PD-L1 Scores Research Articles

Temporal Effect on PD-L1 Detection and Novel Insights Into Its Clinical Implications in Non-Small Cell Lung Cancer.

Several studies rely on archived tissue blocks to assess the PD-L1 scores; however, a detailed analysis of potential variations of scores between fresh and archived tissue blocks still lacks. In addition, the prognostic implications of PD-L1 in lung cancers have not yet been completely understood. Here, we aimed to investigate the temporal variation in PD-L1 scores from clinical samples and the clinical implications of PD-L1 in non-small cell lung cancer (NSCLC). NSCLC cases from January 2005 to June 2023 were considered for this study, and PD-L1 scores in archived and fresh tissue blocks were analyzed. Association of PD-L1 with various driver mutations was explored, and implications of PD-L1 in progression-free survival (PFS) and overall survival (OS) were analyzed. Our study revealed a significant disparity in PD-L1 scores between archived and fresh tissue blocks, and a temporal variation in scores within 6 months of tissue acquisition. Advanced-stage primary tumors, metastatic lymph nodes, and visceral pleural invasion revealed higher PD-L1 expression as presented by tumor proportion score (TPS). Notably, in fully resected stage I/II NSCLC cases, OS was better in the high PD-L1 (≥ 50% TPS) cohort with driver mutations compared to cases without driver mutations (hazard ratio-0.5129, 95% confidence interval 0.2058-1.084, p = 0.0779). In contrast, high PD-L1 was associated with worse OS compared to no PD-L1 (< 1% TPS) (hazard ratio-2.431, 95% confidence interval 1.144-6.656, p = 0.0242) in the cohort without driver mutations. Furthermore, the presence of a KRAS mutation favored the outcome of anti-PD-L1/PD1 immunotherapy in advanced NSCLC. PD-L1 detection from tissue blocks was found to vary temporally, urging for a prioritized consideration for patients with marginal scores when archived blocks are employed for its detection. Prognostic roles of PD-L1 were associated with driver mutations, and KRAS mutations favored the outcome of anti-PD-L1/PD1 therapy in advanced NSCLC.

Impact of PROphet Test in Changing Physicians' Therapeutic Decision-Making for Checkpoint Immunotherapy in Non–Small-Cell Lung Cancer

PurposeImmune Checkpoint Inhibitor (ICI) regimens are approved for first-line treatment of metastatic nononcogene-driven NSCLC. Guidelines do not differentiate which patients with PD-L1 ≥ 50% should receive ICI monotherapy. The clinically validated PROphet NSCLC plasma proteomic-based test is designed to inform this therapeutic decision. MethodsOne hundred oncologists were presented with 3 “virtual” metastatic NSCLC cases with PD-L1 scores and asked to recommend an approved first-line regimen. They then watched an online educational webinar on the PROphetNSCLC test. Postwebinar, the same cases were represented with the addition of a PROphet result, and oncologists again recommended a first-line regimen. Responses were compared to assess the impact on first-line treatment selection. ResultsTreatment recommendation changed in 39.6% of PROphet-tested cases, with 93% of physicians changing at least 1 case. In the PD-L1 ≥ 50% group, 89% of physicians changed their recommendation, followed by 77%, in PD-L1 < 1%, and 36% in PD-L1 1% to 49%. ​In the PD-L1 ≥ 50%, PROphet POSITIVE group, the recommendation for ICI monotherapy increased from 60% to 89%. ​For the PD-L1 ≥ 50%, PROphet NEGATIVE group, the recommendation for monotherapy dropped from 60% to 9%. In the PD-L1 < 1%, PROphet NEGATIVE group, 35% of patients were spared toxicity from ICI compared to 11% in PROphet untested cases. ConclusionAdding PROphet to PD-L1 expression impacted therapeutic decision making in first-line NSCLC. PROphet identifies those predicted to have an overall survival benefit from ICI monotherapy versus combination versus chemotherapy, improving the probability of efficacy and reducing toxicity for some patients.

PD-L1 and PD-L2 Expression in Different Tumor Stages and Types of Malignant Salivary Gland Neoplasms: A Single-center Experience.

There is a limited amount of data on the role of programmed cell death ligand (PD-L) -1 and PD-L2 in salivary gland carcinomas. We aimed to evaluate the prognostic value of PD-L1 and PD-L2 expressions, which are closely related to immune mechanisms, with respect to salivary gland tumor types and stages. Data from patients with salivary gland masses surgically removed between 2006 and 2021, diagnosed with a malignant salivary gland neoplasm, were retrospectively analyzed. Immunoreactivity for PD-L1 and PD-L2 was performed on resection materials. The mean age of 90 patients was 52.1±18.8 and 46.7% were male. Overall, 55.6% of patients were diagnosed with adenoid cystic carcinoma (ACC), 23.3% with mucoepidermoid carcinoma (MEC), 16.7% with acinic cell carcinoma (AciCC), 3.3% with ductal carcinoma (DC), and 1 patient with pleomorphic adenoma ex carcinoma (PA-ex-CA). In all, 52% of ACC, 12% of AciCC, 24% of MEC, and 12% of DC cases were at stage IV. The tumor diameter, frequencies of lymphovascular invasion, metastasis, positive surgical margin, recurrence, and mortality rates of patients at stages III and IV were significantly larger than those at stages I and II ( P <0.05). The percentages of tumor cell score (TCS) and immune cell score (ICS) for PD-L1 were significantly higher among patients with MEC compared with those with other types of tumors ( P =0.0011). However, the percentages of combined score (CS) for PD-L1 and tumor cell score for PD-L2 were comparable among tumor types ( P >0.05). No significant difference was found in these scores for PD-L1 between tumor stages ( P >0.05), but for PD-L2, all patients at stage I had TCS <1% for PD-L2, while all patients at stages II and III, and 92% of patients at stage IV had TCS ≥1% ( P <0.0001). High expression of PD-L1 was mostly observed in MEC cases ( P =0.0016), while all patients with AciCC had a low PD-L1 expression level ( P =0.0206). The mean tumor diameter, rate of lymphovascular invasion, perineural invasion, metastasis, positive surgical margin, recurrence, type of treatment, mortality, and TILs ratio did not differ significantly according to PD-L1 expression level ( P >0.05). The percentage of tumor-infiltrating lymphocytes was comparable among negative and positive PD-L1 scores according to both 1% and 5% threshold values ( P >0.05). High PD-L1 expression is rare in AciCC, while PD-L1 expression is high in MEC. Our findings underline the importance of future screening for PD-L1 and PD-L2 before patients undergoing immunotherapies in all salivary gland tumors.

Multi-institution real world analysis of patients with non-small cell lung cancer (NSCLC) treated with standard of care (SOC) neoadjuvant chemo-immunotherapy (chemoIO).

8049 Background: The approval of neoadjuvant chemoIO transformed the treatment paradigm for patients (pts) with early-stage, resectable NSCLC. Here we present a multi-institution, real world analysis of pts treated with neoadjuvant chemoIO as SOC. Methods: This was a retrospective analysis performed across 3 academic institutions. Pts were eligible if they received neoadjuvant chemoIO as SOC therapy for NSCLC with intent to proceed to surgery. Results: 107 pts were included. Median age was 69 (range, 45-84); 47% were male and 83% were white. 84% were former/current smokers with median 35 pack year history. ECOG scores of 0/1/2 at treatment start were 55%/44%/1%, respectively. Histologies were 65% adenocarcinoma, 27% squamous, and 7% other. PD-L1 scores (in 93% of pts) were ≥ 50% in 30%, and median TMB (in 73% of pts) was 8.2 (range, 0-56). Of tumors with next generation sequencing (81%), 45% had KRAS mutation and 18% had other oncogenic driver present. Pts were clinical stage IB (0.9%), IIA (6%), IIB (24%), IIIA (60%), or IIIB (8%); 47% had N2 disease and 3% had N3. 78% of pts underwent ≥ 3 cycles of chemoIO, with most pts receiving carboplatin (70%). 11 pts (10%) experienced grade ≥ 3 immune-related adverse events, 72% of which were pre-op. 82 pts (77%) proceeded to surgical resection. Of those who did not undergo surgery, 68% were stage III. The most common reason for no surgery was progression (PD)/unresectability (52%), followed by operability (36%) and pt decision (8%). 2 pts (2%) proceeded to surgery but had unresectable tumors. Median interval between last cycle of induction and surgery was 43 days (d) (range, 17-210); median length of stay was 4d (range, 1-21). 5 of 79 (6%) were readmitted within 30d. For pts without resection (27), 48% underwent chemoradiation (CRT), 26% RT alone, and 15% systemic therapy. 29 tumors (36%) demonstrated major pathologic response (MPR); of these, 16 (20%) had pathologic complete response (pCR). 2 pts had no decrement in viable tumor. Of 79 pts with adjuvant treatment data, 11% underwent RT alone, in combination with chemo, or in sequence with systemic therapy; 8% were recommended for RT but declined. 24% had systemic treatment including 14% IO, 6% targeted therapy, and 4% chemo. 18/107 pts (17%) had PD; 14/18 had stage III disease. PD occurred in 10/25 pts (40%) who did not have surgery; 6 of those had undergone definitive CRT or RT. PD occurred in 8/82 pts (10%) after surgery; of these, 5 had undergone adjuvant treatment at median 58d (range, 33-103). Conclusions: Real world analysis demonstrated older, frailer pts with more advanced disease than those enrolled on trial; lower chemoIO completion rates but comparable pCR and resection rates were seen. Further analyses will focus on outcomes of pts who did not undergo surgery.

Digital versus manual pathology scoring of PD-L1 in patients with esophageal squamous cell carcinoma (ESCC).

e16023 Background: Immune checkpoint inhibitors (ICIs) are standard-of-care for patients with ESCC, but the association between PD-L1 and clinical outcomes of patients with ESCC receiving ICIs treatment is unclear. Currently, different PD-L1 antibody clones, scoring, and cutoffs were used for different ICIs development in ESCC. Thus, we develop a digital pathology (DP) algorithm to generate multiple scores at the same time with high reproducibility. We evaluate DP algorithm by: (1) comparing the digital and manual scores of SP263 and 22C3 assays; (2) evaluating the association with OS. Methods: Tumor tissue samples were collected from ESCC patients treated at National Taiwan University Hospital between 2017 and 2020. Slides were tested for PD-L1 using SP263 and 22C3 assays at a College of American Pathologists accredited central laboratory. H&amp;E and PD-L1 slides were scanned for whole slide images (WSIs) using DP200 scanner at 20x. HALO (v3.5) by indica labs was used for DP development by deep learning neural network algorithms. WSIs of SP263 were annotated for region of interest (ROI), tumor/stroma area, and tumor, immune and fibroblast cells by board certified pathologists. After the algorithm is developed and run on SP263 WSIs, a board-certified pathologist reviewed the images and provided an agreement score of DP markups. The images were excluded if an agreement was less than 80%. This algorithm was then applied to 22C3 WSIs. Data of areas and cells were exported for subsequent statistical analysis using R. We calculated the DP scores based on the definition of manual readouts of TAP, CPS and s/iTILs in ROI and tumor and stroma regions in both clones’ WSIs. These DP scores were compared with manual scores using Spearman’s rank-order correlation. Association with OS were evaluated using cox proportional hazard regression. Results: Of 235 evaluable images, twelve (5.1%) were excluded due to the low agreement with the pathologist's judgment. DP generated scores showed a strong correlation with pathologists scoring (rho = 0.80 for SP263 and rho = 0.74 for 22C3). DP generated CPS-like and TAP-like scores in ROI, and tumor and stroma regions from SP263 and 22C3 also demonstrated a strong correlation between each other (rho: 0.77 - 0.86). However, s/iTILs correlation between DP and pathologist score is negligible (rho: 0.002 - 0.34). While examining the association of PD-L1 scores and s/iTILs with OS, no clear associations were found with either DP or pathologists’ scores, after adjusting for stage, ECOG, age and gender. Conclusions: DP algorithms developed in this study demonstrated the efficiency of generating multiple DP scores of CPS, TAP, and s/iTILs from one image at the same time, which performed comparable to manual scores. DP algorithm developed based on SP263 is generalizable to 22C3 and showed a good performance. Further algorithm adjustment will be made to overcome the misclassification in the 12 failed cases.

Abstract CT191: First-in-human, dose escalation and expansion study of MT-6402, a novel engineered toxin body (ETB) targeting PD-L1, in patients with PD-L1 expressing relapsed/refractory advanced solid tumors: Interim data

Abstract MT-6402 is a PD-L1-targeted ETB capable of directly killing PD-L1 expressing cells by internalization of a de-immunized Shiga-like toxin A subunit (SLTA), resulting in ribosomal destruction. Targeting PD-L1 expressing tumor cells may directly drive tumor regression, whereas targeting PD-L1 expressing immune cells may release immunosuppression and drive tumor immune recognition. MT-6402 also delivers an HLA-A*02 restricted cytomegalovirus (CMV) class I antigen into PD-L1 expressing cells (antigen seeding) that can be recognized by existing CMV-specific cytotoxic T cells. A first-in-human dose escalation (Part A) and expansion (Part B) study with a weekly intravenous administration of MT-6402 in 4-week cycles was initiated in 2021. Safety and tolerability are primary objectives. Efficacy, pharmacokinetics, and pharmacodynamic data are also collected. As of 10Dec2023, 53 total patients (pts) with PD-L1-expressing advanced solid tumors received ≥ 1 dose of MT-6402: 48 pts in Part A ranging in dose 16 to 100 µg/kg and 5 pts in Part B at 63 µg/kg, just below the maximum tolerated dose (MTD) of 83 µg/kg. MT-6402 was well tolerated in Part A, completed Sept2023. 5 dose limiting toxicities have been reported: a Grade (G)3 maculopapular rash at 24 µg/kg, a G3 infusion related reaction at 63 µg/kg, a G1 cardiac troponin T increase leading to dosing hold for ≥2 weeks at 83 µg/kg, a G3 maculopapular rash at 100 µg/kg, and a G1 cardiac troponin T increase at 100 µg/kg. There were no G4 or G5 drug-related adverse events. In Part B, pts with PD-L1 scores of TPS ≥ 50% will be treated at MTD or lower and 5 pts have been treated at 63 µg/kg. Signs of antitumor activity have been reported including a long-lasting PR with 70% tumor reduction have been observed in a pt with CPI-refractory nasopharyngeal carcinoma on treatment 14+ cycles at 63 µg/kg. This pt has a PD-L1 TPS score of 2% indicating immune cell clearance may have driven the response. The most common cancer in Part A is Squamous Cell Carcinoma of the Head and Neck (10 pts). Pharmacodynamic effects include a significant reduction in peripheral PD-L1+ immune cells, indicating a PD-L1 target-specific effect. Substantial alterations in serum cytokines associated with the tumor microenvironment were noted after each dose in most pts in a dose dependent manner. Most pts display increased CD8/CD4 ratios with IL-2 and TNFα cytokine release, indicative of immune activation due to elimination of immune cells. CMV-specific CD8 T-cells exhibited a pattern of extravasation from the periphery in HLA-A*02+/CMV+ pts suggesting antigen seeding. These results describe a novel approach to checkpoint modulation by MT-6402 with targeted cell killing and altered tumor immunophenotype, particularly in HLA-A*02/CMV+ pts. These results provide rationale for continued development of MT-6402. Citation Format: Brian A. Van Tine, Eugene Ahn, Rebecca A. Redman, Minal A. Barve, William J. Edenfield, Julio Peguero, Victoria M. Villaflor, Drew W. Rasco, Steven F. Powell, Admasu T. Mamuye, Chris B. Moore, Joshua M. Pelham, Soratree Charoenthongtrakul, David R. Spigel. First-in-human, dose escalation and expansion study of MT-6402, a novel engineered toxin body (ETB) targeting PD-L1, in patients with PD-L1 expressing relapsed/refractory advanced solid tumors: Interim data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(7_Suppl):Abstract nr CT191.

Abstract 3629: Prevalence of claudin18.2 expression in gastric/gastroesophageal junction adenocarcinoma among patients in TranStar101 and TranStar102 clinical trials

Abstract Background: Claudin18.2 (CLDN18.2) is a tight junction protein highly specific to gastric mucosa, and a validated target for gastric cancer (GC) treatment. Immune checkpoint therapy targeting PD-1 combined with chemotherapy has been approved as the first line therapy of GC. Understanding the expression profiles of CLDN18.2 and PD-L1 could guide the development of combination therapies to maximize the benefits of these two agents. This study investigated the prevalence of CLDN18.2 expression in gastric/gastroesophageal junction adenocarcinoma (G/GEJC) screening samples from studies Transtar101 (NCT04396821 in US) and TranStar102 (NCT04495296 in China), and its correlation with various clinical characteristics and PD-L1 expression. Methods: CLDN18.2 expression in formalin-fixed, paraffin-embedded (FFPE) G/GEJC tissue samples was prospectively detected by an immunohistochemistry-based LDT using an in-house anti-CLDN18.2 antibody (Clone14G11) on the Leica Bond III stainer. CLDN18.2 expression was assessed by scoring the staining intensity (0, 1+, 2+, 3+) and the percentage of positive tumor cells. Positive CLDN18.2 expression is defined as the cutoff at ≥10% of tumor cell with ≥1+ staining intensity for this evaluation. PD-L1 expression was assessed by combined positive score (CPS) using Agilent’s PD-L1 IHC 28-8 pharmDx. Both assays were conducted in CAP/CLIA certified LabCorp central lab. Results: 547 G/GEJC patient samples had CLDN18.2 results as part of the screening procedures for the clinical trials. Of these patients (454 GC/93 GEJC), 439 (80%) were Asian, 29 (5%) were Caucasian, 2 (0.4%) were other ethnic group, and 77 (14%) were not recorded; 434 (79%) were primary tumors and 113 (21%) were metastasis; 201 (37%) were core needle biopsies (CNB), 108 (20%) were surgical resections (SR) and 238 (44%) had no information. 311 (57%) samples had positive CLDN18.2 expressions (≥10%/≥1+). No significant difference in CLDN18.2 positive rates were found between TranStar101 and TranStar102 studies (p=0.801), Asian and Caucasian (p=0.612), GC and GEJC (p=0.373), primary and metastatic specimens (p=0.055) or CNBs and SRs (p=0.715). Out of 83 TranStar102 specimens that had both CLDN18.2 and PD-L1 results, 15 (18%) had PD-L1 CPS≥5, 59 (71%) had positive CLDN18.2, and 10 (15%) had both positive CLDN18.2 and PD-L1 CPS≥5. No correlation (p=0.393) was observed between PD-L1 scores (CPS&amp;lt;5 or CPS≥5) and CLDN18.2 expression (≥10%/≥1+ or &amp;lt;10%/≥1+). Conclusions: Data suggested CLDN18.2 expression levels were independent of PD-L1 status, and support the use of Transcenta 14G11 antibody for CLDN18.2 detection regardless of sample collection methods, location, and patient demographics. An anti-CLDN18.2 companion diagnostic device based on 14G11 is being developed (CLDN18.2 IHC 14G11 pharmDx, Agilent Technologies, Inc.). Citation Format: Li Shen, Carol Mao, Alan Lin, Ying Gu, Jenny Milata, Lijuan Zhang, Chuan Qi, Jenny Yao, Steven Yu, Changjun Yue, Lin Shen, Yelena Janjigian, Xueming Qian, Wen-I Chang, Caroline Germa. Prevalence of claudin18.2 expression in gastric/gastroesophageal junction adenocarcinoma among patients in TranStar101 and TranStar102 clinical trials [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3629.

Network meta-analysis of first-line systemic regimens for older patients with advanced NSCLC.

Various immunotherapy treatments have received approval for the treatment of advanced non-small cell lung cancer (NSCLC), either as standalone or in conjunction with chemotherapy, contingent upon the extent of PD-L1 expression. These treatments are commonly utilized in clinical practice. However, a specific gap exists in direct comparisons of these regimens in elderly patients. The aim of this network meta-analysis (NMA) was to examine the effectiveness of PD-1/PD-L1 inhibitors, either alone or in conjunction with chemotherapy, as the initial treatment for elderly patients diagnosed with advanced NSCLC. We extensively searched PubMed, EMBASE and the Cochrane Library to gather randomized clinical trials that utilized PD-1/PD-L1 inhibitors as the first-line therapy for advanced NSCLC. By means of Bayesian NMA, we conducted an analysis on hazard ratios (HRs) related to overall survival (OS). A total of 5240 patients were included in the 21 trials. Across all studies, cemiplimab exhibited a noteworthy superiority to chemotherapy in terms of OS [HR = 0.48, 95% confidence interval (CI): 0.3-0.77]. In the subgroup analysis, it was observed that patients with PD-L1 expression of 50% or higher experienced the greatest OS benefit from cemiplimab (HR = 0.48, 95% CI: 0.3-0.77). Conversely, the cohort with unselected PD-L1 scores (>1 or any score) exhibited the greatest OS benefit when treated with pembrolizumab combined with chemotherapy, as indicated by a HR of 0.69 (95% CI: 0.52-0.9). Chemotherapy combined with pembrolizumab and cemiplimab monotherapy may represent the reference regimens for older patients with NSCLC with unselected and >50% PD-L1 expression, respectively.

Comprehensive analysis of integrin αvβ3/α6β1 in prognosis and immune escape of prostate cancer.

Integrin αvβ3/α6β1 are crucial in the transduction of intercellular cancer information, while their roles in prostate cancer (PCa) remain poorly understood. Here, we systematically analyzed the transcriptome, single nucleotide polymorphisms (SNPs) and clinical data of 495 PCa patients from the TCGA database and verified them in 220 GEO patients, and qPCR was used to validate the expression of the model genes in our patients. First, we found that integrin αvβ3/α6β1 was negatively correlated with most immune cell infiltration and immune functions and closely associated with poor survival in TCGA patients. Then, we divided these patients into two groups according to the expression level of αvβ3/α6β1, intersected differentially expressed genes of the two groups with the GEO dataset and identified eight biochemical recurrence-related genes (BRGs), and these genes were verified by qPCR in our patients. Next, these BRGs were used to construct a prognostic risk model by applying LASSO Cox regression. We found that the high-risk (HR) group showed poorer OS, PFS, biochemical recurrence and clinical characteristics than the low-risk (LR) group. In addition, the HR group was mainly enriched in the cell cycle pathway and had a higher TP53 mutation rate than the LR group. More importantly, lower immune cell infiltration and immune function, higher expression of PD-L1, PD-1, and CTLA4, and higher immune exclusion scores were identified in the HR group, suggesting a higher possibility of immune escape. These findings suggested the key role of integrin αvβ3/α6β1 in predicting prognosis, TP53 mutation and immune escape in PCa.

AI-based pathologic biomarker for pathologic downstaging in patients with muscle-invasive bladder cancer undergoing cystectomy after neoadjuvant nivolumab, gemcitabine, and cisplatin: BLAAST-1 Trial.

e16566 Background: The BLASST-1 study is a multi-center phase II trial evaluating the combination of neoadjuvant nivolumab with gemcitabine-cisplatin (N+GC) for muscle-invasive bladder cancer (MIBC) patients undergoing radical cystectomy (RC). The primary endpoint was pathologic down staging (PaR; ≤pT1N0). We previously reported a PaR rate of 65.8% (Gupta S et al. ASCO GU 2020). Given the lack of validated and optimal biomarkers to predict PaR, we studied the association of an AI-based pathologic biomarker measuring pre-treatment morphological features with PaR. Methods: Forty-one patients with MIBC (cT2-T4a, N≤1, M0) and candidates for RC were enrolled between Feb 2018 and June 2019 (cT2N0 90%, cT3N0 7%, cT4N1 3%). Thirty-six patients had transurethral resection of bladder cancer (TURBT) with pre-treatment diagnostic specimens available for analysis. Patients received four cycles of N+GC followed by RC within 8 weeks. A board-certified pathologist selected diagnostic regions from TURBT-derived representative H&amp;E diagnostic slides for each patient. To compute the pathological biomarker, a proprietary deep-learning algorithm (Valar Labs, Palo Alto, CA) first segmented nuclei from digital whole-slide images of the H&amp;E specimens to extract quantitative histological features. The Valar pathologic biomarker was then computed from features associated with immune infiltration and morphological characteristics of neoplastic cells. The Valar biomarker was split based on unsupervised clustering into two groups: Valar-High was associated with PaR and Valar-Low with no PaR. PD-L1 cutoff of 1% was used for dichotomization. To compare the Valar biomarker with PD-L1, a subcohort of 33 patients with available PD-L1 scores were analyzed. T-tests and diagnostic performance metrics were used to distinguish between PaR response rates in each cluster group across tests. Results: Patients designated Valar-High (n=23) had higher PaR compared to no PaR in the Valar-Low (n=13) group (PAR 78.3% vs 38.4%, respectively, p&lt;0.008). Compared to PD-L1, the Valar biomarker had higher sensitivity (85.7 vs 57.1%), specificity (66.6 vs 66%), positive predictive value (81.8 vs 75.0%), negative predictive value (72.7 vs 47.0%), and accuracy (78.8 vs 60.6%) across both cohorts. The Valar biomarker was not correlated to PD-L1 (r^2=0.029) or sex (22 male, 14 female). The combined Valar-High or PD-L1 High test had a high sensitivity for PaR (95.2%). Conclusions: In this hypothesis-generating study, pre-treatment morphological features measured by the AI-based Valar pathologic biomarker identified responders to neoadjuvant N+GC in MIBC. Further prospective studies are needed to study the prognostic vs. predictive utility of this biomarker.

Cemiplimab for kidney organ transplant recipients with advanced cutaneous squamous cell carcinoma: CONTRAC-1.

9519 Background: Solid organ transplant recipients are often excluded from immunotherapy trials given the risk of allograft rejection and loss. We report the results of the first prospective study using the PD-1 inhibitor Cemiplimab (Cemi) for kidney transplant recipients (KTR) with advanced, incurable cutaneous squamous cell carcinoma (cSCC), adopting a standardized approach to immunosuppression (IS) with mTOR inhibition and dynamic prednisone (NCT04339062). Methods: This single-arm, open-label prospective clinical trial enrolled KTRs (eGFR ≥30 mL/min without proteinuria) with advanced cSCC, ECOG ≤2, measurable disease (RECIST v1.1), with no prior immunotherapy exposure. KTRs received mTOR inhibition (target trough 4-6 ng/mL) with a prednisone taper each cycle (40 mg on day -1 to 3, 20 mg days 4-6, 10 mg days 7-20) along with Cemi 350 mg IV every 21-days. Primary endpoint: rate of rejection (futility defined as ≥2/3 or 4/6 KTRs with rejection events). Secondary endpoints: overall response rate (ORR), duration of response, progression-free survival (PFS), overall survival (OS), infection rates. Exploratory: baseline tumor PD-L1 score, molecular and immunologic predictors of response. Results: From 11/2020 to 1/2023, 10 KTRs (median years from transplant: 8, range: 3-31) enrolled including 8 (80%) men, median age 64 (range: 43-86), median eGFR 48 (range: 32-60) often with head and neck primaries (9, 90%) and distant metastases (7, 70%). Six (60%) had prior systemic therapy. For mTOR inhibition, 7 (70%) received sirolimus and 3 (30%) everolimus. At a median follow-up of 6.3 months (range: &lt; 1-24.9), no patients experienced kidney allograft rejection or loss. Of 8 evaluable patients, ORR was 50% (2 CR, 2 PRs), while 4 had PD. At data cutoff no responder had progressed, with 2/4 in response &gt; 18 months (range: &lt; 1-22.7+). One patient is pending first restaging; 1 was unevaluable (died before first restaging). One KTR with initial PD experienced a subsequent durable response to cetuximab. Fatigue (40%) and limb edema (30%) were the most common treatment-related adverse events (TRAEs). Grade 3+ TRAEs occurred in 5 (50%) patients including diarrhea, infections (n = 3), and electrolyte derangements; there were no Cemi-related deaths. Median PFS was 7.9 mos (95%CI: 1.2-not reached [NR]); the 3-month OS estimate was 61% (95%CI: 27-83). Baseline tumor PD-L1 scores ranged from 0-5%; median TMB was 49 muts/Mb (range: 10-97). Tumor mutations in TP53, CDKN2A, and NOTCH1 were common. Exploratory tumor/circulating multiparametric immune profiling and circulating tumor (ct)DNA findings will be presented. Conclusions: Using IS with mTOR inhibition and dynamic prednisone resulted in no kidney allograft rejection among KTRs treated with Cemi for advanced cSCC. Durable anti-tumor efficacy was observed. mTOR inhibition with prednisone should be the preferred IS regimen when treating KTRs with anti-PD-1 therapy. Clinical trial information: NCT04339062 .

Comparing PD-L1 scores for predicting pembrolizumab efficacy in first-line treatment of unresectable cutaneous squamous cell carcinomas.

e21593 Background: In the CARSKIN study (NCT02883556), 1L pembrolizumab demonstrated promising activity with durable responses and manageable safety in patients (pts) with unresectable locally advanced or metastatic CSCC. Here we report best ORR and survival endpoints of the entire cohort by PD-L1 status. Methods: The imaging assessment per RECIST v1.1 was blinded with independent central review. Best ORR, PFS, DOR OS, in the overall sample and by PD-L1 status was centrally assessed by 2 blinded independent pathologists, one using the anti–PD-L1 E1L3N clone (TPSE1L3N), the other using the 22C3 antibody (TPS22C3 and CPS22C3). Among 57 pts, 6 were excluded from the analysis: 2 untreated pts, 1 early unrelated death and 3 pts tested only with 1 assay). ROC curve analysis was used to compare assays. Results: With a median follow-up of 26 mo, best ORR was 51% with 14 PR (27%) and 12 CR (24%). Relapse occurred in 3 PR and 1 PD-L1 – CR pts. Respective median PFS, DOR, and OS were 13.7, NR, and 25.3 mo; respective 1-year PFS, DOR and OS were 50% [38-66], 80% [65-97] and 73% [61-86]. With a cutoff of 1%, best ORR was higher either for TPSE1L3N + pts vs TPSE1L3N – pts (60% vs 18%, p = 0.02) and for CPS22C3 + pts (59% vs 20%, p = 0.038), but not TPS22C3 + pts (p = 0.76). The optimal cutoff of CPS22C3 for ORR using a ROC curve was estimated to be ≥ 7% with a sensitivity of 0.70 and a specificity of 0.75. For best ORR, area under the ROC curve was higher for CPS22C3 (0.72) than TPSE1L3N (0.62) but the difference was not significant (P = 0.21). Pts with TPSE1L3N ≥ 1% had a higher 1-y PFS (57% vs 27%, p = 0.004) and OS (82% vs 42%, P = 0.01). Pts with CPS22C3 ≥ 1%, had a better 1-y OS (79% vs 50%, p = 0.03), but not 1y-PFS. Conclusions: CPS22C3 ≥ 7% appeared equivalent to TPSE1L3N ≥ 1% for predicting ORR and OS in pts receiving pembrolizumab in 1L treatment of advanced or metastatic CSCCs. Replications are warranted for validation. Pts with low scores might beneficiate from combined treatments. Clinical trial information: NCT02883556 .

Artificial intelligence algorithm developed to predict immune checkpoint inhibitors efficacy in non–small-cell lung cancer.

9132 Background: Many non-small cell lung cancer (NSCLC) patients with high PD-L1 IHC expression and/or high TMB level do not respond to immune-check point inhibitors (ICIs). Discovery of novel predictive biomarkers for ICI response in NSCLC continues to be a critical unmet medical need, given the expense, potential toxicity, and therapeutic delay due to ICI treatment without benefit. Methods: We developed a supervised deep learning algorithm (Deep-IO) to predict therapeutic response to ICIs in NSCLC patients from standard histology whole slide hematoxylin and eosin (H&amp;E) images, which was trained based on ICI objective response rate. We utilized a convolutional neural network (CNN) implemented in Pytorch. The efficiency of the Deep-IO was evaluated through classification performance metrics and the area under the receiver operating characteristic curve (AUROC). The algorithm was trained and tested on 85218 tiles (size: 512x512 pixels) in 446 advanced stage NSCLC patients from the Dana Farber Cancer Institute (DFCI), all treated with ICI monotherapy. Results: The objective response rate, comprising complete and partial response, was 26% (n = 114) in the overall cohort. The DFCI patient cohort was randomly split into a training set (N=379, 85%) and a test set (N=67, 15%). The classifier’s predicted class (responder vs non-responder) probabilities at the tiles level were averaged for each patient. Quantitative evaluation of whole section test dataset shows that the developed model achieves the accuracy of 0.72, weighted average (WAVG) precision of 0.77, WAVG recall of 0.72 and WAVG F-score of 0.74 on classifying ICI non-responders from responders. Comparing with PD-L1 expression (TPS%) and TMB (mu/Mb), Deep-IO had superior predictive power for ICI response (AUROC=0.75) in the test set. The combined Deep-IO+PD-L1 scores resulted in a significantly greater AUROC value of 0.82 compared to single tests and a 32% improvement in specificity (0.88 vs. 0.56) compared to PD-L1 evaluation alone (Table). Conclusions: This proof-of-concept study shows that an artificial intelligence–powered H&amp;E image classifier can predict ICI effectiveness in NSCLC. Deep-IO outperformed both established predictive biomarkers. We will assess Deep-IO in additional external NSCLC data sets for validation and confirmation. [Table: see text]

Changes in PD-L1 and tumor mutational burden between paired samples and relationship to immune checkpoint inhibitor outcomes in non-small cell lung cancer.

9127 Background: Programmed cell death receptor ligand 1 (PD-L1) tumor proportion score (TPS) and tumor mutational burden (TMB) are key biomarkers of response to immune checkpoint inhibitors (ICI) in non-small cell lung cancer (NSCLC). How often PD-L1 and TMB values change across samples within individual patients (pts) and the impact of this variation on ICI outcomes are limited. Methods: Pts with NSCLC and multiple PD-L1 or TMB assessments at the Dana-Farber Cancer Institute were included. Clinicopathologic and genomic data were analyzed in association with PD-L1 and TMB changes. In patients with &gt; 2 samples, each sample was paired with the subsequent and each pair was analyzed independently. Minor and major changes in PD-L1 were defined as absolute changes of 10-49% and 50-100%, respectively. Results: The PD-L1 cohort included 402 sample pairs (median ΔPD-L1: 0; range: -90,+95%) and the TMB cohort included 413 sample pairs (median ΔTMB: 0 mut/Mb; range: -20.6,+24.5). Concordance between pairs was high for both PD-L1 ( R= 0.53, P &lt; 0.01) and TMB ( R= 0.80, P &lt; 0.01). Samples taken within &lt; 1 year had higher concordance for PD-L1 than samples taken ≥1 year apart (P &lt; 0.01), but length of time between biopsies did not impact TMB concordance. Minor or no changes in PD-L1 were observed in 82.3% of pairs; the frequency of major increases was 9.7% and major decreases was 8%. After dividing TMB into tertiles, 72.6% sample pairs showed no tertile change, 16% increased to a higher tertile, and 11.4% decreased to a lower tertile. PD-L1, but not TMB, decreased with intervening ICI (P = 0.02), but PD-L1 and TMB did not change significantly with other intervening systemic therapies. Acquired copy number losses of the CD274 gene (encoding for PD-L1) were enriched in pairs with a major PD-L1 decrease (Q &lt; 0.05). Similarly, acquired loss of chromosome 9p (which contains the CD274 locus) correlated with a reduction in PD-L1, while 9p gains correlated with an increase in PD-L1 (P = 0.02). A total of 142 pts had multiple PD-L1 scores assessed before ICI initiation. Among pts with at least one PD-L1-positive sample (TPS ≥1%), the presence of another sample that was PD-L1 negative ( &lt; 1%) correlated with worse outcomes to ICI compared to pts with no PD-L1-negative samples: response rate was 11% vs 39% (P &lt; 0.01), median progression-free survival (mPFS) 2.7 vs 8.0 months (P &lt; 0.01), and median overall survival 13.6 vs 21.2 months (P = 0.05), respectively. Among pts with at least one PD-L1 TPS ≥1% and one &lt; 1%, cases where the last PD-L1 TPS before ICI was ≥1% achieved longer mPFS than if the pre-ICI sample was PD-L1 negative (3.6 vs 1.8 months, P = 0.03). Conclusions: Although PD-L1 and TMB generally had high intrapatient concordance across samples, significant changes in values occurred in some patients. Variations in PD-L1 influenced ICI outcomes, warranting an assessment proximal to ICI initiation whenever feasible.

Abstract 5706: ImmunoPROFILE: A prospective implementation of clinically validated, quantitative immune cell profiling test identifies tumor-infiltrating CD8+ and PD-1+ cell densities as prognostic biomarkers across a 2,023 patient pan-cancer cohort treated with different therapies

Abstract Tumor-infiltrating lymphocyte (TIL) density has been identified as a prognostic and predictive biomarker in select tumors treated with defined therapies. These observations suggest that TILs may be general markers of patient outcomes, but evidence in support of this hypothesis has been limited by small cohorts. We validated ImmunoPROFILE, a multiplexed immunofluorescence (MIF)-based assay coupled with machine-learning-based image analysis, to identify and quantify tumor cells (cytokeratin, PAX5, PAX8, SOX10), T cells (CD8), T-regulatory cells (FOXP3), exhausted cells (PD-1) and immunosuppressive tumor and immune cells (PD-L1). We applied the MIF panel to specimens from patients collected prospectively over three years and analyzed 2,023 cases across 27 tumor types. The association between biomarkers and overall survival (OS) was investigated using Cox models controlling for patient risk factors such as cancer type, metastatic vs. primary disease, age, and gender. Multivariable biomarker selection was based on likelihood ratios. The assay was highly robust (success rate 97%), reproducible (inter-scanning and intra-staining density controls within 1 SD, inter-staining PD-L1 scores ≤11% CV), and operator-independent (R2 &amp;gt;0.7 to &amp;gt;0.9 for each biomarker and 95% concordance in PD-L1 score-based interpretation between technicians). From whole slide images, a total of 11,932 individual regions of interest were analyzed across the cohort, resulting in &amp;gt;50 million spatially-resolved single cells which were summarized into cell population densities and PD-L1 scores. High densities of CD8+ (&amp;gt;64/mm2, p&amp;lt;0.0001), PD-1+ (&amp;gt;50/mm2, p&amp;lt;0.0001), and FOXP3+ (&amp;gt;30/mm2, p&amp;lt;0.0001) T cells were associated with longer overall survival (OS) irrespective of therapy and across all cancer types. PD-L1 metrics were not associated with OS (p=0.43). Compared to patients with low densities of CD8+ and PD-1+ cells, high densities of at least one of these cell types had better OS (Both high, HR: 0.49, 95% CI: 0.41 - 0.59; CD8+ high, HR: 0.63, (0.48 - 0.82); PD-1+ high, HR: 0.71, (0.54 - 0.93)). The results were consistent in the subset of patients (N=1572) who did not receive immunotherapy (IO). In patients who received IO therapy (N=451), only PD-1+ T-cell density associated with OS (HR: 0.48, (0.36 - 0.65)). To our knowledge, this is the first enterprise-level immune biomarker assay using multiplexed staining, digital imaging, and machine learning to be applied in a prospective manner to clinical specimens at scale. We found that select immune cell densities are prognostic across cancer types and therapies and demonstrated that quantification of multiple cell populations yields better prognostic power than single marker analyses. Citation Format: James Lindsay, Bijaya Sharma, Kristen D. Felt, Anita Giobbie-Hurder, Ian Dryg, Jason L. Weirather, Jennifer Altreuter, Tali Mazor, Priti Kumari, Joao V. Alessi, Ajit J. Nirmal, Michael P. Manos, Ananth R. Kumar, William Lotter, Ethan Cerami, Burce E. Johnson, Neil I. Lindeman, Lynette M. Sholl, Jonathan A. Nowak, Scott J. Rodig. ImmunoPROFILE: A prospective implementation of clinically validated, quantitative immune cell profiling test identifies tumor-infiltrating CD8+ and PD-1+ cell densities as prognostic biomarkers across a 2,023 patient pan-cancer cohort treated with different therapies. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5706.

CD169+ sinus macrophages in regional lymph nodes do not predict mismatch-repair status of patients with colorectal cancer.

Mismatch-repair deficiency and microsatellite instability-high (dMMR/MSI-H) colorectal cancer (CRC) is treated with programmed death (PD)-1 antibody regardless of PD-ligand (L)1 expression in tumor cells. We previously found that abundant CD169+ macrophages in regional lymph node (RLN) sinuses and CD8+ tumor-infiltrating lymphocytes (TILs) positively correlated in CRC and were associated with a favorable prognosis. However, associations between dMMR/MSI-H CRC and CD8+ TILs or prognoses vary among studies. In this study, we attempted to compare the association between MMR status, CD169+ macrophages in RLNs, CD8+ TILs, PD-L1 scores, and prognoses in CRC. We immunostained 83 surgically resected CRC tumors that we previously analyzed for MMR proteins, and identified 9 that were dMMR. The number of CD169+ macrophages in RLNs and CD8+ TILs significantly correlated with overall survival, whereas MMR status did not. The number of cells positive for the TIL markers CD3, CD4, CD8, and TIA-1, and macrophage markers CD68 and CD169 in RLNs did not significantly differ between groups according to MMR status. Furthermore, combined positive scores (CPS) for PD-L1 expression in five of nine dMMR CRCs were all <1. We found that dMMR in CRC did not correlate with numbers of CD169+ macrophages in RLNs or CD8+ TILs. CRC with CD169+ macrophages in RLNs and abundant CD8+ TILs indicates a better prognosis and it should be immunologically classified as a different antitumor group from dMMR CRC.

A Comprehensive Biomarker Analysis of Microsatellite Unstable/Mismatch Repair Deficient Colorectal Cancer Cohort Treated with Immunotherapy

The search for immunotherapy biomarkers in Microsatellite Instability High/Deficient Mismatch Repair system (MSI-H/dMMR) metastatic colorectal cancer (mCRC) is an unmet need. Sixteen patients with mCRC and MSI-H/dMMR (determined by either immunohistochemistry or polymerase chain reaction) treated with PD-1/PD-L1 inhibitors at our institution were included. According to whether the progression-free survival with PD-1/PD-L1 inhibitors was longer than 6 months or shorter, patients were clustered into the IT-responder group (n: 9 patients) or IT-resistant group (n: 7 patients), respectively. In order to evaluate determinants of benefit with PD-1/PD-L1 inhibitors, we performed multimodal analysis including genomics (through NGS panel tumour-only with 431 genes) and the immune microenvironment (using CD3, CD8, FOXP3 and PD-L1 antibodies). The following mutations were more frequent in IT-resistant compared with IT-responder groups: B2M (4/7 versus 2/9), CTNNB1 (2/7 versus 0/9), and biallelic PTEN (3/7 versus 1/9). Biallelic ARID1A mutations were found exclusively in the IT-responder group (4/9 patients). Tumour mutational burden did not correlate with immunotherapy benefit, neither the rate of indels in homopolymeric regions. Of note, biallelic ARID1A mutated tumours had the highest immune infiltration and PD-L1 scores, contrary to tumours with CTNNB1 mutation. Immune microenvironment analysis showed higher densities of different T cell subpopulations and PD-L1 expression in IT-responders. Misdiagnosis of MSI-H/dMMR inferred by discordances between immunohistochemistry and polymerase chain reaction was only found in the IT-resistant population (3/7 patients). Biallelic ARID1A mutations and Wnt signalling activation through CTNNB1 mutation were associated with high and low T cell immune infiltrates, respectively, and deserve special attention as determinants of response to PD-1/PD-L1 inhibitors. The non-MSI-H phenotype in dMMR is associated with poor benefit to immunotherapy. Our results suggest that mechanisms of resistance to immunotherapy are multi-factorial.

Hypoxia-associated prognostic markers and competing endogenous RNA coexpression networks in lung adenocarcinoma

Lung adenocarcinoma (LUAD) is the most common form of non-small cell lung cancer (NSCLC). Hypoxia has been found in 50-60% of locally advanced solid tumors and is associated with poor prognosis in various tumors, including NSCLC. This study focused on hypoxia-associated molecular hallmarks in LUAD. Fifteen hypoxia-related genes were selected to define the hypoxia status of LUAD by ConsensusClusterPlus based on data from The Cancer Genome Atlas (TCGA). Then, we investigated the immune status under different hypoxia statuses. Subsequently, we constructed prognostic models based on hypoxia-related differentially expressed genes (DEGs), identified hypoxia-related microRNAs, lncRNAs and mRNAs, and built a network based on the competing endogenous RNA (ceRNA) theory. Two clusters (Cluster 1 and Cluster 2) were identified with different hypoxia statuses. Cluster 1 was defined as the hypoxia subgroup, in which all 15 hypoxia-associated genes were upregulated. The infiltration of CD4+ T cells and Tfh cells was lower, while the infiltration of regulatory T (Treg) cells, the expression of PD-1/PD-L1 and TMB scores were higher in Cluster 1, indicating an immunosuppressive status. Based on the DEGs, a risk signature containing 7 genes was established. Furthermore, three differentially expressed microRNAs (hsa-miR-9, hsa-miR-31, hsa-miR-196b) associated with prognosis under different hypoxia clusters and their related mRNAs and lncRNAs were identified, and a ceRNA network was built. This study showed that hypoxia was associated with poor prognosis in LUAD and explored the potential mechanism from the perspective of the gene signature and ceRNA theory.

Prognosis analysis of necroptosis-related genes in colorectal cancer based on bioinformatic analysis

Background: Colorectal cancer (CRC) is one gastrointestinal malignancy, accounting for 10% of cancer diagnoses and cancer-related deaths worldwide each year. Therefore, it is urgent to identify genes involved in CRC predicting the prognosis. Methods: CRC’s data were acquired from the Gene Expression Omnibus (GEO) database (GSE39582 and GSE41258 datasets) and The Cancer Genome Atlas (TCGA) database. The differentially expressed necroptosis-related genes (DENRGs) were sorted out between tumor and normal tissues. Univariate Cox regression analysis and least absolute shrinkage and selectionator operator (LASSO) analysis were applied to selected DENRGs concerning patients’ overall survival and to construct a prognostic biomarker. The effectiveness of this biomarker was assessed by the Kaplan–Meier curve and the receiver operating characteristic (ROC) analysis. The GSE39582 dataset was utilized as external validation for the prognostic signature. Moreover, using univariate and multivariate Cox regression analyses, independent prognostic factors were identified to construct a prognostic nomogram. Next, signaling pathways regulated by the signature were explored through the gene set enrichment analysis (GSEA). The single sample gene set enrichment analysis (ssGSEA) algorithm and tumor immune dysfunction and exclusion (TIDE) were used to explore immune correlation in the two groups, high-risk and low-risk ones. Finally, prognostic genes’ expression was examined in the GSE41258 dataset. Results: In total, 27 DENRGs were filtered, and a necroptosis-related prognostic signature based on 6 DENRGs was constructed, which may better understand the overall survival (OS) of CRC. The Kaplan–Meier curve manifested the effectiveness of the prognostic signature, and the ROC curve showed the same result. In addition, univariate and multivariate Cox regression analyses revealed that age, pathology T, and risk score were independent prognostic factors, and a nomogram was established. Furthermore, the prognostic signature was most significantly associated with the apoptosis pathway. Meanwhile, 24 immune cells represented significant differences between two groups, like the activated B cell. Furthermore, 32 immune checkpoints, TIDE scores, PD-L1 scores, and T-cell exclusion scores were significantly different between the two groups. Finally, a 6-gene prognostic signature represented different expression levels between tumor and normal samples significantly in the GSE41258 dataset. Conclusion: Our study established a signature including 6 genes and a prognostic nomogram that could significantly assess the prognosis of patients with CRC.

Abstract CT112: AI-powered and manual assessment of PD-L1 are comparable in predicting response to neoadjuvant atezolizumab in patients (pts) with resectable non-squamous, non-small cell lung cancer (NSCLC)

Abstract Background: PD-L1 expression evaluated by immunohistochemistry (IHC) is a well-established predictor of anti-PD-L1/PD-1 cancer immunotherapy (CIT). The Phase II LCMC3 (NCT02927301) study evaluated pre-operative treatment (tx) with atezolizumab (anti-PD-L1) in pts with untreated early stage resectable NSCLC, achieving a 20% major pathologic response (MPR) rate (primary efficacy pts, n=143). A digital PD-L1 scoring method was developed to assess PD-L1 expression as a potential predictive marker for MPR in squamous and non-squamous tumor samples from LCMC3. Methods: Manual scoring was used to determine PD-L1 status on pre-tx biopsy samples using the tumor proportion score (TPS) with a positive threshold of TPS≥50 (22C3). Binary results were correlated with MPR and stratified by squamous/non-squamous histology. A digital pathology workflow for automated PD-L1 scoring was developed to yield a precise continuous PD-L1 TPS. Deep convolutional neural networks trained using pathologist annotations were used to detect individual cells within the tumor and tumor microenvironment and quantify their PD-L1 expression. These cell type predictions were used to compute a digital PD-L1 TPS. LCMC3 pts with available digital and manual PD-L1 scores were then used to assess the role of PD-L1 expression in predicting MPR. Results: PD-L1 scores were available for pre-tx biopsies from 108 pts. No significant difference in scores was seen between histological subtypes. At cutoff (Oct 15, 2021), TPS≥50 was seen in 41 (non-squamous, n=26 [39%]; squamous, n=15 [36%]) of 108 pts and was associated with MPR in non-squamous (odds ratio [OR], 28.6; P&amp;lt;0.001; Fisher’s exact test) but not squamous histology (OR, 1.3; P=1.0). Continuous digital PD-L1 scores (range: 0-100) were highly correlated with local manual PD-L1 scores (range: 0-100) for squamous (n=42, Pearson r=0.90, P&amp;lt;0.001) and non-squamous stained histology slides (n=66, Pearson r=0.90, P&amp;lt;0.001). Continuous digital and manual PD-L1 TPS on pre-tx biopsies (n=108) were predictive of MPR (digital: area under the receiver operating curve (AUROC)=0.678, logistic regression [LR] P=0.01; manual: AUROC=0.675, LR P=0.003). Strikingly, when pts were stratified by histology, PD-L1 scores were predictive of MPR from pre-tx biopsies for non-squamous samples (n=66; digital: AUROC=0.821, LR P=0.002; manual: AUROC=0.819, LR P=0.001) but not for squamous samples (n=42; digital: AUROC=0.519, LR P=0.93; manual: AUROC=0.506, LR P=0.90), despite no significant difference in MPR rate between the 2 groups. Conclusions: These findings support using digitally assessed PD-L1 IHC as a centralized and standardized scoring system and suggest that tumor histological subtype could be an important factor in the utility of PD-L1 as a predictive biomarker for neoadjuvant CIT in early stage NSCLC. Citation Format: John Abel, Christopher Rivard, Filip Kos, Guillaume Chhor, Yi Liu, Jennifer Giltnane, Sara Hoffman, Murray Resnick, Cyrus Hedvat, Amaro Taylor-Weiner, Farah Khalil, Alan Nicholas, Gregory A. Fishbein, Lynette M. Sholl, Natasha Rekhtman, Stephanie Hennek, Ilan Wapinski, Ann Johnson, Michael Montalto, Katja Schulze, Bruce E. Johnson, David P. Carbone, Konstantin Shilo, Andrew H. Beck, Sanja Dacic, William D. Travis, Ignacio Wistuba. AI-powered and manual assessment of PD-L1 are comparable in predicting response to neoadjuvant atezolizumab in patients (pts) with resectable non-squamous, non-small cell lung cancer (NSCLC) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr CT112.