SESSION TITLE: COPD
SESSION TYPE: Original Investigation Slide
PRESENTED ON: Wednesday, October 30, 2013 at 02:45 PM - 03:45 PM
PURPOSE: Cigarette smokers are at risk for developing IPF. Identifying at-risk individuals with subclinical interstitial lung disease could provide a better understanding of disease progression and earlier therapeutic interventions. As such, we hypothesized that a peripheral blood biomarker signature can identify patients with radiographic interstitial lung abnormalities (ILA).
METHODS: CT scans and plasma samples were obtained from the Univ. of Pittsburgh Lung Screening Study (PluSS). CT scans were scored on a 0-3 scale (0=no evidence;1=indeterminate;2=ILA;3=fibrosis). We then examined plasma samples using multiplex technology for the expression of 126 proteins. Univariate analyses were performed to identify a specific signature in patients with ILA. In addition, the random forest method was used to assess the performance of classification based on overall patterns of all clinical parameters and bioanalytes. We chose the subset of smallest number of variables whose error rate was within the standard error of the minimum error rates of all forests. During each step, the least important variable would be deleted successively and stopped when the reduced and full models did not have similar performance. Out-of-bag error was reported.
RESULTS: Of 415 patients, 148 had CT scores of 2 or 3 of which 80 were selected as cases. 150 controls with CT scores of 0 or 1 were matched for age, gender, race, and semi-quantitative assessment of CT emphysema. An analysis of the protein biomarker data with unadjusted ANCOVA identified 35 analytes that had significant differential expression in cases versus controls. After Bonferonni correction, 6 analytes remained significant: PARC, MIP-3α, SP-D, ICAM-1, E-selectin, CRP. The adjusted model for the above analytes yielded similar p values. Using random forest, a signature with 7 variables—age, emphysema score, MMP-7, PARC, SP-D, resistin, TNFRSF1A—is able to identify patients with ILA on CT with an 18.4% out-of-bag error. The performance of this model was similar to the model using all variables.
CONCLUSIONS: Smokers with ILA on CT can be identified using a specific peripheral blood biomarker signature. Of significance, a number of proteins in the signature have also been shown to have increased expression and predict survival in patients with IPF.
CLINICAL IMPLICATIONS: This biomarker signature, if validated, could serve as the basis for identifying subclinical pulmonary fibrosis in at-risk individuals. These individuals could then be monitored closely for development and progression of IPF.
DISCLOSURE: The following authors have nothing to disclose: Avignat Patel, Tracy Doyle, Yushi Liu, Hiroto Hatabu, Mizuki Nishino Hatabu, Yuka Okajima, Cristobal Risquez, Yuanyuan Shi, Juan Osorio, Maria Golzarri, James Lederer, Souheil El-Chemaly, Victor Pinto-Plata, Bartolome Celli, Gary Hunninghake, George Washko, Frank Sciurba, Naftali Kaminski, Joseph Leader, Jill Siegfried, Joel Weissfeld, Ivan Rosas
No Product/Research Disclosure Information