Rationale Severe pulmonary hypertension (PH) is very uncommon in COPD, and a distinct phenotype has been hypothesised. We aimed to evaluate whether CT can help to recognise this condition non-invasively by measuring small pulmonary vessels.
Material and methods Patients with COPD who underwent pulmonary function tests, unenhanced CT of the chest and right heart catheterisation (RHC) during a period of stability were included in the study. From 105 included patients, 20 patients with COPD with severe PH (mean pulmonary arterial pressure, mPAP>35 mm Hg) were compared with 20 FEV1-matched and age-matched patients with COPD with mild or without PH (mPAP<35 mm Hg). The percentage of total cross-sectional area of vessels less than 5 mm2 normalised by lung area (%CSA<5) and 5–10 mm2 (%CSA5–10), the mean number of cross-sectioned vessels (CSNs) and bronchial wall thickness (WT) were measured on CT examination and compared between groups. Paw scores combining PaO2 measurement and CT parameters best correlated with mPAP were compared by receiver operating characteristic analysis to predict severe PH in COPD.
Results Patients with severe PH COPD had higher %CSA and CSN values than those of patients with COPD without severe PH. Using multiple regression analysis, %CSA<5 and WT were the best predictors of mPAP in patients with and without severe PH, respectively. A score combining %CSA<5, PaO2 and WT best predicted severe PH in patients with COPD.
Conclusions CT measurements of small vessels support a distinct vessel-related phenotype in patients with COPD with severe PH, and combined with WT and PaO2 parameters in the paw score, which may offer a non-invasive tool to select patients for RHC.
- Imaging/CT MRI etc
- COPD Pathology
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
What is the key question?
In a population of patients with COPD, are in vivo morphometric changes of the vascular bed observed by CT able to predict severe pulmonary hypertension?
What is the bottom line?
To date, no study has reported in vivo assessment of pulmonary vessels in a very uncommon COPD phenotype associated with severe pulmonary hypertension, and a non-invasive tool for better characterising this population might be useful.
Why read on?
CT measurements of small vessels support a distinct vessel-related phenotype in patients with COPD with severe pulmonary hypertension, and combined with wall thickness and PaO2 parameters in the paw score, which may offer a non-invasive tool to select patients for right heart catheterisation or for non-invasive follow-up in longitudinal study.
The development of pulmonary hypertension (PH) in COPD is associated with poor prognosis on mortality and quality of life.1 An increase of 10 mm Hg in mean pulmonary arterial pressure (mPAP) increases mortality by more than fourfold.2 Most of the time, PH in COPD remains mild to moderate. However, a small proportion of patients develop severe PH, which has been defined by mPAP higher than 35 mm Hg.3 Therefore, severe PH is very uncommon in patients with COPD,3–6 ranging from 5% to 13%,4 ,7 and prone to display severe hypoxaemia, hypocapnia and decrease in carbon oxide diffusing capacity.5 Though PH in COPD is classified by the WHO in group 33 ,8 ,9 (ie, PH associated with lung disease and/or hypoxia), the development of severe PH in some patients with COPD has been supposed to involve a distinct phenotype related to vessels rather than airways.10 CT of the chest has long been used to study non-invasively chronic obstructive diseases. Quantification of bronchi wall areas (WAs) and voxel attenuation of parenchyma has been demonstrated to correlate with airway remodelling and emphysema, respectively.11 ,12 Very recently, we demonstrated a close relationship between pulmonary arterial pressure and bronchial wall thickening, but the vast majority of patients did not present severe PH.13 Although there is a paramount literature around quantitative CT of airways in COPD, few data have been reported to quantitative CT of small pulmonary vessels.14–17 A single study aimed at evaluating CT assessment of small vessels in COPD subjects with PH and severe emphysema.16 In that study, the degree of mPAP did not exceed the threshold of 40 mm Hg, so data from CT examinations that could support a distinct vessel alteration among COPD subjects with severe PH are still lacking. We aimed at evaluating whether CT can provide in vivo evidence about changes of the vascular bed in a population of patients with COPD with severe PH, in order to select patients for right heart catheterisation. For this purpose, we compared the pulmonary function tests (PFTs), CT parameters and mPAP by right heart catheterisation between COPD populations with and without severe PH. Then, we finally aimed to predict the presence of severe PH in patients with COPD using combined score. Preliminary results of this study have been presented in the form of abstracts.18–20
Material and methods
Patients with COPD were referred between January 2008 and December 2014 to our institution, a tertiary medical centre for complete examination of PH, before initiation of any treatment. All patients underwent within 1 week: medical questioning, physical examination, 6-minute walk tests (6-MWTs), arterial blood gases, blood tests (including C-reactive protein (CRP), antinuclear antibodies and HIV serology), PFT, transthoracic echocardiography, ventilation/perfusion scintigraphy (V/Q scan), right heart catheterisation (RHC), and unenhanced CT within a minimal period of 1 month of disease stability.
Body plethysmography (BodyBox, Medisoft, Belgium) was used to perform PFT. We chose reference values from the American Thoracic Society21 and the European Respiratory Society guidelines.22
CT scans were performed on a Somatom Sensation Definition 64 (Siemens, Erlangen, Germany) at full inspiration. Quantitative analysis was performed by using dedicated and validated softwares.23 ,24 Automatic quantification of bronchi WA, lumen area, WA per cent (WA%) and wall thickness (WT) was obtained on orthogonal bronchial cross sections by using the Laplacian-of-Gaussian algorithm and homemade software.24 ,25 Automatic quantification of emphysema was assessed with Myrian software (Intrasense, Montpellier, France) using low attenuation area per cent (LAA%), as previously described.13
Automated measurement of small vessel area from CT images was obtained by using the Image J software V.1.40g (a public domain Java image programme available at http://rsb.info.nih.gov/ij/), method detailed elsewhere.14–16 The following measurements were obtained: the cross-sectional area of small pulmonary vessel less than 5 mm2 (%CSA<5), and between 5 and 10 mm2 (%CSA5-10), the mean number of cross-sectioned vessels CSN<5 and CSN5–10 normalised by the corresponding lung section area at each CT slice (figure 1). Manual measurements of large vessels were performed on multiplanar reconstruction strictly orthogonal to the main axis of the pulmonary arterial troncus (AP) and the ascending aorta (AO).
NCSS software (NCSS 2001, Kaysville, Utah, USA) was used to assess statistical analyses. p Values <0.05 were considered significant. Results were expressed as mean with SD. Comparisons were performed by using t tests. Parameters that were not normally distributed were expressed as median with IQRs and analysed by Mann–Whitney tests. Categorical variables were tested with Fisher's exact tests. Univariate correlations were assessed using Pearson product–moment coefficients. Identification of the strength of the association between mPAP and other variables was assessed using forward stepwise multiple regression analyses. To predict severe PH, we built four scores (ie, paw scores), combining two to three variables, best correlated to mPAP, the higher value indicating a higher risk of severe PH (table 1). Scores were compared by using areas under the curve (AUCs) of receiver operating characteristic (ROC) curves.
For details about PFT, RHC, CT protocols or echocardiography and statistical analysis, see the online supplemental methods section.
From a total of 198 selected patients, 105 patients with COPD were included in the study, with no other condition susceptible to explain PH (see figure 2 and table 2). Among the 105 patients with COPD, 20 patients demonstrated severe PH, as defined by mPAP at RHC superior to 35 mm Hg, as the last world PH symposium proposed this cut-off.3 These severe patients with PH were FEV1-matched and age-matched to 20 other patients with COPD without severe PH (mPAP <35 mm Hg) (table 2). The 65 remaining patients unable to be matched were presented separately (see online supplementary table S1).
Demographic characteristics, pulmonary functions and clinical, hemodynamic and biological data of both study populations are shown in table 2. There was no difference in demographic characteristics, CRP levels and lung function parameters between patients with COPD with and without severe PH, except for transfert lung capacity of carbon monoxide (TLCO). The patients with COPD with severe PH had significantly lower values of both TLCO and 6-MWT, deeper hypoxia and higher brain natriuretic peptide (BNP) levels. There was no left ventricular dysfunction in both groups (data not shown). All patients with COPD carried out RHC. The mPAP ranged from 35 to 58 mm Hg and from 13 to 34 mm Hg in patients with severe PH and without severe PH, respectively. Moreover, a precapillary PH was demonstrated in all patients with COPD with PH, as assessed by pulmonary capillary wedge pressure (PCWP) values less than 15 mm Hg and gradient values superior to 10 mm Hg.
Relationship between mPAP and quantitative CT parameters
Morphological CT parameters were assessed in both groups of patients with COPD (table 3). At alveolar level, emphysema (LAA%) was not significantly different between the two groups, and there was no significant correlation between emphysema and mPAP (see online supplementary table S2).
At the bronchial level, no significant difference was observed between patients with COPD with and without severe PH (table 3). However, for both groups, bronchial WT was positively correlated with mPAP (see online supplementary table S2).
At the vascular level, the ratio of the main pulmonary artery to the ascending aorta diameter (AP/AO) was greater in patients with COPD with severe PH than in patients without severe PH (table 3). AP/AO was not correlated to mPAP in patients with COPD with severe PH (see online supplementary table S2). However, a positive correlation was found between mPAP and AP/AO in the group of patients with COPD without severe PH (see online supplementary table S2, figure 3A, C). All parameters related to small vessels (ie, %CSA<5, %CSA5–10, CSN<5 and CSN5–10) were greater in patients with COPD with severe PH than in patients without severe PH (table 3). In patients with severe PH, a positive correlation was found between mPAP and %CSA<5. Conversely, in patients without severe PH, mPAP was negatively correlated to %CSA<5 (figure 2B, D).
Factors influencing mPAP
Regarding variables from PFT, arterial blood gases, biology, CT and 6-MWT, in COPD without severe PH, significant univariate correlations were found between mPAP and various variables including PaO2, TLCO%, BNP, AP/AO ratio, bronchial WT, CSAs and 6-MWT (see online supplementary table S2). Conversely, in patients with COPD with severe PH, significant correlations were found only between mPAP and both WT and CSAs (see online supplementary table S2).
Significant variables that were correlated with mPAP and not cross-correlated (see online supplementary tables S3 and S4) were entered in stepwise multiple regression analyses in order to find the best model fitting mPAP within the two groups of patients with COPD (table 4). In patients with severe PH, the best model associated WT, %CSA<5 and PaO2 and explained 65% of the mPAP variations. Among these variables, mean %CSA<5 explained 36% of the mPAP variations alone, whereas WT and PaO2 explained only 29% and 3%, respectively. In patients with COPD without severe PH, the best model explained 69% of mPAP variations, WT was the main factor explaining mPAP (26%), whereas PaO2 and %CSA<5 explaining 13% and 9% of mPAP variations, respectively.
Factors predicting severe PH
Since %CSA<5 was correlated to mPAP in opposite ways in patients with and without severe PH, it seems hard to predict the presence of severe PH in patients with COPD using %CSA<5 only. We thus built four combined paw scores using two to three variables best correlated to mPAP (ie, %CSA<5, WT, PaO2) and assessed their corresponding ROC curves for the 40 matched patients and for all the 105 included patients (see online supplementary figures S1 and S2). Characteristics of these ROC curves for each combination are reported in table 5. For the 40 matched patients, AUCs from each combination including %CSA<5 were significant and greater than 0.5. The best paw score associated the three variables %CSA<5, PaO2 and WT and allowed a high negative predictive value (NPV) of 95% with a sensitivity and specificity of 75% and 70%, respectively. These results were confirmed with all the 105 patients initially included in the study: AUCs from each combination were also significant and greater than 0.5, and the paw score that associates the three variables allowed even greater NPV and specificity, and the sensitivity is unchanged.
Taken together, this study demonstrated that mPAP can be predicted differently in patients with COPD with and without severe PH by using CT. In patients with severe PH, increased small pulmonary vessels area (%CSA<5) is the factor mostly associated with mPAP elevation, whereas it is increased WT in patients without severe PH. The %CSA<5, WT and PaO2were the three best correlated parameters with mPAP. These three variables could be combined to build a paw score helpful for predicting the absence of severe PH in patients with COPD.
Identifying clinically meaningful COPD phenotypes is of critical importance for patient care,26 ,27 since COPD is a complex and heterogeneous disease that the sole FEV1/FVC cut-off of 70% cannot describe adequately.28 To better characterise patients with COPD on the basis of attributes that differ between individuals and relate with clinically relevant outcomes, CT has long been shown to provide invaluable clues by allowing non-invasive and quantitative assessment of structural alterations from bronchi and lung parenchyma.29–31 However, few studies have focused on vessel modifications in COPD subjects,14–17 ,32 though there are compelling evidence that COPD is also mediated by a pulmonary vascular disease.33–35 It has been shown that the toxicity of tobacco smoke affects bronchi and parenchyma, but can also directly alter vessels in both animals and humans.36 We hypothesised that CT measurement of intrapulmonary vessels, especially the more distal ones, could be a tool to characterise COPD subjects in addition to standard measurements of emphysema and bronchial thickness.13
In the present study, we focused our attention towards a population of patients with COPD suffering from severe PH. We paid a special attention to select patients without severe PH as close as possible to those with severe PH in terms of age and FEV1 to limit potential bias on demographic and functional differences. Thus, the clinical features of patients with severe PH shared close similarities to those without severe PH. For instance, the tobacco smoke consumption was not different between groups, which suggest a different susceptibility against tobacco toxicity. However, severe PH in COPD is occasional, as observed previously,4 ,5 and there is, up to now, no clear evidence to explain its aetiology.10 In the study of Chaouat et al,5 none of the PFT parameters, emphysema or hypoxaemia were found to be predictors of PH in an univariate analysis of mPAP in a population of 11 patients with COPD with severe PH and no other condition susceptible to explain PH. Thus, it was supposed that severe PH in COPD was more likely explained by a vascular alteration rather than an airway disease. In patients with severe PH from the present study, %CSAs and WT were found to be predictors of mPAP, while PaO2, FEV1%, TLCO% and LAA% were not. The main morphological parameter able to explain mPAP variation in this population was %CSA<5, suggesting a striking difference in the pathophysiology of increased mPAP between the two populations of COPD. By applying a previously reported method for quantifying small vessels on CT examinations,16 we have found that both %CSA<5 and %CSA5–10 were higher in patients with COPD with severe PH compared with those without severe PH. In this latter population, at univariate analysis, both PaO2 and TLCO% were found to be correlated with mPAP, a finding consistent with the central role played by hypoxaemia in the development of PH.5 ,37 There was a lack of correlation between emphysema and PH, which is in line with previous studies.13 ,16 ,37 In addition, we recently reported that the main parameter able to explain mPAP in patients with COPD without severe PH was WT,13 which was confirmed in this study.
Correlations between %CSAs and mPAP found herein deserve further comments. In the literature evaluation of %CSAs values in COPD subjects with PH was significantly negatively correlated, as in our population of COPD without severe PH. This result was ascribed to reduced distensibility of small vessels, hypoxic vasoconstriction and vessel destruction secondary to emphysema.16 ,38 However, this study was performed based on a population of COPD subjects with severe emphysema, as assessed by a visual of percentage of emphysema on CT scans superior to 75%,7 ,16 and corresponding to total automatic quantification (LAA%) of 25%, and in which the degree of mPAP elevation did not exceed the threshold of severe PH. This difference between subjective visual grading and objective automatic quantification with LAA% of emphysema is already described in the literature.11 In our population with severe PH by contrast, the correlation between %CSAs and mPAP was positive. Thus, our results are complementary from the previous report of Matsuoka et al,16 and bring additional insight into the ability of CT to discriminate COPD phenotypes on a small vessel-based analysis. Moreover, a marker for PH in COPD (ie, AP/AO)39 was not correlated to mPAP in our COPD population with severe PH.
In this COPD population, the absence of severe PH could be well predicted by the paw score (ie, combination of PaO2, WT and %CSA<5), as demonstrated by the ROC curve analysis. The distal vascular parameter %CSA<5 was needed, whatever the combination of parameters tested to significantly distinguish between severe PH and others, emphasising the role of distal vessels changes in severe PH of COPD. Interestingly, the best paw score allowed us to discriminate, non-invasively, matched patients with COPD with severe PH from patients with COPD without severe PH. From the perspective of clinical conditions, patients with paw score higher than 5 should be selected to perform RHC. Indeed, whereas the NPV was very high (95%–98%), the positive predictive value was low (12%–27%). We confirmed the clinical interest of this paw score in a larger group of patients with COPD (n=105). However, these results could be interestingly confirmed in a new prospective cohort. Since the prevalence of severe PH in COPD is low,3–5 ,7 it is thus necessary to better characterise these patients whom reports are still limited in number of subjects.4 ,5 ,40
Several limitations can be discussed in this study. The study was observational and retrospective. It was not possible to draw causal relationship between increased size and number of small vessels and severe degree of PH in COPD. However, mild-to-moderate PH in COPD is commonly thought to be caused by the destruction of small pulmonary vessels as a consequence of emphysema, or secondary to vasoconstriction in response to hypoxia. To support these hypotheses, mild-to-moderate PH in COPD has been consistently reported to be related with a reduction of small vessel areas16 or volume38 at CT examination. Our study is the first to report in vivo data on the distal vascular bed in patients with COPD with severe PH. Moreover, this study indicates that, in this subpopulation of COPD, development of PH is not related to a decrease but to an increase in pulmonary vasculature area. However, veins and arteries were not distinguished by our methods of small vessel measurements. Nevertheless, in our population, pulmonary venous pressures (PCWP) at RHC were not different between groups, as well as the left heart functions at echocardiography. PH was always confirmed to be precapillary at RHC, and %CSA measurements correlated with all values of pulmonary arterial pressures (systolic PAP, diastolic PAP, mean PAP), but not with pulmonary venous pressures (PCWP). Thus, we are reliably confident that postcapillary venous pressures cannot account for %CSA variations. In addition, %CSA measurements allowed a global quantification of vessel areas, irrespective of wall or lumen changes. We cannot speculate about the balance between vessel wall thickening and lumen narrowing in the pathophysiology of severe PH. Finally, patients with COPD without severe PH were a mix of patients with and without PH, since the aim of the study was to discriminate severe PH among the whole population of COPD.
CT measurements of small vessels support a distinct vessel-related phenotype in patients with COPD with severe PH, and combined with WT and PaO2 parameters in the paw score, which may offer a non-invasive tool to select patients for RHC or for non-invasive follow-up in longitudinal study.
FC and GD equal contribution (co-first author)
PB and FL equal contribution (co-last author)
Contributors Conception and design: FC, GD, PB and FL; analysis and interpretation: FC, GD, PB and FL; drafting the manuscript for important intellectual content: FC, GD, PB and FL; revising the manuscript for important intellectual content: FC, GD, PB and FL; final approval of the manuscript: FC, GD, CD, EB, VF-M, P-OG, MM, FB, FP, RM, PB and FL.
Funding This study was achieved within the context of the Laboratory of Excellence TRAIL ANR-10-LABX-57.
Competing interests PB reports grants from Fondation du Souffle—Fonds de dotation Recherche en Santé Respiratoire, personal fees and non-financial support from Novartis, grants and personal fees from Pierre Fabre, personal fees and non-financial support from Chiesi, personal fees and non-financial support from Boehringer Ingelheim, personal fees and non-financial support from Takeda, personal fees and non-financial support from AstraZeneca, outside the submitted work. In addition, PB has a patent new compositions and methods of treating and/or preventing COPD. Patent submitted 28 Jan 2015 (EP N°15152886.6) pending.P-OG reports personal fees and non-financial support from Novartis, personal fees from Pierre Fabre, non-financial support from Chiesi, non-financial support from Takeda, personal fees and non-financial support from Boehringer Ingelheim, outside the submitted work.
Patient consent Obtained.
Ethics approval Our institutional review board and the local ethic committee.
Provenance and peer review Not commissioned; externally peer reviewed.