To the Editor:

Lung function testing is the gold standard for diagnosing and classifying chronic obstructive pulmonary disease (COPD), although it may not capture the heterogeneity of the disease completely (1). Different quantitative methods based on computed tomography (CT) of the lungs have been proposed to untangle the different pathological mechanisms involved (2).

Recently, a new CT-based biomarker was reported that provides visualization and quantification of COPD phenotypes in patients with COPD (3). This so-called parametric response mapping (PRM) is a voxel-based image-analysis technique for evaluating both inspiratory and expiratory CT scans. It provides a color map, which represents normal lung tissue, small airway disease, or emphysema, and gives insight into the extent and localization of disease.

Thus far, only one study has described the association of PRM with lung function, showing moderate to good correlations (3). However, whether PRM adds additional value to current CT-based biomarkers in diagnosing COPD is unknown. We hypothesized that PRM could have added value because of its ability to distinguish between air trapping related to emphysema and air trapping related to small airway disease. The aim of this study was to evaluate the added diagnostic value of PRM in evaluating COPD based on CT.

We analyzed participants taking part in the Dutch and Belgian Lung Cancer Screening Trial (NELSON). Only participants from the University Medical Center Utrecht were included because they underwent both inspiratory and expiratory CT. Inclusion and exclusion criteria have been reported before (4). In short, all participants were current or former smokers with a smoking history of at least 16.5 pack-years and age between 50 and 75 years.

All participants underwent prebronchodilator pulmonary function tests according to the European Respiratory Society/American Thoracic Society guidelines (5). FEV₁ and FVC were recorded. COPD was defined according to the Global Initiative for Chronic Obstructive Lung Disease (GOLD) criteria (6).

Low-dose volumetric CT scans were made with 16 × 0.75 mm collimation (Brilliance 16P; Philips Medical Systems, Cleveland, OH). Emphysema was assessed by the 15th percentile (Perc15) technique (7). Perc15 represents the Hounsfield unit (HU) point below which 15% of the voxels are distributed. The closer Perc15 is to −1,000, the more emphysema is present. Airway wall thickness was calculated by the square root of wall area for a theoretical airway with an internal perimeter of 10 mm (Pi10) (8–10). Air trapping was assessed using the mean expiratory to inspiratory ratio of the mean lung density (E/I-ratio_MLD) (11). PRM measurements were implemented as a joint histogram analysis between inspiratory and expiratory scans after image registration (12, 13). Lung segmentation was visually checked. Applying the standard thresholds of −950 HU for emphysema in the inspiratory acquisition and −856 HU in the expiratory acquisition to define air trapping, PRM is summarized in three quantities: all voxels with HU below −950 in the inspiratory CT and below −856 in the expiratory CT represent emphysema (PRM^Emph), all voxels above −950 in the inspiratory CT and below −856 in the expiratory CT represent air trapping (PRM^fSAD), and all voxels above both thresholds in both scans represent normal lung (PRM^Norm) (see Figure 1). Measurements were corrected for lung volume.

Figure 1. Parametric response mapping (PRM) color map. Inspiratory computed tomography scans with PRM color maps from three different participants, representing normal lung tissue (green), small airway disease (yellow), and emphysema (red). (A) Fifty-eight-year-old man without chronic obstructive pulmonary disease (COPD). PRM values were 0.02% PRM^Emph and 4.02% PRM^fSAD. (B) Fifty-nine-year-old man with COPD, Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage 3. The color map shows that airway obstruction is mainly a result of small airway disease, with PRM values of 4.2% PRM^Emph and 37.3% PRM^fSAD. (C) Sixty-one-year-old man with COPD GOLD stage 3. As this PRM map shows, his airway obstruction is the result of both emphysema and small airway disease. PRM measures were 19.4% PRM^Emph and 29.4% PRM^fSAD. All voxels with Hounsfield units below −950 in the inspiratory CT and below −856 in the expiratory CT represent emphysema (PRM^Emph), all voxels above −950 in the inspiratory CT and below −856 in the expiratory CT represent air trapping (PRM^fSAD), and all voxels above both thresholds in both scans represent normal lung (PRM^Norm).

Download Figure | Download Powerpoint

[More] [Minimize]

A multivariate logistic regression model was developed with COPD status (absent or present) as outcome. Prior to this analysis, PRM measurements were log transformed to obtain normally distributed data. The baseline model included age, body mass index, smoking status, and pack-years. In subsequent models, Pi10, Perc15, E/I-ratio_MLD, and PRM measurements were added one at a time. Models were compared by the log-likelihood ratios test. The ability to discriminate between participants with and without COPD was assessed by calculating the area under the receiver operating characteristic curve (i.e., C-statistic). Reclassification of participants’ COPD status was calculated. Additional gas transfer analyses were performed by using diffusing capacity of the lung for carbon monoxide and Kco values in a subgroup of our cohort to gain insight into the possible clinical relevance of PRM.

A total of 1,140 male participants were included; 87 participants were excluded because of lung segmentation errors, and 6 participants were excluded because of negative inspiratory–expiratory differences, resulting in 1,047 eligible participants for analysis. Excluded participants (n = 93) had a mean (SD) age of 64.7 (6.3) years, and included participants (n = 1,047) had a mean (SD) age of 62.4 (5.1) years (P < 0.001). Median (25–75th percentile) pack-years of excluded participants were 46.4 (32.2–54.2), and median (25–75th percentile) pack-years of included participants were 40.5 (28.0–49.5; P = 0.01). There were no significant differences in body mass index, smoking status, FEV₁, and COPD status (P = 0.47, P = 0.43, P = 0.81, and P = 0.10, respectively). Three hundred ninety-four (37.6%) participants were classified as having COPD, of whom 251 (23.9%) were classified as GOLD 1, 121 (11.6%) as GOLD 2, and 22 (2.1%) as GOLD 3.

Compared with the baseline model, the model improved significantly when CT-derived biomarkers (Pi10, Perc15, E/I-ratio_MLD, and PRM) were added one by one (Table 1). The baseline model with PRM had the highest diagnostic value (C-statistic of 0.827). When adding all three current CT-derived biomarkers (Pi10, Perc15, and E/I-ratio_MLD), the resulting C-statistic was 0.877. The addition of PRM to the latter model led to a further significant increase in the ability to discriminate between participants with and without COPD (C-statistic: 0.883, P < 0.001). Reclassification analyses showed that this model had the least falsely reclassified participants of all models (193; 18,4%). Reclassification results are shown in Table 1.

Of 1,047 participants, 493 underwent gas transfer measurements. When selecting the upper quintile of PRM^Emph (n = 99), 27 (27.3%) participants of this subgroup had normal gas transfer. This could mean that a substantial group of participants with normal gas transfer could already have signs of smoking-related changes on CT measured with PRM. In Table E1 in the online supplement, all CT-derived measurements of participants with normal gas transfer are shown as a reference.

Our results show that in this cohort, PRM exceeded other measures in diagnostic value when added as single biomarkers. When adding PRM to all currently available CT-derived biomarkers, this resulted in a significant increase in C-statistic.

Galbán and colleagues (3) is the only group that has studied the use of PRM in quantifying COPD on CT. They found that with decreasing FEV₁, levels of PRM measurements increased. Next to information about the correlation with FEV₁, it is important to evaluate whether PRM assessment contributes to the current CT assessment of COPD. We evaluated this in a large cohort with generally healthy smokers. With the extensive characterization of a high-risk cohort, this study provided valuable information regarding the role of PRM in the assessment of smoking-related changes underlying COPD. We showed that an expiratory scan could provide information about small and large airways in participants with emphysema.

Although the added value of PRM to currently used CT measurements of COPD is important in evaluating PRM as a diagnostic tool, it should be recognized that the clinical relevance of this measurement is still unknown and that it is not yet confirmed by histology. In addition, PRM should be analyzed in a healthy cohort with nonsmoking subjects to provide a normal range of this measurement. With a normal reference standard, the influence of inspiratory–expiratory volume differences could be evaluated and PRM measurements could be compared with both spirometry and gas transfer to assess its clinical relevance.

In conclusion, PRM adds significantly to the diagnostic value of current CT quantification methods in assessing the presence of COPD. It could be helpful in identifying smoking-related changes and might contribute to the effect of lung cancer screening. Therefore, this novel quantification technique deserves further investigation.

References

Section:

ChooseTop of pageReferences <<CITING ARTICLES