Citations

e13552 Background: Real-world data from Electronic Health Records (EHR) have been widely used for patient identification to build study cohorts for clinical research. Traditionally, diagnosis codes in the EHR, such International Classification of Diseases (ICD), are used to identify the target patients. However, the accuracy of this approach is dependent on the accuracy of ICD coding, with potential errors especially for tumor types that are frequent locations for metastases (which may contribute to mis-coding). In this study, we attempted to develop a Machine Learning (ML) based approach on EHR data to improve the accuracy of identification of patients with lung cancer. Methods: We used survey respondents in the Enhanced, EHR-facilitated Cancer Symptom Control (E2C2, NCT03892967) cluster-randomized trial at Mayo Clinic as our initial pan-cancer cohort. E2C2 includes adults receiving Medical Hematology/Oncology care for a solid or liquid tumor at Mayo Clinic. We collected cancer diagnoses from the individually abstracted Mayo Clinic Cancer Registry to annotate the cancer type for the patients. Lung cancer related ICD-9 (162.X) and ICD-10 (C34.X) codes were used to build a search query on the Mayo’s EHR to find target patients from the E2C2 cohort, and to investigate the ICD-based lung cancer patient identification performance. Diagnosis, radiation oncology treatment (CPT 77261 - 77799), and antineoplastic drug administration data were collected from EHR as variables. Logistic Regression (LR), Support vector machine (SVM), Random Forest (RF), and eXtreme Gradient Boosting (XGB) were selected to build models for lung cancer patient identification. 10-fold cross-validation was implemented to assess the models. Precision, Recall, F1 Score, and area under the curve (AUC) were selected to measure the performance. Results: We collected 13,893 patients with a specific cancer diagnosis, and 1,394 were identified as having lung cancer. The identification performance across different methods is shown. The ICD-based method only had 0.65 precision. It means we collected a lot of false positive cases (other cancer but no lung cancer patient), as we mentioned in the background. SVM gets the best precision results, but its recall and F1 score are not good enough. XGB shows the best F1 Score and AUC, which also means this method achieved the best and most balanced performance. Conclusions: In this study, we found that XGB-based methods achieved the best identification performance for lung cancer. In the future, we will investigate if this is also true for the identification of other cancer types. [Table: see text]

1592 Background: While ongoing efforts to decrease opioid use have dramatically altered the prescribing landscape, the impact on patients with cancer is not well understood. We examined trends in new and persistent opioid use in opioid-naïve older patients with and without cancer in the United States. Methods: This retrospective cohort study used SEER-Medicare linked data for patients with solid tumors malignancies diagnosed from 2012 through 2017, and a random sample of patients without cancer. We included patients who did not receive opioids for 3 months prior to the diagnosis or randomly assigned index date. We identified new opioid use (claim in 0-90 days after diagnosis/index date), early persistent opioid use (claims in days 0-90 and 91-180), and long-term persistent opioid use (claims in days 0-90 and 91-180 and months 13-15). Further analysis focused on 4 clinical strata: non-cancer, metastatic cancer, cancer treated with surgery alone, and cancer treated with surgery plus chemotherapy and/or radiotherapy ("surgery+"). We calculated relative changes in the predicted probability of opioid use from 2012 to 2017 based on a logit model adjusted for patient demographics. Results: We identified 238,470 eligible patients with cancer (mean age 76.3 years). Among patients with cancer, 46.4% had new opioid use; use was higher among those receiving surgery+ (70.6%), those treated with surgery alone (58.8%) and those with metastatic cancer (50.0%), compared to patients without cancer (6.9%). From 2012 to 2017, the predicted probability of new opioid use declined in all groups, with lower relative decline in the cancer subgroups (6.8% surgery+; 9.5% surgery alone; 10.2% metastatic cancer) compared to the non-cancer comparator (24.7%). For instance, in patients with metastatic cancer, predicted probability of new use decreased from 52.4 to 47.1% (10.2%) vs. 8.0 to 6.0% (24.7%) in patients without cancer. Relative decline in predicted probability of long-term persistent opioid use was lower in the surgical cancer strata (44.2% surgery+; 46.1% surgery alone) vs. the non-cancer comparator (54.0%). Conclusions: While opioid use remains common in patients with cancer, new and persistent opioid use declined over time in all patients, though with steeper relative declines in non-cancer compared to cancer populations. Our study suggests important but tempered effects of the changing opioid climate on patients with cancer.[Table: see text]

1552 Background: Immune checkpoint inhibitor therapy-related (ICI) pneumonitis poses a significant challenge in patients with cancer. There is a need for predictive modeling to identify those at high risk of developing serious adverse events of ICI pneumonitis. Methods: A newly created database of cancer patients diagnosed with ICI pneumonitis seen at Mayo Clinic from 2014-2022 was utilized for exploratory data analysis. Modeling was used to determine if three clinical outcomes could be predicted, namely risk of death from: pneumonitis prior to starting immunotherapy, pneumonitis at time of diagnosis of ICI pneumonitis, and any cause at time of ICI pneumonitis diagnosis. The data consisted of 170 patients with annotation for 71 clinical features. Data was divided into features available at time of cancer diagnosis and at time of ICI pneumonitis diagnosis. Exploratory modeling was performed using the gradient boosting technique Xgboost (Chen, 2015) and conducted using k-fold balanced cross validation best practices as well as a train/test/validate schema with 70%/20%/10% data proportions, respectively. Model reverse engineering was done with Shapley statistics (Lundberg, 2017) to determine which features had the largest contribution per model. Once identified, only those highly weighted features were used for logistic regression analysis providing more reproducible predictions by decreasing model variance. Results: Risk of death from pneumonitis prior to starting immunotherapy was modeled with an area under the curve of the receiver operator characteristic (AUC-ROC) of 0.79 with the most contributory features including lymphocyte count, oxygen dependence, PFT values, and PD-L1 expression. Logistic regression produced an AUC-ROC of 0.87 (95% CI 0.72-1.0, p < 0.0013). Risk of death from pneumonitis at the time of ICI pneumonitis diagnosis was modeled with an AUC-ROC of 0.85 and the most contributory features were similar to the list described in the previous endpoint. Logistic regression produced an AUC-ROC of 0.89 (95% CI 0.81-0.96, p < 0.0001). Risk of death from any cause at the time of ICI pneumonitis diagnosis until follow up concluded by 12/2022 produced a model with an AUC-ROC of 0.75 with the most contributory features including supplemental oxygen, PFT values, basic laboratory values, and PD-L1 expression, among others. Logistic regression produced an AUC-ROC of 0.85 (95% CI 0.76-0.93, p < 0.0001). Conclusions: We demonstrate that commonly available clinical data can be used to identify patients at high risk of death from ICI pneumonitis. This study identified clinical features that were predictive in each scenario from which further concerted effort could produce a new clinical model to provide clinician decision support when considering immunotherapy. Further studies should be done to further elucidate feature interdependence and generalizability.

1541 Background: Frailty is common in older adults undergoing major cancer surgery. Mounting evidence indicates multimodal interventions to optimize patient functional status by addressing vulnerabilities associated with frailty before surgery result in decreased complications and enhanced functional recovery. This study tested the effectiveness of OPTI-Surg, a surgical prehabilitation toolkit for improving surgical care and outcomes in older patients undergoing major cancer surgery. Here, we report results from the planned futility interim analysis (IA). Methods: Thoracic (T), gastrointestinal (GI), and urologic (GU) oncological surgery practices affiliated with NCI’s Community Oncology Research Program were randomized 1:1:1, stratified by practice type, to usual care (control), OPTI-Surg toolkit, or OPTI-Surg with an implementation coach. The toolkit was implemented at the practice level. It consisted of the 12-item Edmonton Frail Scale and guided recommendations for interventions directed at identified vulnerabilities. Eligible patients were adults ≥70 years old undergoing curative intent surgery for lung, gastrointestinal, or urologic cancers. Primary outcome was 8-week post-operative function (weekly caloric expenditure) derived from the validated Community Healthy Activities Model Program for Seniors (CHAMPS) questionnaire. A linear mixed model was used to compare usual care to the 2 OPTI-Surg arms combined, controlling for baseline function. Results: From July 2019 to October 2021, 225 patients, 50% of the total accrual goal, were enrolled from 29 practices, triggering the IA. Study was suspended to new patient accrual March to June 2020 due to the COVID-19 pandemic. There were no differences in demographics by arm (mean age 76.2 years; 46.4% women; 79.6% white; 20.4% lung resection, 69.8% major GI surgery, 9.8% total cystectomy). Of the 225 patients, 140 (47 Usual Care, 93 OPTI-Surg) were evaluable for the IA. There was no significant difference in post-operative function between OPTI-Surg and usual care (mean decrease of 0.59 kCal/week for OPTI-Surg patients and 0.19 kCal/week for usual care patients, p = 0.762, intraclass correlation = 0.033). Based on the futility analysis, the DSMB recommended closure of recruitment on September 16, 2022. Conclusions: Implementation of the OPTI-Surg toolkit did not significantly improve functional outcomes 8 weeks after surgery for patients aged 70 and older undergoing major cancer surgery compared to usual care. Future analysis of implementation and ethnographic data may identify reasons why significant differences between the intervention and usual care arms were not observed, potentially including disruptions due to the COVID-19 pandemic. Clinical trial information: NCT03857620 .

1535 Background: Ambulatory management of select patients with febrile neutropenia (FN) may substantially decrease patients’ burden of illness and improve their quality of life, and is recommended by American Society of Clinical Oncology (ASCO) guidelines. However, in the United States, inpatient management remains the universal standard. We hereby report the health care utilization associated with the DEFeNDR remote patient monitoring (RPM) program, which leverages technology and virtual centralized nurse monitoring to facilitate ambulatory management of FN as an alternative to hospital-based care. Methods: Adult patients with cancer hospitalized with FN were assessed daily for eligibility to participate in the DEFeNDR program. Eligibility criteria included MASCC score >21 (per ASCO guidelines), and clinical judgment of the attending provider. Patients receiving cellular therapy (CAR-T, BMT) in the prior 100 days were excluded. Eligible patients agreeing to participate in the 30-day program were provided a kit containing cellular-enabled, pre-connected devices to facilitate monitoring, including twice-daily vital signs and electronic symptom questionnaires. Data were EHR-integrated in real-time and embedded decision trees facilitated alerts for adverse trends. RPM nurses responded to alerts and escalated care as indicated by pre-specified care pathways. Program assessment was performed at 20 weeks after full implementation. The primary outcome was the mean proportion of evaluable days spent inpatient (mPEDI) within 30 days of eligibility compared between those who participated in the program vs. those who declined. Results: 35 patients were offered participation in the DEFeNDR program. Of these, 17 participated and 18 declined. Baseline characteristics are provided. Participating patients experienced mPEDI of 6.9% (Standard Error [SE], 3.7%), compared with 17.7% (SE, 3.6%) for patients who declined participation. Due to small sample size, this difference (10.8%) did not reach statistical significance (Wilcoxon Rank Sums p-value, 0.15). 30-day mortality did not differ between groups. Mean inpatient days were also lower for patients participating in the program. Additionally, when readmitted, patients participating in RPM experienced a trend toward shorter length of stay and lower rates of intensive care unit utilization. Conclusions: A trend towards decreased burden of inpatient care is evident among patients participating in the DEFeNDR program. Further studies are warranted to assess comparative effectiveness of this model vs usual care.[Table: see text]

1533 Background: FDA-approved chimeric antigen receptor T-cell (CAR-T) and bispecific T cell engager (TCE) therapy are most commonly managed inpatient. Here we report our experience with hospital-based outpatient (HBO) management of patients (pts) receiving FDA-approved CAR-T or TCE and a predictive model for increased risk of early hospitalization (≤3 days post CAR-T). Methods: We reviewed the outcomes of pts who received FDA-approved CAR-T (axi-cel, brexu-cel, ide-cel, cilta-cel) or TCE (teclistamab) in the HBO unit between 3/2021 and 1/2023 at Mayo Clinic, Rochester. Results: HBO is closely integrated with inpatient practice and includes the same specialty trained clinical team. It is the first point of contact 24/7 for pts and triage evaluations. CAR-T and TCE are given on HBO followed by monitoring per REMS and as clinically needed until admission criteria is met. Remote patient monitoring (RPM) tools were used as part of outpatient monitoring. Out of 169 CAR-T pts, 87 (51%) were dosed with new indication/products (follicular lymphoma, mantle cell lymphoma, myeloma). Among those, 80 (92%) pts were dosed outpatient, among whom 21 (27%) were never hospitalized. CRS and ICANS incidences were comparable to that reported in registration studies. Median time to first admission from CAR-T infusion was 1 day (range, 0-9), with 31% (n=18/59) of admissions after day 3 post infusion. The most common cause of hospitalization was fever (n=40, 59%). Median length of hospital stay was 4 days (range, 1-49). Median time to tocilizumab (n=45, 80%), steroid (n=1, 2%), escalation of oxygen support (n=7, 12%), vasopressor (n=2, 3%), ventilator (n=1, 2%), hemodialysis (n=1, 2%) and ICU transfer (n=3, 5%) was 1 day (range, 0-31) after first admission. Use of bridging therapy, elevated CRP and LDH were predictive of early admission (ROC AUC 0.62). Four pts received teclistamab on HBO and enrolled in RPM. Data from more pts will be reported at ASCO. Conclusions: Our expanded study in CAR-T and TCE confirms that HBO is feasible and that CAR-T and TCE can be safely dosed outpatient. Approximately 31% of pts were admitted after 3 days from CAR-T infusion, and 27% of our pts were never admitted. Variables associated with aggressive disease can predict early hospitalization in pts receiving FDA-approved CAR-T. Extension of HBO practice to TCE is feasible.[Table: see text]

1509 Background: Medicaid expansion under the Affordable Care Act has been shown to increase insurance coverage and early-stage cancer diagnoses for young adults (YAs), the age group with the highest uninsured rate. We examined whether Medicaid expansion was associated with increased receipt of timely, guideline-concordant treatment and survival among YA women newly diagnosed with breast cancer, the most common YA cancer diagnosis. Methods: Using the National Cancer Database, we identified 51,675 women aged 18-39 years who were diagnosed with breast cancer in 2011-2018. We applied the difference-in-differences (DD) method to estimate outcome changes pre vs. post Medicaid expansion, in expansion- vs. non-expansion states. Linear probability models estimated associations of Medicaid expansion with receipt of guideline-concordant treatment – including any endocrine therapy among women with ER-positive or PR-positive breast cancer, and any chemotherapy or targeted therapy among women with ER-negative and PR-negative breast cancer – and initiation of guideline-concordant treatment (surgery, chemotherapy, or targeted therapy for stage I-III diagnoses, and systemic therapy for stage IV diagnoses) <60 days after diagnosis. Flexible parametric survival models were used to estimate 2-year overall survival. DD models also adjusted for age, race/ethnicity, rurality, zip code-level income, comorbidity, and diagnosis year, with standard errors clustered at the state level. Analyses were stratified by stage at diagnosis (stage I-III vs. IV). Results: Of the subset of women with stage I-III ER-positive or PR-positive breast cancer, the percentage receiving endocrine therapy increased in expansion states (85.20% pre-expansion to 86.46% post-expansion), but decreased in non-expansion states (84.32% to 82.79%), resulting in a net increase of 2.42 percentage points (ppt; 95% CI = 0.56 to 4.28) associated with Medicaid expansion in the adjusted DD model. Among all women with stage I-III breast cancer, the percentage with treatment initiation <60 days after diagnosis decreased less in expansion states (84.10% pre-expansion to 82.36% post-expansion) than in non-expansion states (86.95% to 83.98%), resulting in a net reduction of 1.61 ppt (95% CI = 0.05 to 3.18) in treatment delays associated with Medicaid expansion. Notably, the 2-year overall survival increased in expansion states (96.84% pre-expansion to 96.99% post-expansion), but decreased in non-expansion states (97.24% to 96.30%), resulting in a net survival increase of 1.00 ppt (95% CI = 0.21 to 1.79) associated with Medicaid expansion. Conclusions: Among YA women with stage I-III breast cancer, Medicaid expansion was associated with increased receipt of guideline-concordant treatment, reduced delays in treatment initiation, and improved 2-year survival.

e13100 Background: Aurora-A mitotic kinase (AURKA) plays a pivotal role in Triple Negative Breast Cancer (TNBC) progression through activation of epithelial to mesenchymal transition (EMT)-mediated cancer cell plasticity. Cancer cells that undergo EMT acquire a CD44high/CD24low and/or ALDHhigh cancer stem-like phenotype (Cancer Stem Cells) characterized by high self-renewal capacity and intrinsic drug resistance. Cancer Stem Cells (CSCs) also shows high PD-L1 expression that plays a critical role in inducing tumor immune escape through CD8+ T-cells exhaustion. For these reasons, molecular characterization of oncogenic pathways that promote stemness and intra-tumoral PD-L1 expression is essential to evaluate whether novel combinatorial therapeutic strategies aimed to inhibit cancer cell plasticity, immune evasion capacity and organ metastasis may impact cancer progression. Methods: (1) TNBC Models: MDA-MB 231 and variant MDA-MB 231/LM cells, and unique TNBC-M40 cells established from metastatic PDXs; (2) In Vitro Studies: TNBC cells were treated with the AURKA inhibitor alisertib (50nM). PD-L1 expression was evaluated by immunoblotting assay. Apoptosis was measured by immunofluorescence using Cleaved-PARP antibodies. To assess stemness capacity, TNBC cells were cultured under non-adherent conditions to form mammospheres. ALDH activity (functional stemness marker) was measured using the Aldefluor Kit. (3) In Vivo Studies: 1x106 MDA-MB 231/LM cells were injected into the mammary fat pad of humanized NSG-CD34+ female mice. Animals were randomized into four groups (5 mice per group) and treated with anti-PD-L1 ICI atezolizumab (20 mg/Kg) and alisertib (25 mg/Kg) as monotherapy or in combination. Control groups were treated with saline solution placebo. PD-L1, CD44, Vimentin expression and tumor infiltration of CD8+ T-cells were assessed by immunofluorescence. Results: Treatment of MDA-MB 231/LM xenografts with alisertib decreased the levels of CD44 and PD-L1 expression and increased CD8+ T-cells tumor infiltration. Because PD-L1 genetic targeting enhanced in vitro alisertib-induced apoptosis, MDA-MB 231/LM xenografts were treated with alisertib and the anti-PD-L1 ICI atezolizumab. Combination of atezolizumab and alisertib resulted in the reduced expression of the EMT marker vimentin that was linked to lack of organ metastasis. Remarkably, combination of atezolizumab and alisertib also induced a significant reduction of ALDHHigh CSC in ex-vivo 3D-Organoids established from TNBC-M40 cells. Conclusions: This study provides innovative pre-clinical rationale for combining AURKA inhibitors with ICIs to impair cancer cell plasticity, immune evasion capacity and halt the progression of metastatic TNBCs that currently has limited effective therapeutic options.

e13003 Background: The emergence of somatic KMT2C mutations has been associated with resistance to various treatments, including chemotherapy and endocrine therapy in patients with estrogen receptor-positive breast cancers. The impact of these mutations in patients treated with poly-ADP ribose polymerase inhibitors (PARPi) and phosphatidylinositol-3-kinase inhibitors (PI3Ki) is unknown. Methods: Circulating tumor DNA (ctDNA) was obtained from plasma and peripheral blood at different time points from 38 patients prior to treatment with the combination of the PARPi olaparib with the alpha-specific PI3Ki alpelisib (NCT01623349). CtDNA was also collected immediately before the first cycle and at the time of disease progression. CtDNA underwent deep targeted sequencing (tumor-normal pairs, 397 genes (2Mb) at 25,000x somatic and 10,000x germline) at the Broad Institute of MIT and Harvard. Variants that were exonic, non-synonymous, and predicted to lead to loss-of-function were considered mutations. Cohorts were stratified according to the presence or absence of the genomic alteration, and progression-free survival was compared with the Kaplan-Meier method and log-rank test. A Cox regression was performed using age, tumor type, germline BRCA1/2 mutation status, performance status, and number of previous lines of treatment as covariates. Results: The KMT2C gene was one of the 3 genes most frequently mutated in our cohort, affecting 5 patients (15%) after the exclusion of those with variants with non-deleterious alterations. The other 2 commonly mutated genes were TP53 (76%) and PIK3CA (15%). In the comparison of patients based on these most common mutations, we identified that patients with somatic KMT2C mutations (n=5) have longer mPFS than those that were wild type (n=28): mPFS 386 vs. 151 days, log-rank p = 0.025. Cox proportional hazards model accounting for age, tumor type, germline BRCA1/2 mutation status, performance status, and number of previous lines revealed a hazard ratio of 0.16; p = 0.007. All somatic KMT2C mutations were proximal to the SET domain critical for methyltransferase activity. Upon progression, the KMT2C-mutated clones were suppressed to below the detection level of mutation calling algorithms. Analysis of the raw files revealed that the clones were suppressed but not eliminated. Conclusions: While the emergence of KMT2C has been reported to be associated with cancer progression and resistance to different therapies, patients with deleterious KMT2C mutations in the ctDNA may benefit from therapy with PARPi and PI3Ki.

TPS1115 Background: The current first-line (1L) standard of care (SOC) for human epidermal growth factor receptor 2-positive (HER2+) metastatic breast cancer (MBC) is trastuzumab (T) plus pertuzumab (P) and a taxane. Despite advances in 1L SOC, most patients progress during maintenance therapy with T + P. Tucatinib is a tyrosine kinase inhibitor (TKI) approved in combination with T and capecitabine for adults with HER2+ MBC, with and without brain metastases (BM). In HER2CLIMB, the addition of tucatinib significantly prolonged progression-free survival (PFS) and overall survival (OS) in patients with HER2+ MBC and was well tolerated. Adding tucatinib also reduced the risk of disease progression or death in patients with untreated and/or active BM (Murthy et al. 2020, Curigliano et al. 2021). HER2CLIMB-05 investigates whether adding tucatinib to 1L SOC as maintenance therapy will extend PFS while maintaining quality of life (QOL). Methods: HER2CLIMB-05 (NCT05132582) is a phase 3, randomized, double-blind study evaluating tucatinib plus T + P as maintenance therapy for HER2+ MBC. Approximately 650 patients will be enrolled. Eligible patients will have advanced HER2+ disease, no progression on 4–8 cycles of prior 1L SOC, Eastern Cooperative Oncology Group Performance Status of 0 or 1, and no or asymptomatic BM. Exclusion criteria include prior treatment with anti-HER2 and/or anti-epidermal growth factor receptor TKI (prior SOC for early BC is permitted) or inability to undergo contrast magnetic resonance imaging of the brain. Patients will be randomized 1:1 to receive either tucatinib or placebo twice daily, with T + P once every 21 days. Patients with hormone receptor-positive disease may receive endocrine therapy. The primary endpoint is investigator-assessed PFS. Secondary endpoints include OS (key endpoint), PFS by Blinded Independent Central Review (BICR), time to deterioration of health-related QOL, central nervous system PFS, safety, and pharmacokinetic (PK) parameters. PFS and OS will be compared using a 2-sided stratified log-rank test between treatment groups. Time-to-event endpoints will be summarized using the Kaplan–Meier method. PK and safety data will be summarized using descriptive statistics. Enrollment is ongoing in the US, Canada, Brazil, APAC and EU countries with additional sites planned. Clinical trial information: NCT05132582 .