Validation In Larger Datasets Research Articles

Haploidentical vs HLA-matched sibling donor HCT with PTCy prophylaxis: HLA factors and donor age considerations

AbstractHLA-matched sibling donors (MSDs) are preferred for hematopoietic cell transplantation (HCT). However, the use of alternative donors, especially haploidentical, is increasing, as is our understanding of the impact of HLA factors such as B-leader and DRB1-matching on its outcomes. Yet, data comparing these donor types, particularly considering these HLA factors, is lacking. Herein, we compared haploidentical-HCT (n = 1052) with MSD-HCT (n = 400), both with posttransplant cyclophosphamide (PTCy)-based graft-versus-host disease prophylaxis. In multivariate analysis, haploidentical group had similar overall survival (OS; hazard ratio (HR), 0.94; 95% confidence interval [CI], 0.78-1.14; P = .54), nonrelapse mortality (HR, 0.98; 95% CI, 0.72-1.32; P = .87), and relapse (HR, 0.87; 95% CI, 0.70-1.08; P = .20) as the MSD group. Younger donor age was a significant predictor of improved OS. Next, we directly compared the outcomes of “younger” haploidentical (donor age <35 years, n = 347) vs an “older” MSD (donor age ≥50 years, n = 143) in older recipients (patient age ≥50 years). Patients with younger haploidentical B-leader–matched donors had significantly superior OS (HR, 0.65; 95% CI, 0.48-0.90; P = .009) than the older MSD group. Additionally, patients with younger DRB1-mismatched haploidentical donors (HR, 0.63; 95% CI, 0.46-0.87; P = .004) had significantly lower risk of relapse than older MSDs. Our study suggests that haploidentical-HCT may offer comparable outcomes to MSD-PTCy HCT. Moreover, among older patients, a younger haploidentical B-leader–matched donor might be preferable to an older MSD. These findings need validation in larger data sets.

Molecular profiling of serial liquid biopsy specimens utilizing cell free DNA (cfDNA) and circulating tumor cells (CTCs) in TBCRC041: A phase II study of alisertib in endocrine resistant metastatic breast cancer (MBC).

1037 Background: TBCRC041, a randomized phase 2 trial of alisertib (A) ± fulvestrant (F) in women with HR+/HER2- MBC demonstrated promising clinical activity for alisertib (1). Interrogation of baseline (PreC1) and end of cycle 1 (EOC1) cfDNA and CTCs may identify biomarkers associated with response to A±F. Methods: Plasma cfDNA was sequenced using the Guardant INFINITY platform, which includes genomic and methylated tumor fraction (mTF) analysis. Pathogenic variants were determined using publicly available databases and Mayo Clinic annotation pipelines. CTCs were identified as nucleated, EpCAM+/ cytokeratin+/ CD45- cells and assessed for ER/HER2 staining (RareCyte, Seattle). PFS was defined as the time from randomization to progression or death. EOC1-PFS differs from PFS as time starts at C2D1. Stratified log-rank tests and conditional Cox modeling were used to assess whether PFS or EOC1-PFS differed by mutations in ESR1 or PIK3CA, pre-CTC ≥ 5, or mTF&gt;1%. Results: Of the 91 eligible pts, 84 progressed, and 73 were deceased. In the 86 pts with PreC1-cfDNA results, variants were identified in ESR1 (42; 48.8%), PIK3CA (36; 41.9%), AKT1 (7; 8.1%), PTEN(7; 8.1%), BRCA1 (4; 4.7%), and BRCA2 (7; 8.1%). Pts with a PIK3CAmut had significantly decreased PFS (HR 1.8; 95%CI: 1.1 -2.9, p=0.014) but not by ESR1mut status (p=0.583). PreC1 and EOC1-mTF were determined for 78 and 76 pts, respectively. PreC1-mTF was 0-1% in 21; 1.1-10.0% in 27; and 10.1-100% in 30. mTF levels remained between 0-1% in 17 pts fell to 0-1% in 11 pts, and was 0-1% in 5 pts without a preC1-mTF determination. PFS did not differ with respect to PreC1-mTF (p=0.210), but EOC1-PFS was worse in pts with mTF &gt;1% (HR 3.0; 95%CI: 1.6 -5.2, p&lt;0.001). PreC1 and EOC1-CTC were determined for 78 and 76 pts. 36 (46.2%) pts had a PreC1 CTC count ≥ 5. EOC1-CTC levels remained at (n=17) or increased to (n=1) ≥ 5 and was ≥ 5 in 2 pts without a preC1-CTC determination. EOC1-PFS was not significantly different for EOC1-CTC ≥ 5 (p=0.065), but was significantly decreased in pts with PreC1-CTC≥5 (HR=1.8; 95%CI: 1.1-3.0; p=0.018). Broader cfDNA NGS panel and CTC-derived ER/HER2 staining results will be presented at the meeting. Conclusions: Among patients receiving A±F, PFS varied based on the PIK3CAmutation status. cfDNA and CTC assessment PreC1 and EOC1 provided complementary information. Further validation in larger datasets is needed. 1. JAMA Oncol 2023;9:815. Clinical trial information: NCT02860000 .

Abstract PO3-13-02: Genomic and intrinsic subtype correlates of serum thymidine kinase activity in patients with metastatic breast cancer treated with palbociclib and fulvestrant in the PYTHIA trial

Abstract Background: Serum thymidine kinase activity (sTKa) is a novel non-invasive proliferation biomarker providing prognostic, predictive and monitoring information in patients (pts) with metastatic breast cancer (MBC). We have shown in PYTHIA that high baseline sTKa and lack of sTKa clearance at day 15 of the first cycle of treatment with palbociclib (P)+ fulvestrant (F) are poor prognostic factors in terms of PFS. Here we characterized the molecular features of tumor samples from PYTHIA, and their relation with sTKa. Methods: PYTHIA (IBCSG 53- 14/BIG 14-04; NCT02536742) is a phase II biomarker discovery trial downstream of the AURORA molecular program (BIG 14-01; NCT02102165) which enrolled 122 pts with endocrine-resistant, ER+ and HER2 - MBC receiving P+F. Intrinsic subtypes of metastatic samples from 24 pts were estimated using RNA-Seq. Somatic mutations (mut) from 66 pts were identified using an NGSpanel of 411 cancer-related genes (36 primary and 48 metastatic samples). sTKa was measured using DiviTumÒ TKa (Biovica International) before treatment (D0) and at day 15 (D15) of cycle 1 of P+F. sTKa clearance at D15 was defined as sTKa below the limit of detection (LOD). Intrinsic subtypes and ESR1 mut were assessed in metastatic samples, PIK3CA and TP53 mut were derived from a combined cohort, wheremetastatic samples were used when available and primary samples otherwise. Association of mut and continuous sTKa with PFS was assessed using Cox regression. Results: In metastatic samples, sTKa levels at D0 tended to be higher in luminal B and HER2-enriched subtypes. Lack of sTKa clearance at D15 was observed in 14% of the cases, all classified as luminal B. PIK3CA and TP53 were the most frequently altered genes (38% and 27%, respectively). ESR1 mut were found in 8% of primary and 25% of metastatic samples. Only TP53 was associated with worse PFS. sTKa levels at D0 tended to be higher in pts with TP53 mut vs wild-type (wt) and lower in PIK3CA mut vs wt. Interestingly, lack of sTKa clearance at D15 was more frequent in TP53 mut vs wt and less frequent in PIK3CA mut vs wt. Levels of sTKa were similar in pts with ESR1 mut vs wt both at D0 and at D15 (Table 1). The inclusion of TP53 as a predictor of PFS in bivariate Cox models with sTKa at either D0 (likelihood ratio test p=0.44) or D15 (p=0.21) did not add to the prognostic value provided by sTKa alone. Conclusions: Tumors with high proliferation and poor prognostic features (luminal B, HER2-enriched and TP53 mut) tend to have higher baseline sTKa and less sTKa response on-treatment, in line with sTKa being a circulating dynamic biomarker of tumor cell proliferation. After accounting for sTKa values, TP53mutational status did not contribute to prognosis. This suggests that high sTKa at D0 and lack of sTKa clearance at D15 might be associated with a worse prognosis independently of TP53 status. Further validation in larger datasets is warranted. Table 1: Results summary Citation Format: Svitlana Tyekucheva, Thayane Crestani, Michail Ignatiadis, Patrick Neven, Marco Colleoni, Stéphanie Henry, Konstantinos Papadimitriou, Antonio Bernardo, Elena Seles, Francois Duhoux, Iain Macpherson, Alastair Thomson, David Mark Davies, Mattias Bergqvist, Ilenia Migliaccio, Gabriele Zoppoli, Roswitha Kammler, Heidi De Swert, Barbara Ruepp, Angel Guerrero, Arnau Llinas, Danai Fimereli, Amal Arahmani, David Cameron, Sherene Loi, Martine Piccart, Meredith Regan, Matteo Benelli, Luca Malorni. Genomic and intrinsic subtype correlates of serum thymidine kinase activity in patients with metastatic breast cancer treated with palbociclib and fulvestrant in the PYTHIA trial [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO3-13-02.

Identifying depression-related topics in smartphone-collected free-response speech recordings using an automatic speech recognition system and a deep learning topic model

BackgroundPrior research has associated spoken language use with depression, yet studies often involve small or non-clinical samples and face challenges in the manual transcription of speech. This paper aimed to automatically identify depression-related topics in speech recordings collected from clinical samples. MethodsThe data included 3919 English free-response speech recordings collected via smartphones from 265 participants with a depression history. We transcribed speech recordings via automatic speech recognition (Whisper tool, OpenAI) and identified principal topics from transcriptions using a deep learning topic model (BERTopic). To identify depression risk topics and understand the context, we compared participants' depression severity and behavioral (extracted from wearable devices) and linguistic (extracted from transcribed texts) characteristics across identified topics. ResultsFrom the 29 topics identified, we identified 6 risk topics for depression: ‘No Expectations’, ‘Sleep’, ‘Mental Therapy’, ‘Haircut’, ‘Studying’, and ‘Coursework’. Participants mentioning depression risk topics exhibited higher sleep variability, later sleep onset, and fewer daily steps and used fewer words, more negative language, and fewer leisure-related words in their speech recordings. LimitationsOur findings were derived from a depressed cohort with a specific speech task, potentially limiting the generalizability to non-clinical populations or other speech tasks. Additionally, some topics had small sample sizes, necessitating further validation in larger datasets. ConclusionThis study demonstrates that specific speech topics can indicate depression severity. The employed data-driven workflow provides a practical approach for analyzing large-scale speech data collected from real-world settings.

Development of Cytokine Release Syndrome Is Associated with Superior Progression-Free Survival in DLBCL Patients Receiving Axi-Cel or Liso-Cel

Introduction: Chimeric antigen receptor T-cell therapy (CAR-T) targeting CD19 is a new standard of care for diffuse large B-cell lymphoma (DLBCL) patients with primary refractory disease, early relapse after chemoimmunotherapy, or multiply relapsed disease. Three CD19-directed CAR-T products, axicabtagene ciloleucel (axi-cel), tisagenlecleucel (tisa-cel), and lisocabtagene maraleucel (liso-cel), are FDA approved for relapsed/refractory (R/R) DLBCL and may cure approximately 30-40% of patients. Cytokine release syndrome (CRS), a systemic inflammatory response associated with CAR-T expansion, is a common early toxicity occurring in 42-93% of DLBCL patients in the pivotal phase 2 trials. We hypothesized that development of CRS may be associated with post-CAR-T efficacy outcomes. Methods: We retrospectively evaluated post-CAR-T outcomes for 57 consecutive patients with R/R DLBCL who received axi-cel or liso-cel infusion at the University of California San Francisco between June 2018 and December 2022. Early post-CAR-T toxicities including CRS and immune effector cell-associated neurotoxicity syndrome (ICANS) were recorded using ASTCT consensus grading. Post-CAR-T efficacy endpoints including overall response rate (ORR), complete remission (CR) rate (best response), progression-free survival (PFS), and overall survival (OS) were evaluated for the entire cohort and compared between patients who developed any grade CRS or no CRS after CAR-T infusion. Of note, patients with high-grade B-cell lymphoma or who received tisa-cel were excluded from this study to eliminate potential confounding given poorer outcomes in these subgroups and imbalance in their relative proportions between the CRS and no CRS cohorts. Results: For the entire cohort, the median age was 63 (range 26-81) and 59% of patients were male. Diagnoses included de novo DLBCL (65%), transformed follicular lymphoma (30%), and primary mediastinal B-cell lymphoma (5%). Patients had a median of 2.5 systemic therapies (range 1-8) prior to CAR-T infusion. CAR-T products included axi-cel (70%) or liso-cel (30%). 82% of patients received fludarabine and cyclophosphamide lymphodepletion, and the remaining 18% received bendamustine (during the national fludarabine shortage). Forty-one patients (72%) experienced CRS (grade 1: 35%, grade 2: 37%), and 16 patients (28%) never developed CRS. Seventeen patients (30%) developed ICANS (grade 2: 12%, grade 3: 16%, grade 4: 2%), which occurred exclusively among patients in the CRS cohort. 93% of patients in the CRS cohort received at least one dose of tocilizumab. Baseline characteristics were well balanced between the CRS and no CRS groups with no significant differences in age, sex, performance status, DLBCL subtype, number of prior therapies, CAR-T product administered, lymphodepletion administered, or lactate dehydrogenase at the time of infusion. For the entire cohort, the ORR and CR rate were 86% and 70%, respectively. With a median follow-up of 1.1 years (range 0.2-4.5 years), the 1-year PFS and OS estimates were 61% and 91%, respectively. Median PFS was 1.7 years and median OS was not reached. Outcomes were similar for patients receiving axi-cel and liso-cel (1-year PFS 62% vs 60%, p=0.94). For patients in the CRS cohort vs no CRS cohort, the ORR was 90% vs 75% (p=0.14) and CR rate was 73% vs 63%, respectively (p=0.44). Patients in the CRS cohort had superior PFS compared to patients with no CRS (1-year PFS 70% vs 38%, HR 0.36, 95% CI 0.14-0.98, logrank p=0.011), but there was no significant difference in OS ( Figure 1). There was no significant difference in PFS between patients with grade 1 vs grade 2 CRS (1-year PFS 64% vs 75%, p=0.46). Conclusions: Although limited in sample size, our study suggests that development of CRS may be associated with more favorable post-CAR-T efficacy including superior PFS compared to patients with no CRS after axi-cel or liso-cel infusion. The more favorable outcomes in patients experiencing CRS may reflect better CAR-T proliferation and cytotoxicity compared to patients with no CRS. These findings suggest a relationship between early post-CAR-T toxicity and efficacy outcomes but require validation in larger datasets.

Cerebrospinal Fluid Proteomic Changes after Nusinersen in Patients with Spinal Muscular Atrophy.

Disease-modifying treatments have transformed the natural history of spinal muscular atrophy (SMA), but the cellular pathways altered by SMN restoration remain undefined and biomarkers cannot yet precisely predict treatment response. We performed an exploratory cerebrospinal fluid (CSF) proteomic study in a diverse sample of SMA patients treated with nusinersen to elucidate therapeutic pathways and identify predictors of motor improvement. Proteomic analyses were performed on CSF samples collected before treatment (T0) and at 6 months (T6) using an Olink panel to quantify 1113 peptides. A supervised machine learning approach was used to identify proteins that discriminated patients who improved functionally from those who did not after 2 years of treatment. A total of 49 SMA patients were included (10 type 1, 18 type 2, and 21 type 3), ranging in age from 3 months to 65 years. Most proteins showed a decrease in CSF concentration at T6. The machine learning algorithm identified ARSB, ENTPD2, NEFL, and IFI30 as the proteins most predictive of improvement. The machine learning model was able to predict motor improvement at 2 years with 79.6% accuracy. The results highlight the potential application of CSF biomarkers to predict motor improvement following SMA treatment. Validation in larger datasets is needed.

#4694 AUTOMATIC SEGMENTATION OF GLOMERULAR SUBSTRUCTURES BY DEEP LEARNING

Abstract Background and Aims Electron microscopy (EM) complements light microscopy (LM) evaluation of the kidney biopsy. Foot process effacement, as assessed by EM, helps in diagnosing podocytopathies. However, human interpretation of EM images is time-intensive and often subjective. In this pilot, we investigate how deep learning techniques can help in adequate segmentation of the glomerular basement membranes (GBM) and podocytes in EM images. The ultimate goal would be to design an automatic tool for reliable and fast assessment of foot process effacement. Method Podocytes and glomerular basement membranes (GBM) of 10 patients, 5 with podocyte disease and 5 without glomerular changes (where biopsy showed tubulointerstitial pathology on LM) were annotated with a combination of manual annotation and thresholding for providing ground truth. After data preprocessing including splitting, flipping, rotating and tiling, a modified U-net architecture was applied (‘baseline model’). This baseline model was compared to a combination of U-net and a self-supervised contrastive learning approach with pretraining on 100 additional images (SimCLR framework, ‘fine-tuned model’). Segmentation performance was measured by IoU score. Results Segmentation of the glomerular basement membrane was best achieved by the baseline model and resulted in an IoU score of 0.711±0.089 compared to a IoU score of 0.675 ±0.093 in the fine-tuned model Segmentation of the podocytes was most successful in the fine-tuned model, with a IoU score of 0.609±0.118 compared to a IoU score of 0.591 ±0.118 in the baseline model Conclusion This study pioneers in segmenting glomerular substructures on EM images by means of a modified U-net architecture. The next step is training and validation in larger datasets. Data annotation remains a challenge. Inclusion of more images is expected to greatly improve the performance of the model. Guideline 1500-3600 characters

DCE-MRI-derived microvasculature mapping and outcome prediction in patients with breast cancer.

e12614 Background: Tumor response to neoadjuvant therapy (NAT) in patients with breast cancer is related to perfusion in the tumor microenvironment and surrounding tissue. Direct and indirect measures of microvascular (MV) permeability could improve the ability to predict short- and long-term patient outcomes. Detection and quantification of circulating tumor DNA (ctDNA) is one such measurement that depends on microvasculature and has been shown to strongly correlate with both response to therapy and overall prognosis in a wide range of cancers. We hypothesized that a combination of ctDNA and numerical computations of MV permeability using pretreatment dynamic contrast-enhanced (DCE) MRI images (MV maps) can predict patient response to NAT and long-term survival outcomes. Here, we show how MV maps — computed from standard-of-care (SOC) imaging — improve upon ctDNA measurements in predicting pathologic complete response (pCR) following NAT as well as distant recurrence free survival (DRFS). Methods: DCE-MRIs were obtained for a sub-sample of I-SPY2 patients (n = 81) whose ctDNA levels were previously reported (Magbanua et al 2021). Tumors were segmented from pretreatment images by a convolutional neural network with manual verification. MV maps were computed using a Tofts-like model to describe the transfer kinetics between microvasculature and extracellular extravascular space. Microvasculature parameters (MVPs), such as median and skewness, were extracted from MV maps to describe the distribution of kinetic parameters in the intra- and peritumoral regions. Logistic regression was used to predict pCR with leave-one-out cross validation. Univariate Cox models were used to stratify DRFS. Results: MVPs exhibited a moderate correlation with pretreatment ctDNA levels (Spearman rho = -0.26; p = 0.018) and with pCR status in this cohort (rho = 0.25; p = 0.026). In a series of logistic regression models, a combination of SOC clinical features and ctDNA levels correctly classified 23% of true pCRs, with an F1 score of 0.58. This predictive power increased when informed by MV maps; 45% of true pCRs were identified after the addition of MVPs, and the F1 score increased to 0.68. In univariate Cox models, 24 MVPs significantly stratified DRFS (p &lt; 0.05) when uncorrected for multiple hypothesis testing bias. The top MVP carried a hazard ratio 3.8 (95%CI 1.6 - 9.0) compared to 11.9 (95%CI 2.7 - 53.6) for ctDNA levels. Conclusions: DCE-MRI-derived kinetic features are calculated from SOC images with no additional clinical intervention. We found that MVPs substantially improve pCR prediction when added to models trained on ctDNA levels and clinical features. ctDNA levels and MVPs were predictive of DRFS to differing degrees of confidence. Although further validation in larger datasets is needed, this exploratory analysis uncovered a promising role for MV maps in predicting response to NAT as well as patient outcomes.

Paramagnetic rim lesions predict the development of clinical MS in radiologically isolated syndrome: results from a prospective cohort study (S27.003)

Paramagnetic rim lesions predict the development of clinical MS in radiologically isolated syndrome: results from a prospective cohort study (S27.003)

Association of primary tumor radiomic phenotypes and outcomes in patients (pts) with metastatic renal cell clear cell carcinoma (mRCC) treated with cytoreductive nephrectomy (CN).

622 Background: Despite the advances in systemic therapy for mRCC, CN may benefit a subset of pts. We investigated the correlation of clinical outcomes with primary tumor radiomics and radiology measurements of primary and metastatic lesions in mRCC pts treated with CN. Methods: We searched our institutional database for mRCC pts referred for CN from July 2011 to October 2019. Key eligibility criteria included clear cell histology, measurable metastatic disease, and OS &gt;60 days from CN, to avoid surgery-related deaths. Demographics and IMDC risk strata were collected from medical charts. Volumetry and radiomic features (first and second order) were assessed in the arterial phase abdominal computerized tomography (CT) of the primary lesion. Metastatic disease foci were segmented from venous phase CT and measured. Bone disease and &lt;1cm visceral metastases were disregarded for 3-axial measurement. After dimensionality reduction of the radiomic features with Uniform Manifold Approximation and Projection, an unsupervised Gaussian Mixture Model (GMM) clustering was performed. Characteristics of different clusters were compared with Mann-Whitney U test and univariate Cox regression estimated the effect on outcomes. Results: From 54 mRCC pts treated with CN, 39 were included. Median age was 60 years-old. Most patients were male (74%) and intermediate-risk by IMDC (64%). Sites of metastasis included lung (44%), bone (38%), and adrenal (21%). GMM identified 4 distinct clusters from radiomic features, which were grouped according to median OS relative to all pts (2.8 years): Group A (below mOS) and Group B (above mOS). The table shows the association of different factors with OS. Group B had longer median OS (3.4 vs. 1.8 years, p=0.009), progression-free survival (16.8 vs. 9.8 months, p=0.024), systemic treatment-free survival (9.5 vs. 3.8 months, p=0.028), and survival during first-line tyrosine kinase inhibitor (2.2 vs. 0.9 years, p=0.016). Group men had a lower primary-to-whole-body burden ratio compared to Group A (median 0.83 vs. 0.99, p&lt;0.01). Only IMDC risk criteria (HR 1.97, p&lt;0.01) and radiomic group B (HR 0.35, p=0.01) correlated with OS in univariate Cox analysis. Conclusions: Our study suggests that radiomics of the primary lesion may predict OS in mRCC pts after CN, despite limited by its retrospective nature and small sample size. Further validation in larger datasets is warranted. Univariate Cox proportional hazards of prognostic factors on OS. [Table: see text]

Histologically interpretable clot radiomic features predict treatment outcomes of mechanical thrombectomy for ischemic stroke.

Radiomics features (RFs) extracted from CT images may provide valuable information on the biological structure of ischemic stroke blood clots and mechanical thrombectomy outcome. Here, we aimed to identify RFs predictive of thrombectomy outcomes and use clot histomics to explore the biology and structure related to these RFs. We extracted 293 RFs from co-registered non-contrast CT and CTA. RFs predictive of revascularization outcomes defined by first-pass effect (FPE, near to complete clot removal in one thrombectomy pass), were selected. We then trained and cross-validated a balanced logistic regression model fivefold, to assess the RFs in outcome prediction. On a subset of cases, we performed digital histopathology on the clots and computed 227 histomic features from their whole slide images as a means to interpret the biology behind significant RF. We identified 6 significantly-associated RFs. RFs reflective of continuity in lower intensities, scattered higher intensities, and intensities with abrupt changes in texture were associated with successful revascularization outcome. For FPE prediction, the multi-variate model had high performance, with AUC = 0.832 ± 0.031 and accuracy = 0.760 ± 0.059 in training, and AUC = 0.787 ± 0.115 and accuracy = 0.787 ± 0.127 in cross-validation testing. Each of the 6 RFs was related to clot component organization in terms of red blood cell and fibrin/platelet distribution. Clots with more diversity of components, with varying sizes of red blood cells and fibrin/platelet regions in the section, were associated with RFs predictive of FPE. Upon future validation in larger datasets, clot RFs on CT imaging are potential candidate markers for FPE prediction.

Explainable Artificial Intelligence for Identification of Dosimetric Predictors of Radiation Toxicity: A Secondary Analysis of RTOG 0617

<h3>Purpose/Objective(s)</h3> Dosimetric predictors of toxicity in patients treated with definitive chemoradiation (CRT) for locally advanced non-small cell lung cancer have been identified, though these are often identified through trial and error or difficult to understand statistical methods. This study utilizes machine learning (ML) evaluated using explainable artificial intelligence (EAI) techniques to empirically characterize dosimetric predictors of toxicity in patients with LA-NSCLC treated with CRT as part of a prospective clinical trial. <h3>Materials/Methods</h3> A secondary analysis was performed to utilize ML EAI to identify dosimetric predictors of treatment-related toxicity in the RTOG 0617 trial. We trained a machine-learning based, gradient-boosted tree (GBT) models of ≥ grade 3 pulmonary, cardiac, and esophageal toxicities using Zubrod performance status, T stage, N stage, group stage, radiation technique, receipt of cetuximab, receipt of consolidative chemotherapy, and available dosimetric parameters. Shapley additive explanation (SHAP) values based on the GBT models were used to identify inflection points to select dosimetric thresholds. Identified thresholds were validated using univariate (UVA) and multivariable (MVA) logistic regression. <h3>Results</h3> Of the patients included on the trial, 471 patients had available dosimetry data. A total of 27 (5.7%), 56 (11.9%), and 45 (9.6%) patients experienced ≥ grade 3 pulmonary, esophageal, and cardiac toxicities, respectively. For pulmonary toxicities, using SHAP plots, lung V5≥57% and ≥82% were identified as an inflection points predictive of pneumonitis, which was validated using MVA, with odds ratios (OR) of 4.04 (p=0.004) and 5.11 (p=0.009), respectively. Upon inspection of lung V20, values ≥30% (OR: 3.23; p=0.011) and ≥39% (OR: 5.42; p=0.003) were predictive. Using mean lung dose, values ≥21 Gy were associated with increased toxicity (OR: 3.52; p=0.004). Use of 3DCRT was significantly associated with increased pneumonitis only in the V5 ≥57%, V5 ≥82%, V20 ≥39% MVAs. When assessing esophagitis, esophagus V20 ≥36% (OR: 4.76; p=0.010) and ≥74% (OR: 4.82; p=0.002) were associated with increased risk, as were mean values ≥21 Gy (OR: 4.66; p=0.004) and ≥32 Gy (OR: 3.30; p<0.001). In all of these analyses, radiation technique did not have a significant impact but receipt of consolidative chemotherapy was protective against esophagitis. On evaluation of cardiac dosimetric parameters, only heart V40 ≥57% (OR: 2.04; p=0.032) was predictive of cardiotoxicity. Only group stage was also significantly associated with cardiotoxicity. <h3>Conclusion</h3> Using validation with logistic regression, we demonstrated the feasibility of ML EAI approaches to identify dosimetric predictors of toxicity. Using this method, new dosimetric thresholds were identified, pending validation in larger datasets. Our findings support the utilization of ML EAI for empiric determination of radiation treatment planning considerations.

Predictive Models for Recurrent Membranous Nephropathy After Kidney Transplantation.

Background.Recurrent membranous nephropathy (MN) posttransplantation affects 35% to 50% of kidney transplant recipients (KTRs) and accounts for 50% allograft loss 5 y after diagnosis. Predictive factors for recurrent MN may include HLA-D risk alleles, but other factors have not been explored with certainty.Methods.The Australian and New Zealand Dialysis and Transplant registry was used to develop 3 prediction models for recurrent MN (Group Least Absolute Shrinkage and Selection Operator [LASSO], penalized Cox regression, and random forest), which were tuned using tenfold cross-validation in a derivation cohort with complete HLA data. KTRs with MN but incomplete HLA data formed the validation cohort. Model performance was evaluated using area under the receiver operating characteristic curve (AUC-ROC).Results.One hundred ninety-nine KTRs with MN were included, and 25 (13%) had recurrent MN (median follow-up 5.9 y). The AUC-ROCs for Group LASSO, penalized Cox regression, and random forest models were 0.85 (95% confidence interval, 0.76-0.94), 0.91 (0.85-0.96), and 0.62 (0.57-0.69), respectively, in the derivation cohort, with moderate agreement in selected variables between the models (55%-70%). In their validation cohorts, the AUC-ROCs for Group LASSO and penalized Cox regression were 0.60 (0.49-0.70) and 0.73 (0.59-0.86), respectively. Variables of importance chosen by all models included recipient HLA-A2, donor HLA-DR12, donor-recipient HLA-B65, and HLA-DR12 match.Conclusions.A penalized Cox regression performed reasonably for predicting recurrent MN and was superior to Group LASSO and random forest models. These models highlighted the importance of donor-recipient HLA characteristics to recurrent MN, although validation in larger datasets is required.

An exploratory application of machine learning methods to optimize prediction of responsiveness to digital interventions for eating disorder symptoms.

ObjectiveDigital interventions show promise to address eating disorder (ED) symptoms. However, response rates are variable, and the ability to predict responsiveness to digital interventions has been poor. We tested whether machine learning (ML) techniques can enhance outcome predictions from digital interventions for ED symptoms.MethodData were aggregated from three RCTs (n = 826) of self‐guided digital interventions for EDs. Predictive models were developed for four key outcomes: uptake, adherence, drop‐out, and symptom‐level change. Seven ML techniques for classification were tested and compared against the generalized linear model (GLM).ResultsThe seven ML methods used to predict outcomes from 36 baseline variables were poor for the three engagement outcomes (AUCs = 0.48–0.52), but adequate for symptom‐level change (R 2 = .15–.40). ML did not offer an added benefit to the GLM. Incorporating intervention usage pattern data improved ML prediction accuracy for drop‐out (AUC = 0.75–0.93) and adherence (AUC = 0.92–0.99). Age, motivation, symptom severity, and anxiety emerged as influential outcome predictors.ConclusionA limited set of routinely measured baseline variables was not sufficient to detect a performance benefit of ML over traditional approaches. The benefits of ML may emerge when numerous usage pattern variables are modeled, although this validation in larger datasets before stronger conclusions can be made.

Novel Post-Translational Modifications and Molecular Substrates in Glioma Identified by Bioinformatics.

Glioma is the most common type of brain cancer that originates from the glial cells. It constitutes about one-third of all brain cancers. Recently, transcriptomics, proteomics, and multiomics approaches have been harnessed to discover potential biomarkers and therapeutic targets in glioma. Moreover, post-translational modifications (PTMs) of proteins play a major role in cell biology and function and offer new avenues of research in cancer. Using unbiased multi-PTM bioinformatics analyses of two proteomic datasets of glioma available in the public domain, we identified 866 proteins with common PTMs from both studies. Out of these 866 proteins, 19 proteins were identified with the common PTMs, with the same site modifications pertaining to glioma. Importantly, the identified PTMs belonged to proteins involved in integrin PI3K/Akt/mTOR, JAK/STAT, and Ras/Raf/MAPK pathways. These pathways are essential for cell proliferation in tumor cells and thus involved in glioma progression. Taken together, these findings call for validation in larger datasets in glioma and brain cancers and with an eye to future drug discovery and diagnostic innovation. Bioinformatics-guided discovery of novel PTMs from the publicly available proteomic data can offer new avenues for innovation in cancer research.

Simplifying the ISCHEMIA trial algorithm for clinical practice: Identifying left main coronary artery disease using coronary artery calcium scans

Several recent trials have evaluated invasive versus medical therapy for stable ischemic heart disease. Importantly, patients with significant left main coronary stenosis (LMCS) were excluded from these trials. In the ISCHEMIA trial, these patients were identified by a coronary CT angiogram (CCTA), which adds time, expense, and contrast exposure. We tested whether a coronary artery calcium scan (CACS), a simpler, less expensive test, could replace CCTA to exclude significant LMCS. We hypothesized that patients with ≥50% LMCS would have a LM CACS score > 0. As a corollary, we postulated that a LM CACS=0 would exclude patients with LMCS. To test this, we searched Intermountain Healthcare's electronic medical records database for all adult patients who had undergone non-contrast cardiac CT for quantitative CACS scoring prior to invasive coronary angiography (ICA). Patients aged <50 and those with a heart transplant were excluded. Cases with incomplete (qualitative) angiographic reports for LMCS and those with incomplete or discrepant LM CACS results were reviewed and reassessed blinded to CACS or ICA findings, respectively. Among 669 candidate patients with CACS followed by ICA, 36 qualifying patients were identified who had a quantitative CACS score and LMCS ≥ 50%. Their age averaged 71.8 years, and 81% were men. Angiographic LMCS averaged 72% (range 50%-99%). Median time between CACS and ICA was 6 days. Total CACS score averaged 2,383 Agatston Units (AU), range 571-6,636. LM CACS score averaged 197 AU, range 31-610. Importantly, no LMCS patient had a LM CACS score of 0 vs 57% (362/633) of non-LMCS controls (P < .00001). Our results support the hypothesis that an easily administered, inexpensive, low radiation CACS can identify a large subset of patients with a very low risk of LMCS who would not have the need for routine CCTA. Using CACS to exclude LMCS may efficiently allow for safe implementation of an initial medical therapy strategy of patients with stable ischemic heart disease in clinical practice. These promising results deserve validation in larger data sets.

Can quantitative analysis of multi-parametric MRI independently predict failure of focal salvage HIFU therapy in men with radio-recurrent prostate cancer?

ObjectivesFocal salvage HIFU is a feasible therapeutic option in some men who have recurrence after primary radiotherapy for prostate cancer. We aimed to determine if multi-parametric quantitative parameters, in addition to clinical factors, might have a role in independently predicting focal salvage HIFU outcomes. MethodsA retrospective registry analysis included 150 consecutive men who underwent focal salvage HIFU (Sonablate500) (2006-2015); 89 had mpMRI available. Metastatic disease was excluded by nodal assessment on pelvic MRI, a radioisotope bone-scan and/or choline or FDG PET/CT scan. All men had mpMRI and either transperineal template prostate mapping biopsy or targeted and systematic TRUS-biopsy. mpMRI included T2‐weighted, diffusion‐weighted and dynamic contrast‐enhancement. Pre-HIFU quantitative mpMRI data was obtained using Horos DICOM Viewer v3.3.5 for general MRI parameters and IB DCE v2.0 plug-in. Progression-free survival (PFS) was defined by biochemical failure and/or positive localized or distant imaging results and/or positive biopsy and/or systemic therapy and/or metastases/prostate cancer‐specific death. Potential predictors of PFS were analyzed by univariable and multivariable Cox-regression. ResultsMedian age at focal salvage HIFU was 71 years (interquartile range [IQR] 65–74.5) and median PSA pre-focal salvage treatment was 5.8ng/ml (3.8-8). Median follow-up was 35 months (23-47) and median time to failure was 15 months (7.8–24.3). D-Amico low, intermediate and high-risk disease was present in 1% (1/89), 40% (36/89) and 43% (38/89) prior to focal salvage HIFU (16% missing data). 56% (50/89) failed by the composite outcome. A total of 22 factors were evaluated on univariable and 8 factors on multivariable analysis. The following quantitative parameters were included: Ktrans, Kep, Ve, Vp, IS, rTTP and TTP. On univariable analysis, PSA, prostate volume at time of radiotherapy failure and Ve (median) value were predictors for failure. Ve represents extracellular fraction of the whole tissue volume. On multivariable analysis, only Ve (median) value remained as an independent predictor. ConclusionsOne pharmacokinetic quantitative parameter based on DCE sequences seems to independently predict failure following focal salvage HIFU for radio-recurrent prostate cancer. This likely relates to the tumor microenvironment producing heat-sinks which counter the heating effect of HIFU. Further validation in larger datasets and evaluating mechanisms to reduce heat-sinks are required.

Cardioplegia Solution Related Ventricular Pacing Following Orthotopic Heart Transplantation

<h3>Purpose</h3> Multiple solutions are used for heart allograft preservation. University of Wisconsin (UW) remains the most widely used, although there has been increased utilization of histidine-tryptophan-ketoglutarate (HTK). This study investigated differences in the use of postoperative ventricular pacing (VP) at a single center following a change in allograft preservation from UW to HTK . <h3>Methods</h3> We conducted a retrospective study of 42 consecutive patients who underwent orthotopic heart transplantation between January 2017 and December 2018. 21 patients received UW-preserved allografts followed by 21 patients who received HTK-preserved allografts. Patient characteristics, comorbidities and complications were reviewed. Duration of VP and inotropic infusions in the postoperative period were compared between the two groups, UW and HTK. <h3>Results</h3> Table 1 shows pretransplant use of anti-arrhythmics and mechanical support and duration of inotropic support posttransplantation. Figure 1 graphs the number of patients requiring VP from postoperative day 0 to 15. Despite supplemental inotropy in all patients, there was a higher incidence of VP in the first 24 hours in the HTK group compared to UW (43% vs 5%, p=0.01). Preoperative amiodarone use did not correlate with incidence of VP. No patients required a permanent pacemaker. <h3>Conclusion</h3> Multiple studies have been performed investigating the impact of preservation solutions on heart transplant outcomes with mixed results. This hypothesis generating study suggests a higher incidence of bradyarrhythmias requiring VP in HTK-preserved allografts, warranting validation in larger datasets.

Pretreatment CT and 18 F-FDG PET-based radiomic model predicting pathological complete response and loco-regional control following neoadjuvant chemoradiation in oesophageal cancer.

To develop a radiomic-based model to predict pathological complete response (pCR) and outcome following neoadjuvant chemoradiotherapy (NACRT) in oesophageal cancer. We analysed 68 patients with oesophageal cancer treated with NACRT followed by esophagectomy, who had staging 18F-fluorodeoxyglucose (18 F-FDG) positron emission tomography (PET) and computed tomography (CT) scans performed at our institution. An in-house data-chjmirocterization algorithm was used to extract 3D-radiomic features from the segmented primary disease. Prediction models were constructed and internally validated. Composite feature, Fc =α*FPET +(1-α)*FCT , 0≤α≤1, was constructed for each corresponding CT and PET feature. Loco-regional control (LRC), recurrence-free survival (RFS), metastasis-free survival (MFS) and overall survival (OS) were estimated by Kaplan-Meier analysis, and compared using log-rank test. Median follow-up was 59months. pCR was achieved in 34 (50%) patients. Five-year RFS, LRC, MFS and OS were 67.1%, 88.5%, 75.6% and 57.6%, respectively. Tumour Regression Grade (TRG) 0-1 indicative of complete response or minimal residual disease was significantly associated with improved 5-year LRC [93.7% vs 71.8%; P=0.020; HR 0.19, 95% CI 0.04-0.85]. Four sepjmirote pCR predictive models were built for CT alone, PET alone, CT+PET and composite. CT, PET and CT+PET models had AUC 0.73±0.08, 0.66±0.08 and 0.77±0.07, respectively. The composite model resulted in an improvement of pCR predicting power with AUC 0.87±0.06. Stratifying patients with a low versus high radiomic score showed clinically relevant improvement in 5-year LRC favouring low-score group (91.1% vs. 80%, 95% CI 0.09-1.77, P=0.2). The composite CT/PET radiomics model was highly predictive of pCR following NACRT. Validation in larger data sets is warranted to determine whether the model can predict clinical outcomes.

Abstract 13551: Identifying Left Main Coronary Artery Disease Using Coronary Artery Calcium Scans

Introduction: The ISCHEMIA trial tested an invasive vs. an initial medical strategy in patients (pts) with stable coronary disease and evidence of ischemia. No significant difference between strategies in cardiovascular events was found at 3.2 years. However, pts were screened before randomization by coronary CT angiography (CCTA) to exclude ≥50% left main coronary stenosis (LMCS). CCTA adds complexity to routine medical practice, including time delays, expense, and safety concerns. We tested whether a coronary artery calcium scan (CACS), a simpler, less expensive test, could replace CCTA to exclude significant LMCS. Methods: We hypothesized that pts with ≥50% LMCS would have a LM CACS score&gt;0. As a corollary, we postulated that a LM CACS=0 would exclude pts with LMCS. To test this, we searched Intermountain Healthcare’s electronic medical records database for all adult pts who had undergone non-contrast cardiac CT for quantitative CACS scoring prior to selective coronary angiography (SCA) and were found to have a LMCS ≥50%. Pts aged &lt;50 and those with a heart transplant were excluded. Cases with incomplete (qualitative) angiographic reports for LMCS and those with incomplete or discrepant LM CACS results were reviewed and reassessed blinded to CACS or SCA findings, respectively. Results: Among 674 candidate pts with CACS followed by SCA, 24 qualifying pts were identified who had a quantitative CACS score and LMCS ≥50%. Their age averaged 71 ± 11 years, and 83% were men. Angiographic LMCS averaged 77% (range 50%-99%). A heavy burden of both total CAC and LM CAC was typically present. Total CACS score averaged 2,545 Agatston Units (AU), range 571-6,636. LM CACS score averaged 214 AU, range 47-610. Importantly, no LMCS pt had a LM CACS score of 0 vs. 57% (368/650) of non-LMCS controls (p&lt;0.00001). Conclusions: Our results support the hypothesis that an easily administered, inexpensive, low radiation CACS can identify a large subset of pts with a very low risk of LMCS without the need for routine CCTA. Using CACS to exclude LMCS may efficiently allow for safe implementation of an initial medical therapy strategy in clinical practice for ISCHEMIA trial-like pts with at least moderate ischemia on stress testing. These promising results deserve validation in larger data sets.