B44-26 Assessing Severe Chronic Obstructive Pulmonary Disease (COPD) With Advanced Imaging Analysis

Rationale Chronic obstructive pulmonary disease (COPD) is a heterogeneous disease and its clinical presentation varies. There are no tools to predict development of complications which negatively affect quality of life and mortality, and often prompt a lung transplant referral. In patients with severe COPD, hypercarbia correlates with lower percentage of emphysema and potentially higher percentage of small airways disease, suggesting that imaging features may predict COPD disease severity. The goal of this study was to use machine learning high resolution computed tomography (HRCT) analysis techniques to determine if imaging features in severe COPD patients correlate with hypercarbia. We hypothesized that small airways disease would be associated with hypercarbia. Methods This was a single center pilot study evaluating 8 patients with severe COPD who had contiguous HRCTs prior to lung transplant. Parametric Response Mapping (PRM) via Lung Density Analysis software (4D Medical, Melbourne, Australia) was used to quantify the percentage of small airways disease (FSAD), emphysema (EMPH), and high attenuation areas (HA). Additional measurements included 3 pulmonary vasculature measurements: the volume of pulmonary vasculature (TBV), length of pulmonary vasculature, and the ratio of vessels with less than 5cm2 cross sectional area to TBV (BV5/TBV) as a measure of vascular pruning. We then analyzed the association of each imaging parameter with arterial carbon dioxide levels (pCO2). Results Of the 8 individuals with severe COPD, 5 had hypercarbia (defined as pCO2 ≥ 45). None of the 6 HRCT parameters correlated with pCO2 in univariate analyses (Table 1), but each parameter was associated with sex differences. After controlling for sex, only BV5/TBV was inversely associated with pCO2. To determine whether this was independent of forced expiratory volume in 1 second (FEV1), we performed a multivariable regression controlling for sex and FEV1 and found that BV5/TBV remained inversely associated with pCO2. Conclusions Precision medicine has become increasingly important, and if we understand subtypes of COPD and how to identify them, targeted therapeutics become possible. Further, advanced imaging analysis could become an alternative to invasive techniques for diagnostics. We found no significant correlation between small airways disease and pCO2 but there was an inverse association between vascular pruning and pCO2. This suggests that the loss of pulmonary vasculature may be associated with hypercarbia. Given our small sample size, we need to evaluate more patients in our cohort to see if this trend remains in those with severe COPD as well as at earlier stages of the disease. This abstract is funded by: T32 NIH grant