Search

Nonoperative management with short-course radiation therapy (SCRT) as a component of definitive therapy for oligometastatic rectal cancer has not been previously reported. This single-institution retrospective analysis evaluates treatment with SCRT in combination with chemotherapy (SCRT-CTX) with nonoperative intent for patients with a locoregional clinical complete response (cCR).Thirty-six patients with newly diagnosed oligometastatic rectal cancer were treated with SCRT-CTX between January 1, 2018, and May 31, 2020. Digital rectal examination, endoscopy, and imaging (computed tomography or magnetic resonance imaging) were used to determine cCR. Medically operable patients without cCR underwent surgical resection of the primary rectal tumor. Patients with cCR who experienced a local failure received salvage surgery. Rates of hospitalization related to primary tumor disease and pelvic symptoms were reviewed. Overall survival (OS) and progression free survival were evaluated.Seventeen percent (6/36) of patients achieved cCR after SCRT-CTX. Eleven percent (4) of patients experienced a local failure. OS for all patients was 83% (71%-96%) at 12 months and 57% (41%-80%) at 24 months. Progression free survival for all patients was 56% (41%-74%) at 12 months and 10% (3.1%-35%) at 24 months. On multivariate analysis, having received more than 4 months of chemotherapy (hazard ratio = 0.21; 95% confidence interval, 0.06-0.71; P = .01) and definitive treatment of metastatic site (hazard ratio = 0.17; 95% confidence interval, 0.05-0.66; P = .01) predicted for improved OS. The number of patients requiring hospitalization due to obstruction (8/36, 22%), rectal bleeding (5/36, 14%), or need for permanent ostomy placement (5/36, 14%) was low, and there was a decrease in endorsement of obstructive symptoms and rectal bleeding after completion of SCRT-CTX.SCRT-CTX with nonoperative intent for patients with a locoregional cCR may be a reasonable treatment option for patients with newly diagnosed oligometastatic rectal adenocarcinoma and demonstrates excellent control of pelvic disease and symptoms. Increased duration of chemotherapy within the treatment paradigm may improve oncologic outcomes.

Purpose:Targeting focal adhesion kinase (FAK) renders checkpoint immunotherapy effective in pancreatic ductal adenocarcinoma (PDAC) mouse model. Defactinib is a highly potent oral FAK inhibitor that has a tolerable safety profile.Patients and Methods:We conducted a multicenter, open-label, phase I study with dose escalation and expansion phases. In dose escalation, patients with refractory solid tumors were treated at five escalating dose levels of defactinib and gemcitabine to identify a recommended phase II dose (RP2D). In expansion phase, patients with metastatic PDAC who progressed on frontline treatment (refractory cohort) or had stable disease (SD) after at least 4 months of standard gemcitabine/nab-paclitaxel (maintenance cohort) were treated at RP2D. Pre- and posttreatment tumor biopsies were performed to evaluate tumor immunity.Results:The triple drug combination was well-tolerated, with no dose-limiting toxicities. Among 20 treated patients with refractory PDAC, the disease control rate (DCR) was 80%, with one partial response (PR) and 15 SDs, and the median progression-free survival (PFS) and overall survival (OS) were 3.6 and 7.8 months, respectively. Among 10 evaluable patients in the maintenance cohort, DCR was 70% with one PR and six SDs. Three patients with SD came off study due to treatment- or disease-related complications. The median PFS and OS on study treatment were 5.0 and 8.3 months, respectively.Conclusions:The combination of defactinib, pembrolizumab, and gemcitabine was well-tolerated and safe, had promising preliminary efficacy, and showed biomarker activity in infiltrative T lymphocytes. Efficacy of this strategy may require incorporation of more potent chemotherapy in future studies.

Supplemental Table 2 Patient demographics

This selection from the NCCN Guidelines for Rectal Cancer focuses on management of malignant polyps and resectable nonmetastatic rectal cancer because important updates have been made to these guidelines. These recent updates include redrawing the algorithms for stage II and III disease to reflect new data supporting the increasingly prominent role of total neoadjuvant therapy, expanded recommendations for short-course radiation therapy techniques, and new recommendations for a “watch-and-wait” nonoperative management technique for patients with cancer that shows a complete response to neoadjuvant therapy. The complete version of the NCCN Guidelines for Rectal Cancer, available online at NCCN.org, covers additional topics including risk assessment, pathology and staging, management of metastatic disease, posttreatment surveillance, treatment of recurrent disease, and survivorship.

Modeling outdoor environmental sound levels is a challenging problem. This paper reports on a validation study of two continental-scale machine learning models using geospatial layers as inputs and the summer daytime A-weighted L50 as a validation metric. The first model was developed by the National Park Service while the second was developed by the present authors. Validation errors greater than 20 dBA are observed. Large errors are attributed to limited acoustic training data. Validation environments are geospatially dissimilar to training sites, requiring models to extrapolate beyond their training sets. Results motivate further work in optimal data collection and uncertainty quantification.

Ablative radiation therapy for borderline resectable or locally advanced pancreatic ductal adenocarcinoma (BR/LA-PDAC) may limit concurrent chemotherapy dosing and usually is only safely deliverable to tumors distant from gastrointestinal organs. Magnetic resonance guided radiation therapy may safely permit radiation and chemotherapy dose escalation.We conducted a single-arm phase I study to determine the maximum tolerated dose of ablative hypofractionated radiation with full-dose gemcitabine/nab-paclitaxel in patients with BR/LA-PDAC. Patients were treated with gemcitabine/nab-paclitaxel (1000/125 mg/mThirty patients enrolled (March 2015-February 2019), with 26 evaluable patients (2 progressed before radiation, 1 was determined ineligible for radiation during planning, 1 withdrew consent). One DLT was observed. The DLT rate was 14.1% (3.3%-24.9%) with a maximum tolerated dose of gemcitabine/nab-paclitaxel (1000/100 mg/mFull-dose gemcitabine/nab-paclitaxel with ablative magnetic resonance guided radiation therapy dosing is safe in patients with BR/LA-PDAC, with promising LPFS and DMFS.

Short-course radiotherapy and total neoadjuvant therapy (SCRT-TNT) followed by total mesorectal excision (TME) has emerged as a new treatment paradigm for patients with locally advanced rectal adenocarcinoma. However, the economic implication of this treatment strategy has not been compared with that of conventional long-course chemoradiotherapy (LCCRT) followed by TME with adjuvant chemotherapy.To perform a cost-effectiveness analysis of SCRT-TNT vs LCCRT in conjunction with TME for patients with locally advanced rectal cancer.A decision analytical model with a 5-year time horizon was constructed for patients with biopsy-proven, newly diagnosed, primary locally advanced rectal adenocarcinoma treated with SCRT-TNT or LCCRT. Markov modeling was used to model disease progression and patient survival after treatment in 3-month cycles. Data on probabilities and utilities were extracted from the literature. Costs were evaluated from the Medicare payer's perspective in 2020 US dollars. Sensitivity analyses were performed for key variables. Data were collected from October 3, 2020, to January 20, 2021, and analyzed from November 15, 2020, to April 25, 2021.Two treatment strategies, SCRT-TNT vs LCCRT with adjuvant chemotherapy, were compared.Cost-effectiveness was evaluated using the incremental cost-effectiveness ratio and net monetary benefits. Effectiveness was defined as quality-adjusted life-years (QALYs). Both costs and QALYs were discounted at 3% annually. Willingness-to-pay threshold was set at $50 000/QALY.During the 5-year horizon, the total cost was $41 355 and QALYs were 2.21 for SCRT-TNT; for LCCRT, the total cost was $54 827 and QALYs were 2.12, resulting in a negative incremental cost-effectiveness ratio (-$141 256.77). The net monetary benefit was $69 300 for SCRT-TNT and $51 060 for LCCRT. Sensitivity analyses using willingness to pay at $100 000/QALY and $150 000/QALY demonstrated the same conclusion.These findings suggest that SCRT-TNT followed by TME incurs lower cost and improved QALYs compared with conventional LCCRT followed by TME and adjuvant chemotherapy. These data offer further rationale to support SCRT-TNT as a novel cost-saving treatment paradigm in the management of locally advanced rectal cancer.

Background: To date, immunotherapy has not advanced the initial care of locally advanced rectal cancer (LARC) where a combination of radiation and chemotherapy remains the standard of care. In evaluation of other targets, we identified that the immune-oncology target indoleamine 2,3 dioxygenase-1 (IDO1) is universally overexpressed in rectal cancers after radiation therapy (RT) regardless of MSI status. IDO1 initiates tryptophan metabolism along the kynurenine pathway (KP) and activation of the KP exerts immune-suppressive effects in the tumor microenvironment. Our preclinical data identified that epacadostat, a potent inhibitor of IDO1, reduced CRC growth in vitro and in vivo. Furthermore, we found that epacadostat augmented the anti-tumor effects of RT by directly increasing apoptosis in neoplastic cells as well as by relieving RT induced immune-suppression. Based on these findings, we initiated a Phase I study in patients with LARC where epacadostat was used in combination with short-course RT (5 Gy x 5) and chemotherapy. This study is ongoing with the primary endpoint of determining the recommended phase II dose (RP2D) of epacadostat in this population. The study is a dose escalating design including 300, 400, and 600 mg BID. Aim: To present preliminary results from the corollary studies on sequentially acquired biospecimens. Methods: We collected tumor tissue and blood pre-treatment, post-radiation/pre-chemo, sequentially throughout chemo, and at treatment completion or surgery. Analyses include global metabolomics, focused evaluation of tryptophan pathway metabolites, single cell RNAseq, flow cytometry, ctDNA and tissue analysis for IDO1 and other immune checkpoint targets. We also established tumor-derived organoids/tumoroids and have established a “tumor-on-a-chip” model to conduct ex vivo tumor modeling. Results: Preliminary studies confirm an elevation of kynurenine to tryptophan ratio after SCRT which is reversed by epacadostat. Changes in the global metabolomics profiles and immune profiles shift from baseline through the stages of SCRT and chemotherapy revealing consistent patterns. Changes in ctDNA were assessed using the CAPP-Seq based AVENIO platform providing a personalized oncogenomic analysis and demonstrated of rapid clearance from post-SCRT which persisted throughout chemotherapy. Studies with tumoroids and the analysis of scRNAseq data are ongoing and will be presented. Conclusions: Preliminary analysis of biospecimens from this actively accruing study of epacadostat plus RT and chemotherapy reveals distinct changes in metabolomic, immunologic, and genomic profiling across the stages of treatment. These results have the potential to inform a precision design for the Phase II study and will also inform the most promising disease and treatment related biomarkers. Citation Format: Matthew A. Ciorba, Katrina Pedersen, Lauren Henke, Aadel Chaudhuri, Matthew Mutch, Hyun Kim, Haeseong Park. Biomarkers of IDO1 inhibition in patients with locally advanced rectal cancer treated with radiation and chemotherapy [abstract]. In: Proceedings of the AACR Special Conference on Colorectal Cancer; 2022 Oct 1-4; Portland, OR. Philadelphia (PA): AACR; Cancer Res 2022;82(23 Suppl_1):Abstract nr B020.

The NCCN Guidelines for Rectal Cancer provide recommendations for the diagnosis, evaluation, treatment, and follow-up of patients with rectal cancer. These NCCN Guidelines Insights summarize the panel discussion behind recent important updates to the guidelines. These updates include clarifying the definition of rectum and differentiating the rectum from the sigmoid colon; the total neoadjuvant therapy approach for localized rectal cancer; and biomarker-targeted therapy for metastatic colorectal cancer, with a focus on new treatment options for patients with BRAF V600E- or HER2 amplification-positive disease.

This paper describes the development of an automated classification algorithm for detecting instances of focused crowd involvement present in crowd cheering. The purpose of this classification system is for situations where crowds are to be rewarded for not just the loudness of cheering, but for a concentrated effort, such as in Mardi Gras parades to attract bead throws or during critical moments in sports matches. It is therefore essential to separate non-crowd noise, general crowd noise, and focused crowd cheering efforts from one another. The importance of various features—both spectral and low-level audio processing features—are investigated. Data from both parades and sporting events are used for comparison of noise from different venues. This research builds upon previous clustering analyses of crowd noise from collegiate basketball games, using hierarchical clustering as an unsupervised machine learning approach to identify low-level features related to focused crowd involvement. For Mardi Gras crowd data we use a continuous thresholding approach based on these key low-level features as a method of identifying instances where the crowd is particularly active and engaged.

Outdoor ambient acoustical environments may be predicted through supervised machine learning using geospatial features as inputs. Previous work used K-Means clustering applied to the geospatial features to identify geographic regions that are geospatially distinct. The clustering results help provide physical insights regarding which features are likely to play the largest roles in the supervised learning model and which locations are impacted by different acoustic training data. However, these results may be sensitive to details of the geospatial data, such as how the data are scaled or the presence of similar redundant features. This work builds on previous results by constructing a reduced feature set by removing redundant geospatial features and by using a physically motivated scaling scheme. Clustering analysis applied to this new dataset indicates that the contiguous United States can be naturally clustered into eight human-interpretable geographic regions. Hierarchical clustering is used to further subdivide these eight clusters into more fine-grained regions. One finding of interest is that no geospatial layer in the present soundscape model uniquely identifies rivers. These results will guide further geospatial layer development and acoustical data collection for more accurate soundscape models. [Work supported by a U.S. Army SBIR.]

TPS793 Background: Pancreatic ductal adenocarcinoma (PDAC) is predicted to be the second leading cause of cancer-related death by 2030, and is characterized by resistance to chemo- and radiotherapy and a highly fibrotic tumor microenvironment. Front-line therapies for advanced PDAC include FOLFIRINOX and gemcitabine/nab-paclitaxel with median overall survival ranging from 8.5 to 11 months. After progression on gemcitabine-containing therapy, 5-FU/LV/liposomal irinotecan is a standard second-line option, however outcomes are still poor. Retrospective studies demonstrate superior survival of advanced PDAC in patients with high serum levels of 25(OH) vitamin D. Notably, the PDAC tumor microenvironment is enriched in cancer-associated fibroblasts that favorably respond to vitamin D, prolonging survival in combination with chemotherapy in mouse models. Furthermore, vitamin D suppresses catabolism of irinotecan in gastrointestinal cancer cells, potentiating its efficacy. Therefore, we are conducing an investigator-initiated study of 5FU/LV/liposomal irinotecan with paricalcitol as second-line therapy in advanced PDAC. Methods: This is a pilot study of 5FU/LV/liposomal irinotecan combined with paricalcitol in patients with advanced PDAC progressed on gemcitabine-based therapy. All patients receive liposomal irinotecan, LV, 5-FU and paricalcitol. Liposomal irinotecan is given at 70 mg/m2 IV over 90 minutes, LV at 400 mg/m2 IV over 30 minutes, and 5-FU at 2400 mg/m2 continuous IV infusion over 46 hours, on Day 1 of each 14-day cycle. Paricalcitol IV infusion will precede the above, given according to assigned cohort (75 mcg weekly or 7 mcg/kg weekly). The primary objective of this study is to determine the tolerability between two different dose levels of paricalcitol added to the combination regimen of 5-FU/LV/liposomal irinotecan in patients with advanced PDAC. Secondary objectives are measures of efficacy (ORR, PFS, OS, CA19-9 biochemical response rate). Clinical trial information: NCT03883919.

TPS593 Background: Biliary tract cancers (BTCs) are rare and aggressive malignancies. The current standard of care for advanced BTC is gemcitabine (GEM) plus cisplatin . Although there is no established second-line treatment, regimens such as FOLFOX, XELOX, FOLFIRI, XELIRI, GEM, and capecitabine have activity. Nal-IRI contains IRI free base encapsulated in liposome nanoparticles which shelter IRI from conversion to its active metabolite (SN-38) and increase intratumoral levels of SN-38 compared with IRI. FU/LV/nal-IRI has shown overall survival benefit and acceptable toxicity in patients (pts) with metastatic pancreatic adenocarcinoma following GEM-based therapy in the NAPOLI-1 trial. Methods: This is a single arm, open label, multicenter phase II study of pts with advanced BTC previously treated with gemcitabine plus platinum chemotherapy. Pts will receive nal-IRI 70 mg/m2 IV over 90 minutes, LV 400 mg/m2 IV over 30 minutes, and FU 2400 mg/m2 over 46 hours, every 14 days. The primary objective is to determine progression-free survival (PFS) rate at 4 months (4mo) using RECIST v. 1.1 criteria and central radiology review. Response assessments will occur using imaging every 8 weeks. All pts who receive at least 1 dose of the study treatment will be eligible for the primary analysis. We will substitute pts who screen fail or do not begin treatment. Median PFS reported for pts receiving second-line 5-FU doublet chemotherapy is 3 months with a PFS4mo of 30%. FU/LV/nal-IRI would be of interest if it could increase the PFS4mo to 50% or higher. We will use a 2-stage Simon Minimax design. Using a one-sided α of 0.05 and 80% power, 39 pts will be required to detect a difference in PFS4mo between 30% and 50%. Assuming a dropout rate of 10%, 44 pts will be enrolled across the 5 study sites. Enrollment began in Q2 2019. Clinical trial information: NCT04005339.

The relationship between crowd noise and crowd behavioral dynamics is a relatively unexplored field of research. Signal processing and machine learning (ML) may be useful in classifying and predicting crowd emotional state. This paper describes using both supervised and unsupervised ML methods to automatically differentiate between different types of crowd noise. Features used include A-weighted spectral levels, low-level audio signal parameters, and Mel-frequency cepstral coefficients. K-means clustering is used for the unsupervised approach with spectral levels, and six distinct clusters are found; four of these clusters correspond to different amounts of crowd involvement, while two correspond to different amounts of band or public announcement system noise. Random forests are used for the supervised approach, wherein validation and testing accuracies are found to be similar. These investigations are useful for differentiating between types of crowd noise, which is necessary for future work in automatically determining and classifying crowd emotional state.

Outdoor ambient acoustical environments may be predicted through machine learning using geospatial features as inputs. However, collecting sufficient training data is an expensive process, particularly when attempting to improve the accuracy of models based on supervised learning methods over large, geospatially diverse regions. Unsupervised machine learning methods, such as K-Means clustering analysis, enable a statistical comparison between the geospatial diversity represented in the current training dataset versus the predictor locations. In this case, 117 geospatial features that represent the contiguous United States have been clustered using K-Means clustering. Results show that most geospatial clusters group themselves according to a relatively small number of prominent geospatial features. It is shown that the available acoustic training dataset has a relatively low geospatial diversity because most training data sites reside in a few clusters. This analysis informs the selection of new site locations for data collection that improve the statistical similarity of the training and input datasets.

Outdoor ambient sound levels can be predicted from machine learning-based models derived from geospatial and acoustic training data. To improve modeling robustness, median predicted sound levels have been calculated from an ensemble of tuned models from different supervised machine learning modeling classes. The ensemble is used to predict ambient sound levels throughout the contiguous United States. The training data set consists of 607 unique sites, where various acoustic metrics, such as overall daytime L50 levels and one-third octave frequency band levels, have been obtained. Data for 117 geospatial features, which include metrics such as distance to the nearest road or airport, are used. The spread in the ensemble provides an estimate of the modeling accuracy. Results of an initial leave-one-out and leave-four-out validation study are presented.

This paper discusses methods used to identify the relative importance of audio features in a supervised machine learning (ML) model for predicting crowd behavior at collegiate basketball games. This work builds upon previous research done at Brigham Young University in which ML classifiers were trained using audio recordings of crowds at collegiate basketball games. Previous classifiers were built using several hundred features. A future goal is real-time classification of crowd behavior. This requires reduced computational time for calculating features and classifying audio data. Feature reduction should decrease the computational time for both of these tasks. The audio features can be separated into two categories: (1) spectral features and (2) low-level signal parameters. This paper discusses feature reduction methods—such as random forest Gini importance and p-value feature selection—and compares reduced feature sets. The number of features used in crowd noise classification can be reduced to 10-15 features before seeing a significant decrease in prediction accuracy.

Manifold learning is a type of nonlinear dimensionality reduction that helps identify a minimal combination of features to characterize data. This presentation discusses the use of diffusion maps, a type of manifold learning, on a dataset of 68 geospatial features which cover the contiguous United States. These geospatial features have been used previously to make predictions of ambient soundscapes through an ensemble of machine learning models. As the current model capabilities are expanded to predict the ambient soundscape outside of the contiguous United States, decisions must be made about which geospatial features are required for accurate models. In particular, some of these 68 features are not available or are expensive to obtain for regions outside of the contiguous United States. Diffusion maps can assist in identifying the features, or combinations of features, that best characterize the data space. [Work supported by a U.S. Army SBIR.]Manifold learning is a type of nonlinear dimensionality reduction that helps identify a minimal combination of features to characterize data. This presentation discusses the use of diffusion maps, a type of manifold learning, on a dataset of 68 geospatial features which cover the contiguous United States. These geospatial features have been used previously to make predictions of ambient soundscapes through an ensemble of machine learning models. As the current model capabilities are expanded to predict the ambient soundscape outside of the contiguous United States, decisions must be made about which geospatial features are required for accurate models. In particular, some of these 68 features are not available or are expensive to obtain for regions outside of the contiguous United States. Diffusion maps can assist in identifying the features, or combinations of features, that best characterize the data space. [Work supported by a U.S. Army SBIR.]

11586 Background: Hand-foot skin reaction (HFSR) is the most common regorafenib adverse event. However, no intervention has demonstrated efficacy in preventing or palliating this adverse event, short of modulating the regorafenib dose. Methods: The Regorafenib Dose Optimization Study (ReDOS) was a randomized trial ( https://clinicaltrials.gov/ct2/show/NCT02368886 ); here we describe an a priori secondary analysis of ReDOS with the goal of assessing whether clobetasol 0.05% cream (a potent corticosteroid) applied to the palms and soles twice per day is more effective when prescribed pre-emptively (before the development of HFSR) versus reactively (after the development of HFSR). Patients were assessed during the 1st 2 cycles of regorafenib. Results: Among 116 evaluable patients, 61 received pre-emptive clobetasol, and 55 reactive clobetasol. Baseline demographics were comparable between groups. During the 1st regorafenib cycle, 46% and 52% of patients developed HFSR with pre-emptive and reactive clobetasol, respectively (p = 0.52). However, during the 2nd cycle, 56% (of 47 total patients) and 80% (of 41 total patients), developed this toxicity with pre-emptive and reactive clobetasol, respectively (p = 0.02). Of note, during the 2nd cycle, rates of grade 1,2, and 3 HFSR were 40%, 11%, and 4% (pre-emptive) and 45%, 25%, and 10% (reactive) (p = 0.07). A sensitivity analysis to examine rates of no HFSR over all 2 regorafenib cycles showed that no HFSR occurred in 33% (of 61 total patients) with pre-emptive clobetasol versus 15% (of 55 total patients) with reactive clobetasol (p = 0.02). Patient-reported outcomes showed HFSR compromised activities of daily living, including getting dressed, preparing meals, walking, and even driving, with seemingly worse descriptive outcomes in patients who received reactive therapy. No adverse events from clobetasol were reported. Conclusions: Compared to reactive therapy, pre-emptive topical clobetasol might lessen regorafenib-induced HFSR. Clinical trial information: NCT02368886.

This paper describes using both supervised and unsupervised machine learning (ML) methods to improve automatic classification of crowd responses to events at collegiate basketball games. This work builds on recent investigations by the research team where the two ML approaches were treated separately. In one case, crowd response events (cheers, applause, etc.) were manually labeled, and then, a subset of the labeled events were used as a training set for supervised-ML event classification. In the other, (unsupervised) k-means clustering was used to divide a game’s one-twelfth octave spectrogram into six distinct clusters. A comparison of the two approaches shows that the manually labeled crowd responses are grouped into only one or two of the six unsupervised clusters. This paper describes how the supervised ML labels guide improvements to the k-means clustering analysis, such as determining which additional audio features are required as inputs and how both approaches can be used in tandem to improve automated classification of crowd noise at basketball games. This paper describes using both supervised and unsupervised machine learning (ML) methods to improve automatic classification of crowd responses to events at collegiate basketball games. This work builds on recent investigations by the research team where the two ML approaches were treated separately. In one case, crowd response events (cheers, applause, etc.) were manually labeled, and then, a subset of the labeled events were used as a training set for supervised-ML event classification. In the other, (unsupervised) k-means clustering was used to divide a game’s one-twelfth octave spectrogram into six distinct clusters. A comparison of the two approaches shows that the manually labeled crowd responses are grouped into only one or two of the six unsupervised clusters. This paper describes how the supervised ML labels guide improvements to the k-means clustering analysis, such as determining which additional audio features are required as inputs and how both approaches can be used in tandem to improve au...

486 Background: Total Neoadjuvant Therapy (TNT), or delivery of all radiation and chemotherapy prior to surgery, has improved complete response and downstaging rates compared to adjuvant therapy in patients with rectal cancer. Data regarding the use of short course radiation in the setting of TNT (SC-TNT) are limited. This study compares the pathologic complete response rate (pCR), Neoadjuvant Rectal (NAR) Score – a validated predictor of outcome based on tumor downstaging, and recurrence rates for patients receiving SC-TNT versus chemoradiation (CRT). Methods: Patients who underwent neoadjuvant therapy followed by total mesorectal excision for Stage II or III rectal cancer from 2009 to 2018 were included in this retrospective cohort study. CRT recipients (50-55Gy/25-28 fx with concurrent 5-FU or capecitabine) comprised one cohort; the other included SC-TNT recipients (25-35Gy/5 fx followed by CAPOX or FOLFOX chemotherapy). The primary outcome of pCR rate was assessed in univariate analysis; the secondary outcome of NAR score was calculated and categorized as “Low” ( < 8), “Intermediate” (8–16), and “High” ( > 16). Finally, recurrence rates were measured and classified as local, distant, or both. Results: Of 388 eligible patients, 236 (60.8%) were treated with CRT and 152 (39.2%) underwent SC-TNT. On univariate analysis, the SC-TNT cohort had more advanced disease (77% Stage III disease vs. 67%, p = 0.04) and longer elapsed time between radiation completion and surgery (Median 131 vs. 63 days; p < 0.01). SC-TNT achieved a numerically higher pCR rate compared to CRT (25.0% vs. 19.1%, p = 0.16). Odds of achieving a “low” NAR Score trended higher among the SC-TNT cohort (OR 1.49, 95% CI 0.96 – 2.31). Recurrence rates were also similar (14.3% vs. 14.9%, p = 0.87) over comparable follow-up (CRT = 30.5 months [IQR 11.1 – 49.0]; SC-TNT = 22.3 months [IQR 10.8 – 61.0]; p = 0.82). Conclusions: SC-TNT yielded a pCR rate of 25% and overall recurrence rate of 14.9% among patients with locally advanced rectal cancer. Short course radiation with neoadjuvant multiagent chemotherapy is at least as effective as long-course CRT. [Table: see text]

Ambient soundscape data are a critical component of numerous applications such as evaluating the environmental noise impacts of proposed activities and studying the physical and psychological impacts of a person’s acoustic environment. The current approach to characterize ambient soundscapes requires intensive field measurement programs. A soundscape prediction tool capable of modeling the soundscape over large spatial regions and timeframes is needed to provide an alternative to intensive field measurements. To address this need, a machine learning based soundscape model trained with acoustic data and geospatial layers has been developed. The model’s capability to generate A and flat-weighted exceedance levels across space, time (hourly, daytime, and nighttime), and frequency (one-third octave bands) will be demonstrated. [Work funded by an Army SBIR.]

Outdoor ambient sound levels can be predicted from machine learning-based models derived from geospatial and acoustic training data. To improve modeling robustness, median predicted sound levels have been calculated using tuned models from different supervised machine learning modeling classes. The ensemble-based model reduces errors at training sites for both overall levels and spectra, and produces more physically reasonable predictions elsewhere. Furthermore, the spread in the ensemble provides an estimate of the modeling accuracy. An initial analysis of feature importance metrics suggests that the number of geospatial inputs can be reduced from 120 to 15 without significant degradation of the model's predictive error, as measured by leave-one-out cross validation. However, the predictions from the reduced-feature modeling may be less physical in certain regions when all differentiating geospatial features are removed. These results suggest the need for more sophisticated data collection and validation methods.Outdoor ambient sound levels can be predicted from machine learning-based models derived from geospatial and acoustic training data. To improve modeling robustness, median predicted sound levels have been calculated using tuned models from different supervised machine learning modeling classes. The ensemble-based model reduces errors at training sites for both overall levels and spectra, and produces more physically reasonable predictions elsewhere. Furthermore, the spread in the ensemble provides an estimate of the modeling accuracy. An initial analysis of feature importance metrics suggests that the number of geospatial inputs can be reduced from 120 to 15 without significant degradation of the model's predictive error, as measured by leave-one-out cross validation. However, the predictions from the reduced-feature modeling may be less physical in certain regions when all differentiating geospatial features are removed. These results suggest the need for more sophisticated data collection and validation...

276 Background: BL-8040 is a novel CXCR4 antagonist being developed for multiple oncology indications. Preclinical studies demonstrated that BL-8040 increases the number of immune cells in peripheral blood and promotes CD8+ T cell infiltration into orthotropic pancreatic mouse tumors, reducing tumor load. BL-8040 is being evaluated in a Phase 2a, multicenter, open label trial in patients with metastatic pancreatic cancer (the COMBAT study). Patients are undergoing a 5-day period of monotherapy in which they receive daily doses of BL-8040, followed by 21-day cycles in which patients receive one dose of pembrolizumab and 3 doses/week of BL-8040 until disease progression or discontinuation. To date, 32 patients have been enrolled. Methods: On Day 1 and Day 5, blood samples were taken at pre- and post-dosing, to evaluate peripheral immune cell subset frequency by flow cytometry. In addition, core biopsies were taken from liver metastases, where possible, to assess immune cell infiltration into tumors and the tumor microenvironment (TME). Results: Here we present interim PD biomarker data from the BL-8040 monotherapy portion of the trial. Flow cytometry shows that BL-8040 monotherapy caused an approximately two-fold reduction in frequency of peripheral T regulatory cells, but had no effect on the frequency of T cells, NKT cells or cell populations that contain B cells (CD3- CD56-). Additionally, BL-8040 remained bound to CXCR4 on peripheral immune cells throughout the period of monotherapy. Analysis of available biopsies (N = 7) shows an up to 15-fold increase in the CD3+ population, and up to two-fold increase of CD8+ cells, in the tumor periphery and TME of 43% (3/7) of the patients after five days of BL-8040 monotherapy compared to baseline. Conclusions: In summary, the PD biomarker results in humans support the proposed mechanism of action for BL-8040 that was based on preclinical mouse models. Analysis of tumor biopsies is ongoing, with an emphasis on investigating the effects of BL-8040 on tumor-resident immune cells and the TME. Clinical trial information: NCT02826486.

611 Background: Regorafenib is an oral multikinase inhibitor with survival benefit in refractory mCRC patients (pts). Toxicities such as Palmar-plantar erythrodysesthesia syndrome (PPES), fatigue and hypertension (HTN) have limited its use. Despite absence of supportive data, various dosing or interval scheduling have been implemented into clinical practice. Methods: A randomized phase II study of regorafenib dose-escalation (Arm A: 80 mg/day, weekly dose escalation if no significant drug-related toxicities, up to 160 mg/day) vs. standard dose (Arm B: 160 mg/day) in pts with mCRC for 21 days of a 28-day cycle. Pts were randomized 1:1:1:1 to arms A1 and B1 (Pre-emptive Clobetasol for PPES); A2 and B2 (Reactive Clobetasol). The primary endpoint was the proportion of patients who completed 2 cycles of treatment and initiated the 3rd in Arm A (Pooled A1 + A2) vs. Arm B (Pooled B1 + B2). Superiority for Arm A was to be declared if the one-sided p-value calculated using Fisher’s exact method was less than 0.2. Results: From June 2015 to June 2017, 123 pts were randomized with 116 (A = 54, B = 62) evaluable for the primary endpoint. Demographic data were well balanced with overall median age of 61yrs (range: 29-81), M/F (61/39%) and ECOG PS 0/1 (37/63%) and KRAS MT/WT/UNK (47/44/9%).The study met its primary endpoint with 43% of pts on Arm A initiating the 3rd vs. only 25% of pts Arm B [one-sided p-value 0.028]. Median Overall Survival (OS) was improved in Arm A vs. Arm B (9.0 mos vs. 5.9 mos; p = 0.094). Median Progression Free Survival (PFS) was 2.5 mos for Arm A vs. 2.0 mos for Arm B (p=0.553). Overall rates of grade 3/4 toxicity were more favorable for Arm A vs. Arm B (% PPES 15 vs. 16, HTN 7 vs. 15 and fatigue 13 vs. 18, respectively). Multiple QOL parameters were improved in A vs. B primarily at week 2 of the first cycle. Conclusion: A strategy with weekly dose escalation of regorafenib from 80 mg to 160 mg/day was found to be superior to a starting dose of 160 mg/day. These results establish a new standard for optimizing regorafenib dosing. Further data on outcomes of preemptive vs. reactive clobetasol strategies are undergoing analysis and will be presented later. Clinical trial information: NCT02368886.

746 Background: Peritoneal carcinomatosis (PC) is the intraperitoneal spread of cancer. Optimal treatment for PC is controversial. Systemic chemotherapy offers limited benefit (Franko, 2011). Intraperitoneal (IP) chemotherapy following cytoreductive surgery (CRS) improves outcomes (Verwaal, 2008). Oxaliplatin (Ox), capecitabine, and bevacizumab are standard agents for the treatment of metastatic colorectal cancer (CRC). Evidence suggests benefit of IP Ox at high doses. However, the optimal dose of IP Ox combined with standard systemic therapy is unclear. Methods: We conducted an IRB-approved phase I dose-escalation study of IP Ox D1 at 25mg/m2-100mg/m2, with systemic bevacizumab D1 at 5mg/kg, and capecitabine 850mg/m2 BID for 7 days (cycle = 14 days), in patients with PC from appendiceal or CRC after CRS. The primary aim was to determine the recommended phase II dose (RP2D)/maximum tolerated dose (MTD) for this regimen. Dose limiting toxicities (DLTs) were assessed during cycle 1. DLTs included grade 3 or 4 non-hematological toxicities, or grade 4 hematological toxicities. Results: 18 patients (12 females, median age 56) were enrolled on the study. No DLTs were observed during cycle 1 in the first 4 cohorts. One DLT (abdominal pain) was observed in cohort 5. Another patient in cohort 5 experienced grade 3 abdominal pain soon after cycle 2, thus limiting repeated treatment for this cohort. Other toxicities included fatigue (72%), nausea (61%), peripheral neuropathy (50%), constipation (50%), mucositis (39%) and dizziness (39%). See table below for average # of cycles per cohort and responses. Conclusions: IP Ox combined with capecitabine and bevacizumab is feasible. Our recommended dose for IP Ox is 85 mg/m2 with systemic therapy (cohort 4). An expansion cohort is underway for this dose level. Based on these data, further investigation of IP Ox with systemic chemotherapy for PC is warranted. Clinical trial information: NCT01061515. [Table: see text]

410 Background: Recent advances in front-line therapy have improved survival in patients with advanced PDAC. A fluorouracil-based regimen is recommended for patients who progress on a gemcitabine-based therapy. Tosedostat is an oral aminopeptidase inhibitor that disrupts the cleavage of amino acids from peptides downstream of proteasomal degradation. It prevents the recycling of free amino acids, leads to intracellular depletion of amino acids, and triggers an amino acid deprivation response that subsequently results in apoptosis. Because PDAC cells frequently upregulate expression of these aminopeptidases, tosedostat offers therapeutic promise, particularly in combination with fluoropyrimidines. Methods: This is a single institution phase I/II trial to evaluate the safety and toxicity of tosedostat plus capecitabine in patients with metastatic PDAC progressed on a gemcitabine-based therapy. The phase I portion is being conducted in a dose de-escalation fashion, with two dose levels of tosedostat (120 mg or 60 mg) p.o. daily on days 1 to 21 with capecitabine 1000 mg/m2 p.o. BID on days 1 to 14 of a 21-day cycle. If more than one out of 6 patients in the tosedostat (120 mg) cohort experience a dose limiting toxicity (DLT), then 6 more will be enrolled to the tosedostat (60 mg) cohort. The primary objective of the phase I portion is to determine the optimal phase II dose. The primary objective of the phase II portion is to determine the progression-free survival at 3 months. Secondary objectives include the overall response rate, overall survival and CA 19-9 response. To avoid futility, interim analysis is planned after 10 evaluable patients enrolled. Results: Up to date, a total of 11 patients have been enrolled in the study, and 10 patients are evaluable. No DLT have been observed. Tosedostat at a dose of 120 mg with capecitabine is extremely well tolerated. Prolonged stable disease has been observed in 4 (40%) patients with a time on treatment of 10 months, 7.5 months, 5.5 months and 4 months, and 3 of the 4 patients remain on the trial. Conclusions: The combination of tosedostat and capecitabine is a well-tolerated regimen with impressive clinical activity in the subset of patients studied thus far. Clinical trial information: NCT02352831.

3619 Background: SBAs are rare, resulting in a paucity of data on effective treatments. Highly active immune surveillance within all parts of the intestine may explain the rarity. In such an enviro...