Citations

AbstractObjectivesAlthough hemispheric surgeries are among the most effective procedures for drug‐resistant epilepsy (DRE) in the pediatric population, there is a large variability in seizure outcomes at the group level. A recently developed HOPS score provides individualized estimation of likelihood of seizure freedom to complement clinical judgement. The objective of this study was to develop a freely accessible online calculator that accurately predicts the probability of seizure freedom for any patient at 1‐, 2‐, and 5‐years post‐hemispherectomy.MethodsRetrospective data of all pediatric patients with DRE and seizure outcome data from the original Hemispherectomy Outcome Prediction Scale (HOPS) study were included. The primary outcome of interest was time‐to‐seizure recurrence. A multivariate Cox proportional‐hazards regression model was developed to predict the likelihood of post‐hemispheric surgery seizure freedom at three time points (1‐, 2‐ and 5‐ years) based on a combination of variables identified by clinical judgment and inferential statistics predictive of the primary outcome. The final model from this study was encoded in a publicly accessible online calculator on the International Network for Epilepsy Surgery and Treatment (iNEST) website (https://hops‐calculator.com/).ResultsThe selected variables for inclusion in the final model included the five original HOPS variables (age at seizure onset, etiologic substrate, seizure semiology, prior non‐hemispheric resective surgery, and contralateral fluorodeoxyglucose–positron emission tomography [FDG‐PET] hypometabolism) and three additional variables (age at surgery, history of infantile spasms, and magnetic resonance imaging [MRI] lesion). Predictors of shorter time‐to‐seizure recurrence included younger age at seizure onset, prior resective surgery, generalized seizure semiology, FDG‐PET hypometabolism contralateral to the side of surgery, contralateral MRI lesion, non‐lesional MRI, non‐stroke etiologies, and a history of infantile spasms. The area under the curve (AUC) of the final model was 73.0%.SignificanceOnline calculators are useful, cost‐free tools that can assist physicians in risk estimation and inform joint decision‐making processes with patients and families, potentially leading to greater satisfaction. Although the HOPS data was validated in the original analysis, the authors encourage external validation of this new calculator.

AbstractBackgroundDeformable Image Registration (DIR) is an essential technique required in many applications of radiation oncology. However, conventional DIR approaches typically take several minutes to register one pair of 3D CT images and the resulting deformable vector fields (DVFs) are only specific to the pair of images used, making it less appealing for clinical application.PurposeA deep‐learning‐based DIR method using CT images is proposed for lung cancer patients to address the common drawbacks of the conventional DIR approaches and in turn can accelerate the speed of related applications, such as contour propagation, dose deformation, adaptive radiotherapy (ART), etc.MethodsA deep neural network based on VoxelMorph was developed to generate DVFs using CT images collected from 114 lung cancer patients. Two models were trained with the weighted mean absolute error (wMAE) loss and structural similarity index matrix (SSIM) loss (optional) (i.e., the MAE model and the M+S model). In total, 192 pairs of initial CT (iCT) and verification CT (vCT) were included as a training dataset and the other independent 10 pairs of CTs were included as a testing dataset. The vCTs usually were taken 2 weeks after the iCTs. The synthetic CTs (sCTs) were generated by warping the vCTs according to the DVFs generated by the pre‐trained model. The image quality of the synthetic CTs was evaluated by measuring the similarity between the iCTs and the sCTs generated by the proposed methods and the conventional DIR approaches, respectively. Per‐voxel absolute CT‐number‐difference volume histogram (CDVH) and MAE were used as the evaluation metrics. The time to generate the sCTs was also recorded and compared quantitatively. Contours were propagated using the derived DVFs and evaluated with SSIM. Forward dose calculations were done on the sCTs and the corresponding iCTs. Dose volume histograms (DVHs) were generated based on dose distributions on both iCTs and sCTs generated by two models, respectively. The clinically relevant DVH indices were derived for comparison. The resulted dose distributions were also compared using 3D Gamma analysis with thresholds of 3 mm/3%/10% and 2 mm/2%/10%, respectively.ResultsThe two models (wMAE and M+S) achieved a speed of 263.7±163 / 265.8±190 ms and a MAE of 13.15±3.8 / 17.52±5.8 HU for the testing dataset, respectively. The average SSIM scores of 0.987±0.006 and 0.988±0.004 were achieved by the two proposed models, respectively. For both models, CDVH of a typical patient showed that less than 5% of the voxels had a per‐voxel absolute CT‐number‐difference larger than 55 HU. The dose distribution calculated based on a typical sCT showed differences of ≤2cGy[RBE] for clinical target volume (CTV) D95 and D5, within ±0.06% for total lung V5, ≤1.5cGy[RBE] for heart and esophagus Dmean, and ≤6cGy[RBE] for cord Dmax compared to the dose distribution calculated based on the iCT. The good average 3D Gamma passing rates (> 96% for 3 mm/3%/10% and > 94% for 2 mm/2%/10%, respectively) were also observed.ConclusionA deep neural network‐based DIR approach was proposed and has been shown to be reasonably accurate and efficient to register the initial CTs and verification CTs in lung cancer.

AbstractBackgroundDeep learning auto‐segmentation (DLAS) models have been adopted in the clinic; however, they suffer from performance deterioration owing to the clinical practice variability. Some commercial DLAS software provide an incremental retraining function that enables users to train a custom model using their institutional data to account for clinical practice variability.PurposeThis study was performed to evaluate and implement the commercial DLAS software with the incremental retraining function for definitive treatment of patients with prostate cancer in a multi‐user environment.MethodsCT‐based target organs and organs‐at‐risk (OAR) delineation of 215 prostate cancer patients were utilized. The performance of three commercial DLAS software built‐in models was validated with 20 patients. A retrained custom model was developed using 100 patients and evaluated on the remaining data (n = 115). Dice similarity coefficient (DSC), Hausdorff distance (HD), mean surface distance (MSD), and surface DSC (SDSC) were utilized for quantitative evaluation. A multi‐rater qualitative evaluation was blindly performed with a five‐level scale. Visual inspection was performed in consensus and non‐consensus unacceptable cases to identify the failure modes.ResultsThree commercial DLAS vendor built‐in models achieved sub‐optimal performance in 20 patients. The retrained custom model had a mean DSC of 0.82 for prostate, 0.48 for seminal vesicles (SV), and 0.92 for rectum, respectively. This represents a significant improvement over the built‐in model with DSC of 0.73, 0.37, and 0.81 for the corresponding structures. Compared to the acceptance rate of 96.5% and consensus unacceptable rate (i.e., both reviewers rated as unacceptable) of 3.5% achieved by manual contours, the custom model achieved a 91.3% acceptance rate and 8.7% consensus unacceptable rate. The failure modes of retrained custom model were attributed to the following: cystogram (n = 2), hip prosthesis (n = 2), low dose rate brachytherapy seeds (n = 2), air in endorectal balloon(n = 1), non‐iodinated spacer (n = 2), and giant bladder(n = 1).ConclusionThe commercial DLAS software with the incremental retraining function was validated and clinically adopted for prostate patients in a multi‐user environment. AI‐based auto‐delineation of the prostate and OARs is shown to achieve improved physician acceptance, overall clinical utility, and accuracy.

8567 Background: Adjuvant osimertinib (osi) improves disease-free survival (DFS) in patients (pts) with resected, early-stage, EGFR-mutant (EGFRmut) NSCLC, yet the benefit of osi after CRT in pts with unresectable locally advanced NSCLC is unknown. Post-hoc analysis of the PACIFIC trial showed a lack of survival benefit with consolidation durva versus placebo in pts with EGFRmut NSCLC. Comparisons between consolidation durva and EGFR TKIs in unresectable EGFRmut NSCLC following CRT are lacking. Methods: We conducted a multi-institutional retrospective analysis of pts with stage III unresectable EGFRmut NSCLC (exon19 deletion, exon21 L858R), who received EGFR TKI or durva after ≥ 2 cycles of platinum-based chemotherapy plus definitive radiation therapy between 2015-2022. Baseline characteristics including age, sex, smoking history, PD-L1 status, and outcomes on DFS, overall survival (OS), and safety were collected. Multivariable (MVA) cox regression analysis was used for statistical analysis. Treatment-related adverse events (trAE) were defined using CTCAE 5.0. Results: Seventeen pts from 12 institutions received an EGFR TKI (osi, n=15; erlotinib, n=2), and 13 pts received consolidation durva. Median follow-up was 23 months. Median age of all pts was 61 years (IQR:52-72) and 76.7% were female. Most pts in both groups had never-smoked and had adenocarcinoma. All pts received ≥ 60 Gy of radiation with concurrent chemotherapy. PD-L1 expression was ≥ 50% in 1/10 (10%) pts treated with durva vs 6/13 (46.2%) treated with TKI, p=0.09. Median duration on treatment for EGFR TKI and durva was 12.2 months and 4.8 months, respectively. Pts treated with EGFR TKI had significantly longer 24-month DFS versus pts treated with durva after adjusting for stage (3A vs 3B vs 3C) (Table). Any grade tRAE occurred in 58.8% (10/17) of pts treated with EGFR TKI vs 38.5% (5/13) with durva. Grade > 3 tRAE occurred in 15.4% (1 pneumonitis and 1 AST/ALT elevation) of pts treated with durva but in none of pts treated with EGFR TKI. Three pts within each arm came off treatment due to toxicity (EGFR TKI: 1 pneumonitis and 2 dermatitis; durva: 1 of each pneumonitis, Type 1 diabetes, AST/ALT elevation). Conclusions: EGFR TKI therapy after definitive CRT was associated with significantly longer DFS compared to durva in this retrospective study of pts with stage III unresectable EGFRmut NSCLC, without unanticipated safety signals. Follow-up is ongoing and OS outcomes will be evaluated subsequently. Further investigation is warranted to define the optimal therapy for locally advanced EGFRmut NSCLC. [Table: see text]

7569 Background: B-cell lymphoma 2 inhibitors (BCL-2i) are well tolerated and effective in treating B-cell lymphoproliferative disorders such as CLL. Investigational agent lisaftoclax is a novel, oral, highly selective and potent BCL-2i. Preclinical data show a strong synergistic effect of lisaftoclax plus ibrutinib in an ibrutinib-insensitive PDXWM1 model. Methods: This global, open-label, multicenter study evaluates the safety and efficacy of lisaftoclax alone and combined with ibrutinib or rituximab in pts with WM. Pts were enrolled in 3 arms: Arm A (lisaftoclax), BTKi-resistant/intolerant pts; Arm B (lisaftoclax plus ibrutinib), treatment-naïve pts; and Arm C (lisaftoclax plus rituximab), BTKi-naïve relapsed/refractory pts. Lisaftoclax was escalated from 400 to 1,200 mg using a modified toxicity probability interval design. A 3-day ramp-up was used for pts at high TLS risk. Primary objectives were pharmacokinetics (PK), safety, and efficacy assessments; responses were assessed per IWWM criteria. Results: As of January 25, 2023, 46 pts were enrolled (Arm A [n = 14]; up to 1,000 mg; Arm B [n = 24]; up to 1,200 mg; Arm C [n = 8]; up to 800 mg). Baseline characteristics are shown in the table. Median (range) treatment duration was 6 (1-18; Arm A), 14 (1-27; B), and 8 (2-26; C) months. Overall response rates (ORRs) were: Arm A, 25% (median time to response [MTTR], 4.3 months); Arm B, 90.9% (MTTR, 1.9); Arm C, 37.5% (MTTR, 4.4). No significant difference was observed between pts with and without CXCR4-mut. At 1,200 mg, 1 DLT (grade 3 clinical TLS), due to pre-existing renal impairment, was reported. At 1,000 mg, 1 grade 3 laboratory TLS occurred in Arm B because of dehydration and active symptomatic therapies; abnormal electrolytes resolved after 1 day of drug interruption, without recurrence. Grade ≥ 3 lisaftoclax-related AEs included neutropenia (13%), leukocytopenia (4.3%), anemia (2.2%), weight loss (2.2%), and septic shock (2.2%). Ventricular arrhythmias were not observed. Only 1 pt required dose reduction (due to neutropenia). MTD has not been reached. A total of 14 pts discontinued study treatment because of disease progression (n = 5) and AE (n = 1). Lisaftoclax plus ibrutinib showed a PK exposure comparable to ibrutinib alone, indicating no potential DDI. Conclusions: Lisaftoclax alone or combined with ibrutinib/rituximab demonstrated measurable effects in pts with naïve or BTKi-treatment-failed WM. Internal study identifier: APG2575WU101. *SA and ML are co-first authors. Clinical trial information: NCT04260217 . [Table: see text]

5094 Background: In prior studies, basal vs. luminal differentiation portends differential response to a) androgen suppression (AS) duration for high-risk prostate cancer (PCa) (NRG-GU-TS006) and b) docetaxel chemotherapy (CT) and AS for metastatic hormone-sensitive disease (CHAARTED). In NRG-GU-TS006, basal tumors appeared to benefit the most from longer duration of AS while those with luminal subtype did not whereas, in CHAARTED, luminal subtype had greater benefit from addition of CT. Herein, we examined basal-luminal subtyping to predict differential response to the addition of CT in the setting of high-risk PCa treated with AS and definitive radiotherapy (AS + RT) in NRG/RTOG 0521. Methods: Per pre-specified and approved NCI analysis plan (Navigator #1061), we obtained available NRG biobank biopsy specimens from patients enrolled on the NRG/RTOG 0521 phase III trial, which randomized patients to AS + RT +/- CT. The highest-grade tumors were profiled on clinical-grade whole-transcriptome arrays (Veracyte, San Diego, CA). Basal-luminal subtypes were generated using a 215-gene PCa subtyping classifier (PSC), a second generation PAM50 basal-luminal model. Treatment effects in Luminal Proliferating (LP) and non-LP subtypes were examined with Cox (for overall survival, OS, and metastasis-free survival, MFS) and Fine & Gray (for distant metastasis, DM) multivariable (MVA) models adjusted for age and trial risk stratification, as well as with restricted mean survival times (RMST). Results: Of 183 samples available for analysis (91 from control and 92 from the CT arm), 52 (28%) were classified as LP subtype. Median follow-up for OS was 9.9 years. No significant differences in clinical or pathological variables were observed between LP and non-LP subtypes. In MVA, we did not detect any significant LP subtype-by-CT treatment interaction but observed evidence of non-proportional hazards (Grambsch-Therneau test, p<0.05). In unselected patients, 5- and 10-year RMST difference between the arms was 0.9 (95% CI [-1.5, 3.4], p=0.46) and 5.7 (95% CI [-2.5, 13.9], p=0.17) months of OS. Stratifying by tumor subtype, at 5 years the RMST difference in months was 3.8 (95% CI [0.3, 7.2], p=0.03) for LP and -0.2 ([-3.4, 2.9], p=0.89) for non-LP. The RMST difference in months at 10-years was 13.7 (95% CI [-0.2, 27.5], p=0.053) for LP and 2.5 ([-7.6, 12.5], p=0.63) for non-LP. Conclusions: In pre-treatment biopsy samples from a Phase 3 trial cohort of localized high-risk men, luminal-basal subtyping of tumors demonstrated differential response to docetaxel CT. Similar to results from the CHAARTED trial of metastatic disease, LP tumors may derive greater benefit from addition of CT. Validation of these findings may enable effective earlier use of CT in men with PCa.

4028 Background: T-cell immunoreceptor with immunoglobulin and immunoreceptor tyrosine-based inhibitory motif domains (TIGIT) inhibitor in combination with an anti-programmed cell death protein 1 (PD-1) antibody has shown antitumor activity in solid tumors. AdvanTIG-105 (NCT04047862) is a phase 1/1b open-label study designed to assess the safety and preliminary antitumor activity of OCI, an anti-TIGIT monoclonal antibody (mAb), + TIS, an anti-PD-1 mAb, in pts with unresectable, locally advanced or metastatic solid tumors. In the dose-escalation part, OCI + TIS was well tolerated with preliminary antitumor activity observed, and the recommended phase 2 dose (RP2D) of OCI 900 mg intravenously (IV) every three weeks (Q3W) + TIS 200 mg IV Q3W was established. We report results from the dose-expansion part (GC/GEJC Cohort 9) of the AdvanTIG-105 study. Methods: Eligible pts had histologically/cytologically confirmed stage IV GC/GEJC. Pts were excluded if they had squamous cell, undifferentiated or other histological types of GC, had GC/GEJC with positive HER2 expression, or if they had received any prior therapy for metastatic disease. Pts received either the RP2D of OCI + TIS with oxaliplatin + capecitabine Q3W for 6 cycles (C), followed by maintenance therapy with the RP2D of OCI + TIS, + capecitabine Q3W, or the RP2D of OCI + TIS with cisplatin + 5-fluorouracil Q3W for 6 C. Treatment continued until disease progression, intolerable toxicity, or withdrawal of consent. The primary endpoint was investigator-assessed overall response rate (ORR) per RECIST v1.1. Secondary endpoints included progression-free survival (PFS), duration of response (DoR), disease control rate (DCR) per RECIST v1.1, and safety. Results: As of September 29, 2022, 60 pts with a median age of 61.5 years (range 35-82) were enrolled; 59 were efficacy evaluable. Median study follow-up was 31.1 weeks (range 1.4-78.4). ORR was 50.8% (95% CI: 37.5, 64.1); DCR was 84.7% (95% CI: 73.0, 92.8) with a median DoR of 8.1 months (95% CI: 4.5, not evaluable [NE]). Median PFS was 8.2 months (95% CI: 5.8, NE). In a subgroup analysis, ORR in pts with PD-L1 tumor area positivity (TAP) score ≥5% (n = 27) was 59.3% (95% CI: 38.8, 77.6), and 50.0% (95% CI: 30.7, 69.4) in pts with PD-L1 TAP < 5% (n = 28). All 60 pts reported ≥1 treatment-emergent adverse event (TEAE), the most common being anemia (43.3%) and platelet count decreased (41.7%). In total, 43 pts (71.7%) had ≥grade 3 TEAEs and 28 (46.7%) had serious adverse events. TEAEs leading to discontinuation of TIS and OCI occurred in 5 (8.3%) pts. TEAEs led to death in 2 (3.3%) pts; one event (neutropenic sepsis) was related to chemo, while the other (pulmonary embolism) was not treatment related. Conclusions: OCI 900 mg + TIS 200 mg + chemo was generally well tolerated and showed encouraging antitumor activity in pts with stage IV GC/GEJC. Clinical trial information: NCT04047862 .

Accurate prediction of progression in subjects at risk of Alzheimer's disease is crucial for enrolling the right subjects in clinical trials. However, a prospective comparison of state-of-the-art algorithms for predicting disease onset and progression is currently lacking. We present the findings of "The Alzheimer's Disease Prediction Of Longitudinal Evolution" (TADPOLE) Challenge, which compared the performance of 92 algorithms from 33 international teams at predicting the future trajectory of 219 individuals at risk of Alzheimer's disease. Challenge participants were required to make a prediction, for each month of a 5-year future time period, of three key outcomes: clinical diagnosis, Alzheimer's Disease Assessment Scale Cognitive Subdomain (ADAS-Cog13), and total volume of the ventricles. The methods used by challenge participants included multivariate linear regression, machine learning methods such as support vector machines and deep neural networks, as well as disease progression models. No single submission was best at predicting all three outcomes. For clinical diagnosis and ventricle volume prediction, the best algorithms strongly outperform simple baselines in predictive ability. However, for ADAS-Cog13 no single submitted prediction method was significantly better than random guesswork. Two ensemble methods based on taking the mean and median over all predictions, obtained top scores on almost all tasks. Better than average performance at diagnosis prediction was generally associated with the additional inclusion of features from cerebrospinal fluid (CSF) samples and diffusion tensor imaging (DTI). On the other hand, better performance at ventricle volume prediction was associated with inclusion of summary statistics, such as the slope or maxima/minima of patient-specific biomarkers. On a limited, cross-sectional subset of the data emulating clinical trials, performance of the best algorithms at predicting clinical diagnosis decreased only slightly (2 percentage points) compared to the full longitudinal dataset. The submission system remains open via the website https://tadpole.grand-challenge.org, while TADPOLE SHARE (https://tadpole-share.github.io/) collates code for submissions. TADPOLE's unique results suggest that current prediction algorithms provide sufficient accuracy to exploit biomarkers related to clinical diagnosis and ventricle volume, for cohort refinement in clinical trials for Alzheimer's disease. However, results call into question the usage of cognitive test scores for patient selection and as a primary endpoint in clinical trials.