Changes in Airway Dimensions on Computed Tomography Scans of Children with Cystic Fibrosis

Chronic bacterial infection and an excessive neutrophil-mediated inflammatory response are important characteristics of pulmonary disease in cystic fibrosis (CF) (1). These processes lead to structural damage of the airways, including wall thickening and lumen dilatation (2, 3). At birth, airway dimensions are believed to be normal (3), but structural abnormalities develop early in life and are progressive (4,5). Quantitative measurement of airway dimensions could be a sensitive and relevant endpoint for clinical management and for therapeutic interventions in CF-related lung disease. Airway abnormalities can be assessed in a reproducible fashion using computed tomography (CT) scoring systems (6-9). Previous studies using these scoring systems have shown that bronchiectasis together with mucus plugging are the most sensitive components of CF lung disease (4, 10). Hence, CT scoring systems have been shown to be more sensitive for the detection of airway abnormalities in patients with CF between 5 and 28 years old when compared with conventional pulmonary function tests (PFTs) (4, 10-13). Similarly, plain chest radiographs were found to be more sensitive than PFTs for patients aged between 0.5 months and 43 years (6, 14-18). Furthermore, airway wall thickening and mucus plugging, as can be observed on CT scans, was found to be reversible with antibiotic therapy in two studies of patients with CF with ages ranging from 19 to 43 and 9 to 33 years (19, 20). In diseases like asthma and chronic obstructive pulmonary disease, airway wall thickness (AWT) is related to airway inflammation (21).

A disadvantage of present scoring systems is that the total scores are a composite of various structural abnormalities, including the following: bronchiectasis, airway wall thickening, mucus plugging, mosiac perfusion, atelectasis, consolidation, bulla and cysts, acinar nodules, alveolar consolidation, emphysema, air trapping, overinflation, septal thickening, ground glass opacities, and sacculations or abcesses. The composite CT score makes it difficult to discern the most important structural change in individual patients. In addition, the sensitivity of scoring systems to detect small changes in luminal dilatation and AWT is unknown. Quantitative measurements of airway dimensions have been developed and applied in chronic obstructive lung disease (22,23). Therefore, a quantitative measurement of AWT and lumen dilatation could be an important addition to the time-consuming and subjective scoring systems. Although semiautomated airway measurements are still relatively time consuming, software is in development that will allow airways to be measured substantially faster in the near future.

We hypothesized that children with CF have thicker airway walls and more dilated lumens than control subjects and that airway wall thickening and bronchial dilatation would be progressive over 2 years in CF. To test this hypothesis, airway wall and lumen dimensions and dimensions of the accompanying pulmonary artery were measured on CT using an image analysis system that can measure airways with lumen diameter of 1 mm and above (22, 23). Some of the results of these studies have been previously reported in the form of abstract (24).

METHODS

Study Population

CF. Children with CF who had two biennial routine CT scans (CT^sub cf1^ and CT^sub cf2^), during a clinically stable period, in combination with two PFTs (PFT^sub cf1^ and PF^sub cf2^) were included in this study.

Control. Control subjects were patients without a diagnosis of heart or chronic lung disease who had a thin-slice thoracic CT scan (CT^sub controls^) performed for a variety of clinical indications (Table 1). The lungs of the control subjects were normal as assessed by the radiologists. Additional details of the study population are provided in an online supplement. The ethical review boards of Erasmus MC/Sophia (Rotterdam, the Netherlands) and Columbus Children's Hospital (Columbus, OH) approved this study.

CT Scans

CF. CT scans were acquired using 1-mm-thick images and 10-mm intervals from lung apex to base during full-suspended inspiration. CT images were archived and printed on film blinded for patient name, age, and date of the scan.

Control. Volumetric CT scans were obtained from lung apex to base during suspended full inspiration and reconstructed at 1.25-mmthick slices.

CT Scoring

CF. All printed scans were scored using scoring systems published by Bhalla and coworkers (6), Brody and coworkers (7), Helbich and coworkers (8), and Santamaria and coworkers (9) in random order by a single experienced observer.

Inspiration during CT Scanning

Lung volume during CT scanning was estimated using a previously described method (25, 26) and expressed as a percentage of predicted total lung capacity (TLC) (27).

Quantitative Analysis of Airways and Arteries

Airway-artery pairs were measured using a previously described method (22, 23). The following dimensions were measured: airway wall area (WA), airway lumen area (LA), airway lumen perimeter (Pi), and arterial area (AA; see Figure 1). AWT and WA/AA were used as markers of airway wall thickening and LA/A A as a marker of bronchial dilatation.

Two observers measured 86 randomly selected airway-artery pairs, and one observer remeasured the same airway-artery pairs after 3 months to determine inter- and intraobserver variability.

Two observers matched cross-sectioned airway-artery pairs on CT^sub cf1^ and CT^sub cf2^. The matched airway-artery pairs in subjects with CF and all visible cross-sectioned airway-artery pairs in control subjects were measured by one observer. Additional details of scanning protocols and CT analysis are provided in an online supplement.

PFTs

CF. The following lung function parameters were measured and expressed as percent predicted: FEV^sub 1^ (28), FVC (28), forced expiratory flow between 25 and 75% of expiratory VC (FEF^sub 25-75%^) (29), airway resistance (Raw) (27), residual volume (RV) (27), and TLC (27). Two ratios were calculated and expressed as a percentage: RV/TLC% and FEV^sub 1^/FVC%. Two patients were too young to perform spirometry at the time of the first scan. FEF^sub 25-75%^ was measured in 16 and Raw, RV, and TLC were measured in 18 of 23 children. Additional details of the PFT protocol are provided in an online supplement.

Statistical Analysis

Full statistical description is given in an online supplement. A p value of less than 0.05 was considered significant, and data are presented as mean

Comparison with Published Data

We employed the methods described in this manuscript on infant data which has been previously published by Long et al. to enable comparison between children and infants (5).

RESULTS

Study Population

The age of the 23 children with CF (12 male) at CT^sub cf1^ was 11.1 ± 3.7 (4.0-15.9) years, and at CT^sub cf2^, 12.9 ± 3.7 (6.2-17.9) years. The age of the 21 (11 male) control subjects was 11.6 ± 4.7 (3.6-17.2) years. The height of the 23 children with CF at CT^sub cf1^ was 147 ± 16 (122-174) cm, and at CT^sub cf2^, 155 ± 16 (130-180) cm. The height of the 21 control subjects was 144 ± 25 (99-181) cm. PFTs and CT scores at CT^sub cf1^ of children with CF are shown in Table 2.

Inspiration during CT

There was no significant difference between the lung volume measured on CT^sub cf1^ (61.7 ± 11.1% predicted TLC) and CT^sub cf2^ (63.6 ± 8.7% predicted TLC, p = 0.6) and CT^sub cf1^ and CT^sub controls^ (56.2 ± 21.4% predicted TLC, p = 0.2), but it did show a trend toward a higher lung volume in CT^sub cf2^ compared with CT^sub controls^ (p = 0.08). CT estimated gas volume was 2,429 ± 1,211 (506-5,533), 2,327 ± 787 (1,037-3,606), and 2,823 ± 1,038 (1,111-4,785) ml for CT^sub controls^, CT^sub cf1^, and CT^sub cf2^, respectively.

Quantitative Analysis of Airways and Arteries

In children with CF, 619 airway-artery pairs were matched. Mucus plugs were observed in 49 airways on CT^sub cf1^ and in 127 airways on CT^sub cf2^. Therefore, more airways were excluded for analysis on CT^sub cf2^ (20.5%) versus CT^sub cf1^ (7.9%), and 443 airway-artery pairs were measured on CT^sub cf1^ as well as on CT^sub cf2^. In the control subjects, 397 airway-artery pairs were measured, and no mucus plugs were observed.

Reproducibility of airway measurements. No systematic intraor interobserver differences in measurements were observed. Intraclass correlation coefficients for intra- and interobserver variability for WA and LA were greater than 0.99 and greater than 0.97 and greater than 0.99 and greater than 0.98, respectively. The mean interobserver difference was 0.78 ± 3.6% (-10 to 15%) for LA and 0.87 ± 3.2% (-10 to 11%) for WA. The mean intraobserver difference was -0.17 ± 2.2% (-5 to 6%) for LA and 1.02 ± 3.9% (-8 to 12%) for WA. Differences expressed as a percentage decreased with increasing airway size.

Airway lumen dilatation. Figure 2 shows the relation between airway lumen-artery dimensions and age for the children of the present study and the infants of Long and coworkers (5). The regression equations for these relations are given in Table 3. LogLA/AA showed no significant relationship with age for children with CF (CT^sub cf1^, CT^sub cf2^) and control subjects (p > 0.31). Intercepts for logLA/AA versus age were significantly (p < 0.0001) higher for children with CF (CT^sub cf1^, CT^sub cf2^) relative to control subjects. LA/AA was 1.92 times higher in children with CF compared with control subjects. There was no difference in intercept between CT^sub cf1^ and CT^sub cf2^.

The data of Long and coworkers (5) in infants showed that the mean airway lumen/artery ratio between infants with CF and control subjects increased from 0.92 at age 1 month to 1.60 at age 5 years (p = 0.02).

AWT. Figure 2 shows the relation between airway wall-artery dimensions and age for the children of the present study and the infants of Long and others (5). The regression equations for these relations are given in Table 3. LogWA/AA showed no significant relationship with age for the children with CF (CT^sub cf1^, CT^sub cf2^) and control subjects (p > 0.31). Intercepts for logWA/ AA versus age (p < 0.0001) were significantly higher for children with CF (CT^sub cf1^, CT^sub cf2^) relative to control subjects. WA/AA was 1.45 times higher in children with CF compared with control subjects. There was no difference in intercept between CT^sub cf1^ and CT^sub cf2^.

The data of Long and coworkers (5) in infants showed that the mean airway wall/artery ratio between infants with CF and control subjects increased from 1.05 at age 1 month to 1.46 at age 5 years (p = 0.02).

Figure 3 shows the relation between logAWT and logPi (perimeter). The regression equations of these correlations are given in Table 3. LogAWT (Figure 3) showed a significant correlation with logPi (p < 0.001). The slope was not significantly different for CT^sub cf1^ versus CT^sub cf2^ (p > 0.16). The intercept was significantly higher (p = 0.02) for CT^sub cf2^ relative to CT^sub cf1^. The difference in intercept was 2%, which corresponds to 0.03 mm.

Airway-artery dimensions, CT scores, lung function. LA/AA correlated (p < 0.01) and WA/AA did not correlate (p > 0.25) with CT scores (Bhalla, Helbich, Santamaria, Brody). LA/AA and WA/AA did not correlate significantly (p > 0.12) with PFTs (FEV^sub 1^, FVC, FEF^sub 25-75%^, FEV^sub 1^/FVC, Raw, RV, TLC, RV/TLC). For all comparisons, there were no differences between CT^sub cf1^ and CT^sub cf2^.

Changes in airway-artery dimensions, CT scores, CT component scores, and lung function. The change in AWT (AAWT) showed a significant negative correlation with the change in FEF^sub 25-75%^ (p = 0.002) in the 15 children who were able to perform spirometry and for whom we could calculate percent-predicted values for the FEF^sub 25-75%^. For each 0.01 mm increase in AWT, the FEF^sub 25-75%^ decreased by 0.45% predicted. Changes in the other airway parameters (?LA/AA and ?WA/AA) did not correlate significantly with changes in lung function tests (?FEV1, ?FVC, ?FEF^sub 25-75%^, ?FEV^sub 1^/FVC, ?Raw, ?RV, ?TLC, ?RV/TLC) or changes in CT scores (?Bhalla, ?Helbich, ?Santamaria, ?Brody). Changes in bronchiectasis (?bronchiectasis score) and airway wall thickening scores (?AWT score) did not correlate significantly with changes in FEV^sub 1^ (?FEV^sub 1^, p = 0.47) or changes in FEF^sub 25-75%^ (AFEF^sub 25-75%^, p = 0.11).

All CT scores worsened significantly over time (p < 0.02). Mean changes expressed as percentage of the maximal possible scores were 4, 4, 3, and 4% for Brody, Helbich, Santamaria, and Bhalla, respectively. Of the CT component scores, only the bronchiectasis score worsened significantly (p = 0.007), whereas all other component scores remained unchanged (p > 0.15; see also Table 2). Most PFTs (FEV^sub 1^, FVC, FEF^sub 25-75%^, FEV^sub 1^/FVC, Raw, RV, TLC) remained unchanged with time (p > 0.06). Only RV/TLC% improved significantly (p = 0.03) by -12% over the 2-year interval (see Table 2).

DISCUSSION

This longitudinal study aimed to compare airway wall and lumen dimensions on CT scans of children with CF with those of control subjects and to examine changes in these measures over a 2-year interval in CF. We hypothesized that children with CF would have thicker airway walls and more dilated lumens than control subjects and that airway wall thickening and airway dilatation would be progressive over 2 years in children with CF. To test the hypothesis, dimensions of airway wall and lumen and accompanying pulmonary artery were measured.

The most striking finding of our study was that the mean airway lumen-to-artery ratio was almost double in children with CF compared with control subjects. This increased ratio can be caused by dilatation of the airway and/or by a reduction in size of the accompanying artery. The ratio did not show an increase with age in control subjects or children with CF and remained unchanged over 2 years in CF. This result contrasts with a recent publication that showed a significant increase with age in the airway lumen-to-artery ratio of infants with CF (5). Therefore, we remeasured that cohort using our method. The reanalysis showed that the airway lumen-artery ratio in infants increased from 0.92 at age 1 month to 1.6 at age 5 years. This suggests progressive bronchial and/or arterial disease in the infants. Beyond age 5, there was a flattening of the curve in children for which several explanations are possible. First, bias could have been introduced by the exclusion of more mucus-plugged airways in the older children. Similarly, more mucus-plugged airways were excluded from the analysis on CT^sub cf2^ than on CT^sub cf1^. This, together with different methods of airway sampling because of differences in scanning protocols between the groups, could have influenced our results. Second, inflation techniques differed between the infants and the children, which may have influenced the results. A third explanation could be that the differences are caused by a center effect because it is known that important differences exist between centers and patient populations (30). Finally, it could be that there was a stabilization of the disease process leading to stable airway lumen dimensions after age 5. It might be that treatment is more effective in children of 6 years and older. Most drugs in CF are only used in children 6 years and older because they are able to perform lung function. For example, in our center, most children 6 years and older are treated with rhDNase, whereas few children younger than 6 years are treated with rhDNase. Our finding, however, appears at contrast with our observation that the scoring systems show a steady significant increase in bronchiectasis in the children with CF.

There are a number of possible explanations for the discrepancy between changes in bronchiectasis score and quantitative LA/AA ratio. First, it is known that peripheral airways are important in CF lung pathology (2, 13). The scoring systems include assessment of peripheral bronchiectasis even when a visible artery does not accompany the airways. Because our quantitative method does not include airways not accompanied by arteries, it is possible that it is insensitive to peripheral airway pathology/ peripheral bronchiectasis. Second, sample size could play a role because more airways could be evaluated in the qualitative CT assessment of bronchiectasis than in the quantitative analysis.

It is important to realize that, for the radiologic diagnosis of bronchiectasis in CT scoring systems, airway size is often compared with the adjacent artery. This comparison assumes that the arterial diameter is not changed in relation to the disease process. The increased airway/artery ratio observed in our study and in the study by Long and coworkers (5) might be in part the result of hypoxia-induced vasoconstriction (24, 31). If arterial size is changed in CF, it cannot be used as the reference structure to define bronchiectasis or airway wall thickening as is done both in the scoring systems and in quantified ratios. If the arterial size is reduced in CF, this would result in an overestimation of the degree of dilatation and airway wall thickening. Although we believe that this is unlikely to be a full explanation for our finding, it could contribute to the difference in airway/arterial ratio. In theory, comparing the mean absolute arterial size of all vessels in the infants and children with CF and control infants and children could test this assumption. However, because the vessels for analysis were selected based on the identification of airways and the identification of airways is dependent on them being big enough to be reliably measured on CT, this approach is flawed in this study. Future studies may want to compare absolute values for airway lumen and airway wall directly at well-defined, specific locations of the airway tree without use of the accompanying pulmonary artery.

A second important finding in our study was a 45% higher WA/AA ratio in the subjects with CF compared with the control subjects. This ratio did not increase with age and remained unchanged from CT^sub cf1^ to CT^sub cf2^. This could be because of relatively low inflation, exclusion of mucus-plugged airways, or differences in airway sampling between infants and children and control children and CF children. However, AWT as a function of airway size was higher for CT^sub cf2^ compared with CT^sub cf1^, reflecting a mean increase in AWT over the 2-year follow-up period. The reason why this change from CT^sub cf1^ to CT^sub cf2^ was not reflected in the WA/AA ratio might be because the WA/AA ratio is more variable than AWT since the ratio includes variation in the measurement of airways and arteries. Long and coworkers found an increase in the WA/AA ratio in infants with CF from 1.05 at 1 month to 1.46 at 5 years, which was not significant. Therefore, their data correspond well to our data, which showed a mean value of 1.45 at the age of 5 years.

The increase in AWT from CT^sub cf1^ to CT^sub cf2^, although significant, was only 0.03 mm. A 0.03 mm increase in AWT is modest as is its potential effect on airway resistance. Despite this, the change in AWT correlated significantly with the change in FEF^sub 25-75%^ predicted, a lung function test believed to be sensitive to abnormalities in the peripheral airways (32, 33). This could be because only modest changes in CT-measured AWT reflect modest changes in small airway caliber (34) that could have a striking effect on airflow. Alternatively, or in addition, the modest increase in AWT could be a surrogate for other pathologic abnormalities, including, for example, obliteration of small airways. It could also be that the 0.03 mm is an underestimation of the real difference between CT^sub cf1^ and CT^sub ct2^ because more mucusplugged airways were excluded on CF^sub CT2^.

The quantitative analysis of AWT was more sensitive to change than the semiquantitative scoring of AWT. The semiquantitative evaluation of AWT as done in the scoring systems is subjective and not very reproducible. The fact that, in this study, AWT in the scoring systems was not progressive, but quantitatively assessed AWT increased over 2 years, may suggest an added value of quantitative analysis of AWT to scoring in children with CF.

The PFT findings in a larger sample of this cohort have previously been described (4). The significant improvement in RV/ TLC seen in this study was also observed in the larger cohort. This increase could be from a learning effect because this cohort includes many young patients. In the larger sample, both bronchiectasis score and mucus-plugging score worsened significantly. There are a number of potential reasons for dissociation between changes in structure and function: PFTs are a measure of global lung function, they depend on patient cooperation, and they are expressed as percent predicted using equations derived from a large population sample. In contrast, CT can detect (small) localized areas of bronchiectasis and individual mucus plugs, which may, over the short term, have little influence on lung function (4).

There are several limitations of our study. First, for ethical reasons, we used control scans that were performed for a clinical indication. However, none of the clinical indications would be expected to cause changes in airway wall or vascular dimensions. In addition, the airway/arterial ratio in the control subjects was very comparable to the published normal values in adults (35-37). However, even if our control scans were not completely normal, any bias would likely result in an underestimation of true difference between patients with CF and control subjects.

A second limitation is that children with CF and control subjects were scanned using different CT scanners and protocols. However, the CT scanners were made by the same manufacturer and a high-enough beam current was used to provide good-quality images. Although differences in scanners and scanning technique might affect quantitative measurements of airway dimensions, such as AWT, it is unlikely that it will affect the ratio between dimensions. Therefore, we restricted our between-group analyses to comparisons of ratios rather than absolute airway dimensions versus age.

Differences in protocols also meant that airways were sampled differently in the various groups. Infants were scanned at only four levels, older children with CF were scanned at approximately 25 levels, and for older control subjects, full-lung scans were analyzed. Although we sampled a similar number of airways per CT in the older children with CF and control children, there were fewer airways per subject for infants with CF and control infants. Despite these sampling differences, there was no offset of the regression lines for age versus LA/AA and WA/AA between the infants and children for both control subjects and patients with CF.

Another important difference between this study and that of Long and coworkers (5) is that we did not control for lung volume during scanning. Lung volume is an important determinant of airway lumen (38) and arterial dimensions. The differences in the slopes of the relationships between age and LA/AA and WA/AA between our children and the infants of Long and coworkers could be explained by differences in inflation level between the studies. To overcome the potential for such differences, future studies should be done using CT scanning with spirometer gating (39, 40).

A third limitation is that we could have introduced a selection bias because a higher percentage of airways on CT^sub cf2^ were excluded from the analysis because of mucus plugging. The exclusion of these occluded (possibly more severely affected) airways might have led to an underestimation of the progression of AWT as a function of airway size and an underestimation of the changes in the LA/AA and WA/AA ratios of CT^sub cf2^ relative to CT^sub cf1^. In addition, it may be that fewer airways were excluded in the infants with CF compared with the children with CF, which might have led to an underestimation of the difference in the slopes of age versus WA/AA and LA/AA ratios in children with CF relative to infants with CF.

A fourth limitation is that some of the patients were too young to perform PFTs or were excluded from comparisons because their age was younger than the specified age for Wang and coworkers' (29) predicted equations for FEF^sub 25-75%^ results, and therefore some comparisons were made with a smaller number of subjects, which reduced the power of our statistical analysis.

To conclude, quantitative measurements of airways are highly reproducible. In children with CF, the ratio between airway LA and AA and between airway WA and AA is increased and the airway walls are thickened. AWT, but not the ratio between airway WA and airway LA and AA, increased over 2 years of follow-up. Although lung function did not change significantly over the 2-year follow-up period, there were significant changes in quantitative airway wall thickening and qualitative CT scores as estimates of airway structure. Some of our findings are potentially influenced by differences in airway sampling between the groups, by differences in lung inflation during the CT procedure, and by the possibility of changes in arterial size in patients with CF. Future quantitative studies should address these issues by measuring absolute values for airway wall, airway lumen, and artery at specified locations of the airway tree using volumetric CT scanning at a standardized well-inflated lung volume. Our data extend previous findings regarding CT imaging and quantitative assessment of CT findings in CF (5, 11, 41), and further indicate that CT scanning and quantitative CT measures may provide valuable information on the presence and progression of lung disease in CF.

Conflict of Interest Statement: P.A.d.J. does not have a financial relationship with any commercial entity that has an interest in the subject of this manuscript; Y.N. does not have a financial relationship with any commercial entity that has an interest in the subject of this manuscript; W.C.H. does not have a financial relationship with any commercial entity that has an interest in the subject of this manuscript; F.R.L. does not have a financial relationship with any commercial entity that has an interest in the subject of this manuscript; H.O.C. received U.S. $2,500 in 2002 and £1,500 in 2003 for serving on an advisory board for GlaxoSmithKline (CSK). In addition, he is the coinvestigator on two multicenter studies sponsored by GSK and has received travel expenses to attend meetings related to the project. A percentage of his salary between 2003 and 2006 (U.S. $15,000/year) derives from contract funds provided to a colleague (P.D.P.) by GSK for the development of validated methods to measure emphysema and airway disease using CT; P.D.P. has support from GSK to test and validate software for measuring emphysema and airway wall dimensions. The support is for Can $408,000 over 3 years, 2001-2004. He is applying to GSK for an additional ~ Can $200,000 to continue this work. He has received funding from GSK to partake in a multicenter study of the genetics of chronic obstructive pulmonary disease, ~ Can $1.1 million between 2000 and 2003. H.A.T. acted from 2002 to 2004 as consultant for Chiron Biopharmaceuticals RStD (U.S. $22,050/year). In 2002, he received U.S. $2,542 from Chiron as a member of an advisory board on aerosolized antibiotics and received within the last 3 years honoraria and travel expenses for lectures and workshops from Chiron, Hoffman-La-Roche, and Genentech, and holds no stock options. The BV Kindergeneeskunde of the Erasmus MC/Sophia Children's Hospital has in the last 3 years received research grants from Hoffman-La-Roche.

Acknowledgment: The authors thank Dr. Maarten Lequin for assistance in retrieving all CT scans from the CF cases of this study.

© 2005 American Thoracic Society Provided by ProQuest LLC. All Rights Reserved.