Rakshit, Sagar

8549 Background: First-line treatment of early stage non-small cell lung cancer (NSCLC) is surgical resection, but with a 5-year survival of only 54%, rates of future relapse are high. Identifying patients at high risk of relapse can help guide adjuvant treatment decisions. Deep convolutional neural network (DCNN) AI models trained on tumor histology have shown incredible flexibility as potential biomarkers, both in lung cancer and more generally across many malignancies. While DCNN models can obtain extremely accurate results when used for routine purposes such as diagnosis, subtyping, and grading of malignancies, most models trained for prognostication or treatment response prediction do not reach performance sufficient for clinical application. Uncertainty quantification (UQ) – a family of techniques that give DCNN models the ability to report confidence alongside predictions – is an underexplored avenue in cancer AI that may help further improve performance and clinical application of models designed to provide clinicians with estimations of risk. Methods: To explore the potential use of UQ in clinically-oriented DCNN models, we trained models on a single-institution, retrospective digital tumor histology slide cohort from patients with Stage I-III NSCLC who underwent surgical resection to predict risk of future relapse. Estimation of uncertainty was performed using dropout as a Bayesian approximation, and uncertainty thresholds were calculated from training sets using a novel method to identify and remove low-confidence predictions. For comparison, a separate multivariate logistic regression was trained in cross-validation using known clinical risk factors. Results: We trained DCNN models on slides from 198 patients (40 relapsed, 158 without relapse). In this cohort, 130 patients had stage I disease (65.6%), 42 had stage II (21.2%), and 26 had stage III (13.2%). The average age was 69 years, 85% were current or previous smokers, and 38 received guideline-concordant adjuvant chemotherapy. Without UQ, a DCNN model predicted risk of future relapse with an average area under receiver operator curve (AUROC) of 0.74 across three-fold cross-validation. Using UQ estimation, 63% of slides were reported with high confidence. Relapse prediction was significantly improved in the high-confidence cohort, with an average AUROC of 0.83 in cross-validation. With a specificity of 70%, this corresponds to an average sensitivity of 86.8% across the three cross-folds (79.1%, 85.5%, and 95.7%). In comparison, a clinical-only multivariate regression model predicted relapse with a cross-validated AUROC of 0.67. Conclusions: This method of uncertainty quantification appears to be a powerful tool to predict lung cancer recurrence risk from digital histopathology while simultaneously providing clinicians with a measurement of algorithm trustworthiness.

Abstract Introduction Multiple myeloma (MM) is frequently accompanied by renal impairment (RI) in 30-40% patients at presentation (Dimopoulos et al, IMWG 2010). As per International Myeloma Working Group (IMWG), RI is defined as Creatinine clearance (CrCl) <40 mL/min. Patients with RI at diagnosis have inferior overall survival (OS), however, over the years their outcome has improved due to treatment with novel agents. We examined the baseline clinical characteristics, and outcomes in patients with a new diagnosis of multiple myeloma at our institution to identify factors associated with baseline RI and study prognostic trend over the last 2 decades. Methods We retrospectively reviewed records for MM patients with a new diagnosis of MM who were seen at Mayo Clinic, Rochester between 1/2004 and 12/2018. RI was defined as per IMWG criteria. Categorical variables were compared using chi-square test and continuous variables using Wilcoxon test. Overall Survival (OS) was estimated using Kaplan-Meier method and difference between groups was assessed using log-rank test. Results In our cohort of 2702 patients, median age at diagnosis was 65 (range, 22-96) years and 61% were males. Number of patients with CrCl <15, 15-29, 30-59 and ≥60 ml/min at diagnosis was 176 (6.5%), 204 (7.6%), 685 (25.3%) and 1637 (61%) respectively. RI was present in 557 (21%) of patients at the time of presentation. On univariate analysis, factors significantly associated with RI at diagnosis were higher age, ECOG PS ≥2, hemoglobin< 10 g/dL, calcium >11 mg/dL , LDH >222 U/L, β-2-microglobulin >5.5 mg/L, free light chain ratio >100, uric acid >8 mg/dL, higher ISS stage and cytogenetics [gain 1q, t (14;20), absence of any trisomy] as shown in table 1. However, on multivariate analysis, only elevated serum calcium >11 g/dL (OR: 11.5; 95%CI:1.78-74.9; p=0.01) and free light chain ratio>100 (OR: 4.17; 95%CI:1.29-13.4; p=0.016) had significant association with RI. The RI group had inferior OS when compared to patients with no RI at diagnosis (4.5 years vs 7.7 years, p-value <0.001). However, use of novel agents was associated with improvement in OS in patients with RI (4.9 years vs 2.6 years; p=0.0001) (Figure 1). Looking at trends over time, the rate of RI at presentation had not changed. Over the last two decades, OS for the CrCl<30 group had significantly improved in the last 5-year period (3.8 years in 2004-2014 vs not reached for 2014-2018; p= 0.0064) as shown in figure 2. Conclusion In our cohort of 2702 patient, frequency of RI at baseline was 21%, which corroborates with previous studies. Hypercalcemia and serum free light chain ratio were found to be independent factors associated with RI. RI at diagnosis is associated with inferior survival, however, use of novel agents in this population has led to improvement of patient outcomes. Figure 1 Figure 1. Disclosures Kapoor: Karyopharm: Consultancy; Cellectar: Consultancy; BeiGene: Consultancy; Pharmacyclics: Consultancy; Sanofi: Consultancy; Amgen: Research Funding; Ichnos Sciences: Research Funding; Regeneron Pharmaceuticals: Research Funding; Glaxo SmithKline: Research Funding; Karyopharm: Research Funding; Sanofi: Research Funding; Takeda: Research Funding; AbbVie: Research Funding. Gertz: Aurora Biopharma: Other: Stock option; Ionis Pharmaceuticals: Other: Advisory Board; Akcea Therapeutics, Ambry Genetics, Amgen Inc, Celgene Corporation, Janssen Biotech Inc, Karyopharm Therapeutics, Pfizer Inc (to Institution), Sanofi Genzyme: Honoraria; AbbVie Inc, Celgene Corporation: Other: Data Safetly & Monitoring; Akcea Therapeutics, Alnylam Pharmaceuticals Inc, Prothena: Consultancy. Dingli: Alexion: Consultancy; Novartis: Research Funding; Apellis: Consultancy; GSK: Consultancy; Sanofi: Consultancy; Janssen: Consultancy. Kumar: Roche-Genentech: Consultancy, Research Funding; Adaptive: Membership on an entity's Board of Directors or advisory committees, Research Funding; Beigene: Consultancy; Novartis: Research Funding; Merck: Research Funding; Takeda: Consultancy, Membership on an entity's Board of Directors or advisory committees, Research Funding; Amgen: Consultancy, Research Funding; Bluebird Bio: Consultancy; Oncopeptides: Consultancy; Carsgen: Research Funding; Celgene: Membership on an entity's Board of Directors or advisory committees, Research Funding; Tenebio: Research Funding; Janssen: Consultancy, Membership on an entity's Board of Directors or advisory committees, Research Funding; Astra-Zeneca: Consultancy, Membership on an entity's Board of Directors or advisory committees, Research Funding; BMS: Consultancy, Research Funding; Abbvie: Consultancy, Membership on an entity's Board of Directors or advisory committees, Research Funding; Antengene: Consultancy, Honoraria; KITE: Consultancy, Membership on an entity's Board of Directors or advisory committees, Research Funding; Sanofi: Research Funding.

PURPOSE: The benefit of routine pre-emptive screening for severe acute respiratory syndrome coronavirus 2 infections in patients with cancer before cancer-directed therapies is unclear. Herein, we characterize the outcomes of a cohort of patients with cancer who were diagnosed with COVID-19 by routine screening (RS) in comparison with those diagnosed on the basis of clinical suspicion or exposure history (nonroutine screening [NRS]). METHODS: A multisite prospective observational study was conducted at three major and five satellite campuses of the Mayo Clinic Cancer Center between March 18 and July 31, 2020. The primary outcome was COVID-19–related hospital admission. Secondary outcomes included intensive care unit admissions and all-cause mortality. RESULTS: Five thousand four hundred fifty-two patients underwent RS in the outpatient setting only, and 44 (0.81%) were diagnosed with COVID-19. RS detected 19 additional patients from the scheduled inpatient admissions for surgical or interventional procedures or inpatient chemotherapy. One hundred sixty-one patients were diagnosed with COVID-19 on the basis of NRS. COVID-19–related hospitalization rate (17.5% v 26.7%; P = .14), intensive care unit admission (1.6% v 5.6%; P = .19), and mortality (4.8% v 3.7%; P = .72) were not significantly different between the RS and NRS groups. In the multivariable analysis, age ≥ 60 years (odds ratio, 4.4; P = .023) and an absolute lymphocyte count ≤ 1.4 × 109/L (odds ratio, 9.2; P = .002) were independent predictors of COVID-19–related hospital admission. CONCLUSION: The COVID-19 positivity rate was low on the basis of RS. Comparing the hospital admission and mortality outcomes with the NRS cohort, there were no significant differences. The value of routine pre-emptive screening of asymptomatic patients with cancer for COVID-19 remains low.

Abstract Abstract Importance: The benefit of routine screening for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infections in patients with cancer before cancer-directed therapies is unclear. Herein, we characterize the outcomes of a cohort of cancer patients diagnosed with Coronavirus Disease 2019 (COVID-19) by routine screening in comparison with those diagnosed based on clinical suspicion or exposure history (non-routine screening). Objective: To describe and compare the outcomes of cancer patients diagnosed with COVID-19 on routine screening vs. non-routine screening at a multi-site tertiary cancer center. To identify risk factors for COVID-19-related hospital admission. Design: A multi-site prospective observational study was conducted between March 18 and July 31, 2020. Setting: Three major and 5 satellite campuses of the Mayo Clinic Cancer Center. Participants: Adult patients diagnosed with active cancer within the past five years and confirmed SARS-CoV-2 infection were included. Primary Outcomes and Measures: Clinical and laboratory data were assessed as independent variables. The primary outcome was COVID-19-related hospital admission. Secondary outcomes included intensive care unit (ICU) admissions and all-cause mortality. Results: Between March 18 and July 31, 2020, 5452 patients underwent routine screening in the outpatient setting, 44 (0.81%) were diagnosed with COVID-19. Routine screening detected additional 19 patients from inpatient and pre-procedural settings; 161 patients were diagnosed with COVID-19 based on non-routine screening. The median age of the entire cohort at diagnosis was 54 years, and 95 patients (42.2%) were female. COVID-19 related-hospitalization rate (17.5% vs. 26.7%, p=0.14), ICU admission (1.6% vs. 5.6%, p=0.19), and mortality (4.8% vs. 3.7%, p=0.72) were not significantly different between routine screening and non-routine screening groups. In the multivariable analysis, age ≥ 60 years (odds ratio: 4.4, p=0.023) and an absolute lymphocyte count ≤1.4 × 109/L (odds ratio: 9.2, p=0.002) were independent predictors of COVID-19-related hospital admission. Conclusions and Relevance: The COVID-19 positivity rate was low based on routine screening. Comparing the outcome with the non-routine screening cohort, there was no significant difference. These results led to an important practice change at our cancer center. We currently follow a testing strategy based on symptoms, exposure, risk factors, and clinical judgment. Citation Format: Zhuoer Xie, Antoine N. Saliba, Jithma Abeykoon, Umair Majeed, Daniel Almquist, Julia Wiedmeier-Nutor, Evandro Bezerra, Xavier Andrade-Gonzalez, Ashley Hickman, Karl Sorenson, Sagar Rakshit, Christopher Wee, Sri Tella, Anuhya Kommalapati, Nadine Abdallah, Joshua Pritchett, Mariza De Andrade, Dipesh Uprety, Andrew Badley, Joleen Hubbard, Naseema Gangat, Carrie A. Thompson, Thomas Witzig, Robert R. McWilliams, Konstantinos Leventakos, Thorvardur R. Halfdanarson. Outcomes of COVID-19 in patients with cancer: Results of a prospective observational comparison of routine screening strategy versus testing based on clinical suspicion [abstract]. In: Proceedings of the AACR Virtual Meeting: COVID-19 and Cancer; 2021 Feb 3-5. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(6_Suppl):Abstract nr S06-03.

Introduction Immune-mediated hematologic disorders are relatively rare but potentially life-threatening immune-related adverse events (irAE) with immune checkpoint inhibitor (ICI) therapy (Shiuan, Beckermann et al 2017). The management strategy often relies on systemic corticosteroids and other immunosuppressant agents in steroid-refractory cases (Brahmer, Lacchetti et al 2018). We hereby describe our institution's experience in diagnosing and managing immune-related hematologic adverse events (ir-h-AE). Methods We retrospectively searched the electronic health record for all ir-h-AE diagnosed by treating providers and confirmed by study authors between 2015 and 2020. The search included patients on ICI and the following conditions: immune-related neutropenia (IRN), autoimmune hemolytic anemia (AIHA), cold agglutinin disease (CAD), immune thrombocytopenia (ITP), immune-related pancytopenia, aplastic anemia (AA), and pure red cell aplasia (PRCA). For AIHA, PRCA, and CAD, partial response (PR) was defined as hemoglobin greater than 10 g/dL but < 12 g/dL and 2 g/dL above the nadir without blood product transfusion. Complete response (CR) was defined as hemoglobin of ≥ 12 g/dL and 2 g/dL above the nadir without blood product transfusion (Peñalver, Alvarez-Larrán et al. 2010). For ITP, CR was defined as platelet count ≥ 100 × 109/L and absence of bleeding. PR was defined as platelet count ≥ 30 × 109/L and ≥ two-fold increase from the baseline count and absence of bleeding (Gómez-Almaguer, Tarín-Arzaga et al. 2013). For IRN, CR was defined as absolute neutrophil count (ANC) ≥ 1500/ mm3 on 2 consecutive measurements separated by 7 days; PR was defined as ANC ≥ 500/ mm3 sustained over 7 days. JMP® 14.1 was used for data analysis. Results Seventeen patients were identified, 9 (52.9%) of whom were female (table 1). Median age was 68 years (range 26-78). The most common pre-existing malignancies were melanoma (6 patients, 35.3%), non-small cell lung cancer (4 patients, 23.5%), and gastrointestinal cancer (3 patients, 17.6%). The ir-h-AE included 7 patients (41.2%) with warm AIHA, 4 (23.5%) with ITP (1 had immune-related pancytopenia that resolved with ITP persisting), 3 (17.6%) with IRN, 1 with AA, 1 with CAD, and 1 with PRCA and mild ITP. All patients had received an ICI dose within 60 days of diagnosis (range 3-52 days, median 18 days). Only 1 patient had a concomitant chronic autoimmune disorder (antinuclear antibody-positive arthritis). Four patients (23.5%) were receiving chemotherapy. Sixteen patients (94.1%) required treatment and received corticosteroids. Eight out of 15 evaluable patients (53.3%) responded to corticosteroids only. Another 4 responders (26.7%) required additional therapies including intravenous immunoglobulins (IVIG) (4, 23.5%) and/or rituximab (4, 23.5%). Overall response rate was 71.4% for AIHA and 100% for IRN. ITP was likely underrepresented in our cohort, and the response rate of 66.7% is likely lower than the response rate in clinical practice. This could be related to the fact that many mild thrombocytopenia cases are not further evaluated in clinical practice. None of our patients died from the sequelae of an ir-h-AE. Two patients (11.8%) died within 30 days of ir-h-AE diagnosis (disseminated intravascular coagulation likely due to the underlying malignancy and sepsis from bacterial pneumonia). At the time of ir-h-AE diagnosis, the malignancy was showing evidence of response in 9 patients (52.9%), progression in 6 patients (35.3%), and stability in 2 patients (11.8%). ICI therapy was permanently discontinued in 9 patients (52.9%) and temporarily held in 6 patients (35.3%). Three of the 5 patients (60%), who were rechallenged with ICI therapy, had recurrence of the same ir-h-AE (AIHA) with adequate response to therapy. Non-hematologic irAE (7 patients, 41.2%) included 3 patients with hepatitis, 2 with thyroiditis, 2 with pneumonitis, 1 with colitis, and 1 with dermatitis. Conclusion Our clinical experience with ir-h-AE highlights the importance of exercising vigilance in assessing the hematologic parameters of patients receiving ICI therapy. While relatively rare, ir-h-AE may impact the clinical course of patients with cancer and their eligibility for therapies that have the potential of offering durable control of the underlying malignancy. ir-h-AE respond to classical therapies including corticosteroids, IVIG, and rituximab. Disclosures Block: Merck: Research Funding; Bristol-Myers Squibb: Research Funding; Genentech: Research Funding; Pharmacyclics: Research Funding; Marker Therapeutics: Research Funding; Transgene: Research Funding; Immune Design: Research Funding. Kottschade:Bristol-Myers Squibb: Consultancy, Research Funding.

Background: Lytic bone lesions are one of the most common clinical characteristics of patients with multiple myeloma (MM) and identification of bone lesions help distinguish between patients with smoldering multiple MM and active MM. Given this, the most recent update of the diagnostic criteria for MM incorporates advanced imaging approaches for distinguishing between the two entities. Several small retrospective studies have compared conventional skeletal survey (SS) with whole-body low dose computed tomography (WBLDCT) scan or the CT portion of a positron emission tomography (PET) scan. Conducting prospective studies comparing these two modalities side by side is limited by the radiation exposure. We undertook this study to provide a comparison between these two imaging modalities in terms of their ability to recognize lytic bone lesions in patients with MM. Patients and methods: This was a retrospective study of consecutive patients with the diagnosis of MM treated at Mayo Cinic Hospital during 2004- 2018. Patients were included if they had a conventional skeletal survey no more than 3 months before MM diagnosis or any time after diagnosis and also had a WBLDCT or PET-CT within a 12-month period following the skeletal survey. We chose this approach since it is unlikely that a patient would have had both modalities done at the same time as part of standard of care (SOC). This approach was also facilitated by a gradual change in our SOC for skeletal imaging from conventional skeletal survey to WBLDCT. To measure the robustness of our findings, we performed a sensitivity analysis in patients who had the exam ≤6 months apart. Proportions were compared using a chi-square test and survival estimates were calculated using the Kaplan- Meier methodology and compared using log rank test. Results: The overall study cohort had 1040 patients, median age was 62 years at diagnosis (range, 24-94), 61% were male and 39% were female. The median time to the skeletal survey was 8 months from diagnosis (range, 0-160) and the median time to WBLDCT or PET-CT from SS was 2 months (0-12). A PET-CT was available in 789 (76%) of patients and WBLDCT in 251 (24%) of patients. Among the 300 patients with no lesions identified by SS, 188 (63%) had lytic lesions identified by CT, while in 60/740 (8%) patients with a positive SS did not have lytic lesions observed on the CT (p<0.0001). Overall, CT identified lytic lesions in 868 (84%) patients compared with 740 (71%) patients with lytic lesions seen on SS. Although the proportion of lytic lesions was slightly higher with PET compared to WBLDCT, the analysis did not reach statistical significance (65% vs. 56%; p = 0.17; Table 1a). After restricting the analysis to those with ≤6 months gap between low dose CT and SS (n=737), the result did not change (Table 1b). Further examination demonstrated that presence of lytic lesions by SS had no impact on OS from diagnosis, detection of lesions on the CT exam was associated with an inferior survival (112 vs. 81 months, p<0.0001). (Figure 1) Conclusion: Skeletal survey has been the screening technique of choice for evaluation of bone involvement in MM for decades. However, CT based approaches have higher sensitivity with lytic lesions identified in a higher proportion of patients. The prognostic value of lytic disease is evident with CT based detection, again suggesting that this provides a more accurate estimate of the skeletal disease burden. Low dose CT rather than conventional radiographs should be the modality of choice for monitoring disease-progression and diagnosing active multiple myeloma. Disclosures Kapoor: Cellectar: Consultancy; Janssen: Research Funding; Sanofi: Consultancy, Research Funding; Amgen: Research Funding; Takeda: Honoraria, Research Funding; GlaxoSmithKline: Research Funding; Celgene: Honoraria. Gertz:Abbvie: Other; Celgene: Other; Physicians Education Resource: Other: personal fee; Medscape: Other: personal fee, Speakers Bureau; Janssen: Other: personal fee; Spectrum: Other: personal fee, Research Funding; Johnson and Johnson: Speakers Bureau; Teva: Speakers Bureau; Annexon: Other: personal fee; Research to Practice: Other; Sanofi: Other; Proclara: Other; DAVA oncology: Speakers Bureau; Springer Publishing: Patents & Royalties; Appellis: Other: personal fee; Amgen: Other: personal fee; Prothena: Other: personal fee; Aurora Bio: Other; Alnylam: Other: personal fee; Ionis/Akcea: Other: personal fee. Dispenzieri:Takeda: Research Funding; Pfizer: Research Funding; Janssen: Research Funding; Intellia: Research Funding; Alnylam: Research Funding; Celgene: Research Funding. Dingli:Alexion: Consultancy; Janssen: Consultancy; Sanofi-Genzyme: Consultancy; Millenium: Consultancy; Bristol Myers Squibb: Research Funding; Rigel: Consultancy; Apellis: Consultancy; Karyopharm Therapeutics: Research Funding. Lin:Merck: Research Funding; Takeda: Research Funding; Legend BioTech: Consultancy; Juno: Consultancy; Bluebird Bio: Consultancy, Research Funding; Celgene: Consultancy, Research Funding; Novartis: Consultancy; Vineti: Consultancy; Janssen: Consultancy, Research Funding; Kite, a Gilead Company: Consultancy, Research Funding; Sorrento: Consultancy, Membership on an entity's Board of Directors or advisory committees; Gamida Cells: Consultancy. Kumar:Dr. Reddy's Laboratories: Honoraria; AbbVie: Other: Research funding for clinical trials to the institution, Consulting/Advisory Board participation with no personal payments; Takeda: Other: Research funding for clinical trials to the institution, Consulting/Advisory Board participation with no personal payments; Sanofi: Research Funding; Novartis: Research Funding; Janssen Oncology: Other: Research funding for clinical trials to the institution, Consulting/Advisory Board participation with no personal payments; Genentech/Roche: Other: Research funding for clinical trials to the institution, Consulting/Advisory Board participation with no personal payments; Merck: Consultancy, Research Funding; Adaptive Biotechnologies: Consultancy; Celgene/BMS: Other: Research funding for clinical trials to the institution, Consulting/Advisory Board participation with no personal payments; Amgen: Consultancy, Other: Research funding for clinical trials to the institution, Consulting/Advisory Board participation with no personal payments, Research Funding; Carsgen: Other, Research Funding; Cellectar: Other; MedImmune: Research Funding; Karyopharm: Consultancy; BMS: Consultancy, Research Funding; Tenebio: Other, Research Funding; Genecentrix: Consultancy; Oncopeptides: Consultancy, Other: Independent Review Committee; IRC member; Kite Pharma: Consultancy, Research Funding.