Thoracic Dimensions at Maximum Lung Inflation in Normal Subjects and in Patients With Obstructive and Restrictive Lung Diseases*

Transcription

1 Thoracic Dimensions at Maximum Lung Inflation in Normal Subjects and in With Obstructive and Restrictive Lung Diseases* Jean-François Bellemare, BSC; Marie-Pierre Cordeau, MD; Pierre Leblanc, MD; and François Bellemare, PhD Objectives: To compare the distribution of lung volume at total lung capacity (TLC) among adult men and women known to have normal lung function or chronic obstructive disease or restrictive lung disease (RLD). Design: Five-year retrospective study. Setting: Review of available clinical pulmonary function testing (PFT) reports and chest radiographs. : Sixty-four patients presenting with normal PFT and chest radiograph findings (normal subjects), 26 patients with severe COPD and increased TLC (COPD group), 29 patients with cystic fibrosis (CF) and increased TLC (CF group), and 19 patients with RLD with a clinical diagnosis of pulmonary fibrosis and a reduced TLC (RLD group). Measurements: Average posteroanterior rib cage diameter (PAave), average lateral rib cage diameter (LAave), and average vertical height of the diaphragm (HDIave) were measured using radiography. Normal prediction equations were generated based on stature, body mass index (BMI), age, and sex as independent variables and then used in between-group comparisons. Results: PAave correlated positively with BMI and age but not with height, whereas LAave correlated positively with BMI and height but not with age. HDIave correlated positively with height and age but negatively with BMI. PAave and LAave were smaller and HDIave was greater in women than men having the same stature. In the COPD group and in male CF group patients, BMI was low and only HDIave was greater than in sex-, age-, and height-matched normal subjects, but in female CF group patients, only the rib cage diameters were greater than normal. In the RLD group, PAave and HDIave were smaller than predicted and inversely related to each other, but LAave was normal. Conclusion: Variations in maximum lung volume caused by gender, growth, or by lung diseases are nonisotropic and entail substantial changes in chest wall shape. (CHEST 2001; 119: ) Key words: chest radiography; chest wall; COPD; cystic fibrosis; pulmonary fibrosis; rib cage dimensions Abbreviations: ANCOVA analysis of covariance; ANOVA analysis of variance; BMI body mass index; CF cystic fibrosis; Dlco diffusing capacity of the lung for carbon monoxide; FRC functional residual capacity; HDI height of the diaphragm measured on posteroanterior chest radiograph; HDIave average vertical height of the diaphragm; LAave average lateral rib cage diameter; LA-R3 LA diameter measured as the maximal internal diameter of the chest wall at the level of the third pair of ribs; LA-R6 LA diameter measured as the maximal internal diameter of the chest wall at the level of the sixth ribs; LA-R9 LA diameter measured as the maximal internal diameter of the chest wall at the level of the ninth pair of ribs; LA-R9 R3 difference between LA-R9 and LA-R3; PA posteroanterior; PAave average posteroanterior rib cage diameter; PA-T5 PA diameter measured on LA radiograph taken at the base of the fifth thoracic vertebral body; PA-T7 PA diameter measured on LA radiograph taken at the base of the seventh thoracic vertebral body; PA-T9 PA diameter measured on LA radiograph taken at the base of the ninth thoracic vertebral body; PA-T9 T5 difference between PA-T9 and PA-T5; PFT pulmonary function testing; RLD restrictive lung disease; RV residual volume; TLC total lung capacity; Vth intrathoracic cavity volume. *From the Research Center (Dr. J. Bellemare), and Departments of Radiology (Dr. Cordeau), Pneumology (Dr. Leblanc), and Anesthesiology (Dr. F. Bellemare), Hôtel-Dieu, Montreal University Hospital Centre, Montréal, Québec, Canada. This study was supported by the Medical Research Council of Canada. Manuscript received June 12, 2000; revision accepted September 25, Correspondence to: François Bellemare, PhD, Centre de recherche, Hôtel-Dieu du CHUM, 3850 rue St-Urbain, Montréal, Québec, Canada H2W 1T8; francois.bellemare@ umontreal.ca 376 Clinical Investigations

2 The volume of the intrathoracic structures can vary substantially among subjects of different build or sex, as well as in those with diseases of the heart and lungs. How variations in intrathoracic air or tissue volume are accommodated by the chest wall in different conditions is a question that has not been completely addressed. Consideration of the volume and shape of the thorax is relevant to lung growth 1 and also to the action of the respiratory muscular pump 2 and to the distribution of pleural pressure and ventilation. 3 The factors that determine the shape of the human thorax and that must be considered are also incompletely defined. Muscular 2,4 6 and gravitational forces 7,8 as well as genetic factors 9 clearly play an important role, but other factors are also likely to be involved. As is the case for the volume of the lungs, anthropologic variables such as height, weight, sex, and age are likely to affect thoracic dimensions. 10,11 However,to our knowledge, there is no quantitative study on which to base predictions of the relative influence of these variables on the shape of the human thorax. Quantitative predictions are required in order to correctly assess variations caused by disease or by other factors. The first objective of this study, therefore, was to determine the relationships between thoracic dimensions at full inflation and anthropologic variables in subjects having normal lung function. Prediction equations based on these relationships were then used to study changes in thoracic dimensions in patients with obstructive lung disease and restrictive lung disease (RLD). Subject Selection Materials and Methods Subjects included in this study were all white and had been referred over a period of 5 years to the pulmonary physiology laboratory for complete pulmonary function testing (PFT) and for whom standard chest radiographs were available. For the initial screening, 2,638 PFT reports were consulted and four groups of subjects were identified using the following criterion. who met American Thoracic Society standards for normal lung function 12 (ie, with values for FVC, FEV 1, singlebreath diffusing capacity of the lung for carbon monoxide [Dlco]), plethysmographically determined residual volume [RV], and total lung capacity [TLC] within the normal range as defined by the 95% confidence intervals for predicted values, and no respiratory diseases detectable by PFT) were identified as normal subjects. In our laboratory, predicted values for FVC and FEV 1 are from Knudson et al, 13 predicted values for lung volumes are from Crapo et al, 14 and predicted values for Dlco are from Ayers et al. 15 classified by the pneumologists to either the COPD group or cystic fibrosis (CF) group and whose TLC was greater than the upper 95% confidence limit for predicted normal values were identified. COPD patients, in addition to having severe airways obstruction and increased TLC, also needed to present a Dlco smaller than the lower 95% confidence limit for predicted normal values, as we wished this group to be representative of those COPD patients having pulmonary emphysema as their primary abnormality. classified by the pneumologists as having pulmonary fibrosis and presenting with a TLC and a Dlco smaller than the lower 95% confidence limits for normal predicted values were also identified. Because a lung biopsy specimen-proven diagnosis was not available in this group, they will be referred to as patients with RLD. This initial screening yielded 494 patients, of whom 258 were classified as normal subjects and 236 were classified to the COPD group, the CF group, or the RLD group. Chest radiographs from these patients, when available, were reviewed along with the radiologist reports to eliminate subjects showing any lung and/or skeletal abnormalities, other than those related to the primary diagnosis, that could modify thoracic dimensions. Of the 494 cases, 130 were discarded because chest radiographs were not available for review and 226 were rejected because of significant abnormalities other than the primary diagnosis. These include patients with a history of thoracic or breast cancer surgery, and patients presenting with evidence of pneumonia, cardiovascular disease, multiple pulmonary nodules, lung or mediastinal mass, abnormalities of the pleural space, or thoracic deformities. One hundred thirty-eight cases were retained, in that the radiographs did not show other significant abnormalities than those related to the primary diagnosis. An average ( SD) delay of days separated the two examinations. Of these 138 patients, 64 were classified as normal subjects, 26 to the COPD group, 29 to the CF group, and 19 to the RLD group. The physical characteristics and pulmonary function data of all patients analyzed are given in Table 1. Measurements All posteroanterior (PA) and lateral (LA) chest radiographs analyzed had been taken at full inspiration and with the same equipment and standardized protocol. The subjects were standing with their arms in front and their torso against the cassette, which was at a distance of six feet from the X-ray tube. No correction was made for the magnification of thoracic structures. All measurements made on these radiographs are illustrated in Figure 1. On PA radiographs, LA diameters were measured as the maximal internal diameter of the chest wall at the level of the third, sixth, and ninth pair of ribs (LA-R3, LA-R6, and LA-R9, respectively) and an average LA rib cage diameter (LAave) was calculated as the mean of LA-R3, LA-R6, and LA-R9 measures. The height of the diaphragm measured on PA chest radiograph (HDI) was measured as the vertical distance from the base of T1 to the silhouette of the left and right diaphragmatic domes midway between the internal aspect of chest wall at the level of the ninth rib and the center of the thoracic spine. Average vertical height of the diaphragm (HDIave) was calculated as the mean of left and right HDI determinations. The vertical height of the thoracic segment between the base of T1 and the base of T10 was also measured as an indication of thoracic spinal height. PA diameters were measured on LA radiographs taken at the base of the fifth, seventh, and ninth thoracic vertebral bodies (PA-T5, PA-T7, PA-T9, respectively); an average PA rib cage diameter (PAave) was calculated as the mean of PA-T5, PA-T7, and PA-T9 measures. For these measurements, the anterior aspect of the spine was taken as the posterior limit, as this could always be precisely defined. The posterior aspect of the sternum was taken as the anterior limit. In the instances where T9 was located below the sternum, the posterior aspect of the anterior chest wall was used as the anterior limit. The craniocaudal gradients in rib cage dimension were calculated for each subject CHEST / 119 / 2/ FEBRUARY,

3 Table 1 Subject Characteristics* Normal COPD Group CF Group RLD Group Variables Male Subjects (n 26) Female Subjects (n 38) Male (n 16) Female (n 10) Male (n 14) Female (n 15) Male (n 11) Female (n 8) Age, yr Height, cm Weight, kg BMI, kg/m FVC, % predicted FEV 1, % predicted FEV 1, % FVC Dlco, % predicted RV, % predicted FRC, % predicted TLC, L TLC, % predicted *Data are presented as mean SD. as the difference between PA-T9 and PA-T5 ( PA-T9 T5) and between LA-R9 and LA-R3 ( LA-R9 R3). Reproducibility of the Measurements Ten sets of radiographs were analyzed by two independent observers. Each observer analyzed five sets of radiographs as first observer, erased all the markings, and then analyzed the other five sets. When expressed as percentage of the values reported by the experienced observer, the mean ( SD) difference for 100 paired observations averaged % and was not significantly different from zero. Data Analysis The analysis was divided in three parts. In part 1, standard linear regression techniques were employed to examine the relationships between measured thoracic dimensions and the anthropologic variables in the normal group. Plots of residuals were examined to identify possible outliers, and their influence on the regression models was evaluated. In part 2, an analysis of covariance (ANCOVA) was employed to see whether the relationships obtained in part 1 could be extended to the other groups. Interaction terms among sex, patient groups, and anthropologic characteristics were included in the models as indicated Figure 1. Radiographic determination of thoracic dimensions. 378 Clinical Investigations

4 in the Results section. In part 3, prediction equations derived in part 1 and validated in part 2 were used to generate normal predicted values based on each subject s anthropologic characteristics. The measured thoracic dimensions in each subject were then each expressed in percentage of these predicted values and compared between groups using an analysis of variance (ANOVA). When ANOVA indicated significant between groups differences, post hoc multiple comparisons were generated based on Tukey s Studentized Range Test to determine those that were significant at a p value of The statistical analysis is described in more detail in the relevant Results section. All computations were made using software programs (SAS, version 6; SAS Institute; Cary, NC; and SPSS, version 10 SPSS; Chicago, IL). All results are presented as mean 1 SD. Results Subject Characteristics The anthropologic characteristics and lung function data of the different groups are summarized in Table 1. The age, height, and body mass index (BMI) of normal subjects varied widely and uniformly and covered the range seen in the study patients. In normal subjects, BMI was significantly and positively correlated with age. For each gender, height was comparable among groups. The CF group, however, was significantly younger, and the COPD group and the RLD group were significantly older than normal subjects. The COPD group and the CF group had significantly smaller BMI than normal subjects. Their BMI was also significantly smaller than the recommended value of 24 kg/m 2. Only in CF group patients did the pulmonary function differ between genders, with male patients having significantly more severe airways obstruction and gas trapping (ie, functional residual capacity [FRC] and RV) than female patients, in spite of a comparable degree of hyperinflation (ie, TLC). Smoking status was noted but was not used in the selection process. Half of the normal control subjects were nonsmokers, 19% were former smokers, and 31% were smokers. ANOVA showed no significant relationship between the smoking status and any of the measurements obtained in this group. The RLD group presented a similar proportion of smokers (22%) and nonsmokers (39%) than normal subjects. Only 3% of the COPD group patients were nonsmokers, whereas this was the case for 92% of the CF group patients. Findings in the Normal Group and Prediction Equations A multiple linear regression model was adjusted to PAave, LAave, and HDIave as dependent variables, with sex, height, BMI, and age as independent variables (model I equations). A separate model was also adjusted to the same dependent variables, in that BMI was excluded from the list of independent variables (model II equations). Interaction terms with sex were initially included in both models to take into consideration possible gender-related differences in the relationships between the dependent and independent variables. For LAave and HDIave, the interaction terms with sex were not significant using either model I or model II equations and were therefore excluded from the models. In the case of PAave, however, there was a significant interaction between height and sex in model I but not in model II. Thus, a different height coefficient was required for each sex. For each model, plots of the residuals against predicted values were examined to identify possible outliers and to evaluate their contribution to the parameter estimates. Two such outliers were identified in the case of HDI, which had significant effects on the model coefficients. These two outliers were eliminated from the HDI models but not from the PA or LA models. For PA and LA diameters, no outliers could be identified and all cases were included in the models. The final coefficients for model I and model II equations are given in Table 2. For HDIave, PAave, and LAave, 29%, 69%, and 72% of the total variance, respectively, was explained by these models. Gender Coefficients: In both models, all sex coefficients were significant, with female subjects having smaller PA and LA diameters but greater HDI values than male subjects with the same anthropologic characteristics. The thoracic index (ratio of PA to LA diameters) was also significantly smaller in female subjects than in male subjects (0.502 vs 0.537, respectively; p 0.03). BMI Coefficients: All BMI coefficients in model I were significant. BMI correlated positively with PA and LA diameters but negatively with HDI. The same regression model was extended to the diameters measured at different axial levels. The coefficients for BMI were found to be greater for lowerthan for upper-rib cage PA and LA diameters. Using the same multiple linear regression model, PA- T9 T5 and LA-R9 R3 were both found to be significantly and positively correlated with BMI but not with age or height. PA-T9 T5 and LA- R9 R3 were also significantly smaller in female subjects than in male subjects having the same BMI. Height Coefficients: In both models, height was positively correlated to LA diameters and to HDI, whereas PA diameters were independent of height. Age Coefficients: Age correlated with PA diameters in both models, but its contribution to the CHEST / 119 / 2/ FEBRUARY,

5 Table 2 Prediction Equations* Variables PAave, mm LAave, mm HDIave, mm Model I RMSE R Intercept Sex (p 0.023) (p ) (p 0.024) Age, yr (p 0.029) (p 0.009) Height, cm (p ) 1.34 (p 0.001) BMI, kg/m (p ) (p ) (p 0.008) Sex height (p 0.036) Model II RMSE R Intercept Sex (p ) (p ) (p 0.012) Age, yr 0.51 (p ) Height, cm (p ) (p 0.001) *RMSE root mean square error. explained variance was markedly reduced in model 1 (2.9% vs 15.6%), reflecting the positive relationship between BMI and age. When BMI was included in the model, HDI was significantly and positively correlated with age. Comparison of Model I and Model II Equations in the Four Groups To compare the coefficients relating thoracic dimensions to anthropologic variables among the four groups of subjects, an ANCOVA was performed, in that all anthropologic variables served as covariates, and interaction terms between the group category and the covariates were included in the model. These interaction terms, if significant, indicate between-group differences in the coefficients relating the measured thoracic dimensions and the covariate. The ANCOVA confirmed all the findings in part 1 above and therefore extended these findings to the four groups. In addition, for PAave and LAave, the interaction terms were not significant, thus showing similar regression coefficients in the four groups. For HDI, however, the interaction term with sex was Table 3 Thoracic Dimensions PAave, LAave, HDIave, Groups Male Female Male Female Male Female Measures, mm Normal subjects COPD CF RLD Model I, percent predicted Normal subjects COPD * * * * * CF * * * RLD * * * * Model II, percent predicted Normal subjects COPD * * CF * * RLD * * * * *Significantly different from normal subjects at p 0.05 using ANOVA. 380 Clinical Investigations

6 significant, thus showing that the coefficients associated with sex differed between groups. The analysis was therefore repeated for each sex separately. In both male and female subjects, there was no significant interaction between the group category and either age, height, or BMI, thus showing similar coefficients between these variables and HDIave in the four groups. Comparison Between and Control Subjects: For this analysis, the various thoracic dimensions in a given subject or patient were each expressed in percentage of the value predicted by model I and model II equations of Table 2. Group results are summarized in Table 3. Between-group comparisons were performed using ANOVA and post hoc t test. Because of the significant sex/group category interaction just described, a separate between-group comparison was carried for each sex. Given the observed coefficients of variation, this analysis could detect a 4% difference in LA diameters and a 10% difference in PA diameters or in HDI between any two groups, with a probability of 0.05 and a power of 90%. values using either model I or model II equations, whereas LA diameters were not different from control subjects with either model. These reductions were comparable in male and female subjects. The reduction in PA diameters was negatively correlated with HDI (r 2 0.3), but no correlation was found between LA diameter and HDI. This negative relationship appeared to be related to the smoking habit, as PA diameters were significantly greater (p 0.045) and HDI tended to be smaller, although not significantly (p 0.08), in smokers or ex-smokers than in nonsmokers. However, none of the pulmonary function variables differed significantly between smokers or ex-smokers and nonsmokers. PA-T9 T5 and LA-R9 R3 after adjustment for BMI, age, height, and sex tended to be greater than in normal control subjects but not significantly so. In RLD group patients, however, the genderrelated difference in these gradients was reduced to a nonsignificant level. Relationships to Lung Volume: The exponents of the power functions relating plethysmographic TLC Findings in COPD: Using model I equations, PAave and HDIave were both significantly elevated in COPD patients as compared to normal control subjects. In male subjects, LAave was also significantly elevated. By contrast, using model II equations, only HDI values were significantly elevated both in male and female subjects. PA-T9 T5 and LA-R9 R3 adjusted for age, sex, height, and BMI were not different from control subjects. Findings in the CF Group: In contrast to COPD group patients, findings in the CF group were gender specific. In male patients in the CF group, PAave and HDIave, but not LAave, were significantly greater than in control subjects using model I equations, whereas only HDI was significantly greater than in control subjects using model II equations. By contrast, in female patients in the CF group, only PAave was greater than in control subjects using either model I or model II equations, whereas HDIave and LAave were not different from normal control subjects using either model. Genderrelated differences in PA-T9 T5, but not LA- R9 R3, disappeared in CF group patients. For both sexes, combined PA-T9 T5 was also significantly greater than in control subjects having the same anthropologic characteristics. Findings in the RLD Group: In RLD group patients, PA diameters and HDI were both significantly reduced, as compared to predicted normal Figure 2. Relationship of TLC to body height (top, A) and to thoracic spinal height (bottom, B) in normal male (closed symbols) and female subjects (open symbols). A power function is fitted through each set of data. CHEST / 119 / 2/ FEBRUARY,

7 extent, to the observed variance in thoracic dimensions among adults and that these relationships are comparable among normal subjects and patients with obstructive lung disease and RLD. Radial rib cage dimensions were more tightly coupled to anthropologic variables than HDI, but the axial dimension contributed more to variations in the volume of the chest wall and lungs. Variations in the volume of lungs associated with sex, height, and with obstructive lung disease and RLD all entailed substantial and systematic variations in the shape of the chest wall and lungs. Figure 3. Relationship between estimated Vth and plethysmographic TLC. The line of linear best fit, the regression equation, and the coefficient of determination are also shown. For further details see text. to height in normal male and female subjects (2.32 vs 2.26) were comparable (Fig 2, top, A). When corrected for height, plethysmographic TLC was 700 ml smaller in female than in male subjects. This gender-related difference in TLC was not reduced when referenced to thoracic spinal height instead of body height (Fig 2, bottom, B). When compared at equal plethysmographic TLC using an ANCOVA, PA and LA diameters were significantly smaller and HDI was significantly greater in female than in male subjects. To estimate the relative contribution of the three thoracic dimensions to lung volume changes among normal subjects, the three thoracic dimensions were first plotted against plethysmographic TLC and the change of each dimension over the lung volume changes was determined by linear regression. The relative contribution was then estimated by multiplying the change in one thoracic dimension by the surface formed by the two remaining dimensions. In male subjects, variations in PA and LA dimensions and in HDI each accounted for 6.9%, 26.1%, and 67%, respectively, of the variations in TLC. The corresponding figures in female subjects were 10.2%, 24.4%, and 65.4%, respectively. Intrathoracic cavity volume (Vth) was approximated by multiplying HDIave by the surface formed by the product of LAave and PAave taken as the major and minor axis of an ellipse. Vth estimated this way was strongly correlated to plethysmographic TLC in all subjects and patients studied (r ; p 0.001; Fig 3). Discussion These results show that all anthropologic variables considered contribute significantly, albeit to a different Influence of Anthropologic Variables on Lung Volume and Thoracic Dimensions TLC is known to vary with height and sex but to be almost invariant with age. 12,14 In our study, HDI and LA diameters both varied with height whereas PA diameters did not. Thus a two-dimensional system such as seen on frontal chest radiographs appeared sufficient to describe the variations of the volume of lungs among adults. Approximately two thirds of this variation was accounted for by axial diaphragm displacements, the remaining one third being essentially accounted for by LA rib cage displacements. The exponent of the power functions relating plethysmographic TLC to stature was close to 2, also suggesting a two-dimensional variation in the maximum volume of adult lungs. This finding is surprising in view of the fact that PA diameters have been shown to increase, along with LA diameters, during the period of growth. 1 In this period, the exponent of the power function relating lung volume to stature is closer to 3, also suggesting a three-dimensional variation in the volume of the lungs during growth. 16 Findings similar to ours were noted by Takahashi and Atsumi, 10 who measured the external thoracic dimensions in the seated posture at FRC in 9,000 inhabitants of the Tohoku area in Japan. The reason for this is unknown. However, because FRC is passively determined, the finding of Takahashi and Atsumi 10 would suggest that muscular factors need not be involved. In accord with the normal prediction equations of Crapo et al, 14 TLC was approximately 700 ml smaller in female subjects than in male subjects having the same stature. A smaller ratio of trunk height to body height in female subjects has been suggested as a factor contributing to this lung volume difference. 17 We tested this possibility by relating TLC to the height of the thoracic spine as measured on chest radiographs, a procedure that failed to reduce the difference in TLC between male and female subjects (Fig 2, bottom, B). In our study, a smaller rib cage size was uniquely responsible for the smaller lung volume of normal female subjects, the 382 Clinical Investigations

8 axial dimension being in fact greater than in male subjects having the same stature. Not only was the rib cage smaller in our female subjects, but the thoracic index was also significantly smaller. Furthermore, PA-T9 T5 and LA-R9 R3 were smaller in female subjects than male subjects, a finding consistent with an earlier study in children 8 to 17 years of age. 18 The reasons for these findings are unknown but may have important physiologic implications, as the small lung volume of female subjects has been shown to limit exercise hyperpnea in fit women, 19 and as exercise-induced arterial hypoxemia has been shown to occur more readily in normal female subjects than in normal male subjects. 20 Whether TLC is affected by obesity is debatable, 21 but Ray et al 22 have shown that TLC can remain constant over a wide range of BMI (up to 60 kg/m 2 ) when obesity is not complicated by other factors. BMI, which was taken as an index of adiposity, showed opposite relationships with rib cage dimensions and HDI. Furthermore, increased BMI was associated with an increased craniocaudal gradient in rib cage dimensions that was greater in male than female patients. Gastric pressure, a surrogate of abdominal pressure, was shown to correlate positively with BMI. 23 The increased abdominal pressure, by displacing the diaphragm in the cranial direction, should improve the action of this muscle on the lower rib cage, while at the same time opposing its axial displacements. An increased abdominal pressure would also tend to inflate that part of the rib cage directly apposed to the diaphragm and to the abdominal viscera. These two mechanisms could explain the opposite relationships of BMI with rib cage dimensions and HDI, as well as the craniocaudal gradient in rib cage dimensions that we observed, including the gender-related difference of this gradient. Although additional studies will be required to test this possibility, the functional consequence of a positive relationship between BMI and rib cage dimensions is predictable, as this would tend to preserve lung volume when the abdomen is distended. 2 Several studies 10,24 of thoracic dimensions have reported a progressive rounding of the chest with aging and have associated this with an increase in PA dimension. In these earlier studies, 10,24 the confounding effect of BMI on PA diameters was not considered. In the present study, the increased PA diameters with aging appeared to be primarily explained by a concomitant increase of BMI and only secondarily by the degree of dorsal kyphosis. In both sexes, age appeared to exert a positive influence on HDI independently of BMI and of stature. Everything else being kept constant, loss of lung elastic recoil 25 could contribute to this by changing the balance between gravitational and lung retractile forces, thereby modifying the relationship between BMI and HDI. The effect of BMI on HDI could explain why TLC does not increase with age. Findings in Disease COPD: Thoracic dimensions have been reported before in COPD patients All studies concur to show a descent and flattening of the diaphragm dome in severe COPD with increased TLC. There is controversy, however, as to whether rib cage expansion also contributes. Two studies 26,27 in upright subjects reported no change in rib cage dimensions, whereas one study 28 in supine subjects reported greater PA diameters at all lung volumes in COPD patients. There are two findings in our study that are relevant to this discussion. First, we found a similar volume distribution in male and female COPD group patients. Therefore, the fact that the majority of subjects studied by Cassart et al 28 were female whereas those of Kilburn and Asmundsson 26 and Walsh et al 27 were all male cannot account for the different results. Secondly, we find, like Kilburn and Asmundsson 26 and Walsh et al, 27 that when compared to sex-, age-, and height-matched normal subjects, the rib cage of these patients remains of normal size and shape, the 40% increase in TLC in our study being entirely accounted for by a caudal displacement of the diaphragm. When BMI was included in the prediction equation, however, both the axial and the radial thoracic dimensions were found to be greater than predicted. We interpret these findings as showing that our COPD group patients had greater rib cage dimensions than normal subjects given their BMI, but because their BMI was low, their rib cage remained of normal size and their diaphragm was displaced more caudally. Thus, an increased PA diameter in COPD group patients does not appear to be a specific marker of emphysema unless this is corrected for BMI. This postulated effect of BMI on diaphragm position and rib cage size may help explain the finding of Rochester and Braun 29 of a shorter diaphragm in COPD patients who were markedly underweight, as compared to normal subjects or normal-weight COPD patients at the same lung volume. Our findings, however, do not help explain the findings of Cassart et al. 28 Although weight loss appeared sufficient to explain the finding of a normal rib cage size in COPD, a mechanical limitation to rib cage expansion cannot be ruled out. The bucket handle motion of ribs is close to maximal in normal subjects at TLC, 30 such that no further expansion of LA rib cage diameters may be expected to occur. In fact, LA diameters CHEST / 119 / 2/ FEBRUARY,

9 were never found to be greater than normal in these patients The PA angulation of ribs relative to the transverse plane (ie, the pump handle angle) is substantial at TLC both in normal subjects 30 and patients with COPD. 27 Thus, substantial increases in PA rib cage dimension should be possible. In fact, it is possible for normal subjects to further expand their rib cage at TLC by compressing the abdomen. 31 Similar determinations, however, do not appear to have been made in COPD patients. Because the diaphragm of these patients is flat, the inwardly directed force would tend to limit rib cage expansion at active TLC. CF: Findings in male CF group patients were comparable to those of COPD group patients. By contrast, female CF group patients showed no increase in HDI, the greater TLC in these female patients being entirely accounted for by greater PA diameters. This increase was such as to abolish the gender-related difference in craniocaudal gradient in PA diameter. The reason for such a marked and systematic difference is unclear. Earlier studies 32 were restricted to the period of growth. Because BMI was low in male and female subjects, a caudal displacement of the diaphragm would have been expected in both. Female CF group patients, as a group, had less severe airways obstruction than male patients. However, the gender differences in chest wall dimensions persisted when corrected for FEV 1. On teleologic grounds, a preferential expansion of the rib cage in female CF group patients could be anticipated on the basis of the smaller rib cage size seen in normal female subjects, which could thus be easier to displace than the already lower diaphragm. This reasoning, however, should also be applicable to female COPD patients, which is not what was found. Whatever the mechanism, however, our finding is relevant to the radiographic evaluation of pulmonary hyperinflation in CF, as different criteria would be required in male and female patients. It is also of interest to note that PA-T9 T5 increased in CF, although the reason for this unknown. This does not appear related to the presence of airways obstruction or hyperinflation, as similar results can be found in RLD group patients (see below). RLD: In RLD group patients, chest wall volume was reduced as shown by the smaller rib cage PA diameters and HDI. However, LA rib cage diameters were normal, thus showing a substantial deformation of the rib cage cross-section that assumed a more elliptical shape. As shown by Agostoni and Mognoni, 4 the rib cage assumes a more elliptical shape during static inspiratory efforts and a more circular shape during static expiratory efforts. They reasoned that respiratory muscle forces act principally on LA rib cage dimension, whereas the rib cage PA dimension is largely determined passively by pleural pressure changes that are negative during inspiratory efforts but positive during expiratory efforts. This mechanism could also explain the elliptical shape of the rib cage in our RLD group patients at full inflation. Indeed, these patients are known to generate markedly negative pleural pressures at TLC, owing to an increased strength of their inspiratory muscles as TLC decreases. 33 In contrast to the other groups studied, in RLD group patients, PA diameters were inversely related to HDI. Surprisingly, this appeared related to the smoking status, as ex-smokers and smokers had greater rib cage PA diameters than nonsmokers. Ex-smokers and smokers also tended to have lower values of HDI. The different chest wall configuration of the smokers could not be related to differences in lung function, and none presented with radiographic evidence of emphysema. Conceivably, the smoking habit might have led these patients to adopt a different respiratory muscle recruitment strategy. It is of interest that, like the CF group patients, PA-T9 T5 tended to increase in RLD group patients and particularly in female patients in whom the values of this gradient approached those found in male patients. In this connection, it is also worth mentioning that upper-rib cage motion was found to be impaired during quiet breathing in a majority of patients with fibrosing alveolitis. 34 Study Limitations Because of the retrospective nature of our study, the ways in which the lung function tests were performed and the chest radiographs were obtained were not under our control. However, these procedures are highly standardized and are unlikely to have varied systematically in the different groups or as a function of the variables considered by this analysis. The comparisons should thus be valid. Furthermore, the Vth estimated from chest radiographs was highly correlated with the plethysmographic TLC (Fig 3), thus showing that the measured thoracic dimensions did reflect the variations in the volume of lungs among individuals. Our approach has the advantage of being more representative of the thoracic dimensions seen clinically by radiologists, and ensures that the changes in thoracic dimensions that we describe can readily be observed using a routine chest radiograph protocol. All measurements were obtained by a single observer but this seemed justified, as none required substantial interpretation, and measurements reported by two independent observers agreed within 384 Clinical Investigations

10 1%. The relationships between measured thoracic dimensions and anthropologic variables were not found to be different between the different groups, a result that must be taken with some reservations, as the number of patients in the study groups was smaller than in the normal group. Our analysis and the conclusions that can be derived from it are limited to the condition of full active lung inflation. A comparison with other lung volumes (particularly FRC, which is determined passively) would have been helpful in discriminating between passive and active muscular forces affecting thoracic dimensions. Such studies are planned. Other Implications of the Findings and Conclusion Rib cage dimensions were highly correlated with the selected anthropologic variables, with coefficients of determination comparable to those found for FVC and FEV Prediction equations for rib cage dimensions could thus be useful clinically, not only in lung diseases but in diseases affecting the rib cage or the respiratory muscles. The axial position of the diaphragm was less well correlated with the selected anthropologic variables. For this axial dimension, uncontrolled factors such as the shape, strength, and nervous activation of the diaphragm probably play a more important role. Nevertheless, because the coefficient of variation of HDIave was comparable to that of PAave, the analysis should be equally sensitive in detecting abnormalities in diaphragm position. Because BMI is a potential target for therapeutic interventions, the comparison of model I with model II predictions may also allow an evaluation of the potential outcome in terms of rib cage dimensions and diaphragm position. For example, the low and inefficient flat diaphragm of patients with severe emphysema is thought to contribute to the high energy cost of breathing of these patients. 35 Our study shows that this can be partly explained by weight loss, which is potentially reversible by appropriate nutritional supplementation strategies. Although well correlated with anthropologic variables, variations in thoracic dimensions all entailed substantial changes in chest wall shape. Depending on the relative ease with which the shape of the lungs can conform to that of the chest wall, the observed variations in chest wall shape could affect the local pleural surface pressure and the topographic distribution of ventilation. 3 Furthermore, as was shown recently in rabbits, nonisotropic lung and chest wall expansion not only affects the distribution of pleural surface pressure but results in an apparent decrease in overall lung compliance and increased work of breathing. 36 Whether the differences in chest wall shape documented in this study also affect the distribution of pleural surface pressure is a question for future investigations. In summary, we have shown that the major thoracic dimensions measured on standard chest radiographs can be predicted with sufficient accuracy using simple anthropologic variables so as to be useful clinically. We showed the following: (1) the small lung volume of normal female subjects is entirely accounted for by a smaller rib cage size; (2) in the presence of weight loss or low BMI, the rib cage of markedly hyperinflated COPD patients can remain of normal size; (3) the distribution of pulmonary hyperinflation at TLC in CF patients is gender specific, being via the diaphragm-abdomen pathway in male patients but via the rib cage pathway in female patients; and (4) the rib cage cross-section is deformed to a more elliptical shape in RLD patients at full inflation, a deformation that is less severe when the cranial displacement of the diaphragm is more severe. Some implications of these findings are discussed. ACKNOWLEDGMENT: The authors thank the personnel of the radiology department, for their assistance in retrieving the radiographs, and Dr. M.C. Guertin for assistance with the statistical analysis. References 1 Howatt WF, DeMuth GR. The growth of lung function: II. Configuration of the chest. Pediatrics 1965; 35: Mead J, Loring SH, Smith JC. Volume displacements of the chest wall and their mechanical significance. In: Roussos C, ed. The thorax: part A. Physiology, lung biology in health and disease series. Vol 85. New York, NY: Marcel Dekker, 1995; Agostoni E. Mechanics of the pleural space. In: Macklem PT, Mead J, eds. Handbook of physiology, section 3: The respiratory system: mechanics of breathing part 2. Bethesda, MD: American Physiological Society, 1986; Agostoni E, Mognoni P. Deformation of the chest wall during breathing efforts. J Appl Physiol 1966; 21: D angelo E. Cranio-caudal rib cage distortion with increasing inspiratory airflow in man. Respir Physiol 1981; 44: Mellisinos CG, Goldman M, Bruce E, et al. Chest wall shape during forced expiratory maneuvers. J Appl Physiol 1981; 50: Paiva M, Estenne M, Engel LA. Lung volume, chest wall configuration, and pattern of breathing in microgravity. J Appl Physiol 1989; 67: Liu S, Wilson TA, Schreiner K. Gravitational forces on the chest wall. J Appl Physiol 1991; 70: Donnelly PM, Yang TS, Peat JK, et al. What factors explain racial differences in lung volumes? Eur Respir J 1991; 4: Takahashi E, Atsumi H. Age differences in thoracic form as indicated by thoracic index. Human Biol 1955; 27: Lennon EA, Simon G. The height of the diaphragm in the chest radiograph of normal adults. Br J Radiol 1965; 38: American Thoracic Society. Lung function testing: selection CHEST / 119 / 2/ FEBRUARY,

11 of reference values and interpretative strategies. Am Rev Respir Dis 1991; 144: Knudson RJ, Lebowitz MD, Holberg CJ, et al. Changes in the normal maximal expiratory flow-volume curve with growth and aging. Am Rev Respir Dis 1983; 127: Crapo RO, Morris AH, Clayton PD, et al. Lung volume in healthy nonsmoking adults. Bull Eur Physiopathol Respir 1982; 18: Ayers LN, Ginsberg ML, Fein J, et al. Diffusing capacity, specific diffusing capacity and interpretation of diffusion defects. West J Med 1975; 123: DeMuth GR, Howatt WF, Hill BM. The growth of lung function: I. Lung volumes. Pediatrics 1965; 35: Schwartz J, Katz SA, Fegley RW, et al. Sex and race differences in the development of lung function. Am Rev Respir Dis 1988; 138: Grivas TB, Burwell RG, Purdue M, et al. A segmental analysis of thoracic shape in chest radiographs of children: changes related to spinal level, age, sex, side and significance for lung growth and scoliosis. J Anat 1991; 178: McClaren SR, Harms CA, Pegelow DF, et al. Smaller lungs in women affect exercise hyperpnea. J Appl Physiol 1998; 84: Harms CA, McClaran SR, Nickele GA, et al. Exerciseinduced arterial hypoxaemia in healthy young women. J Physiol (Lond) 1998; 507: Rochester DF. Obesity and abdominal distension. In: Roussos C, ed. The thorax: part C. Lung biology in health and disease series. Vol 85. New York, NY: Marcel Dekker, 1995; Ray CS, Sue DY, Bray G, et al. Effects of obesity on respiratory function. Am Rev Respir Dis 1983; 128: Sampson MG, Grassino AE. Load compensation in obese patients during quiet tidal breathing. J Appl Physiol 1983; 55: Milne JS, Lauder IJ. Age effects in kyphosis and lordosis in adults. Ann Hum Biol 1974; 1: Knudson RJ, Clark DF, Kennedy TC, et al. Effect of aging alone on mechanical properties of the normal adult human lung. J Appl Physiol 1977; 43: Kilburn KH, Asmundsson T. Anterior chest diameter in emphysema. Arch Intern Med 1969; 123: Walsh JM, Webber CL Jr, Fahey PH, et al. Structural change of the thorax in chronic obstructive pulmonary disease. J Appl Physiol 1992; 72: Cassart M, Genevois PA, Estenne M. Rib cage dimensions in hyperinflated patients with severe chronic obstructive pulmonary disease. Am J Respir Crit Care Med 1996; 154: Rochester DF, Braun NMT. Determinants of maximal inspiratory pressure in chronic obstructive pulmonary disease. Am Rev Respir Dis 1985; 132: Wilson TA, Rehder K, Krayer S, et al. Geometry and respiratory displacement of human ribs. J Appl Physiol 1987; 62: Konno K, Mead J. Measurement of the separate volume changes of the rib cage and abdomen during breathing. J Appl Physiol 1967; 22: Mearns MB, Simon G. Patterns of lung and heart growth as determined from serial radiographs of 76 children with cystic fibrosis. Thorax 1973; 28: De Troyer A, Yernault JC. Inspiratory muscle force in normal subjects and patients with interstitial lung disease. Thorax 1980; 35: Brennan NJ, Morris AJR, Green M. Thoracoabdominal mechanics during tidal breathing in normal subjects and in emphysema and fibrosing alveolitis. Thorax 1983; 38: Pitcher WD, Cunningham HS. Oxygen cost of increasing tidal volume and diaphragm flattening in obstructive pulmonary disease. J Appl Physiol 1993; 74: D Angelo E, Giglio R, Lafontaine E, et al. Influence of the abdomen on respiratory mechanics in supine rabbits. Respir Physiol 1999; 115: Clinical Investigations