78 Exploratory radiomics analysis in unresectable hepatocellular carcinoma treated with durvalumab alone or combined with tremelimumab or bevacizumab
Bibliographic record
Abstract
Background The Phase 2 Study 22 [ NCT02519348] demonstrated the efficacy of durvalumab alone (D) or combined with tremelimumab (D+T) or bevacizumab (D+B) in unresectable hepatocellular carcinoma (HCC).1 2 We analyzed baseline and end-of-treatment (EOT) abdominal CT scans to explore associations between image features and clinical outcomes, including overall survival (OS), progression-free survival (PFS), and lesion-specific responses.Methods Arterial phase CTs from 124 patients, free of artifacts, were reviewed; measurable HCC lesions (>10 mm) and entire liver were delineated by radiologists. Radiomic features from tumor, peri-tumoral regions, and liver were extracted.OS and PFS were modeled using Cox regression with baseline radiomic features of liver or lesions. Performances were reported as concordance index (c-index).For patients with measurable tumors at both baseline and EOT, individual lesion response was defined by volumetric change (growing/shrinking) and exponential decay/growth rates were estimated. Multivariate mixed-effect models were assessed to create a classification model (60-40 train-test split) to distinguish growing/shrinking lesions using baseline radiomic features of lesions, with performance reported as AUC.Results Among 124 patients, 93 (SR-1) had measurable tumors with 490 lesions at baseline; 31 (SR-2) were determined not measurable or ‘diffuse’. Median tumor burdens were 112 cm 3 (D, n=39), 66 cm3 (D+T, n=32), and 39 cm3 (D+B, n=22).OS was shorter for SR-2 than SR-1 for D and D+T (424 and 217 days respectively), with a similar trend observed for B+D. Univariate analysis identified eight liver-based radiomic features associated with SR-2.A survival model using liver radiomic features had a c-index of 0.61 (124 patients, 3-fold cross-validation), surpassing lesion-based OS models. For PFS, the best performance was achieved using radiomic features of the largest lesion (c-index=0.62, SR-1).In 57 patients from SR-1 with suitable EOT scan, 298 lesions were tracked. Combination therapy yielded a higher tumor decay rate than D-monotherapy, with similar growth-rates for non-responding lesions across treatments. In responding lesions, median tumor half-life decreased from ~670 days to ~215 days (combination arms). A baseline mixed-effects model achieved AUC of 0.86 in lesion response classification.Conclusions Deep lesion level analyses revealed the impact of combination therapy in shrinking lesions, complementing RECIST assessment. High-throughput radiomic detection of the negative prognostic feature ‘diffuse’ appears feasible. For immunotherapy response/resistance, baseline radiomic features may predict OS, PFS, lesion-level outcome and could be utilized to identify candidate features. Limitations include sample size and potential overfitting, requiring validation in a larger cohort.Acknowledgements Image quality control, radiology reads, segmentations, and statistical analysis was performed by Radiomics.bioTrial Registration ClinicalTrials.gov identifier: NCT02519348References Kelley RK, Sangro B, Harris W, Ikeda M, Okusaka T, Kang YK, et al. Safety, efficacy, and pharmacodynamics of tremelimumab plus durvalumab for patients with unresectable hepatocellular carcinoma: randomized expansion of a phase I/II study. J Clin Oncol 2021;39:2991-3001.Lim HY, Heo J, Kim T-Y, Tai WMD, Kang Y-K, Lau G, et al. Safety and efficacy of durvalumab plus bevacizumab in unresectable hepatocellular carcinoma: results from the phase 2 study 22 (NCT02519348). J Clin Oncol 2022; 40: abs 436.Ethics Approval This open-label, phase I/II study was conducted at 19 sites in nine countries (ClinicalTrials.gov identifier: NCT02519348) according to the Declaration of Helsinki. All patients provided written informed consent. Protocol approval was obtained from institutional review boards or ethics committees at each site.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.003 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".