INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Emphysema is morphologically defined as the permanent enlargement of airspace distal to the terminal bronchiole and the destruction of its wall with no obvious fibrosis (1). However, the examination of such changes during life has been extremely difficult until high-resolution computed tomography (HRCT) scanners capable of detecting subtle emphysematous changes were developed. Although various ventilatory tests to detect functional abnormalities caused by smoking leading to chronic obstructive pulmonary disease (COPD) have been the center of attention, they are of limited value for direct estimation of structural abnormalities at the lung peripheries (2). HRCT has been shown to be more sensitive than the classic chest X-ray in the detection of lung parenchymal abnormalities (10). HRCT has thus been actively applied for morphological assessment of the extent and severity of COPD, especially caused by emphysematous alterations. Unfortunately, however, the importance and sensitivity of HRCT for estimating structural abnormalities caused by COPD have been analyzed in only a cross-sectional manner (11). Longitudinal changes in FEV₁ in subjects suffering from a variety of diffuse lung diseases have been studied, with the result that annual changes in FEV₁ for nonsmoking healthy subjects, smoking subjects, or patients with COPD appear to be consistently settled (3, 4, 20). To the best of our knowledge, however, there has been little systematic study shedding light on longitudinal changes in CT parameters related to smoking-induced COPD. On the basis of these facts, we attempted to estimate annual changes in various HRCT parameters quantifying the lung tissue destruction obtained for nonsmoking healthy subjects, current smokers, and past smokers during a five-yr follow-up period. In comparison, longitudinal changes in various pulmonary function parameters were also examined for the same groups of subjects. Through this analysis, we examined the quantitative effect of continuing smoking on HRCT findings as well as pulmonary function in a prospective way. The effect of smoking cessation on the parameters of HRCT and pulmonary function was also assessed.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subject Selection

Protocols were approved by the institutional review board for human studies and informed consent was obtained from all subjects. We initially enrolled 269 subjects randomly selected from those voluntarily visiting the Keio University Hospital (Tokyo, Japan) over the period from 1991 through 1994, for a medical examination of lung cancer. A total of 263 subjects were certified to have no lung cancer and considered as possible candidates for the analysis. Among them, 83 subjects (55 males and 28 females) with the mean age of 59 ± 12 yr (mean ± SD) were finally selected, because they were completely examined for pulmonary function tests as well as HRCT on at least three occasions over a 5-yr period of observation. Among them, 36 subjects, who were free of pulmonary disorders and had never smoked, were assigned as nonsmoking controls (NS group). The remaining 47 subjects were current smokers (CS group, n = 35) or past smokers (PS group, n = 12) with significantly long smoking histories but no marked abnormalities in lung fields, such as giant bullae, bronchial asthma, bronchiectasis, pulmonary fibrosis, pulmonary vascular disease, or lung cancer, compromising pulmonary function. The subjects categorized as the current smoker group were not able to quit smoking during the 5-yr follow-up period, although they were always advised to stop smoking when visiting the hospital. There were no differences in age, height, and weight among the nonsmoking, current-smoking, and past-smoking groups (Table 1). Lifetime cigarette consumption did not differ between the current- and past-smoking groups (Table 1). The past smoker group had ceased smoking for at least 2 yr and the average period of smoking cessation was 5.0 ± 2.8 yr at the moment of entry to the study. These subjects were confirmed not to have resumed smoking on the occasion of their visit to the hospital. The subjects belonging to the past smoker group had given up smoking due chiefly to their own recognition of aggravation of productive cough and dyspnea on exertion. Therefore, the extent of airflow obstruction was worse in past smokers than in current smokers. Although some of the subjects in current- and past-smoking groups were continuously medicated with bronchodilators including ₂-agonists, anticholinergic agents, and/or xanthine derivatives, their prescribed medication and the dose were not altered during the observation period unless the disease conditions of the subjects were aggravated. Pulmonary function tests and HRCT were annually performed at nearly the same period of time. Transitional variations of all parameters observed for various tests of pulmonary function and HRCT were assessed by estimating the annual change in each parameter. The annual change was calculated on the basis of the linear regression equation constructed using longitudinal data of each parameter obtained for each subject.

Pulmonary Function Tests

Pulmonary function was tested within 1 wk before or after accomplishing the HRCT examination. To minimize measurement errors, pulmonary function tests were always performed in the morning between 9:00 and 11:00, using the same instruments that were handled by the same technician, who was well trained for all pulmonary function tests. As ventilatory indicators, we measured vital capacity (VC), forced vital capacity (FVC), and the expiratory flow-volume curve by using an electronic spirometer (MFR-8200; Nihon Koden, Tokyo, Japan). Respiratory impedance was examined by the oscillation method (MZR-4000; Nihon Koden). Lung volumes including functional residual capacity (FRC), residual volume (RV), and total lung capacity (TLC) were measured by helium dilution (Chestac-55V; Chest, Tokyo, Japan), while pulmonary diffusing capacity for carbon monoxide (DL_CO) was estimated by 10-s breath-holding (Chestac-55V; Chest). PO₂, PCO₂, and pH in arterial blood samples were examined with electrodes (model 1306; IL, Lexington, MA).

HRCT Scans

HRCT scans were conducted with the subject supine (ProSeed; GE Yokogawa Medical Systems, Tokyo, Japan), holding a breath for 5 s at deep inspiration or deep expiration, obtaining inspiratory and expiratory HRCT images for each subject. Five CT images, with sections 1 mm thick and 2 mm apart, were taken in the upper and lower lung fields at both breath-holding positions. To reduce the total quantity of X-ray exposure for the subjects, CT images in the middle lung field were, however, taken at deep inspiration only. Images of the upper lung field were obtained at the midportion of the intrathoracic trachea, while those of the middle lung field were taken at the inferior end of the carina. Images of the lower lung field were obtained at the portion 1 cm above the diaphragm. HRCT scanning was carried out with 120 kVp, 200 mA, 1-s scan time, 1-mm collimation, 35 reconstruction circle (field of view, FOV), and 512 × 512 matrix, thus yielding a pixel size of 0.7 mm. Scan data were reconstructed using a high-spatial frequency algorithm (11, 13, 16, 19, 25). Images thus obtained were photographed at levels of 500 Hounsfield units (HU) and window widths of 1,500 HU to visually assess lung parenchyma abnormalities.

Quantitative Parameters of HRCT Images

Morphological abnormalities at the lung peripheries were quantified in each CT section of the right lung by estimating three objective CT indices at deep inspiration and deep expiration in the upper and lower lung fields, but only at deep inspiration in the middle lung field: (1) mean lung density (MLD, HU), mean CT value of each section; (2) HIST (HU), CT value with the most frequent appearance in the density histogram constructed for every 10 HU; and (3) percent low attenuation area (%LAA) (%), relative area of low CT density. %LAA was estimated solely at deep inspiration in all lung fields. Excluding large vessels and airways visible in each lung section, the parameters described above were calculated by applying the attenuation mask program proposed by Müller and colleagues (11). Although several CT values defining LAA have been reported in the literature (11, 13, 14, 16, 17, 26, 27), we refrained from applying these values to the present study because the cutoff CT level discriminating the area with abnormally low density from that with normal structure might differ depending on the instrument used (16). Therefore, we preliminarily examined the 95% confidence limit of CT densities in the upper, middle, and lower lung fields of the right lung in nonsmoking healthy young volunteers with little visible abnormalities in their lungs (age, 27 ± 5.4 yr; n = 12). The nonsmoking subjects assigned as the nonsmoking group (NS group) were not included in this analysis, because most of them were more than 50 yr old (Table 1). Namely, we attempted to determine the cutoff CT value defining LAA, on which effects of smoking and aging were nearly removed. HRCT images were taken at deep inspiration. Excluding the regions with visible airways and vessels, five small spots with a diameter of 5 mm and 1 cm apart from each other (three spots at the outer region and two spots at the inner region) were selected in the upper, middle, and lower lung fields, respectively. CT values of these 15 small spots distributed throughout the right lung field were examined. By averaging these 15 values, the mean CT value of the right lung in each nonsmoking young volunteer was estimated. Finally, mean CT values of all volunteers were altogether averaged. Thus, we found the mean CT density of 863 HU and its standard deviation of 24.5 HU for the normal right lung field at deep inspiration, confirming that the lower 95% confidence limit of CT density in the right lung was 912 HU (i.e., mean 2 SD). We therefore defined the areas having CT values of less than 912 HU at deep inspiration as LAA. We did not analyze the LAA at expiration, because the pathological significance of expiratory LAA has not yet been conclusively decided (14, 17, 28, 29). Furthermore, we considered that the cutoff CT value for LAA determined at deep inspiration could not be applied for defining LAA at deep expiration (17, 19).

MLD indicates the CT density obtained by averaging the CT values of all the pixels in a given lung section. Thus, MLD is the approximate, but the most reliable, measurement of the averaged density in the lung periphery produced by substances organizing the acinar structure and also by invisible small vessels and airway walls, suggesting that MLD may act as the comprehensive indicator of acinar density throughout the lung peripheries including acini with normal structure, those with incomplete destruction, and those with serious damage. HIST may not be influenced by relatively large vessels or bronchi but it reflects the densities of peripheral lung regions with normal acini and those with incomplete destruction of the acinar structure but without serious destruction forming LAA, when HIST is located where the CT value is more positive than 912 HU. Meanwhile, HIST may reflect a part of LAA, when it is positioned where the CT value is more negative than 912 HU. Thus, overall acinar densities in regions having significant damage in the acinar structure may be analyzed from HIST, as well. Abnormalities in areas with seriously compromised acini may be examined from LAA data. An example of CT images obtained from a patient with COPD during breath holding at deep inspiration is shown in Figure 1, in which the histogram of CT density and LAA in the right lung are depicted.

View larger version (47K): [in this window] [in a new window]   Figure 1.   Inspiratory CT images in the upper lung field of the right lung obtained for a patient with mild emphysematous alterations. (Left panel ) Histogram of lung CT densities. (Right panel ) Areas with low CT densities below 912 HU (LAA) observed for the same patient. LAA is depicted by highlights.

Statistical Analysis

Values are expressed as means ± SD. One-factor ANOVA followed by multiple comparison analysis with the Schffe examination was used to compare data obtained from the three groups (30). If data did not show equal variance, however, we applied the Kruskal-Wallis test to assess significance. The difference in data at entry and those at 5 yr later was judged with the paired t test or the one-sample Wilcoxon test. The significance of annual changes (i.e., changes in values against follow-up time, which were determined with the linear regression) in various parameters of pulmonary function tests and CT images was estimated on the assumption that the difference in slopes conforms to the t distribution (30). Linear regression analysis was used to determine the correlation between CT indicators and age, cigarette consumption, or pulmonary function parameters. A p value of less than 0.05 was considered significant.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Pulmonary Function Parameters at Entry

The results of various pulmonary function parameters at the entry of this study are summarized as "Initial" in Table 2. There were no differences in FEV₁, FVC, FRC, ₂₅, Pa_O2, and Pa_CO2 among the three groups studied, while FEV₁% (defined as FEV₁/FVC) in the current- or past-smoking group was much lower than that in the nonsmoking group. Although RV/TLC and DL_CO/VA in past smokers differed from those in nonsmokers, these values were not different between current smokers and nonsmokers. There were no pulmonary function parameters showing differences between current and past smokers.

HRCT Parameters at Entry

Initial inspiratory MLD or HIST did not differ in the upper, middle, and lower lung fields in any group studied (Table 3). Although %LAA values in the nonsmoking group estimated at the beginning of the study were larger at middle and lower lung regions than those at upper lung regions, this tendency was obscure in current and past smokers. Initial MLD or HIST at expiration showed no difference in the upper, middle, and lower lung fields in all groups studied (Table 3).

Although initial inspiratory MLD values in the middle lung field were more negative in current smokers than those in nonsmokers, MLD values estimated in the upper and lower lung fields did not differ between the two groups (Table 3). Initial expiratory MLD values at upper and lower lung regions did not differ among the three groups.

The behavior of initial HIST obtained at deep inspiration qualitatively differed in comparison with that of MLD, i.e., a significant difference in initial inspiratory HIST at upper or lower lung regions was observed between past smokers and nonsmokers, but not between current smokers and nonsmokers (Table 3). Inspiratory HIST at middle lung regions did not differ among the three groups. Although expiratory HIST at lower lung regions was more negative in current and past smokers than nonsmokers, expiratory HIST at upper lung regions was not different among the three groups (Table 3).

Initial %LAA in the upper lung field was larger for current and past smokers than nonsmokers, while %LAA in the middle lung field was larger for past smokers than nonsmokers (Table 3). %LAA in the lower lung field did not differ among the three groups.

Cross-sectional Correlation between HRCT and Pulmonary Function Parameters

Although subject age was well correlated with all of the CT indices, the correlation with age was higher in CT parameters obtained in the lower lung field than that in the upper or middle lung field at deep inspiration (Table 4, correlation coefficients [r] between %LAA and age: 0.39 (p = 0.001) for upper, 0.47 [p = 0.0001] for middle, and 0.55 [p = 0.0001] for lower lung field). When contrasting CT parameters at inspiration and those at expiration, correlation coefficients between age and CT parameters were higher at expiration. On the other hand, lifetime cigarette consumption was more prominently correlated with CT parameters obtained in the upper lung field at each breath-holding position (Table 4, r values between %LAA and cigarette consumption: 0.48 [p = 0.0001] for upper, 0.45 [p = 0.0002] for middle, and 0.24 [not significant] for lower lung field). Comparing inspiratory and expiratory CT parameters, the latter revealed higher correlation with cigarette consumption than the former.

In general, pulmonary function parameters were evenly correlated with inspiratory CT parameters in any lung field, with the exception of RV/TLC (Table 4). RV/TLC appeared to have a higher correlation with inspiratory CT parameters at lower lung regions than those at middle or upper lung regions. When examining the expiratory CT parameters, the interrelation between pulmonary function and regional CT parameters was clearly demonstrated (Table 4). Higher correlation was observed between FEV₁%, FRC, or RV/TLC and expiratory CT parameters at lower lung regions (r values between RV/ TLC and expiratory MLD: 0.66 [p = 0.0001] for upper and 0.74 [p = 0.0001] for lower lung field), whereas DL_CO/VA or ₂₅ was highly related to expiratory CT parameters at upper lung regions (r values between DL_CO/VA and expiratory MLD: 0.73 [p = 0.0001] for upper and 0.68 [p = 0.0001] for lower lung field). When comparing inspiratory and expiratory CT images, expiratory CT parameters systematically showed a higher correlation with any pulmonary function than inspiratory CT parameters irrespective of the lung field at which the CT images were taken (r values between FEV₁% and MLD in the lower lung field: 0.62 [p = 0.0001] at inspiration but 0.82 [p = 0.0001] at expiration).

Annual Changes in Pulmonary Function Parameters

None of the pulmonary function parameters, excepting FEV₁, showed significant annual changes in any group during the 5-yr follow-up period (Table 2). Annual decline in FEV₁ was found to be 0.02 L/yr in nonsmokers, 0.06 L/yr in current smokers, and 0.03 L/yr in past smokers, among which the annual decline in current smokers was larger than that in nonsmokers. Although the mean annual decline in FEV₁ for past smokers was intermediate between nonsmokers and current smokers, significant difference was not demonstrated between past smokers and nonsmokers, or between past and current smokers.

Annual Changes in Mean Lung Density of HRCT Images

During the 5-yr follow-up period, inspiratory MLD estimated in the upper lung field did not change in nonsmokers, current smokers, and past smokers (Figure 2 and Table 5). Inspiratory MLD in the middle lung field shifted in a positive direction in current smokers, but in a negative direction in past smokers. In the nonsmoking group, there was little alteration in inspiratory MLD obtained at the middle portion of the lung. In the lower lung field, the MLD in the past-smoking group was found to shift in a negative direction, whereas that in the other two groups did not show any significant change.

View larger version (14K): [in this window] [in a new window]   Figure 2.   Longitudinal changes in inspiratory MLD for upper (UL), middle (ML), and lower (LL) lung fields during 5-yr follow-up period. Values at entry were the same as those presented in Table 3, while values at the end of the study (5 yr later) were those corrected for annual changes estimated by linear regression analysis shown in Table 5. NS (open circles): nonsmokers (n = 36). CS (solid circles): current smokers (n = 35). PS (solid triangles): past smokers (n = 12). Values are means ± SD. *Significant increase over 5 yr (p < 0.05); +significant decrease over 5 yr (p < 0.05).

Expiratory MLD in the upper and lower lung fields exhibited no annual change during the observation period in all the groups studied (Figure 3 and Table 5).

View larger version (16K): [in this window] [in a new window]   Figure 3.   Longitudinal changes in expiratory MLD for upper (UL) and lower (LL) lung fields during 5-yr follow-up period. Values at 5 yr later were those corrected by applying the same procedure as described in Figure 2. NS (open circles): nonsmokers. CS (solid circles): current smokers. PS (solid triangles): past smokers. Values are means ± SD.

Annual Changes in CT Density Histogram of HRCT Images

Although inspiratory HIST at upper lung regions did not change in the three groups, HIST at middle lung regions observed at inspiration shifted in a positive direction in the current-smoking group over 5 yr (Figure 4 and Table 5). Inspiratory HIST at lower lung regions did not change annually in nonsmokers, whereas it moved to a positive value in current smokers but to a negative value in past smokers, resulting in that inspiratory HIST revealed a trend qualitatively similar to that of inspiratory MLD.

View larger version (16K): [in this window] [in a new window]   Figure 4.   Longitudinal changes in inspiratory HIST for upper (UL), middle (ML), and lower (LL) lung fields during 5-yr follow-up period. Values at 5 yr later were those corrected for annual changes given in Table 5. NS (open circles): nonsmokers. CS (solid circles): current smokers. PS (solid triangles): past smokers. Values are means ± SD. *Significant increase over 5 yr (p < 0.05); +significant decrease over 5 yr (p < 0.05).

We found little change in expiratory HIST in all the groups irrespective of the lung field where CT images were taken (Figure 5 and Table 5).

View larger version (16K): [in this window] [in a new window]   Figure 5.   Longitudinal changes in expiratory HIST for upper (UL) and lower (LL) lung fields. Values after 5-yr follow-up were corrected by taking annual changes shown in Table 5 into account. NS (open circles): nonsmokers. CS (solid circles): current smokers. PS (solid triangles): past smokers. Values are given as means ± SD.

Annual Changes in Relative Area with Low Attenuation of HRCT Images

Although appreciable increase in %LAA in the upper lung field was investigated for both current and past smokers, upper lung %LAA for nonsmokers was not altered during the follow-up period (Figure 6 and Table 5). There was little difference in the extent of upper lung %LAA changes between current and past smokers. In the middle or lower lung field, %LAA was significantly enhanced in all three groups over 5 yr. Annual changes in middle lung %LAA did not differ among the three groups, whereas lower lung %LAA changes were larger for past smokers than nonsmokers, in association with no difference between current and past smokers (Figure 6 and Table 5).

View larger version (15K): [in this window] [in a new window]   Figure 6.   Longitudinal changes in %LAA for upper (UL), middle (ML), and lower (LL) lung fields. Values at entry were presented in Table 3, whereas values at 5 yr later were those corrected for annual changes shown in Table 5. NS (open circles): nonsmokers. CS (solid circles): current smokers. PS (solid triangles): past smokers. Values are means ± SD. *Significant increase over 5 yr (p < 0.05); #the slope is larger than that for nonsmokers (p < 0.05).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Critique of Methods

Although several groups of investigators reported the cutoff CT values defining the LAA that discriminates areas with emphysematous alterations from those with normal acinar structure (11, 13, 14, 16, 17, 26, 27), we did not use these values because the cutoff CT level might vary when different CT instruments were used for analysis (16). Instead, we attempted to determine the cutoff CT value based on the data obtained from young healthy nonsmokers. We found that 912 HU would be the most reliable cutoff CT value for estimating LAA as far as the type of CT scanner used in this study. This cutoff CT value is consistent with the values reported by Hayhurst and coworkers (26), Müller and coworkers (11), Kinsella and coworkers (13), and Knudson and coworkers (14), all of whom demonstrated that the cutoff CT value of 900 or 910 HU would be significantly correlated with the morphological destruction of acinar structures. On the other hand, Gevenois and coworkers (16, 17) found the highest correlation coefficient between morphologically detected emphysema and HRCT-related emphysema (i.e., LAA) when 950 HU was used as the cutoff CT level. Gierada and coworkers (27) reported that LAA defined by cutoff CT values of 900 or 910 HU would include areas with incomplete destruction of the acinar structure in addition to those with serious destruction, while LAA defined by a cutoff CT value of 960 HU would solely reflect emphysematous areas with serious damage. Since our cutoff CT value defining LAA was determined from the data obtained for normal healthy young volunteers, but not from the data for patients with established emphysema, it may not be unreasonable to assume that LAA defined in the present study reflects areas with incomplete as well as total destruction of the acinar structure, as discussed by Gierada and coworkers (27).

Correlation between Pulmonary Function and CT Parameters in Cross-sectional Analysis

Gurney and coworkers (15) suggested that the upper lung field acted as a silent zone for overall pulmonary function. This hypothesis appears to be reasonable because the tissue volume in the lower lung is much greater than that in the upper lung, with the result that contribution of the lower lung to overall pulmonary function is proportionally greater. However, we found that the silent-zone hypothesis can be applicable for some, but not for all, of the pulmonary function parameters (Table 4). For instance, DL_CO/VA is found not to conform to the silent-zone hypothesis. This may be attributed to the fact that DL_CO/VA mainly reflects airspace destruction (1, 15, 19), which predominantly occurs in the upper lung field by habitual smoking (Table 3).

Comparing CT findings with the severity of both macroscopically and microscopically quantified emphysema, Gevenois and coworkers (17) reported that quantitative CT parameters estimated at inspiration would be superior to those at expiration for judging morphological emphysema. In contradiction to the relation between inspiratory CT and morphological emphysema, several studies including ours (14, 18, 19) confirmed that CT parameters obtained at expiration had a higher correlation with various functional parameters. Higher correlation between the parameters of expiratory CT and those of airflow obstruction such as FEV₁, FEV₁%, and ₂₅ may be attributed to the fact that airflow obstruction is exaggerated at expiration and reflected as enhanced air trapping in CT images taken at expiration, as discussed by several investigators (17, 18, 28, 29). However, it is not clear why DL_CO-associated parameters also revealed a higher correlation with CT parameters at expiration than those at inspiration. DL_CO has been believed to mainly reflect the extent of emphysema caused by airspace destruction in acini (1, 15, 19). If this is wholly true, DL_CO-related parameters should have a better correlation with CT images estimated at inspiration, because a higher correlation was obtained between inspiratory CT parameters and the extent of morphologically quantified emphysema (17). One possibility to explain the apparent contradiction mentioned above is that DL_CO is significantly influenced by airway abnormalities leading to augmented airflow obstruction (31). Airflow obstruction, especially that induced by peripheral airway abnormalities, may considerably enhance functional inhomogeneities and thus impairs DL_CO (31). Since airflow obstruction is exaggerated at deep expiration and this information is included in expiratory CT images, it is not unreasonable to infer that the DL_CO value is connected with quantitative CT parameters obtained at expiration through qualitatively the same mechanism as that considered for FEV₁ and ₂₅.

Longitudinal Changes in FEV₁

FEV₁ is a representative parameter frequently used in longitudinal studies to detect a difference in overall ventilatory function between nonsmokers and smokers. Burrows and coworkers (4) demonstrated that the follow-up time of more than 5 yr was indispensable in stabilizing the longitudinal data for FEV₁. Bande and colleagues (3) attempted to elucidate longitudinal changes in FEV₁ for many subjects and found that the annual decline of FEV₁ in smokers was much greater than that in nonsmokers. In addition, they reported that the difference in FEV₁ between nonsmokers and smokers was distinct when subjects were more than 40 yr old (3). Qualitatively, the same results were obtained by many investigators including Xu and coworkers (20), Tashkin and coworkers (21), Sherrill and coworkers (22), Anthonisen and coworkers (23), and Burchfiel and coworkers (24), all of whom demonstrated that the rate of loss in FEV₁ was accelerated in continuing smokers as compared with that in subjects who had never smoked, but it became less steep in smokers who had given up smoking. In addition, Xu and coworkers (20) and Burchfiel and coworkers (24) found that the rate of decline in FEV₁ in smokers who had given up smoking diminished to a level similar to that in persons who had never smoked. In contradiction to these studies, Knudson and coworkers (32) failed to discern a significant difference in longitudinal decline of FEV₁ between smokers and nonsmokers over a 5-yr follow-up period. Although the numbers of subjects analyzed in the present study are small because of difficulties in regularly completing annual examinations of both pulmonary function and HRCT at almost the same time, we found qualitatively the identical trends in FEV₁ to those reported in most of the previous studies excepting Knudson and coworkers (32), i.e., the annual decline in FEV₁ was significantly larger for current smokers than nonsmokers (Table 2). On the other hand, the annual decline in FEV₁ did not differ significantly between past and current smokers, although the mean rate of decline in FEV₁ for current smokers was about twofold over that for past smokers. Although our findings for past smokers are inconsistent with those of Xu and coworkers (20) and Burchfiel and coworkers (24) in a statistical sense, this may be attributed to the large difference in subject numbers studied. Interindividual variation of pulmonary function measurements arising from variation among subjects with regard to personal factors such as height, sex, and age is substantially decreased as the total number of subjects analyzed is increased (33). The same holds true for the intraindividual variation of pulmonary function tests performed on occasions apart (33).

Summary of Longitudinal Changes in Quantitative CT Parameters in Three Groups

Summarizing the important findings on longitudinal changes in various CT parameters: (1) nonsmoking healthy subjects showed significant annual changes in %LAA in the middle and lower lung fields, but not in the upper lung field (Figure 6 and Table 5), while other inspiratory CT parameters including MLD and HIST were not altered in any lung field over a 5-yr follow-up period (Figures 2 and 4, and Table 5); (2) in current smoking subjects, %LAA was augmented in all the lung fields during the observation period, in association with a shift of inspiratory MLD or HIST estimated for the middle or lower lung field in a positive direction (Figures 2, 4, and 6, and Table 5). The annual increase in %LAA obtained in the upper lung field was greater for the current-smoking group than for the nonsmoking group, whereas that in the middle or lower lung field did not differ between the two groups; (3) past-smoking subjects exhibited a greater increase in %LAA in the upper and lower lung fields than nonsmokers. There was little difference in the extent of increase in %LAA at any lung region between past and current smokers (Figure 6 and Table 5). In past-smoking subjects, inspiratory MLD in the middle and lower lung fields as well as HIST in the lower lung field shifted in a negative direction (Figures 2 and 4, and Table 5), the tendency being qualitatively different from that observed in current smokers; and (4) expiratory CT indices did not change annually in any lung field for all the groups tested (Figures 3 and 5).

Characteristics of Longitudinal Changes in CT Parameters in Nonsmokers

The findings of annual changes in %LAA obtained for nonsmoking healthy subjects (Figure 6 and Table 5) indicate that effects of aging on the lung peripheries such as airspace enlargement are manifested mainly in the middle and lower lung fields, but the upper lung field is somewhat resistant to aging. This is also supported by the findings observed in cross-sectional analysis, in which nonsmoker %LAA estimated for the upper lung field was much smaller than that for the middle and lower lung fields (Table 3). Furthermore, upper lung %LAA exhibited the lowest correlation with subject age (Table 4). Longitudinal analysis may further suggest that %LAA is a more sensitive CT parameter than MLD or HIST in detecting aging phenomena in the lung.

Characteristics of Longitudinal Changes in CT Parameters in Current Smokers

Longitudinal changes in %LAA observed for current smokers (Figure 6 and Table 5) suggest that the injurious effects of cigarette smoke inducing emphysematous alterations predominantly occur in and affect the upper lung field, but to a relatively lesser extent in the middle and lower lung fields. This is also supported by the cross-sectional results, in which the highest correlation coefficient was obtained between smoking consumption and %LAA in the upper lung field (Table 4). Besides, cross-sectional analysis showed that upper lung %LAA was larger in current smokers than nonsmokers, while middle or lower lung %LAA did not differ between the two groups (Table 3). The other notable findings on changes in CT parameters in current smokers were that inspiratory MLD and HIST in the middle and/or lower lung fields, on the contrary, shifted to more positive values during the 5-yr follow-up period (Figures 2 and 4, and Table 5), indicating an increase in lung density in the middle and lower portions of the lungs of smokers. Taken together with the results of %LAA, these findings indicate that continuous smoking incurs respiratory bronchiolitis and/or microscopic fibrotic changes causing increased lung density, but brings about little serious destruction of acinar structures leading to emphysema in the middle and lower lung fields, for at least 5 yr. Remy-Jardin and coworkers (34) investigated the effects of cigarette smoking on morphological changes in pulmonary parenchyma in a cross-sectional manner. They evaluated, in terms of HRCT, small nodules in the parenchyma, areas with ground-glass opacities, emphysematous alterations, bronchial wall thickening, and septal lines in nonsmoking subjects, current smokers, and past smokers. They demonstrated that parenchymal nodules, indicating the presence of respiratory bronchiolitis, and areas with ground-glass opacities, suggesting microscopic fibrosis or accumulation of inflammatory cells in respiratory bronchioles and/or alveolar spaces, were more prominent in current smokers than those in nonsmoking subjects. Although Remy-Jardin and co-workers (34) did not specify the major regions where smoking-related respiratory bronchiolitis and ground-glass opacities occur, their findings appear to be compatible with the present longitudinal results. The longitudinal HRCT measurements suggest that accelerated decline in FEV₁ in current smokers over 5 yr (Table 2) is predominantly caused by increased emphysematous alterations in the upper lung field as well as enhanced respiratory bronchiolar inflammation in the middle and lower lung fields. It is uncertain, however, why cigarette smoking aggravates bronchiolar impairments in the middle and lower lung fields but not in the upper lung field where airspace abnormalities are accelerated. One of the possible reasons for this peculiar phenomenon is that smoking-related bronchiolar inflammation also occurs in the upper lung field, thus increasing the lung density there, but this increase is concealed by a greater increase in emphysematous lesions which lead to a decrease in the lung density. Supporting this explanation, cross-sectional analysis elucidated that upper lung %LAA was larger in habitual smokers than nonsmokers, whereas upper lung MLD and HIST were not different between the two groups (Table 3). Further study, however, is necessary for settling this question.

Characteristics of Longitudinal Changes in CT Parameters in Past Smokers

Surprisingly, we found that the annual change in %LAA in any lung field did not differ between past and current smokers (Figure 6 and Table 5). In addition, %LAA changes in the upper and lower lung fields were greater for past smokers than for nonsmokers, suggesting that development of airspace abnormalities did not slow down or stopped at least within 5 yr after smoking cessation. On the other hand, in opposition to the tendency obtained in current smokers, lung densities in the middle and/or lower lung fields estimated on the basis of MLD or HIST were significantly decreased in past smokers (Figures 2 and 4, and Table 5). Furthermore, the annual change in lung density in any lung field did not differ between past smokers and nonsmoking healthy subjects (Table 5). These findings may indicate that smoking-related augmentation in lung density due probably to acinar bronchiolitis and/or microscopic fibrosis is reversible and disappears within 5 yr after smoking cessation. Although it is extremely difficult to know the reason for sustained changes in %LAA in subjects who have stopped smoking (Figure 6 and Table 5), one possibility is that longtime smoking habit makes the airspace more susceptible to aging and this does not diminish over years after smoking cessation. Enhanced susceptibility to aging, which leads to increased %LAA, due probably to airspace enlargement, may be masked while smoking is continued, because this is counterbalanced by increased lung density caused by smoking-induced bronchiolar inflammation and/or microscopic fibrosis. On the other hand, smoking cessation may disclose the enhanced susceptibility to aging because of disappearance of increased lung density.

Significance of Expiratory CT Parameters

CT parameters taken at expiration appear to be less informative than those at inspiration for longitudinal assessment of structural abnormalities at the lung peripheries caused by aging and/or smoking (Figures 3 and 5, and Table 5). This may be attributed to the fact that the lung volume at expiration is largely reduced and thus the ratio of tissue mass to air in the lung is increased in parallel. Since relative increase in tissue mass leads to enhancement of the background density throughout the lung field, this may decrease the sensitivity of HRCT in detecting subtle alterations in peripheral lung structures related to airspace abnormalities. However, HRCT taken at expiration has been used as a valuable tool allowing assessment of the extent of air trapping (17, 18, 28, 29). Taken together, our findings obtained at deep expiration may suggest that the extent of air trapping does not change significantly in any group at least during the 5-yr observation period. However, cross-sectional analysis demonstrated that expiratory lung density in the lower lung field estimated from HIST was more negative in smokers than nonsmokers (Table 3). These results may suggest that air trapping is enhanced very gradually in the lungs of smokers and a long period (i.e., more than 5 yr) is necessary for an increase in air trapping in the lungs of smokers to be detected on CT.

In conclusion, HRCT is superior to pulmonary function measurements in detecting subtle longitudinal changes in acinar structures caused by aging and smoking. Aging augments airspace enlargement predominantly in the middle and lower lung fields, while continuous smoking further worsens emphysematous alterations in the upper lung field and bronchiolar abnormalities such as bronchiolitis and/or microscopic fibrosis in the middle and lower portions of the lung, both of which are considered to contribute importantly to deteriorating FEV₁ in habitual smokers. Although inflammation and/or microscopic fibrosis surrounding bronchioles and/or airspaces are significantly improved after quitting smoking, airspace abnormalities accelerated by exposure to cigarette smoke may progress even after stopping smoking.