Muscle-invasive and Metastatic Bladder Cancer

6. MARKERS

6.1. Introduction

Both patient and tumour characteristics guide treatment decisions and prognosis of patients with MIBC.

6.2. Prognostic markers

6.2.1. Histopathological and clinical markers

The most important histopathological prognostic variables after RC and LN dissection are tumour stage and LN status [205]. In addition, other histopathological parameters of the RC specimen have been associated with prognosis.

The value of lymphovascular invasion was reported in a systematic review and meta-analysis including 78,000 patients from 65 studies treated with RC for BC [206]. Lymphovascular invasion was present in 35% of the patients and correlated with a 1.5-fold higher risk of recurrence and CSM, independent of pathological stage and peri-operative chemotherapy. This correlation was even stronger in those patients with node-negative disease [207].

In a systematic review and meta-analysis including 23 studies and over 20,000 patients, the presence of concomitant CIS in the RC specimen was associated with a higher odds ratio (OR) of ureteral involvement (pooled OR: 4.51, 2.59–7.84). Concomitant CIS was not independently associated with OS, recurrence-free survival (RFS) and DSS in all patients, but in patients with organ-confined disease concomitant CIS was associated with worse RFS (pooled HR: 1.57, 1.12–2.21) and CSM (pooled HR: 1.51, 1.001–2.280) [207].

Tumour location has been associated with prognosis. Tumours located at the bladder neck or trigone of the bladder appear to have an increased likelihood of nodal metastasis (OR: 1.83, 95% CI: 1.11–2.99) and have been associated with decreased survival [205, 208-210].

Prostatic urethral involvement at the time of RC was also found to be associated with worse survival outcomes. In a series of 995 patients, prostatic involvement was recorded in 31% of patients. The 5-year CSS in patients with CIS of the prostatic urethra was 40%, whilst the prognosis of patients with UC invading the prostatic stroma was worse with a 5-year CSS of only 12% [211].

Neutrophil-to-lymphocyte ratio (NLR) has emerged as a prognostic factor in UUT tumours [1] and other non-urological malignancies. In a pooled analysis of 21 studies analysing the prognostic role of NLR in BC, the authors correlated elevated pre-treatment NLR with OS, RFS and disease-free survival (DFS) in both localised and metastatic disease [212]. In contrast, a secondary analysis of the Southwest Oncology Group (SWOG) 8710 trial, a randomised phase III trial assessing cystectomy ± NAC in patients with MIBC, suggests that NLR is neither a prognostic nor a predictive biomarker for OS in MIBC [213].

In patients with LN-positive disease, the American Joint Committee on Cancer (AJCC)-TNM staging system provides 3 subcategories. In addition, several other prognostic LN-related parameters have been reported. These include, but are not limited to, the number of positive LNs, the number of LNs removed, LN density (the ratio of positive LNs to the number of LNs removed) and extranodal extension. In a systematic review and meta-analysis, it was reported that LN density was independently associated with OS (HR: 1.45, 95%, CI: 1.11–1.90) [214]. It has been suggested that LN density outperforms the AJCC-TNM staging system for LN-positive disease in terms of prognostic value [215,216]. However, in spite of these studies supporting the use of LN density, LN density relies on the number of LNs removed which, in turn, is subject to surgical and pathological factors. This makes the concept of LN density difficult to apply uniformly [217].

Two studies investigated whether any of the reported LN-related parameters may be superior to the routinely used AJCC-TNM staging system [217,218]. Whilst the conclusion was that the AJCC-TNM staging system for LN status did not perform well, none of the other tested variables outperformed the AJCC system.

6.2.2. Molecular markers

6.2.2.1. Molecular variants based on the Cancer Genome Atlas cohort

The updated Cancer Genome Atlas (TCGA) reported on 412 MIBCs and identified two main groups; luminal and basal-squamous - consisting of five mRNA expression-based molecular variant including luminal- papillary, luminal-infiltrated, luminal; basal-squamous; and neuronal; a variant associated with poor survival in which part of tumours do not have small cell or neuroendocrine histology. Each variant is associated with distinct mutational profiles, histopathological features and prognostic and treatment implications [219].

The basal-squamous variant is characterised by expression of basal keratin markers, immune infiltrates and is felt to be chemosensitive. The different luminal variants are characterised by fibroblast growth factor receptor 3 (FGFR3) alterations (luminal-papillary [LumP]), epithelial-mesenchymal transition (EMT) markers (luminal-infiltrated) and may be associated with chemotherapy resistance [67,68,219,220]. In 2019, a consensus on molecular variant classification was reported [221]. The authors analysed 1,750 MIBC transcriptomic profiles from 18 datasets and identified six MIBC molecular classes that reconcile all previously published classification schemes. The molecular variant classes include LumP, luminal non-specified (LumNS), luminal unstable (LumU), stroma-rich, basal/squamous (Ba/Sq), and neuroendocrine-like (NE-like). Each class has distinct differentiation patterns, oncogenic mechanisms, tumour micro-environments and histological and clinical associations. However, the authors stressed that consensus was reached for biological rather than clinical classes. Therefore, at this time, the classification should be considered as a research tool for retrospective and prospective studies until future studies establish how these molecular variants can be used best in a clinical setting.

Molecular classification of MIBC is still evolving and treatment tailored to molecular variant is not a standard yet. A novel 12-gene signature derived from patients in the TCGA utilising published gene signatures has been developed and externally validated to predict OS in MIBC [222]. Interestingly, an analysis of molecular typing in MIBC demonstrated that although molecular variants reflect the heterogeneity of bladder tumours and are associated with tumour grade, clinical parameters outperformed variants for predicting outcome [223]. In the coming years, new insights into BC carcinogenesis may change our management of the disease and our ability to better predict outcomes [224]. Outside clinical trials, molecular examination, either by expression profiling or immunohistochemistry, is not yet part of routine clinical work-up awaiting more conclusive data.

6.3. Predictive markers

6.3.1. Clinical and histopathological markers

Based on retrospective data only, patients with secondary MIBC have a worse response to NAC compared to patients with primary MIBC [225]. Pietzak et al., retrospectively analysed clinico-pathologic outcomes comparing 245 patients with clinical T2–4a N0M0 primary MIBC and 43 patients with secondary MIBC treated with NAC and RC. They found that patients with secondary MIBC had lower pathologic response rates following NAC than those with primary MIBC (univariable 26% vs. 45%, multivariable OR: 0.4 [95% CI: 0.18–0.84, p = 0.02]). They also found that MIBC patients progressing after NAC had worse CSS as compared to patients treated with cystectomy alone (p = 0.002).

Subtypes and non-UC have also been linked to worse outcomes after NAC, but there is, as yet, insufficient data to conclude that they can be considered as predictive markers [226].

6.3.2. Molecular markers

Several predictive biomarkers have been investigated such as serum vascular endothelial growth factor (VEGF) [227], circulating tumour cells, immune and stromal signatures, as well as expression of or defects in DNA damage repair (DDR) genes including ERCC2, ATM, MRE11, RB1 and FANCC that may predict response to cisplatin-based NAC [228,229] or chemoradiation [230-233]. More recently, alterations in FGFR2/3 including both mutations and gene fusions have been shown to be associated with response to FGFR inhibitors [234,235].

More recent efforts have focused on markers for predicting response to immune checkpoint inhibition. Programmed death-ligand 1 (PD-L1) expression by immunohistochemistry has been evaluated in several studies with mixed results which may in part be related to the use of different antibodies and various scoring systems evaluating different compartments, i.e., tumour cells, immune cells, or both. The major limitation of PD-L1 staining relates to the significant proportion of PD-L1-negative patients that respond to immune checkpoint blockade. For example, in the IMvigor 210 phase II study of atezolizumab in patients with advanced/metastatic UC who progressed after platinum-based chemotherapy, responses were seen in 18% of patients with low/no PD-L1 expression [236]. At present, the only indication for PD-L1 testing relates to the use of immune checkpoint inhibitors as monotherapy in patients with locally-advanced or metastatic UC unfit for cisplatin-containing chemotherapy who have not received prior therapy. In this setting, atezolizumab (the European Medicines Agency [EMA] approval) or pembrolizumab (EMA approval) should only be used in patients unfit for cisplatin-containing chemotherapy whose tumours overexpress PD-L1 (i.e., in case of atezolizumab; tumour-infiltrating immune cells [IC] covering ≥ 5% of the tumour area using the SP142 assay; in case of pembrolizumab, a combined positive score (CPS) of ≥ 10 using the Dako 22C33 platform) [237]. The FDA revised the label for pembrolizumab in patients with advanced UC with approval in first line only for patients not eligible for any platinum-based chemotherapy, however, irrespective of PD-L1 status.

Urothelial cancer is associated with a high tumour mutational burden (TMB) [238]. Both predicted neoantigen burden and TMB have been associated with response to immune checkpoint blockade in several malignancies. High TMB has been associated with response to immune checkpoint inhibitors in metastatic BC [236,239]. Conflicting results have been seen in studies evaluating immune checkpoint inhibitors in the neoadjuvant setting with the Pembrolizumab as Neoadjuvant Therapy Before Radical Cystectomy in Patients With Muscle- Invasive Urothelial Bladder Carcinoma (PURE)-01 study demonstrating an association of high TMB with response while there was no association with atezolizumab in the Phase II study investigating the safety and efficacy of neoadjuvant atezolizumab in MIBC (ABACUS) [240,241].

Other markers that have been evaluated in predicting response to immune checkpoint inhibitors include molecular subtypes as discussed earlier, CD8 expression by immunohistochemistry and other immune gene cell signatures. Recent work has focused on the importance of stroma including the role of transforming growth factors (TGFs) in predicting response to immune checkpoint blockade [242,243]. Powles et al., have reported on the potential for ctDNA to guide the use of adjuvant IO in UC [244]. In 581 patients from a phase III RCT of adjuvant atezolizumab vs. observation in UC, ctDNA testing at the start of therapy identified 214 (37%) patients who were positive for ctDNA and who had poor prognosis (observation arm HR = 6.3, 95% CI: 4.45–8.92; p < 0.0001). Patients who were positive for ctDNA had improved DFS and OS in the atezolizumab arm vs. the observation arm (DFS: HR = 0.58 [95% CI: 0.43–0.79]; p = 0.0024, OS: HR = 0.59 [95% CI: 0.41–0.86]). There was no difference in DFS or OS between treatment arms for patients who were negative for ctDNA. The rate of ctDNA clearance at week 6 was higher in the atezolizumab arm (18%) than in the observation arm (4%) (p = 0.0204). An ongoing clinical trial (IMvigor011) is evaluating atezolizumab as adjuvant therapy in patients with high-risk MIBC who are ctDNA positive following cystectomy [245].

A exploratory analysis in patients with metastatic UC who received pembrolizumab in the first-line (KEYNOTE-052 trial) and salvage (KEYNOTE-045 trial) settings, demonstrated that TMB and T-cell inflammed gene expression profile were significantly associated with improved outcomes, however PD-L1 was associated with improved outcomes and stromal signature with worse outcomes in KEYNOTE-052, but not KEYNOTE-045 suggesting that these biomarkers may perform differently in different clinical disease states i.e. first line versus salvage settings [246]. In a second study, a scoring system (CPT) based on CD39, PD-L1 and TMB was shown to predict response to PD-L1 blockade and platinum-based chemotherapy in patients with MIBC [247].

Although promising, there are currently no validated predictive molecular markers that are routinely used in clinical practice. Further validation studies are awaited.

6.4. Conclusion

The updated TCGA and other efforts have refined our understanding of the molecular underpinnings of BC biology. Molecular variants, immune gene signatures as well as stromal signatures may ultimately have an important role in predicting response to IO. Although PD-L1 expression by immunohistochemistry and TMB have demonstrated predictive value in certain settings, additional studies are needed. Prospectively validated prognostic and predictive molecular biomarkers will present valuable adjuncts to clinical and pathological data, but large phase III RCTs with long-term follow-up will be needed to clarify the many questions remaining.

6.5. Summary of evidence for urothelial markers