Salvage Therapies for Non–muscle-invasive Bladder Cancer: Who Will Respond to Bacillus Calmette-Guérin? Predictors and Nomograms

The best predictors of response to intravesical immunotherapy are tumor grade and stage, tumor recurrence pattern, nomograms, panels of urinary cytokines, and fluorescent in situ hybridization patterns of urine cytology examinations. Future investigations on predictors of Bacillus Calmette-Guérin efficacy are needed to better select those patients who will really benefit from a conservative treatment. Hardly any of the proposed nomograms were designed to precisely predict the outcome of Bacillus Calmette-Guérin immunotherapy. A new nomogram for NMIBC recurrence and progression based on all non–muscle-invasive bladder cancer subgroups would include factors already proven in cancer prognosis and prediction.

Key points

•

The best predictors of response to Bacillus Calmette-Guérin are tumor grade and stage, recurrence pattern, nomograms, panels of urinary cytokines, and fluorescent in situ hybridization patterns of cytology.
•

It has to be emphasized that hardly any of the proposed nomograms were designed to precisely predict the outcome of Bacillus Calmette-Guérin immunotherapy.
•

A nomogram for non–muscle-invasive bladder cancer would be based on all non–muscle-invasive bladder cancer subgroups and would include factors that are already of proven value in prognosis and prediction.

Introduction

Among genitourinary tract tumors, urothelial bladder cancer represents the second most common diagnosis after prostate cancer. In Western countries, incidence rates have been increasing, with a prevalence of 500,000 cases per year in the United States. At diagnosis, most such tumors are non–muscle-invasive bladder cancer (NMIBC). Among NMIBC, T1 high grade is considered to be a high-risk subgroup with an increased risk of progression that is sometimes characterized by the biological behavior of an invasive tumor. Its natural history suggests an unfavorable long-term outcome, as documented by early untreated series reporting a progression rate of 27% to 65% and a cancer-specific death rate of 34%.

Bacillus Calmette-Guérin (BCG) is currently considered the gold standard conservative treatment for this kind of lesion, ^, but up to 30% of patients fail to respond initially and, of the responders, 74% eventually relapse. Patients with BCG failure are unlikely to respond to further BCG therapy; moreover, delayed radical cystectomy is unlikely to be able to cure the disease. A decrease in cancer-specific survival has been shown in patients with high-grade recurrence after BCG therapy who underwent radical cystectomy. According to these findings, a delay in radical cystectomy seems to have a crucial role in the worsening of oncological outcomes.

Predictors

Pathologic features have been the first factors used to predict the prognosis of bladder cancer. Based on these factors, prognostic models have been built, such as the European Organisation for Research and Treatment of Cancer (EORTC) and Club Urológico Español de Tratamiento Oncológico (CUETO) risk tables. Both of these models identify early recurrence and multiplicity as predictors of progression to muscle-invasive disease. Unfortunately, the available predictive models of recurrence and progression (EORTC) are unable to identify those patients who will not respond to intravesical immunotherapy with BCG.

To date, a clear association between recurrent and multiple tumors and BCG response rate has not been found. In a population of 81 patients with stage T1G3 disease treated with BCG, neither multiplicity nor the presence of carcinoma in situ (CIS) was associated with the risk of failure of BCG treatment. Similar results were previously highlighted by Cookson and Sarosdy. According to these authors, tumor grade and stage, tumor multiplicity, and concurrent CIS before BCG did not predict tumor recurrence or progression. It is noteworthy that the role of concurrent CIS as a predictor of BCG response has been widely investigated. A number of studies have corroborated the relationship between the presence of CIS and a shorter recurrence- and progression-free survival in patients treated with BCG. In a series of 146 patients treated with BCG, the presence of CIS was the only predictor of progression, and consequently of poor BCG response, at multivariate analysis.

Patient sex and age have also been investigated as potential predictors of response to BCG therapy. Even if female sex is associated with a worse prognosis, an investigation of a large series of 1021 consecutive patients treated with an induction course of BCG found no significant differences between men and women in terms of initial response, time to recurrence, or progression. Because age could affect the integrity of the immune system, through which BCG acts, it has been supposed that older patients could have an unsatisfactory response to BCG. In confirmation of this hypothesis, a BCG plus interferon-alpha phase II study showed that patients older than 80 years had the poorest recurrence-free survival and that age over 80 years was independently associated with shorter recurrence-free survival.

Even if the available risk tables are useful in the management of these groups of patients and in the prediction of risk of recurrence and progression, they are not equally accurate in predicting BCG response. Mainly owing to the lack of a standard BCG maintenance protocol among the patients used to elaborate the EORTC and CUETO risk tables, these models are unable to identify those patients who will not respond to intravesical immunotherapy with BCG.

Recent studies have shown the role of lymphovascular invasion (LVI) and T1 substaging in predicting BCG response rates. According to Fukumoto and colleagues, high-risk LVI-positive patients treated with BCG had a significantly lower progression-free survival rate compared with those without LVI (8.2% and 65.9%, respectively; P = .015). Similarly, patients with tumors deeply invading the lamina propria (HGT1b) showed a 3-fold increase in risk of progression, with lower response rates to BCG therapy.

Biomarkers, the Immune System, and Cytokines

The host immune system plays a crucial role in the therapeutic effect of BCG. Based on the mechanism of action of BCG, a number of studies have already been undertaken to evaluate the immunohistochemical pattern in tumor tissue as well as the serum and urine cytokine levels in patients with bladder cancer. It has been demonstrated that a higher level of leukocyturia after BCG induction is associated with improved response to BCG. The mechanism of action of BCG seems to be partially related to the activity of antigen-presenting cells, like lymphocytes and natural killer cells. Based on this observation, a number of studies were performed in immunocompetent patients, demonstrating an increase in the number of CD4 ⁺ Th cells after BCG treatment and in the number of CD4 ⁺ T cells (hazard ratio, 0.13; P = .025). Similarly, Pichler and colleagues, found that a decreased density of Th2-predominant CD4 ⁺ T cells contributes to poor recurrence-free survival after BCG therapy. This interesting finding suggests a potential role of the tumor immune pattern in prediction of the therapeutic response to BCG, permitting individualized treatment.

Natural killer cells are considered a crucial element in the immune system activation induced by BCG treatment. Interactions between natural killer cells and tumor ligands have been suggested as potential predictors of BCG response. Yutkin and colleagues evaluated tumor expression of natural cytotoxicity receptor ligands in specimens of transurethral resection of bladder tumor from patients with primary, non–muscle-invasive high-grade bladder cancer who were subsequently treated with BCG. According to their results, primary tumors from favorably responding patients expressed higher levels of ligand for all the tested proteins.

Another turning point in the understanding of BCG activity has been the recognition of the role of macrophages. In this field, 2 different studies found an association between higher numbers of tumor-infiltrating CD68 ⁺ tumor-associated macrophages and higher recurrence rates after BCG and suggested that these cells are involved in the inflammatory circuit that promotes tumor progression. ^,

A number of biomarkers are currently under evaluation as predictors of response to BCG. Among them, the most promising seem to be RB, survivin, bcl-2, E-cadherin, ezrin, FGFR3, and Ki67. To give an example, underexpression of RB and ezrin is a predictor of no response to BCG and increased risk of progression. Conversely, the downregulation of FGFR3 predicts a good response. Unfortunately, the available literature on biomarkers has several limitations, including nonstandard methods of measurement and lack of standardized cutoff and validation.

Urinary cytokines are one of the most promising prognostic markers in BC. They can be found in the urine within 1 to 4 weeks from the start of BCG treatment. After gaining an increased understanding of their functions, cytokine profiles have been analyzed to assess the efficacy of BCG-induced cytotoxic response. IL-2 expression has been most extensively studied. When cytokine levels were evaluated in 39 patients receiving BCG for NMIBC or CIS, absence of urinary IL-2 during the BCG induction course and the first extended induction cycle (6 + 3 schedule) was found to correlate with time to recurrence ( P = .01) and progression ( P = .01). Similarly, Cai and colleagues evaluated the potential role of the ratio between IL-6 and IL-10, comparing 72 patients affected by high-risk NMIBC with 49 controls. At multivariate analysis, IL-6/IL-10 ratio ( P <.003) and the number of lesions ( P <.001) were identified as independent predictors of probability of BCG response. Patients with an IL6/IL10 ratio of greater than 0.10 had a higher risk of recurrence and progression ( Table 1 ).

Table 1

Predictive makers (BCG response)

Predictive Marker	Study	Determination
Ratio IL6/IL10 (23)	Prospective	Urine/ELISA
Leukocyturia (13) (during BCG treatment)	Prospective	Urine (counting)
CD4 ⁺(15)	Prospective	Tissue
Density Th2 (16)	Prospective	Tissue/IHC
Natural killer cells (17)	Retrospective	Tissue/IHC
Macrophages (18)	Prospective	IHC (density positive cells)
FGFR3 (22) (downregulation)	Prospective	Cell culture
FISH (24)	Prospective	Urine
Methylation (tumor suppressor genes)	Retrospective	Tissue
Cytokines (43) CyPRIT	Prospective	Urine
Ezrin (21)	Retrospective	Tissue/IHC

Abbreviation: IHC, immunohistochemistry.

Urinary fluorescent in situ hybridization (FISH) is a molecular cytogenetic test able to detect chromosomal abnormalities. Thanks to its ability to anticipate tumor formation, FISH could be considered a valuable predictor of BCG response. Kamat and colleagues evaluated the role of FISH in predicting BCG failure in a population of 126 patients treated with BCG. According to these authors, patients with a positive FISH result had a 58% risk of recurrence compared with 15% in patients with a negative result ( P <.001) and a risk of progression of 25% compared with 7% in cases of negative FISH ( P <.013).

Genomic Variant

Another theory has emerged from the idea that genomic variations affecting key genes implicated in BCG-induced inflammatory pathways could impact on BCG response rates. Leibovici and colleagues observed that a variant genotype in IL-6 is associated with increased risk of recurrence with BCG. Their results supported the idea that inflammation gene polymorphisms are associated with a modified treatment response to BCG, with a consequent impact on survival.

Additional genes linked to outcomes following BCG treatment include those involved in detoxification (hGPX1), nucleotide exCISion repair and regulation of macrophage susceptibility to intracellular mycobacterial growth (NRAMP1). Similarly, polymorphisms impairing cellular DNA damage repair have been associated with better outcomes after BCG treatment. In a population of 25 high-risk patients with NMIBC, selective next-generation sequencing was performed to compare the genomic profiles of cancers that responded to intravesical therapy and those that progressed to muscle-invasive disease. This analysis found a significant decrease in total mutational burden to be associated with immunotherapy response. This association suggests that more advanced tumors have a decreased neoantigen burden and may explain the mechanism of BCG response in nonprogressors.

Methylation profiles have been examined in several panels of genes. Agundez and colleagues assessed the role of the methylation of 25 tumor suppressor genes in predicting the prognosis in 91 patients with primary T1G3 tumors who were receiving BCG treatment. Their results confirmed the methylation status of tumor suppressor genes to be associated with the clinical outcome. The authors concluded that methylation status can be used to distinguish patients who will respond to BCG from those who will not.

Predicting Bacillus Calmette-Guérin Response in Carcinoma in Situ

Response to BCG is the result of complex molecular interactions between host and tumor. To date, not a single marker or test has been identified that can predict the precise response to BCG, either in CIS or in other NMIBC. Nonetheless, a combination of surrogate biomarkers and predictors of response has been studied in a heterogeneous population of NMIBC and this combination could be cautiously extrapolated to CIS alone.

Clinicopathologic factors have been extensively studied by the EORTC and CUETO groups. Female sex, old age, high grade and stage, tumor multiplicity, early recurrence, and presence of CIS were identified as predictors of recurrence and progression of non–muscle-invasive tumors. One of the major drawbacks of the EORTC and CUETO studies is the disparity in BCG application and schedules. Moreover, the conclusions are related to poor tumor prognosis and caution must be exercised in assuming an association with BCG response itself, although it seems logical that a poor BCG response could be indicative of a worse outcome.

Urinary FISH (Urovysion, Abbot Molecular, Des Plaines, IL), a molecular cytogenetic test, can diagnose molecular failure after BCG and has been found capable of predicting subsequent tumor recurrence in the following 6 to 24 months. In a recent meta-analysis including 6 studies and 442 patients, post-BCG FISH test had a sensitivity of 0.54 (95% confidence interval [CI], 0.38–0.69), a specificity of 0.84 (95% CI, 0.72–0.91) and an area under the curve of 0.78 (95% CI, 0.74–0.81) for predicting recurrences. Patients with a positive post-BCG FISH test were more likely to have recurrences during follow-up (hazard ratio, 3.95; 95% CI, 2.72–5.72). No specific studies have been conducted exclusively for CIS. Many molecular biomarkers have been explored using immunohistochemistry. In a study conducted by Sato and colleagues in 27 patients with CIS, tumors where analyzed before treatment and overexpression of pRb, p16, and p53 was found in 41%, 37%, and 48% of patients, respectively. pRb overexpression had a significant relationship with poor response to BCG therapy, but neither p16 nor p53 seemed to have a predictive value for initial BCG failure.

The lack of predictive value of p53 for CIS was also identified in other studies. ^, Other markers have been assessed for NMIBC (survivin, bcl-2, e-cadherin, ezrin, and Ki-67), but study heterogeneity at multiple levels makes response prediction unreliable when using only molecular biomarkers.

An indirect way of evaluating BCG-induced immunity is to measure the number and type of cells and cytokines involved in the immune response. This factor has been quantified in various samples (tumor, urine, and peripheral blood) and at different timings (before, during, and after treatment). Of all these tests, probably the one that stands out is determination of the polarity of lymphocytes and eosinophils in the tumor’s microenvironment. The goal of BCG is to drive an immunomodulated response by changing immune cells’ polarity from Th2 to Th1, and BCG has been found to have no effect if the microenvironment is already polarized to Th1 before treatment. In this study, the proposed algorithm included 3 immunohistochemistry tests, 2 scores (eosinophil activity index and GATA-3 ⁺ /T-bet ⁺ lymphocyte ratio) and a Th2 signature to assess response in patients with CIS. This tool was able to predict BCG response with 100% sensitivity and 80% specificity at the authors’ proposed cutoff.

Other tools to assess immune response include evaluation of the presence of polymorphonuclear cells, because they are related to direct cytotoxicity through tumor necrosis factor-related apoptosis-inducing ligand. Also, the dual role of macrophages has been assessed by 3 different studies, which concluded that presence of CD68 ⁺ tumor-infiltrating macrophages predicts a higher number of recurrences after BCG. ^, ^,

Antigen presentation-related molecules (heat shock protein 90, expression of major histocompatibility complex I and II) and signaling molecules (such as ICAM1) could also be an indirect sign of increased immune response. It remains unclear whether the presence of these molecules actually contributes in improving outcomes, because they are upregulated by the presence of interferon, which is always increased during BCG treatment.

Urinary cytokines have also been widely assessed. Kamat and colleagues created a nomogram (CyPRIT) using a panel of 9 inducible cytokines that seemed to have an accuracy of 85.5% (95% CI, 77.9%–93.1%) in predicting response. This diagnostic tool is currently under validation. Clinical immune response evaluation is an inconsistent way of drawing conclusions regarding the future response to BCG. Studies assessing adverse effects (eg, fever) have yielded controversial results, and although this could be related to a more energetic immunologic response, better outcomes in terms of recurrence or survival have not yet been proven. Finally, genetic polymorphisms may have a role in BCG response. On the one hand, when impairing genes related to BCG-induced pathways (IL-6, IL-17, IL-2, tumor necrosis factor-α, MCP-1 and tumor necrosis factor-related apoptosis-inducing ligand receptor), they can be related to a worse response to BCG. ^, In contrast, when affecting DNA damage repair genes, they can contribute in increasing the tumor mutational burden, and subsequently to a better response to BCG and other new immunotherapies. ^, Further investigations are required to elucidate further the findings of all these studies.

Using the same NMBIC evaluation tools to predict response involves several shortcomings. First, CIS seems to have a different molecular pathway in its origin compared with Ta/T1, and extrapolation of results of the available tests is probably not completely reliable. Although no differences in BCG response between clusters were identified in this study, groups were small (n = 20) and the statistical power was insufficient to draw consistent conclusions. Moreover, most of the available studies have been applied in a heterogeneous population, mixing intermediate- and high-risk bladder cancer and also Ta/T1 with or without concomitant CIS. This makes it difficult to individualize conclusions for a subgroup of patients such as those with CIS alone.

Ongoing studies exploring CIS molecular origin, bladder cancer immunology, and tumor gene expression will help to resolve this puzzle and improve treatment choices and results.

Nomograms

Precise clinical risk assessment and identification of patients who will not benefit from conservative BCG immunotherapy is vital. Various scientific bodies have tried to facilitate this task by creating nomograms that are easy to use in everyday clinical practice. According to the latest update of the European Association of Urology guidelines, 2 nomograms are recommended for prediction of the NMIBC recurrence and progression risks. The first was constructed by the EORTC on the basis of data from 2596 patients diagnosed with TaT1 tumors and treated with adjuvant intravesical chemotherapy in 7 randomized EORTC trials. The scoring model included 6 factors (tumor stage, grade, focality and size, recurrence rate, and presence of CIS); a weight score was assessed for each factor to allow for the calculation of the total score and, finally, estimation of the risk using risk tables.

The EORTC tables have been externally validated in several populations. Studies from Europe and Asia have shown the EORTC tables to be efficient in predicting recurrence and progression with high concordance indexes. Also, a more recent study has validated the EORTC nomogram in 479 patients who underwent second (restaging) transurethral resection of bladder tumor. The authors found that especially in the intermediate-risk group, the model overestimates the risks of recurrence yet predicts tumor progression risk well. However, it has to be stressed that the EORTC nomogram was mainly based on patients who did not receive BCG immunotherapy. Although the parameters used in the EORTC system are generally related to poor prognosis and reflect an inferior response to BCG, the reliability of the EORTC system in the BCG setting is limited. Seo and colleagues demonstrated the EORTC tables to be effective in predicting recurrence and progression in 251 patients treated with BCG, yet the predicted risks were higher than the actual number of events in the analyzed population. This observation has been confirmed in other studies.

The second nomogram, recommended by the European Association of Urology, was created by CUETO for patients treated with BCG immunotherapy. It is based on an analysis of 1062 patients from four CUETO trials. Patients received 12 instillations over a 5- to 6-month period, and no immediate postoperative instillation or restaging transurethral resection of bladder was performed. The scoring system uses 7 prognostic factors: gender, age, recurrence rate, tumor stage, grade, focality, and CIS presence.

Similar to the EORTC tables, the CUETO system has been externally validated by various authors. A recent study analyzing 414 T1G3 cancers showed that the CUETO scoring model underestimates the risk of tumor recurrence, but predicts well the risk of progression. In another study conducted by Choi and colleagues, similar underestimation for higher risk tumors was observed for recurrence, but not for progression (only 53% patients received BCG). In contrast, Xylinas and colleagues showed, in a large multicenter population, that the CUETO tables overestimate the risks of recurrence and progression. However, only 11% of their patients were treated with BCG. In a recent study analyzing patients with intermediate- and high-risk NMIBC who received BCG with conventional maintenance it was found that the CUETO model underestimated the risk of recurrence in better risk patients and overestimated the risk of progression in poor risk patients. It is worth mentioning that in this study the authors also analyzed prognostic factors for both cancer-specific survival and overall survival (stage and grade for cancer-specific survival and grade and age for overall survival).

It has to be highlighted that from the current perspective, neither the EORTC nor the CUETO model is free from flaws. When constructing both nomograms, only a relatively small number of T1G3/high-grade tumors, the most clinically important entity in salvage setting, were included. Similarly, few CIS cases were considered, limiting the models’ value for both CIS prognosis and BCG response estimation. Neither system assessed prognostic factors known to impair BCG response, such as histologic variants and depth of submucosal or LVI, or analyzed the influence of restaging transurethral resection of bladder, the quality of the transurethral resection of bladder or the value of immediate chemotherapy instillation. Number of BCG instillations, dosage, strain, or other factors that might theoretically be associated with BCG outcome were also not studied.

In the recent literature, new nomograms aiming to meet current requirements are emerging. Ali-El-Dein and colleagues constructed nomograms for tumor recurrence and progression, analyzing the data of 1019 patients treated mainly with adjuvant BCG immunotherapy. They incorporated tumor stage, focality, recurrence rate and method of adjuvant therapy in the recurrence nomogram and tumor grade and size, as well as method of adjuvant therapy, in the progression nomogram.

Kim and colleagues composed a nomogram on the basis of 970 patients with primary NMIBC. LVI, variant histology (VH), gross hematuria, and history of upper tract urothelial carcinoma (UTUC) were incorporated in the mathematical model. The authors proved that gross hematuria and UTUCs are related to tumor recurrence and that UTUC and LVI are related to cancer progression. Similar to classic nomograms, this one is not free from imperfections. The analyzed population included a high number of T1LG tumors and UTUCs and a low number of CIS, LVIs, and VHs. Second, all the tumors were primary and only approximately 60% of patients were treated with some form of BCG immunotherapy.

Another new system was recently presented by D Andrea and colleagues. On the basis of retrospective analysis of 1289 patients with T1G3 NMIBCs treated with more current regimens, the authors identified VH and LVI as associated with disease progression and created a nomogram to facilitate system usage. Although the study population was treated with various therapeutic regimens, CIS cases were excluded and once more the transurethral resection of bladder and restaging of transurethral resection of bladder tumor quality was not assessed, this nomogram strengthens the position of VH and LVI as prognostic and predictive factors. It has to be remembered that the majority of the available nomograms and mathematical models are designed for prediction of the risk of a clinical event (eg, recurrence, progression), and have not primarily been created for prediction of BCG immunotherapy response. From this it follows that although some factors incorporated in the nomograms are associated with the general poor prognosis of bladder cancer, this does not necessarily translate directly into the ability to predict a poor BCG response. A tool designed to predict BCG response was recently proposed by Kamat and colleagues. In a prospective study of 125 patients, the authors measured cytokine levels in urine at various time points. Nine cytokines (IL2, IL-6, IL-8, IL-18, IL-1ra, tumor necrosis factor-related apoptosis-inducing ligand, interferon-γ, IL-12 [p70], and tumor necrosis factor-α) were found to have accuracy in predicting response to BCG immunotherapy in terms of recurrence. Despite the relatively low number of patients and the lack of staging and grading subanalyses, the proposed CyPRIT nomogram shows solid piece of evidence in.

Summary

Despite the number of investigations that have been conducted on this topic, the lack of a validated definition of BCG failure until the recent expert consensus on BCG-unresponsive disease has made it difficult to interpret and compare the results of the available studies. Currently, the best predictors of response to intravesical immunotherapy are tumor grade and stage, tumor recurrence pattern, nomograms such as the EORTC and CUETO tables, panels of urinary cytokines, and FISH patterns of urine cytology examinations. Future investigations on predictors of BCG efficacy are needed to better select those patients who will really benefit from a conservative treatment. It has to be emphasized that hardly any of the proposed nomograms were designed to precisely predict the outcome of BCG immunotherapy. Ideally, a new nomogram for NMIBC recurrence and progression (and cancer-specific mortality) would be based on all NMIBC subgroups treated with modern schedules and would include factors that are already of proven value in cancer prognosis and prediction. This is particularly important in high and very high-risk patients, as well as in the salvage setting.

Disclosures: The authors declare that the development of the manuscript was not supported by an honorarium, a grant, or any other sources of support, including sponsorship or any material sources of support.