A SuStaIn-able Approach to Modeling COPD Progression?

C hronic obstructive pulmonary disease (COPD) is a common disorder characterized by respiratory symptoms in the presence of airflow limitation and develops when an individual with risk factors is exposed to a triggering antigen.1 Disease manifests as a combination of parenchymal destruction and airway damage, which progresses over time and in a variety of patterns.1 As clinicians, our ability to identify and manage advanced disease is well-established. However, our ability to discover and preempt early pathology is lacking. Prior studies have utilized cluster analysis to identify disease subtypes on the basis of computed tomography (CT) abnormalities, but were limited by an inability to parse disease phenotype from severity.2 The authors of this study applied the Subtype and Stage Inference (SuStaIn) model to analyze CT imaging data from the COPDGene cross-sectional data set in an attempt to accurately categorize disease subtype and map disease stage. The model compared baseline imaging characteristics of 3698 smokers with COPD with 3479 smoking controls without COPD. Four main imaging variables were examined: emphysema, functional small airway disease (fSAD), square root wall area, and segmental airway wall thickness. Each subtype was defined by a unique trajectory of imaging features over time, and each stage was defined by the relative position along that trajectory. Patients with COPD were assigned probabilistically to a given subtype and stage on the basis of their individual imaging findings reaching a particular z-score relative to the control group. Notable baseline differences between the COPD group and control group included mean age (63.13 vs. 56.90 y), mean smoking history (51.91 pack-years vs. 37.33 packyears), and mean annual exacerbations (0.64 per year vs. 0.13 per year). The SuStaIn model delineated 2 unique disease subtypes, “tissue → airway” and “airway → tissue.” The tissue → airway subtype (n= 2354, 70.4%) demonstrated early emphysema and fSAD, with large airway involvement occurring later in the disease course. Conversely, the airway → tissue subtype (n= 988, 29.6%) demonstrated early large airway damage, with emphysema and fSAD occurring later in the disease course. Patients with the tissue → airway subtype had a lower mean body mass index than those with the airway → tissue subtype (26.65 vs. 30.54, P< 0.001), lower mean FEV1% predicted (53.63% vs. 58.64%, P< 0.001), lower mean FEV1/FVC (0.49 vs. 0.56, P< 0.001), and lower prevalence of chronic bronchitis (25.1% vs. 31.8%, P <0.001). Over a 5-year follow-up, 87% of individuals remained in their initially assigned subtypes. The authors determined that disease stage could be used as a marker of disease severity. In the tissue → airway subtype, stage correlated with decline in FEV1/FVC (r= –0.63, P< 0.001) and FEV1% predicted (r= –0.66, P< 0.001). The nonlinear relationship indicated that more significant decline in lung function occurred during early stages. In the airway → tissue subtype, stage correlated linearly with decline in FEV1/ FVC (r= –0.58, P< 0.001) and FEV1% predicted (r= –0.51, P< 0.001). GOLD 1-2 patients of both subtypes demonstrated a statistically significant correlation between baseline stage and future decline in lung function. In GOLD 3-4 patients, those with the airway → tissue subtype exhibited a significant correlation between baseline stage and change in FEV1/FVC, while those with the tissue → airway subtype exhibited no significant correlation between baseline stage and change in lung function metrics. Interestingly, 29% of smoking controls had imaging findings that placed them at a stage > 0. Of these controls, 18% had the tissue → airway subtype, where stage correlated with decline in FEV1/FVC (r= –0.099, P= 0.012) but not FEV1% predicted. Similarly, 11% had the airway→ tissue subtype, where stage also correlated with decline in FEV1/FVC (r= –0.19, P< 0.001) but not FEV1% predicted. Importantly, smoking controls with stage > 0 disease were more likely to progress to GOLD 1 at follow-up (23% in tissue → airway subtype, 20.9% in airway → tissue subtype) compared with controls with stage 0 disease (8.7%). These findings suggest that imaging characteristics identified using the SuStaIn model may detect smokers with early COPD at risk for spirometric progression. This study applied the SuStaIn model to compare CT images of smokers with and without COPD and uncovered 2 distinct disease patterns. Disease stage within each subtype correlated with spirometric impairment and offered insight into the timing of functional decline. In smokers without COPD, the presence of stageable disease within both subtypes signifies an opportunity to identify and intervene upon early pathology to stymie disease progression. This study was limited by the assumption that imaging abnormalities either progress or remain stable, but do not remit. In addition, the COPDGene data set includes only patients aged 45 to 80 with at least a 10 pack-year smoking history,3 and therefore excludes younger and lighter smokers, which may confound detection of early disease. Nonetheless, this mechanism of disease modeling represents an innovative method to detect early COPD, classify disease phenotype, and anticipate progression.