POS1055 ASSOCIATIONS AND MORTALITY IMPACT OF MACHINE LEARNING-DERIVED QUANTITATIVE PARENCHYMAL LUNG COMPUTED TOMOGRAPHY FEATURES IN RHEUMATOID ARTHRITIS AND NON-RA COMPARATORS IN A MULTICENTER PROSPECTIVE COHORT OF SMOKERS

Abstract

BackgroundQuantitative computed tomography (QCT) methods have been developed to automatically quantify parenchymal lung features on chest CT imaging. There have been limited investigations of QCT in RA participants and non-RA comparators or studies of the mortality impact of QCT features in RA.ObjectivesDetermine the association and mortality impact of QCT features in RA and non-RA participants.MethodsWe analyzed associations between RA and QCT features in COPDGene, a multicenter cohort study of current or former smokers that excluded participants with known interstitial lung disease or bronchiectasis. We identified participants with and without RA using RA self-report and DMARD use. We assessed the lung parenchyma in each scan by categorizing regions of interest into normal lung, interstitial changes, or emphysema using the local tissue density and distance from the pleural surface. Interstitial changes were subclassified into reticular, subpleural line, linear scar, honeycombing, centrilobular nodule, nodular, and ground glass. Each feature was summed and standardized to total lung volume. We examined associations between QCT features and RA using multivariable linear regression. We dichotomized participants using the 75thpercentile for each QCT feature among non-RA participants and investigated mortality associations by RA status and QCT features using Cox regression. We examined multiplicative interactions between RA and continuous interstitial and emphysema percentages and additive interactions between RA status and >75thpercentile of QCT features using multiplicative interaction terms and attributable proportion.ResultsWe identified 82 RA cases and 8820 non-RA comparators. RA was associated with a lower percentage of normal lung (85.8% vs. 91.0% p=0.0001), increased interstitial changes (7.0% vs. 4.8%, p<0.0001), and no statistical difference in emphysema (2.6% vs. 1.9%, p=0.09) compared to non-RA comparators. In linear regression analyses adjusted for age, sex, smoking status, pack-years, and body mass index, RA was associated with increased interstitial changes (β=1.7±0.5, p=0.0008) but not emphysema (β=1.3±1.7, p=0.44). The combination of RA and >75thpercentile of emphysema had significantly higher mortality compared to both non-RA participants (HR 5.86, 95%CI 3.75-9.13) and RA participants (HR 5.56, 95%CI 2.71-11.38) with <75thpercentile of emphysema. There were statistically significant interactions between RA and emphysema for mortality (multiplicative: p=0.014; additive: attributable proportion 0.53, 95%CI 0.30-0.70, p<0.0001).ConclusionUsing machine learning-derived QCT data in a cohort of smokers, we found that RA was associated with increased interstitial changes, even after adjustment for smoking and other lifestyle factors. The combination of RA and emphysema conferred greater than 5-fold increased mortality.Table 1.Quantitative CT features and multivariable linear regression by RA status in COPDGene (n=8902)Feature (% of lung volume, median, IQR)RA cases (n=82)Non-RA comparators (n=8820)Unadjusted p-valueAdjusted* β±SE (RA vs. non-RA)Adjusted p-valueNormal lung85.8 (77.3-92.4)91.0 (82.7-94.7)0.0001-3.0±1.70.08Interstitial7.0 (4.0-9.7)4.8 (3.0-7.8)<0.00011.72±0.510.0008 Reticular6.0 (3.6-8.8)4.1 (2.6-6.8)<0.00011.55±0.460.0008 Subpleural line0.40 (0.22-0.64)0.31 (0.18-0.51)0.020.084±0.030.009 Linear scar0.17 (0.09-0.30)0.13 (0.06-0.24)0.0060.029±0.020.08 Honeycombing0.05 (0.02-0.11)0.03 (0.009-0.07)<0.00010.058±0.020.003 Centrilobular nodule0.005 (0.00-0.03)0.008 (0.00-0.05)0.47-0.006±0.010.69 Nodular0.008 (0.002-0.03)0.004 (0.001-0.02)0.007-0.005±0.030.88 Ground glass0.005 (0.001-0.01)0.003 (0.00-0.007)0.010.0016±0.020.93Emphysema2.6 (0.8-12.8)1.9 (0.5-8.8)0.091.3±1.70.43* linear regression adjusted for age, sex, smoking status (current/former), pack-years, BMIREFERENCES:NIL.Acknowledgements:NIL.Disclosure of InterestsGregory McDermott: None declared, Keigo Hayashi: None declared, Kazuki Yoshida Consultant of: OM1, Matthew Moll Grant/research support from: Bayer (institutional grant support), Michael Cho Speakers bureau: Illumina, Consultant of: AstraZeneca, Grant/research support from: Bayer, GSK, Tracy Doyle Speakers bureau: AURA, Consultant of: Boehringer-Ingelheim, LEK consulting, Grant/research support from: BMS, Genentech, Paul Dellaripa Grant/research support from: Genentech, Bristol Meyers Squibb, Gregory Kinney: None declared, Zachary Wallace Consultant of: Zenas Biopharma, Horizon, Sanofi, Shionogi, Viela Bio, and MedPace, Grant/research support from: Bristol-Myers Squibb and Principia/Sanofi, Elizabeth Regan: None declared, Gary Hunninghake Consultant of: Boehringer-Ingelhim, Chugai Pharmaceuticals, Gerson Lehrman Group, Edwin Silverman Grant/research support from: Bayer, GlaxoSmithKline, Raul San Jose Estepar Shareholder of: Quantitative Imaging Solutions, Speakers bureau: Chiesi, Consultant of: Leuko Labs, Grant/research support from: Lung Biotechnology, Insmed, Boehringer Ingelheim, Imbio, Samuel Ash Shareholder of: Quantitative Imaging Solutions, George Washko Shareholder of: Quantitative Imaging Solutions, Consultant of: Pulmonx, Janssen, Novartis, Vertex, Grant/research support from: Boehringer Ingelheim, Jeffrey Sparks Consultant of: AbbVie, Amgen, Boehringer Ingelheim, BMS, Gilead, Inova Diagnostics, Janssen, Optum, Pfizer, Grant/research support from: Bristol Meyers Squibb,