking male population. Although CT does not
provide dynamic measurements, airway wall thickening and emphysema explained
respectively 31.1% and 9.5% of the variance in
FEV1%predicted, the most commonly used variable of airflow
limitation. Changes in AWT of more than 0.1 mm reflecting lumen surface
changes over 8% measured at one air lumen level were observed between cases with
and without respiratory symptoms.
Our study confirmed that increased AWT is associated with lower
FEV1%predicted. This lower FEV1%predicted depends on
the airway size in which the measurements of AWT are being performed [12, 17] and on the characteristics of the study population [7, 20]. Our population consisted of rather healthy elderly males from a randomly
recruited Dutch heavy smoking population and still we were able to find significant
associations between thicker airway walls and more severe airflow limitation. In
contrast with the study of Nakano we found a significant negative association
between AWT and FEV1/FVC illustrating the sensitivity of our method .
The significant but weak negative association between airway wall thickness and
emphysema has also been reported in other studies [7, 13, 30] but was not found in the study by Nakano . Loss of elastic recoil may contribute to collapse of the airways
resulting in a more proximal localization of airways with 3.5 mm internal
lumen diameter. As these more proximal airways have thicker airway walls this
phenomenon contributes to the weak negative association. Another possible
explanation for this negative association may be that there are subjects with
predominantly airway wall thickening and others with predominantly emphysematous
changes. Particularly subjects with relatively more airway wall thickening are
responsible for the negative association and subjects with predominantly
emphysematous changes do hardly contribute. Apparently, in our population of
subjects with normal lung function and with mild airflow limitation, the bronchitic
phenotype is already present in the very early stages of smoking-induced lung
disease. Discrepancies between the study of Nakano  and our study may be due to the composition and size of the study
populations, respectively predominantly emphysema versus predominantly healthy
smokers with respiratory symptoms.
Importantly, we observed that the contribution of AWT3.5 to airflow
limitation was larger than the emphysema component. Moreover, AWT3.5 and
emphysema together only explained about 40% of the variance in
FEV1%predicted in this smoking male population. This unexpected low
contribution of AWT3.5 and emphysema to FEV1%predicted may
be that the CT images were obtained at full inspiration, while FEV1
reflects expiratory airflow limitation. One explanation for this observation is that
airflow limitation is not only due to reduced airway diameter at one level but
should be evaluated as an integral of all airways at all lumen diameters. This is
difficult to achieve and therefore we took the smallest measurable lumen diameter
that provides the largest contribution to airflow limitation. A more obvious
physiological explanation may be the presence of the heterogeneity in airway
ventilation interrupting the symmetry in parallel airways leading to large clusters
of poorly ventilated lung units .
In the univariate analysis, increased AWT3.5 was associated with
respiratory symptoms. However, AWT3.5 was not associated with the
presence of any respiratory symptom in the multivariate analysis after adjustment
for FEV1%predicted, emphysema and smoking behavior. This finding
corresponds with other studies  and is consistent with the idea that inflammation and airway remodeling,
associated with chronic bronchitis, is located in the more central airways . The study of Martinez et al. showed a positive association between
airway dimensions and questionnaires, the BODE index  and the St. George’s respiratory questionnaire  including questions about BMI, respiratory symptoms, exercise capacity
and lung function. Also Camiciottoli et al. found a positive association between
BODE and airway wall thickness . Our study also showed that including respiratory symptoms in the
multivariate model with AWT3.5 and emphysema has no impact on airflow
Lung volume depends on height, weight and sex and as a consequence each person has
different airway dimensions. Therefore, airway dimensions should be corrected for
lung volume. Actually, volume-corrected AWT is the best parameter to use. In this
study lung volume does not change the multivariate model because FEV1%predicted
is already corrected for lung volume by correcting for patient height.
It has been shown that the automated imaging program (MEVIS Airway Examiner) based on
a method by Weinheimer et al. performed much better than the often used
“full-width-at-half-maximum” method in a silicon tube phantom, regarding
the blurring effect of CT [25, 36]. Usually it is better to use sharper kernels for airway quantification.
However, it was shown in the study by Schmidt et al. that the MEVIS airway examiner
provides reproducible quantitative results across different reconstruction kernels
(B30f and B50f) and repeated acquisitions . Moreover, the “full-width-at-half-maximum” technique
systematically overestimates AWT, particularly in small airways . Because low-dose CT and the automated imaging program (MEVIS Airway
Examiner) had not been used previously in clinical studies, we firstly optimized our
post processing measurements in smaller cohorts of patients before applying it in
the population study. We demonstrated that the highest number of AWT measurements
could be performed on airways with an internal diameter of 3.5 mm,
irrespective of the presence of airflow limitation or emphysema. In addition we
demonstrated that differences in AWT3.5 are not explained by differences
in airway generations. Finally, we demonstrated that low-dose CT imaging provided
similar AWT results as high-dose CT imaging.
In a non-biased way we were able to evaluate 230 cumulatively assessed perimeter
fractions (APF) per CT, ranging from 27-641 APF. In contrast to the commonly used
pi10 method, in which a secondary derived variable from few, mostly 6 selected
airways is used to estimate the airway wall thickness [5, 6, 13–15], we obtained many direct airway wall measurements. To our opinion direct
measurements assessed over all lobes provide a better overall reflection of AWT than
a limited number of secondary AWT measurements.
A limitation of this study is, inherent to general population-based studies, that
only male smokers with mild emphysema and/or airflow limitation were included.
Strengths of our study is the non-biased way of analyzing a high number of airway
sections in all lobes that makes our approach more suitable for combined airway wall
thickness and emphysema measurements on one low-dose CT scan. Such approach allows
monitoring of intervention effects on both parameters. This is important when new
treatment modalities will become available for clinical testing.
In the future further developments may involve measurements of thickness of airway
walls along the full length of the bronchial tree at in- and expiration scans.
Possibly more airflow variability will be explained.
In conclusion, post processing standardization of large numbers of airway wall
measurements in all lung lobes is feasible, reliable and an useful method to assess
airway wall thickness. We have demonstrated that increased airway wall thickness
contributes more to airflow limitation than emphysema in a smoking male population
even after adjustment for smoking behavior.