Citations

Accurate clinical staging of bladder cancer aids in optimizing the process of clinical decision-making, thereby tailoring the effective treatment and management of patients. While several radiomics approaches have been developed to facilitate the process of clinical diagnosis and staging of bladder cancer using grayscale computed tomography (CT) scans, the performances of these models have been low, with little validation and no clear consensus on specific imaging signatures. We propose a hybrid framework comprising pre-trained deep neural networks for feature extraction, in combination with statistical machine learning techniques for classification, which is capable of performing the following classification tasks: (1) bladder cancer tissue vs. normal tissue, (2) muscle-invasive bladder cancer (MIBC) vs. non-muscle-invasive bladder cancer (NMIBC), and (3) post-treatment changes (PTC) vs. MIBC.

e16550 Background: Improved computational power and modern algorithms have generated significant interest in radiomics for cancer diagnosis and staging. Here we assess the performance of deep learning (DL) models as a means for feature extraction in combination with supervised machine learning (ML) algorithms for accurate staging and chemotherapy response assessment of bladder cancer. Methods: Deidentified grayscale CT images from bladder cancer patients scheduled to undergo radical cystectomy were included in this retrospective study. These images were manually annotated with two regional masks (normal region and cancer region). Five DL models- namely, AlexNet, GoogleNet, InceptionV3, ResNet-50, and XceptionNet pre-trained on the ImageNet dataset, a public dataset, were then fine-tuned on our bladder CT scan data to extract features. Through feature selection process, the subset of the features was used to build ML classifiers for classification. The classification was performed using five different ML classifiers, namely k-Nearest Neighbor (KNN), Naïve-Bayes (NB), Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), and Decision Tree (DT). The classification task was performed with 10-fold cross-validation, and each of the experiments contained a different but not mutually exclusive subset of samples. The evaluation metrics include accuracy, sensitivity, specificity, precision, and F1-score. Results: A total of 200 deidentified grayscale CT images of 100 patients with histologically proven bladder cancer, were included in this study. For experiment (1) normal vs. cancer, the LDA classifier on XceptionNet based features provides the best performance with an accuracy of 86.07%, sensitivity of 96.75%, specificity of 69.65%, precision of 83.07%, and F1-score of 89.39%. For experiment (2) non-muscle invasive Cancer (NMIBC) vs. muscle invasive bladder cancer (MIBC), the LDA classifier on XceptionNet based features provided the best performance with an accuracy of 79.72%, sensitivity of 66.62%, specificity of 87.39%, precision of 75.58%, and F1-score of 70.81%. For experiment (3) T0 lesion vs. MIBC, the LDA classifier on XceptionNet based features provides the best performance with an accuracy of 74.96%, sensitivity of 80.51%, specificity of 70.22%, precision of 69.78%, and F1-score of 74.73%. Conclusions: Our proposed model has shown good results in differentiating normal vs cancer and promising performance in differentiating T0 vs MIBC after chemotherapy treatment. We are expanding our dataset to further improve the performance in differentiating T0 vs MIBC. In addition, we will investigate the applicability of GAN for data augmentation to address data limit. We believe the hybrid DL and ML framework may facilitates radiologists' decisions and clinical decision-making in patients with bladder cancer.

518 Background: Cisplatin (C) based neoadjuvant chemotherapy (NAC) has been the mainstay treatment for muscle invasive bladder cancer (MIBC). However, the optimal choice of NAC is not well-established. We therefore conducted a network-meta-analysis (NMA) to assess comparative efficacy of different treatment options. Methods: Several electronic databases (MEDLINE, and EMBASE) and conference proceedings were systematically used to identify randomized trials evaluating first-line treatments in neoadjuvant MIBC. Outcomes of interest included overall survival (OS), progression-free survival (PFS), objective response rate (ORR) and pathologic complete response (pCR). DerSimonian-Laird random-effects meta-analysis was used to compute direct comparisons between NAC and local therapy (LT) only. Fixed effect NMA was conducted within the frequentist framework for mixed treatment comparisons. P-scores were used to assess relative treatment rankings. All statistical analyses were conducted in R (v 4.1.1). Results: This systematic review included 23 trials (25 references) with a total of 5313 patients with 10 unique treatments. Direct comparisons showed significant benefit in PFS (HR: 0.82; 95% CI: 0.71-0.94), and in OS (HR: 0.85; 95% CI: 0.78-0.92) with NAC when compared to LT only. However, the likelihood of achieving an objective response (RR: 1.18; 95% CI: 0.72-1.95: I2: 85%) or CR (RR: 1.51; 95% CI: 0.87-2.61; I2: 79%) was not different between the two arms. Data available in 15 trials contributed to the network for PFS outcome while only eight trials contributed to OS outcome. Mixed treatment comparisons showed significantly improved PFS with nintedanib-gemcitabine-cisplatin (NinGC; HR: 0.42; 95% CI: 0.20-0.87) and dose-dense-methotrexate-vinblastine-doxorubicin-cisplatin (ddMVAC; HR: 0.61; 95% CI: 0.43-0.88) compared to LT only. Consistent benefit was observed when NinGC (HR: 0.48; 95% CI: 0.24-0.97) and ddMVAC (HR: 0.70; 95% CI: 0.51-0.96) were compared to GC only. However, no significant differences were observed between NinGC (rank 1) and ddMVAC (rank 2) or among other mixed treatment comparisons. Similarly, OS was significantly improved with NinGC (rank 1) relative to MVAC (HR: 0.41; 95% CI: 0.17-0.97), MVC (HR: 0.37; 95% CI: 0.16-0.88), MC (HR: 0.38; 95% CI: 0.16-0.92), AC (HR: 0.36; 95% CI: 0.15-0.91), and GC (HR: 0.39; 95% CI: 0.17-0.90). While similar results were observed with ddMVAC (rank 2) when compared to MVC (HR: 0.63; 95% CI: 0.42-0.94), MC (HR: 0.64; 95% CI: 0.42-0.98), and GC (HR: 0.66; 95% CI: 0.47-0.92), no significant differences were observed with ddMVAC when compared to MVAC (HR: 0.62; 95% CI: 0.38-1.01) and NinGC (HR: 1.70; 95% CI: 0.69-4.19). Conclusions: Current evidence shows improved PFS and OS with the use of neoadjuvant ddMVAC and Nin-GC in patients with MIBC relative to other treatment options.

481 Background: EV is an anti-Nectin-4 antibody-drug conjugate that has been approved by the US FDA for mUC progressing on platinum and PD1/L1 inhibitor therapy and in cisplatin-ineligible pts following prior therapy. However, prognostic and predictive biomarkers in the setting of EV therapy are unclear. We aimed to investigate baseline and early changes in commonly available clinical and laboratory tests that could help predict survival and toxicities in patients with mUC treated with EV. Methods: Data from patients with mUC who were treated with EV at Dana Farber Cancer Institute between 2017 and 2021 were reviewed retrospectively. We assessed the association of demographic (age, sex), clinical (ECOG-performance status [PS], sites of metastasis) and laboratory variables (hemoglobin, RDW, neutrophil count, lymphocyte count, neutrophil-to-lymphocyte ratio, platelet count and serum albumin) at baseline as well as after 4-7 weeks of treatment with Overall Survival (OS) by constructing univariate and multivariable Cox proportional-hazards models. Any variables significantly associated with OS at the univariate level (using an alpha-level of 0.1 to determine statistical significance) were then entered into the multivariable model. Predictors were eliminated from the final multivariable models by a backward selection process, retaining only those with a p-value of ≤ 0.05. Results: A total of 49 patients who received EV were included in the analysis. The median age was 72 years (range 48-88) and 36 (73.5%) were male. The median OS was 13.3 months (range: 2.7-38.7) and grade 3-4 toxicities were seen in 11 pts (22.4%). In the final multivariable models, lower serum albumin at 4-7 weeks (HR = 0.18; 95% CI = 0.05-0.71; p = 0.015), and baseline ECOG-PS of 1 vs. 0 (HR = 2.56; 95% CI = 1.05-6.21; p = 0.038) were significantly associated with worse OS (Table). Patients with higher serum albumin levels at 4-7 weeks also had significantly lower odds of experiencing grade 3-4 adverse events on EV (OR = 0.05; 95% CI = 0.002-0.52; p = 0.02). Conclusions: Lower serum albumin appeared to predict worse survival and severe toxicities in mUC pts receiving EV. This information may guide therapy and assist in the prevention of toxicities. Additionally, ECOG-PS was prognostic while the location of metastatic disease was not prognostic in the setting of EV.[Table: see text]

6549 Background: Representation and outcomes of women, older adults, and racial minorities in ICI trials has not been previously described. Methods: MEDLINE and Embase were searched to identify ICI RCTs. Data for trial characteristics, proportion of trials reporting race, age and sex as well as the proportion of patients by race, age and sex enrolled in ICI trials was collected. Descriptive statistics were reported for trials reporting minority representation and proportion of included patients by race, age and sex. Disparities in representation were calculated using enrollment incidence disparity (EID) and enrollment incidence ratios (EIR) by comparing trial enrollment against U.S. population-based estimates acquired from the SEER 18 incidence database. The relationship of EID to key trial characteristics were compared using standard parametric and non-parametric statistical tests. Trends in EIR were analyzed using the Joinpoint Regression Analysis software. Results: 108 ICI trials from 2009 to 2020 with 48,360 patients were included in this analysis. All RCTs reported sex (101/101). 78 trials reported race (72%), of which only 41 trials (38%) reported data on all 5 U.S. racial categories (Black, White, Asian, Pacific Islander and Native American). Participation of Black patients was reported in 66 trials (61%), White participants in 78 trials (72%), Asians in 69 trials (64%), Native Americans and Pacific Islanders in 41 trials (38%), and Hispanics in 24 trials (22%). Age categories were inconsistently defined, and 80 trials (74%) reported the proportion of patient by age categories. Subgroup analyses of clinical outcomes by race, age and sex were reported in 17 (22%), 62 (79%) and 57 (73%) trials respectively. Women (trial proportion [TP]: 32%; EIR: 0.77), patients aged ≥ 65 years (TP: 42%; EIR: 0.74), Black participants (TP: 1.8%; EIR: 0.17) and Hispanic participants (TP: 5.9%; EIR: 0.67) were largely underrepresented, and Asians were overrepresented (TP: 15.9%; EIR: 2.64). Black patients were underrepresented across all cancer types. Similarly, women, older adults ( > 65 years of age) and Hispanic patients were consistently underrepresented across cancer types with few exceptions. Representation of older adults increased significantly from 2010-2020 (APC: 2.72), while representation of Black patients decreased significantly from 2009-2020 (APC: -23.37). Black patients were found to be significantly underrepresented in phase III trials (p = 0.0005), trials with OS as the primary endpoint (p = 0.004), and PD1 inhibitor trials (p = 0.002). Hispanics were significantly underrepresented in PD1 inhibitor trials (p = 0.003). Conclusions: There is both suboptimal reporting about participation and underrepresentation of women, racial minorities (particularly Black patients) and older adults in ICI trials as compared to their cancer incidence.

Objectives To evaluate outcomes of patients achieving a post‐treatment pathological stage of <ypT2N0 at radical cystectomy (RC) following neoadjuvant chemotherapy (NAC) for muscle‐invasive bladder cancer (MIBC) to identify an optimal definition of pathological response. Patients and Methods Patients from 10 international centres who underwent NAC for cT2–4aN0–1 MIBC and achieved <ypT2N0 disease at RC were included. The primary outcome was time to recurrence, either local or distant. Kaplan–Meier and Cox proportional hazards regression were used to evaluate associations between clinicopathological variables and outcomes. Results A total of 625 patients were included. The median age was 66 years and 80% were male. Gemcitabine and cisplatin (GC, 56%) and methotrexate, vinblastine, doxorubicin and cisplatin (MVAC)/dose‐dense (dd)MVAC (32%) were the most common NAC regimens. ypT0, pure ypTis, ypTa ±ypTis and ypT1 ± ypTis were attained in 58.1%, 20.0%, 7.6% and 14.2% of patients, respectively. The cumulative incidence of recurrence at 5 years was 9%, 16%, 29% and 30%, respectively. Pathological stage was prognostic for recurrence, with ypTa ± Tis (hazard ratio [HR] 3.20, 95% confidence interval [CI] 1.40–7.30) and ypT1 ± Tis disease (HR 4.03, 95% CI 2.13–7.63) associated with a significantly higher recurrence risk. Pure ypTis (HR 1.66, 95% CI 0.82–3.38) and the type of NAC regimen (ddMVAC: HR 1.59, 95% CI 0.55–4.56; MVAC: HR 1.18, 9%% CI 0.25–5.54; reference: GC) were not associated with recurrence. Conclusion We propose that optimal pathological response after NAC be defined as attainment of ypT0N0/ypTisN0 at RC. Patients with ypTaN0 or ypT1N0 disease (with or without Tis) at RC displayed a significantly higher risk of recurrence and may be candidates for trials investigating adjuvant therapy.

472 Background: Enfortumab Vedotin (EV), an antibody-drug conjugate that targets nectin-4, is approved for metastatic urothelial carcinoma (mUC) progressing post-platinum and PD1/L1 inhibitor therapy. Erdafitinib is approved in patients for post-platinum mUC with activating genomic alterations in FGFR2/3, but the activity of EV in this subset is unclear. We investigated the activity of EV in patients (pts) with mUC based on FGFR2/3 genotype to inform management. Methods: In this multi-center, retrospective analysis, we assessed the objective response rate (ORR) to EV in mUC pts with and without FGFR2/3 genomic alterations detected by targeted panel next-generation sequencing. Activating gene fusions and known hotspots mutations in the two genes were considered. Descriptive analysis of ORR and patient characteristics was performed. Fisher’s exact test and binomial test with two-tailed p-value were used. Results: 40 pts were available from 4 institutions. Most pts were male (31/40, 78%) and the median age at start of EV was 74.1 (range 49 – 90) years. Ten patients (25%) had upper tract urothelial carcinoma (UTUC), and 33 (82%) had baseline ECOG performance status of 0-1. 31 of 39 patients had received both platinum-based chemotherapy and PD1/L1 inhibitors. Seven patients had confirmed activating hotspot FGFR3 mutations (p.S249C or p.Y373C). One pt had FGFR2 high-level amplification (HA), and one had FGFR3 HA. Of 36 patients evaluable for ORR, 18 had partial response (PR), 12 had stable disease (SD) and 6 had progressive disease (PD). Patients with FGFR2/3 activating mutations exhibited an ORR that was not statistically different compared to patients without no mutations: 2/7 (29%; 90% CI: 5 – 66%) vs. 16/29 (55%; 90% CI: 38 – 71%) respectively ( p-value = 0.4). 3/7 patients with FGFR3 hotspot mutations received an FGFR2/3 inhibitor and none responded; one of them had a sequential response to EV. Conclusions: In this multi-center retrospective cohort, FGFR2/3 activating genomic alterations did not appear to compromise response to EV in mUC. Larger studies are required to confirm our findings and optimal sequencing of EV and erdafitinib in mUC pts with FGFR2/3 genomic alterations requires further assessment.

5043 Background: Pathologic complete response (pCR) after NAC for MIBC is strongly correlated with long-term overall survival. However, there are sparse data on the risk of recurrence based on depth of pathologic response (pT0, pTa, pTis, pT1), and the differential impact of clinicopathologic factors and NAC regimen on recurrence. Methods: Baseline data on all pts with cT2-4N0-1 MIBC receiving NAC and who achieved < ypT2N0 disease at radical cystectomy (RC) from 9 international centers were obtained. The key outcome was time to recurrence (TTR) – defined as the time to any recurrence in the urinary tract or regional/distant metastasis, with death (in the absence of recurrence) considered a competing risk. Cox regression analysis was used to analyze the impact of clinical factors on recurrence. Results: A total of 506 pts were available. Median age was 66 years (range 33-86) and 78% (n = 396) were male; median follow-up after RC was 2.6 years. The majority of patients had pure urothelial histology (n = 371, 73%), and baseline stage was cT2N0 (n = 368, 73%), cT3-4N0 (n = 95, 19%) and TanyN1 (n = 43, 9%). NAC regimens were gemcitabine-cisplatin (GC, n = 296, 59%), dose-dense methotrexate-vinblastine-doxorubicin-cisplatin (ddMVAC, n = 141, 28%), split-dose GC (n = 29, 6%), MVAC (n = 29, 6%) and non-cisplatin based regimens (n = 11, 2%). At RC, 304 patients (60%) had ypT0N0 disease, 32 (6%) had ypTaN0, 107 (21%) had ypTisN0 and 63 (13%) had ypT1N0. Overall, 43 patients (8%) recurred with a median TTR of 56 weeks (range 7-251); 5-year freedom from recurrence was 87% (95% CI 83-91). The majority (n = 38) recurred outside the urinary tract. On multivariable analysis, ypTa (HR = 3.36 [1.24-9.11]) and ypT1 (HR = 2.88 [1.33-6.22], p = 0.013) disease at RC were predictors of shorter TTR, while female sex was associated with longer TTR (HR = 0.52 [0.27-0.98], p = 0.043). The type of NAC was not predictive of TTR (GC vs. other, HR = 1.49 [0.75-2.97], p = 0.26). Conclusions: To our knowledge, this is the largest study to quantify the risk of recurrence in pts achieving pathologic response after NAC and RC for MIBC. 8% of patients undergoing NAC and achieving < ypT2N0 at RC recurred. Residual ypTa and ypT1 disease conferred a significantly higher risk of recurrence, while ypTis did not; female sex was associated with a lower risk of recurrence. Importantly, the type of cisplatin-based NAC regimen used was not an independent predictor of recurrence.