Citations

9065 Background: Rearranged during transfection ( RET) gene alterations are the oncogenic driver in diverse tumor types, including RET fusions identified in 1-2% of non-small cell lung cancers (NSCLC). While RET-selective tyrosine kinase inhibitors (TKIs) selpercatinib and pralsetinib are effective, acquired drug resistance remains a challenge. Here, we report the initial results from the RETgistry, an international consortium aimed at elucidating mechanisms of resistance to RET TKIs across RET-altered solid tumors. Methods: This was a retrospective analysis performed across 16 institutions. Patients (pts) were eligible if they had advanced solid tumor harboring an oncogenic RET alteration, received ≥1 RET TKI with disease progression, and underwent resistant tumor or liquid biopsy for next-generation sequencing (NGS). Results: A total of 103 time-distinct biopsies (62 tissue, 30 plasma, 11 paired tissue/plasma) were included in analysis, obtained from 88 pts with progression on a RET-selective TKI [selpercatinib (n = 70), pralsetinib (n = 14), selpercatinib and pralsetinib sequentially (n = 4)]. Pts had the following tumor types: 72 NSCLC (69% KIF5B-RET, 21% CCDC6-RET, 10% other RET fusion), 13 medullary thyroid cancer (TC) (54% RET M918T, 46% other RET mutation), and 2 papillary and 1 anaplastic TC (all, RET fusion). Median age was 58 (range, 21-86); 48% were male with 89% never/light smokers. Median duration of RET TKI preceding biopsies [first-line, n = 32 (31%); second-line, n = 42 (41%); third-/greater-line, n = 29 (28%)] was 16.5 months (mos) (95% CI, 14.0-19.6); median PFS was 14.1 mos (95% CI, 9.3-17.0). Resistant biopsies were obtained at median 15.0 mos from TKI initiation (range, 1.8-58.8). Acquired RET mutations were detected in 14 (14%), most common being G810 substitutions in 12 (12%). Potential off-target resistance gene alterations identified in 43 cases (42%) included MET amplification (14%), BRAF V600E or fusion (2%), KRAS gain or mutation (5%), ERBB2 amplification (2%), EGFR amplification (3%), ROS1 fusion (1%), and activating PIK3CA mutation or PTEN loss (4%). No resistant lung cancer biopsy demonstrated small cell transformation. The duration of TKI therapy (HR 0.64, p = 0.11) or PFS (HR 0.75, p = 0.33) did not differ according to the presence of on-target vs off-target resistance. NGS analyses of pre-RET-selective TKI tumors (n = 92) revealed frequent co-occurring alterations in TP53 (29%) and CDKN2A/B (12%). Conclusions: On-target resistance to RET inhibition due to acquired RET mutations was less common than off-target resistance, identified in 14%. The majority of RET TKI resistance is mediated by off-target mechanisms such as bypass receptor tyrosine kinase activation. Further studies are warranted to enable the development of therapeutic strategies to address resistance in pts with RET-altered tumors. The RETgistry accrual and data analyses continue.

TPS5634 Background: The prognosis for women with recurrent or persistent EC after progressing on first-line chemotherapy is poor, with a median overall survival (OS) of 12 months, highlighting the need for more efficacious therapies for this population. Immune checkpoint inhibitors either as monotherapy or when combined with kinase inhibitors have recently demonstrated encouraging response rates in pts with microsatellite instability-high (MSI-H)/mismatch repair-deficient (dMMR) and MS-stable (MSS)/MMR-proficient (MMRp) EC, respectively. Atezo is a humanized monoclonal anti-programmed cell death ligand 1 (PD-L1) inhibitor that has demonstrated monotherapy antitumor activity with an acceptable safety profile in relapsed recurrent EC. Atezo has shown compelling clinical efficacy for pts with certain solid tumors (e.g., urothelial, NSCLC, triple negative breast cancer) as monotherapy and as part of combinatorial therapy. The AFT-50 EndoMAP trial is a platform trial designed to evaluate the efficacy and safety of Atezo in combination with biomarker-defined targeted agents in pts with recurrent or persistent EC. Recognizing pts may have either progressed through or received a checkpoint inhibitor in a prior line of therapy, a non atezo cohort will also be open to enrollment. Methods: This is a phase IB/II non-randomized, multicenter, multicohort, biomarker-driven platform study consisting of two cohorts: A (atezo containing) and B (non atezo). Eligible pts have recurrent or persistent EC and have received no more than 2 prior lines of therapy. FoundationOne CDx (F1CDx) NGS assay will be used for genomic tumor profiling. In cohort A, pts may be eligible for one of the following doublets: Atezo+ipatasertib (PIK3CA/PTEN/AKT1-altered cancers), Atezo+talazoparib (tumors with genomic loss of heterozygosity (LOH) ≥16%), Atezo+Trastuzumab emtansine (ERBB2/HER2 mutated and/or amplified tumors), Atezo+Tiragolumab (MSI-H and/or TMB>10 mut/MB), and Atezo+bevacizumab (biomarker unmatched). Pts will receive Atezo in addition to the targeted agent (at the study approved dosing schedule) until progression, unacceptable toxicity, pt or physician decision to withdraw from the study, death, or study termination. Pts in cohort B, will be eligible for inavolisib (PIK3CA/PTEN/AKT1-altered cancers) + letrozole. Each cohort will enroll ~20 pts; however, the exact size of the cohort may be adjusted. The primary endpoint for Cohort A is confirmed overall response rate (ORR) for each cohort, and for Cohort B is progression free survival at 6 months. Secondary endpoints include disease control rate, duration of response, OS, safety and tolerability, 6-month PFS (cohort A). As a platform study, additional arms may be added, as supported by evolving understanding of EC and molecular targets. EndoMAP is actively enrolling at 17 sites in the US with a target of 25 sites nationwide. Clinical trial information: NCT04486352 .

5512 Background: The multicenter, open-label, randomized phase 2 NCI-9944 study (NCT02595892) demonstrated that addition of ATR inhibitor (ATRi) berzosertib to gemcitabine increased progression-free survival (PFS) compared to gemcitabine alone (HR = 0.57, one-sided log-rank p = 0.044, which met the one-sided significance level of 0.1 used for sample size calculation). Final overall survival (OS) and corresponding biomarker analyses are reported here. Methods: Patients (pts) with platinum-resistant high-grade serous ovarian cancer and unlimited previous lines of cytotoxic therapy in the platinum-sensitive setting but no more than one line of cytotoxic therapy in the platinum-resistant setting, were randomized 1:1 to gemcitabine/berzosertib versus gemcitabine alone. Randomization was stratified based on platinum free interval (PFI), PFI≤3 months versus > 3 months. Crossover from gemcitabine to gemcitabine/berzosertib was allowed upon disease progression by RECIST 1.1. OS was a secondary endpoint while preplanned exploratory correlative studies included assessment of DNA repair pathways and replication stress (RS) alterations by targeted gene sequencing and/or immunohistochemistry (IHC). Results: Seventy pts were randomly assigned to treatment with gemcitabine/berzosertib (34 pts) or gemcitabine alone (36 pts); 15 pts crossed over from gemcitabine to gemcitabine/berzosertib. At the final OS analysis (92.9% maturity), median follow-up was 53.2 weeks in the gemcitabine/berzosertib and 43 weeks in the gemcitabine alone groups. Median OS in the intent-to-treat (ITT) population was 59.4 weeks in the gemcitabine/berzosertib group versus 43.0 weeks in the gemcitabine alone group (HR 0.79, 90% CI 0.52-1.2, one-sided p = 0.18). However, when patients who crossed over to gemcitabine/berzosertib were excluded from analysis, a significant OS benefit was observed with gemcitabine/berzosertib (HR 0.60, 90%CI 0.37−0.97); HR was 0.26 (90% CI 0.1−0.7) in pts with PFI≤3 months and 0.89 (90%CI 0.50−1.59) in pts with PFI > 3 months. Furthermore, significant OS benefit was observed in pts with RS-low tumors (HR 0.39, 90%CI 0.17−0.91, defined as tumors harboring no genomic RS alterations related to loss of RB pathway regulation and/or oncogene-induced RS) but not in pts with RS-high tumors (HR 0.74, 90%CI 0.35−1.56). Additional targeted gene sequencing and IHC analyses as well as analyses adjusting for patient crossover will be reported at the meeting. Conclusions: In the ITT population, gemcitabine/berzosertib did not significantly improve OS versus gemcitabine alone. Pts with PFI≤3 months and pts with RS-low tumors may derive a survival advantage from addition of berzosertib to gemcitabine in the platinum-resistant setting. Clinical trial information: NCT02595892 .

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

3580 Background: TAS-102 (FTD/TPI) is a combination of a nucleoside analogue and a thymidine phosphorylase inhibitor, and has showed activity in 5FU-resistant CRC. Nal-IRI achieves higher intra-tumor concentrations than irinotecan (Iri; 142-fold) and its major metabolite, SN-38 (9-fold). This resulted in superior anti-tumor activity compared to free Iri in multiple tumor xenografts. Furthermore, multiple clinical trials have established the activity of nal-IRI in combination with 5FU in pancreaticobiliary cancers. The combination of nal-IRI with the more potent nucleoside analogue TAS-102 may thus result in a more effective systemic therapy. We conducted this study to define the safety and efficacy of the combination of TAS-102 and nal-IRI in advanced CRC. Methods: Eligible patients (pts) on this investigator-initiated phase II study had advanced CRC and must have had disease progression on at least one prior therapy. Additional inclusion criteria - measurable disease, ECOG PS 0-1 & adequate organ function. Based on previously published dose escalation phase I data, the pts were treated at the recommended phase II dose of TAS-102 (given orally at 35mg/m2 bid on days 1-5) and nal-IRI (60mg/m2 IV on day 1) in 14-day cycles. The primary endpoint was objective response rate (ORR) = complete (CR) + partial (PR) response per RECIST v1.1. Multiple secondary endpoints included safety/tolerability, progression-free survival (PFS), overall survival (OS), disease control rate (DCR) and duration of response (DoR). Results: Twenty-two pts were enrolled, of whom 20 were evaluable for response. Male - 50%; ECOG PS 1 – 75%; African Americans – 38%; median age - 58yrs (interquartile range [IQR] 49-67). Median line of prior therapy was 1 (IQR 1-2) and 13.6% (3 pts enrolled during the dose escalation phase) had received prior Iri. 47.6% received concurrent Bevacizumab. ORR was 15% (3 PR, 0 CR). 60% had stable disease (SD) as best response; DCR of 75%. mPFS - 9.7 mos (95% CI: 5.6, 14.2); mOS - 10.1 mos (95% CI: 7.3, 15.9). 12-month PFS and OS rates were 34.6% and 39.5% respectively. At a median follow up of 11.9 mos, DoR was yet to be determined and 50% of the pts had maintained response for 12 mos. Median duration of progression free among pts with SD was 7.6 mos (95% CI: 4.6, 13.2), with a 12-month survival rate of 47.7% (95% CI: 15.5%, 74.5%). Median time to response (TTR) was 4.8mos. Most common treatment related toxicities included fatigue and neutropenia (G3; 27.3% each); anemia and leucopenia (G3; 18.2% each). Other G3 AEs included febrile neutropenia, diarrhea, hypokalemia, lymphopenia and nausea (4.5% each). Conclusions: We demonstrated that the combination of TAS-102 and nal-IRI had an acceptable safety profile, and showed antitumor activity in patients with advanced CRC. Additional biomarker analyses such as survival correlation with UGT1A1 phenotype and RAS mutational status are ongoing. (NCT03368963). Clinical trial information: NCT03368963 .

3579 Background: Panitumumab (Pmab) is a recombinant monoclonal antibody that binds specifically to the human epidermal growth factor receptor (EGFR) and it is indicated for the treatment of metastatic colorectal carcinoma (mCRC). However, the objective response rates (ORR) for the single agent are historically about 8-11%. EGFR inhibition induces synthetic lethality with poly ADP ribose polymerase inhibitors (PARPi) such as Niraparib, by attenuating DNA repair pathways. Combining PARP and EGFR inhibition has the potential to confer synergistic benefit, while also ameliorating resistance mechanism to EGFRi. We conducted this study to define the safety and efficacy of the combination of Niraparib and Pmab in RAS wildtype (WT) mCRC. Methods: Patients (pts) with RAS WT mCRC who have progressed on at least one line of systemic chemotherapy were eligible for the trial. Other selected inclusion criteria were ≥18 years of age, ECOG PS 0-1 and measurable disease per RECIST 1.1. Pmab was administered at 6 mg/kg IV on days 1 & 15 of each 28-day cycle, while Niraparib was taken orally at 200mg or 300mg (based on body weight and platelet count) daily throughout the cycle. The primary endpoint was clinical benefit rate (CBR) defined as complete (CR) + partial (PR) response + stable disease (SD) per RECIST v1.1. Multiple secondary endpoints included safety/tolerability, ORR, progression-free survival (PFS), overall survival (OS), and duration of response (DoR). Results: A safety run-in cohort of 6 pts was initially enrolled. Following a preplanned safety analysis showing an acceptable toxicity profile, an additional 19 pts were enrolled. Of the total 25 enrolled pts, 24 were evaluable for response. Male - 50%; Whites/African Americans/Asians – 65.2%/21.7%/13%; median age – 58.5yrs. The majority had left sided tumor (92%) and the median line of prior therapy was 2 (range 1-4). CBR was 83.3% (0 CR, 6 PR, 14 SD) and ORR was 25%. mPFS – 5.6mos (95% CI: 3.7, 6.9); mOS – 20.9mos (95% CI: 9.2, NR). Six-month PFS and OS rates were 48.4% and 100% respectively. At a median follow up time of 7.5mos (95% CI: 4.2, 10.4), 3 patients remained on treatment and DoR was yet to be determined. Muco-cutaneous adverse events (AEs) of any grade included rash (58.4%), oral mucositis (20.8%), dry skin (16.7%) and paronychia (4.2%). Most common (grade > 2) treatment related AEs (TRAEs) were anemia (G3; 12.5%), dermatitis, oral mucositis, hypertension and neutropenia (G3; 4.2% each). There were no grade 4-5 TRAEs. Conclusions: The combination of Pmab and Niraparib had an acceptable safety profile, and showed considerable antitumor activity in pts with advanced RAS WT mCRC compared to historical rates. Additional biomarker analyses such as survival correlation with immune cell infiltration in paired skin biopsies and HER2/BRAF mutational status are ongoing. Funding and product (Niraparib) for the study were provided by GSK (NCT03983993). Clinical trial information: NCT03983993 .

3533 Background: The BACCI trial (NCT02873195) randomized 133 mCRC pts to C+B+A (investigational) or C+B+placebo. Longer progression free survival (PFS) was observed in the investigational vs placebo arm. To investigate T-cell functional heterogeneity, we performed single-cell proteomic analyses on pts with either long or short PFS and evaluated T-cell function as a prognostic and predictive biomarker. Methods: Isolated CD8+ and CD4+ cells were analyzed at baseline and first restaging from pts with long ( > 6 mo, N = 14) or short ( < 2 mo, N = 13) PFS. Single-cell, functional proteomic analyses were conducted using the IsoLight system (Isoplexis), which assessed secretion of 32 cytokines from an average of 786 cells per sample. The cumulative readout was polyfunctional strength index (PSI), a composite score of frequency and magnitude of secreted cytokines. Cytokines are organized into effector (E), stimulatory, chemoattractive (CA), regulatory, and inflammatory (I) clusters. Differential expression analyses were done using Wilcoxon rank-sum tests. Potential predictive effects were assessed using logistic regressions with interaction between treatment and biomarkers. No multiple comparison control was done due to the exploratory nature of analyses. Results: Baseline CD8+ and CD4+ cells were stimulated ex vivo and single-cell cytokine analyses performed. Prognostic analyses indicated that CD8+ cells from pts with long PFS had ~3-fold higher PSI compared to pts with short PFS (p = 0.005). Further analysis identified that the E (p = 0.005) and CA clusters (p = 0.005) most impacted PSI, significant individual cytokines are listed in the table. Baseline CD8+ cell function was not predictive of outcome. Baseline CD4+ cell function was not prognostic or predictive of outcome. Analysis of CD8+ cells after treatment identified that changes in CA cluster was prognostic of long PFS (p = 0.024). Cytokine changes from CD4+ cells were not prognostic; however, early changes in the CD4+ CA cluster were predictive of outcome (p-int = 0.055). Similarly, early changes in the CD8+ I cluster were predictive of outcome (p-int = 0.073). Conclusions: These results reveal pts with long PFS had more functional CD8+ T-cells with a higher PSI than pts with short PFS. Predictive analyses suggested that early changes in the CD4+ CA and CD8+ I clusters were associated with benefit from immunotherapy. These preliminary results highlight the importance of single-cell functional proteomic analyses. [Table: see text]

e15131 Background: Elimusertib is a highly selective oral inhibitor of the Ataxia telangiectasia and Rad3 related (ATR) kinase, which is required for efficient end-joining repair of double-stranded breaks, particularly during DNA replication. Many pediatric cancers involve potential ATR susceptibilities such as increased replication stress due to oncogenic fusions, specific DNA damage repair (DDR) deficiencies, oncogenic activation of DNA damage and nucleases such as PGBD5, as well as alternative lengthening of telomeres. We report the initial phase 1 results of PEPN2112 (NCT05071209), an ongoing pediatric phase 1/2 clinical trial of elimusertib monotherapy. Methods: Eligible patients (pts) included children, adolescents, and young adults with relapsed or refractory tumors with: (1) Ewing sarcoma or related EWSR1 fusion positive tumors (EWS), (2) alveolar rhabdomyosarcoma, PAX3-FOXO1 fusion positive (aRMS), or (3) non-CNS primary tumors exhibiting specific DDR pathway defects anticipated to sensitize to ATR inhibition. For pts < 18 years (yrs), a rolling-six design was used to evaluate elimusertib administered BID x3 days/week of a 28-day cycle starting at the adult equivalent recommended phase 2 dose (RP2D) of 24 mg/m2/dose (max 40 mg) with a potential de-escalation to 18 mg/m2/dose (max 30 mg). The pediatric RP2D determination was based upon dose limiting toxicity (DLT) criteria (CTCAEv5) during cycle 1. Pharmacokinetic (PK) studies were performed. Pts > 18 yrs could enroll onto the phase 2 cohorts at the adult RP2D simultaneously with the pediatric phase 1. Once the pediatric RP2D was defined, pts < 18 yrs accrued to the phase 2 cohorts and a PK expansion for pts < 12 yrs was opened. Results: 8 eligible pts were enrolled (4 males) on the phase 1 part of the study. 2 pts were inevaluable for DLT as they received < 75% of protocol therapy. Median (range) age was 13 (10-17) yrs. Diagnoses included: EWS (n = 4) and 1 pt each with aRMS, adrenocortical carcinoma, clear cell sarcoma, and osteosarcoma. At 24 mg/m2/dose, no DLTs were observed. Toxicities were hematologic, with grade 3 toxicities including anemia (n = 3), neutropenia (n = 4), and decreased white blood cells (n = 1). One pt experienced grade 4 lymphopenia. There were no grade 3/4 non-hematologic toxicities. Day 1 mean ± SD values of Tmax, Cmax, AUC0-8h and t1/2 were 1.6 ± 0.5 h, 1532 ± 385 ng/mL, 5863 ± 1841 h*ng/mL and 3.6 ± 0.9 h, respectively. Conclusions: Elimusertib-related toxicities were primarily hematologic. The RP2D for elimusertib monotherapy in pediatric pts is 24 mg/m2/dose (max 40 mg) orally BID x3 days/week continuously, analogous to dosing in adults with similar PK. Phase 2 and PK expansion cohorts as well as pharmacodynamic studies are ongoing. Clinical trial information: NCT05071209 .

2002 Background: Elderly patients with glioblastoma have a poor prognosis with phase III trials reporting median overall survival (OS) and global time to deterioration of 6–9 and 1.2 months respectively. 18F-DOPA (3,4-dihydroxy-6-[18F]-fluoro-L-phenylalanine) positron emission tomography (PET) is sensitive and specific for identifying regions of high density and biologically aggressive glioblastoma. Proton beam therapy (PBT) can spare more healthy tissue that may improve quality of life for this patient population. With improved targeting of disease along with dosimetric advantages of PBT, this phase II trial tested whether hypofractionated PBT can offer both improved survival and quality of life. Methods: Patients with newly diagnosed glioblastoma aged ≥65 years without contraindications to 18F-DOPA were eligible. Target volumes were defined by PET (tumor/normal brain SUV ≥2.0) and MRI areas of contrast enhancement including surgical cavity. Patients received dose escalated hypofractionated PBT of 35-40 GyE to PET/MRI over 5-10 treatments with 10 mm margin. Patients received concurrent/adjuvant temozolomide. With an alpha of 0.1, 18 of 39 patients would be required to be alive at 12 months for the study to be considered a success. Kaplan-Meier/Cox regression analysis was performed for progression free survival (PFS) and OS. Toxicities were evaluated with CTCAE v4.0. Patients completed EORTC BN20, QLQ-C30, and MMSE questionnaires. Time to deterioration, defined as a 10-point decrease in the score of the function domain or 10-point increase in score of symptom domain, was evaluated. Results: Between 5/2019-6/2021, 43 patients were enrolled, with 4 patients withdrawing due to progression/insurance denial prior to beginning protocol therapy. The median age was 70.2 yrs. and median follow-up was 12.5 months. MGMT methylation was present in 13 (33%). Tumors were multifocal in 10 (26%) and unifocal in 29 (74%). The primary endpoint was met with median PFS/OS was 6.5/12.9 months for all patients and for MGMT methylated patients 11.9/29.8 months respectively. Grade 3 baseline adjusted treatment related adverse events included 1 (3%) confusion, 2 (5%) fatigue, and 1 (3%) seizure. There were no grade 4/5 events. Multivariate analysis revealed MGMT unmethylated, ECOG 2 and biopsy only patients had worse survival. PET volume was more predictive than MRI volume for OS with PET volume>26 cc having a hazard ratio of 3.7. The median time to global deterioration per QLC-C30 was 6.3 months. Conclusions: This phase II trial incorporating 18F-DOPA PET-guided dose escalated hypofractionated proton beam treatment for elderly patients with glioblastoma met its primary endpoint for improved OS and patient reported quality of life compared to historical controls. Further investigation is warranted. Clinical trial information: NCT03778294 .

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]