Abstract only
548
Background: Real-world data (RWD) linking clinical outcomes with comprehensive genomic profiling (CGP) may enable identification of biomarkers to guide treatment selection and ...stratification in future trials. The primary objective was to characterize patients with metastatic urothelial carcinoma (mUC) included in a clinic-genomic database (CGDB), comprised of the electronic health record-derived Flatiron Health database with linked FoundationOne CGP results. As secondary objective, a novel Bladder Immune Prognostic Index (BIPI) was developed. Methods: A retrospective exploratory analysis was performed of de-identified RWD, retrieved from the CGDB. Data from mUC patients starting first-line single-agent immune checkpoint inhibitors (ICIs) and an unmatched group treated with front-line platinum-based chemotherapy (CHT) between Jan 1, 2011, and Sept 30, 2019, were analyzed and correlated with overall survival (OS). Known driver alterations, tumor mutational burden (TMB), and PD-L1 expression were described. A BIPI predicting outcome with ICIs was developed using a Cox-LASSO model and validated externally in a phase II trial (NCT02951767). Results: Of the 1021 patients with mUC identified in CGDB, 118 ICI-treated and 268 CHT-treated patients were included. Median follow-up duration was 9.4 and 14.5 months, respectively. Median OS was 5.4 months (95%CI, 3.3–9.2) with ICIs and 8.2 months (95%CI, 6.8–10.0) with CHT. In ICI-treated patients, low albumin and metastatic disease at initial presentation were associated with worse OS HR (95%CI) 2.15 (1.18–3.90), p =.012; 2.58 (1.30–5.10), p =.007, respectively whereas surgery for organ-confined disease and high TMB (≥10 mut/Mb) were associated with improved OS (HR (95%CI) 0.56 (0.36–0.88), p =.012; 0.58 0.35–0.95; p=.03), respectively. In CHT-treated patients, those with high APOBEC had worse OS (HR 1.43 95% CI, 1.06–1.94; p=.02). Neither PD-L1 (HR 0.96 0.37-2.46; p =.93), FGFR3 mutations (HR 0.98 0.65-1.47; p =.92) nor DNA damage-repair pathway alterations (HR 1.06 0.73-1.52; p =.77) were associated with OS. A novel BIPI for ICI-treated patients combining clinical and genomic variables (non-metastatic at initial diagnosis, normal albumin level, previous surgery for organ-confined disease, high TMB) was developed. Patients were categorized in 3 groups (low, intermediate, high risk) which correlated with OS. Median OS (95%CI) for low, intermediate and high risk was 11.7 (8.9−17.7), 4.1 (2.5–NE) and 2.4 months (1.0–4.0), (p <.001). Same results were observed in the validation cohort from an independent phase II immunotherapy trial in mUC (p <.001). Conclusions: This is the first time RWD including CGP were used to develop and validate a novel BIPI in mUC. This prognostic index may help patient selection in everyday practice and inform future trial design.
Abstract only
251
Background: Immune checkpoint inhibitors (CPIs) (anti-CTLA4, anti-PD-1/PD-L1 mAbs) have had limited monotherapy activity in prostate cancer (PC) compared to urothelial cancer (UC). ...Recent biomarker studies revealed molecular features associated w/ better efficacy/resistance to CPIs in UC.
1
Here we examined the tumor immune microenvironment (iTME) of hormone naïve/sensitive (HSPC) & castrate resistant prostate (CRPC) specimens from primary & metastatic sites, to determine if parallels to UC molecular correlates exist that will allow us to predict subsets of PC pts who are more likely to benefit from rational combination therapies w/ CPIs. Methods: Tumor FFPE archival tissues (HSPC, n = 98; CRPC, n = 46; Unk, n = 6) were sourced from a Royal Melbourne Hospital cohort of PC patients. Tissues were evaluated by H&E for TILs & IHC performed for PD-L1 (clones SP142 & SP263) & CD8. RNAseq was conducted & Lund bladder molecular subtyping performed during bioinformatics analyses. TCGA PRAD dataset was used for further validation. Results: We first applied Lund molecular subtyping to broadly categorize this cohort into luminal/basal subtypes. CRPC tumors were almost entirely luminal, whereas HSPC tumors were mainly basal. Luminal PC tumors enriched for Wnt gene sets (p = 0.026) & enriched for a genomically unstable (GU) subtype w/ higher expression of DNA damage repair (DDR) & cell cycle gene sets (p = 0.082). Compared to CRPC, HSPCs had higher numbers of infiltrating CD8s (p = 4.37e-004), increased expression of T-effector, MHC-I antigen presentation machinery, immune checkpoints, macrophages & activated stromal biology, including the fibroblast TGFβ response signature. PD-L1 on immune cells by SP142 & SP263 was 30% & 17% respectively & was enriched in HSPC samples (SP142: CRPC 18% vs. HSPC 34%, p = 0.08). Within patients w/ CRPC, differences were minimal between primary & metastatic samples. Conclusions: These analyses on the iTME contexture of PC reveal potentially actionable biological nodes for targeted therapies. These findings could inform future clinical strategies to improve CPI responses in CRPC & HSPC w/ rationale combinations.1. Mariathasan et al. Nature 2018.
Abstract only
5011
Background: Tumor mutational burden (TMB), PD-L1 expression, T-effector gene expression (GE) and a fibroblast TGF-β–response signature (F-TBRS) are associated with clinical ...outcomes with atezo mono in mUC (Mariathasan, Nature, 2018). Here we explore the potential predictive role of these biomarkers and APOBEC mutagenesis in IMvigor130. Methods: Pts receiving first-line (1L) mUC treatment (tx) were randomized 1:1:1 to atezo + PBC, atezo mono, or placebo + PBC. Coprimary efficacy endpoints were PFS and OS. Planned exploratory biomarker analyses included PD-L1 expression, TMB (FoundationOne), and T-effector GE (RNA-seq). Results: The 851 biomarker-evaluable pts (BEP) were representative of the 1200 ITT pts. Biomarker results are shown in Table. PD-L1 IC2/3 was associated with significantly longer OS for atezo mono vs placebo + PBC and a combination of PD-L1 IC2/3, and high TMB (> 10 muts/Mb) identified a pt subset (≈ 14% of BEP) with particularly favorable outcomes with atezo mono vs placebo + PBC; similar results for PD-L1 and TMB were not seen with atezo + PBC vs placebo + PBC. APOBEC mutagenesis was associated with improved OS with atezo-containing regimens whereas high F-TBRS was associated with inferior OS with atezo mono. Conclusions: These results reinforce the potential predictive nature of biomarkers associated with response/resistance to atezo and highlight potentially distinct biology driving benefit with atezo and atezo + PBC. These findings suggest a possible biomarker-directed approach to 1L mUC tx that warrants mechanistic interrogation and prospective validation. Clinical trial information: NCT02807636 . Table: see text
Abstract only
5565
Background: mCRPC patients (pts) tend to have a poor prognosis and limited treatment (tx) options, especially those with concomitant bone metastases (mets). We explored the ability ...of combination tx with atezo (anti–PD-L1) and r-223 (α-particle emitter) to stimulate anti-tumor immunity in mCRPC pts. Methods: This Phase Ib study evaluated the safety and tolerability of atezo + r-223 in pts with mCRPC and multiple bone mets, visceral mets and/or lymphadenopathy who progressed after androgen pathway inhibitor tx. The initial cohort phase evaluated the safety and tolerability of a concurrent dosing schedule (CDS), in which atezo and r-223 were administered on the same day. Following assessment of CDS, pts were randomized 1:1:1 to CDS or 1 of 2 staggered dosing schedules (atezo or r-223 introduced a full cycle before the other). This was followed by an expansion of enrollment (randomized 1:1:1). Pts got atezo 840 mg IV q2w and r-223 at 55 kBq/kg IV 6 times at 4-wk intervals until unacceptable toxicity or loss of clinical benefit. Exploratory measures of efficacy included investigator-assessed ORR (RECIST 1.1), PSA response rate, time to PSA progression, radiographic PFS (rPFS; PCWG2 criteria) and OS. Biopsy samples were collected at baseline and prior to cycle 2 to evaluate changes in the tumor microenvironment during tx. Results: As of Oct 4, 2019, 45 pts were enrolled and 44 had evaluable data. Baseline characteristics were generally similar across groups. All 44 evaluable pts had ≥ 1 all-cause AE; 23 (52.3%) had Gr 3-4 AE. Eight pts (18.2%) had Gr 5 AE as per protocol reporting of deaths; 4 (9.1%) were from disease progression. Median follow-up was 13.9 mo (range, 1.7–34.2). Confirmed ORR was 6.8% (95% CI: 1.43, 18.66). Confirmed PSA response rate was 4.5% and median time to PSA progression was 3.0 mo (95% CI: 2.8, 3.3). Median rPFS was 3.0 mo (95% CI: 2.8, 4.6) and median OS was 16.3 mo (95% CI: 10.9, 22.3). Changes in PD-L1 and CD8 IHC were consistent with the known mechanism of action of atezo, as were changes in alkaline phosphatase with radium. Conclusions: No dose-limiting toxicities, safety signals, or changes in serum biomarkers were observed beyond the known safety profiles of atezo and r-223. This Phase 1b study did not seem to show clinical benefit from combination tx. Ongoing subgroup and biomarker analyses may provide additional insights. Studies of PD-1/PD-L1 targeted therapies in combination with tumor-directed radiation in molecularly selected mCRPC pts are planned or underway. Clinical trial information: NCT02814669 .
Predictive biomarkers of treatment response are lacking for metastatic clearcell renal cell carcinoma (ccRCC), a tumor type that is treated with angiogenesis inhibitors, immune checkpoint inhibitors, ...mTOR inhibitors and a HIF2 inhibitor. The Angioscore, an RNA-based quantification of angiogenesis, is arguably the best candidate to predict anti-angiogenic (AA) response. However, the clinical adoption of transcriptomic assays faces several challenges including standardization, time delay, and high cost. Further, ccRCC tumors are highly heterogenous, and sampling multiple areas for sequencing is impractical.BackgroundPredictive biomarkers of treatment response are lacking for metastatic clearcell renal cell carcinoma (ccRCC), a tumor type that is treated with angiogenesis inhibitors, immune checkpoint inhibitors, mTOR inhibitors and a HIF2 inhibitor. The Angioscore, an RNA-based quantification of angiogenesis, is arguably the best candidate to predict anti-angiogenic (AA) response. However, the clinical adoption of transcriptomic assays faces several challenges including standardization, time delay, and high cost. Further, ccRCC tumors are highly heterogenous, and sampling multiple areas for sequencing is impractical.Here we present a novel deep learning (DL) approach to predict the Angioscore from ubiquitous histopathology slides. In order to overcome the lack of interpretability, one of the biggest limitations of typical DL models, our model produces a visual vascular network which is the basis of the model's prediction. To test its reliability, we applied this model to multiple cohorts including a clinical trial dataset.ApproachHere we present a novel deep learning (DL) approach to predict the Angioscore from ubiquitous histopathology slides. In order to overcome the lack of interpretability, one of the biggest limitations of typical DL models, our model produces a visual vascular network which is the basis of the model's prediction. To test its reliability, we applied this model to multiple cohorts including a clinical trial dataset.Our model accurately predicts the RNA-based Angioscore on multiple independent cohorts (spearman correlations of 0.77 and 0.73). Further, the predictions help unravel meaningful biology such as association of angiogenesis with grade, stage, and driver mutation status. Finally, we find our model is able to predict response to AA therapy, in both a real-world cohort and the IMmotion150 clinical trial. The predictive power of our model vastly exceeds that of CD31, a marker of vasculature, and nearly rivals the performance (c-index 0.66 vs 0.67) of the ground truth RNA-based Angioscore at a fraction of the cost.ResultsOur model accurately predicts the RNA-based Angioscore on multiple independent cohorts (spearman correlations of 0.77 and 0.73). Further, the predictions help unravel meaningful biology such as association of angiogenesis with grade, stage, and driver mutation status. Finally, we find our model is able to predict response to AA therapy, in both a real-world cohort and the IMmotion150 clinical trial. The predictive power of our model vastly exceeds that of CD31, a marker of vasculature, and nearly rivals the performance (c-index 0.66 vs 0.67) of the ground truth RNA-based Angioscore at a fraction of the cost.By providing a robust yet interpretable prediction of the Angioscore from histopathology slides alone, our approach offers insights into angiogenesis biology and AA treatment response.ConclusionBy providing a robust yet interpretable prediction of the Angioscore from histopathology slides alone, our approach offers insights into angiogenesis biology and AA treatment response.