Background Lung cancer develops progressively from normal lung tissue to the formation of small lesions, often beginning with subtle cellular changes that evolve into imaging detectable abnormalities. Current screening approaches and a ‘wait-and-watch’ strategy often miss the optimal window to intercept lung cancer effectively. We are testing immunoprevention strategies on selected high-risk patients. To enhance prevention strategies, it is essential to comprehensively understand the trajectory of tumor evolution and accurately measure the efficacy of prevention interventions.¹ Addressing this, our study examines lesion progression from normal tissue to precancerous stages and cancer diagnosis in real-world lung cancer patients, and then assess the interception effects on lesion growth in our phase-II Can-Prevent-Lung trial that testing anti-IL-1B antibody in lung precancer setting.²

Methods We gathered longitudinal CT scans from 122 real-world patients before lung cancer diagnosis. Radiomics modeling quantified tumor morphology, texture, and intensity evolution to predict lung cancer risk. Insights from this analysis informed the Can-Prevent-Lung Trial, exploring how Canakinumab interception influences tumor evolution.

Results Based on radiomics analysis of longitudinal CT data in a real-world pre-lung cancer cohort, we achieved a C-index of 0.66 for predicting future lung cancer risk through cross-validation. Significant changes were noted in key radiomics features as patients approached their cancer diagnosis. We observed tumor volume and size progression concurrent with increased cancer risk, with a median length of 12 mm for those diagnosed within a year, compared to 6 mm for those diagnosed more than 3 years later. Moreover, nodules exhibited higher density, reflected in a 37% greater mean Hounsfield unit (HU) among patients diagnosed within a year, compared to those diagnosed after more than 3 years. In the Can-Prevent-Lung trial, similar trends were seen in the untreated control group, which showed a lesion volume growth rate of +17% per year. In contrast, the treated group demonstrated a median volume reduction rate of -6% per year. Furthermore, lesions in the treated cohort exhibited significantly lower energy (P=0.024), less contrast (P=0.049), and reduced mean Hounsfield unit (HU) (P<0.001) compared to those in the control group.

Conclusions Radiomics demonstrates promising potential in identifying high-risk patients for developing lung cancer and assessing the effectiveness of prevention strategies. It provides a valuable tool for stratifying patients for future immunoprevention trials and monitoring early interception response.

References

1. Lambin Philippe, et al. Radiomics: extracting more information from medical images using advanced feature analysis. European Journal of Cancer 2012;48(4):441–446.

2. Yuan Bo, et al. Targeting IL-1β as an immunopreventive and therapeutic modality for K-ras–mutant lung cancer. JCI insight 2022;7(11).