A recent analysis, published in the American Journal of Respiratory and Critical Care Medicine, highlights the potential of quantitative computed tomography (CT) methods to predict disease progression and evaluate drug effectiveness in clinical trials for patients with progressive pulmonary fibrosis (PPF). In interstitial lung diseases (ILD), forced vital capacity (FVC) is widely used to gauge severity and progression, serving as a critical endpoint in clinical trials. However, researchers emphasize the need for more precise and sensitive tools to detect structural lung deterioration. This study builds on previous efforts to develop better methods for measuring and predicting fibrotic outcomes in ILD patients.
Advancements in Lung Imaging Techniques
Improvements in machine learning and adaptive denoising have enhanced quantitative CT methods, enabling them to predict FVC declines and short-term mortality across various ILDs. The study analyzed data from the INBUILD trial (NCT02999178), a double-blind, placebo-controlled study evaluating nintedanib’s efficacy in fibrosing lung diseases. An optional substudy collected high-resolution CT (HRCT) scans at baseline, week 24, and week 52.
Two algorithms were used to assess the scans. The UCLA algorithm analyzed abnormal lung tissue based on pixel density and texture, generating a quantitative ILD (QILD) score. This score comprised the quantitative lung fibrosis (QLF) score for reticular patterns with architectural distortion, the QHC score for honeycomb cysts, and the QGGO score for ground-glass opacities. The e-Lung algorithm, developed using artificial intelligence, measured total disease extent (TDE), ground-glass opacities (GGO), reticulovascular score (RVS), and weighted RVS (WRVS).
Of the 663 INBUILD trial participants, 474 took part in the HRCT substudy. Mean medication exposure was 15.1 months (SD 6.9) in the nintedanib group and 16.3 months (SD 5.4) in the placebo group. At baseline, 403 patients (197 nintedanib, 206 placebo) had evaluable scans for the UCLA algorithm, while 371 patients (179 nintedanib, 192 placebo) had scans for the e-Lung algorithm.
Strong correlations were observed between UCLA and e-Lung parameters at baseline, week 24, and week 52, as well as in changes from baseline at these intervals. While small differences were noted in total lung capacity between the methods, a larger discrepancy was seen in the total extent of ILD measured by the UCLA QILD score and e-Lung TDE, primarily due to differences in ground-glass opacity quantification.
Read Also: Complement inhibitors have pros and cons
Assessing Impact and Outcomes
In both nintedanib and placebo groups, correlations between changes in FVC % predicted and QILD score were –0.35 (95% CI, –0.48 to –0.20) and –0.35 (95% CI, –0.47 to –0.22) at week 24, and –0.37 (95% CI, –0.50 to –0.21) and –0.46 (95% CI, –0.58 to –0.32) at week 52.
Correlations with e-Lung TDE were –0.29 (95% CI, –0.43 to –0.13) and –0.34 (95% CI, –0.46 to –0.20) at week 24, and –0.29 (95% CI, –0.44 to –0.12) and –0.38 (95% CI, –0.51 to –0.23) at week 52. These correlations were described as moderate. The authors noted, “These findings suggest that quantitative CT measures provide complementary information to that provided by FVC and assess structural changes in the lungs that are not completely reflected in changes in FVC.”
Higher baseline QILD, QLF, e-Lung TDE, WRVS, and RVS scores in the placebo group were linked to a faster FVC decline over 52 weeks. Patients with scores above the median faced a higher risk of a 10% or greater absolute FVC decline or death over 52 weeks, with restricted mean survival time differences ranging from 40 to 65 days.
Nintedanib showed significant effects compared to placebo on QILD score and e-Lung TDE changes at week 24 (relative difference, –7% [95% CI, –11% to –2%]; P = .005 and –8% [95% CI, –12% to –4%]; P < .001, respectively) and week 52 (–7% [95% CI, –12% to –2%]; P < .05 for both). These results indicate that quantitative CT measures offer additional insights beyond FVC, capturing structural lung changes not fully reflected in FVC variations.
Study Limitations
The researchers acknowledged several limitations. HRCT scans at weeks 24 and 52 were only available for substudy participants. The sample size was insufficient to analyze subgroups by specific ILD diagnoses. The findings are applicable only to the INBUILD population, which had fibrosis extent greater than 10% on HRCT.