{
  "abstract": "Introduction Progressive pulmonary fibrosis (PPF) is characterized by relentless fibrotic remodeling of the lung parenchyma. Predicting disease progression using baseline high-resolution computed tomography (HRCT) remains challenging. Our group recently created a novel statistical learning model based on Quantum Particle Swarm Optimization-Random Forest to distinguish voxels expected to progress (e.g., normal pattern transitioning to fibrosis; ground glass pattern transitioning to fibrosis) from voxels expected to be stable on subsequent HRCT imaging in patients with idiopathic pulmonary fibrosis (Shi Y, et al. Artif Intell Med. 2019;100:101709). The Single-scan total probability (STP) score represents the probability of progression on future HRCT and is calculated per region.1 This study aimed to test whether the STP score on baseline HRCT of the chest can predict PPF in patients with systemic sclerosis-associated interstitial lung disease (SSc-ILD).Material and Methods Data from the randomized, double-blind, multicenter Scleroderma Lung Study (SLS) II (comparing mycophenolate vs. cyclophosphamide) was used to test whether the STP score on the baseline HRCT predicted PPF by 24 months. Of 142 enrolled, 126 patients (mycophenolate mofetil [MMF]: n=61; cyclophosphamide [CYC]: n=65) had acceptable baseline HRCTs and PPF outcome data. PPF was defined based on ATS guidelines; patients had to meet at least two of the following criteria: (1) Worsening respiratory symptoms by transition dyspnea index; (2) Absolute decline in FVC >= 5% predicted and/or absolute decline in the diffusing capacity for carbon monoxide (DLCO) >=10%; (3) Radiological evidence of disease progression at 24 months. Progression was defined as either PPF or death. A multivariable Cox regression model was applied, adjusting for sex, baseline percent predicted FVC, and treatment arm.Results Of the 126 patients, 35 (28%) experienced PPF (N=22) or died (N=13) within 24 months. A higher STP score of normal appearing patterns showed a trend toward association with progression (HR ± SE: 1.08 ± 0.05; p=0.137) in our multivariate model ( table 1). Using tertile thresholds (12.2% and 16.3%), the risk of progression was 2.6-fold higher in patients in the 2nd or 3rd tertiles compared with those in the 1st tertile (STP+ normal patterns <=12.2%; HR ± SE: 2.61 ± 1.18; p=0.035; figure 1).Conclusions A novel quantitative score (STP) derived from the pre-treatment baseline HRCT predicted PPF in patients with SSc-ILD receiving mycophenolate or cyclophosphamide therapy. Future independent prospective cohort studies are needed to validate these findings.Abstract P.238 Figure 1Probability of PFS based on tertiles of normal lung patterns in single time-point prediction of worsening and the covariates (sex, the percent predicted forced vital capacity, and treatment)Abstract P.238 Table 1Cox proportional hazards model analysis for predicting PPF in patients with SSc-ILD using the STP score",
  "authors": [
    {
      "affiliations": [
        "David Geffen School of Medicine, University of California Los Angeles, Los Angeles, USA"
      ],
      "name": "Kim Grace Hyun"
    },
    {
      "affiliations": [
        "David Geffen School of Medicine, University of California Los Angeles, Los Angeles, USA"
      ],
      "name": "Jonathan Goldin"
    },
    {
      "affiliations": [
        "David Geffen School of Medicine, University of California Los Angeles, Los Angeles, USA"
      ],
      "name": "Donald Tashkin"
    },
    {
      "affiliations": [
        "David Geffen School of Medicine, University of California Los Angeles, Los Angeles, USA"
      ],
      "name": "Pangyu Teng"
    },
    {
      "affiliations": [
        "David Geffen School of Medicine, University of California Los Angeles, Los Angeles, USA"
      ],
      "name": "Bianca Villegas"
    },
    {
      "affiliations": [
        "David Geffen School of Medicine, University of California Los Angeles, Los Angeles, USA"
      ],
      "name": "Elizabeth Volkmann"
    }
  ],
  "title": "P.238 A novel single time point prediction tool for systemic sclerosis-associated interstitial lung disease based on high resolution computed tomography",
  "uid": "fcb5a1d6-b595-5391-a7b1-a3eed202116e"
}