Molecular Subtyping and Neoadjuvant Chemotherapy in Bladder Cancer - Christopher Wallis & Zachary Klaassen
September 27, 2022
Christopher Wallis and Zachary Klaassen discuss a publication entitled, “Different Responses to Neoadjuvant Chemotherapy in Urothelial Carcinoma Molecular Subtypes.” This study is investigating how molecular subtypes impact pathological response and survival in patients receiving preoperative cisplatin-based chemotherapy. This study shows that tumor classification by gene expression profiling and molecular subtyping can identify patients who are more likely to benefit from chemotherapy before radical cystectomy for muscle-invasive bladder cancer.
Biographies:
Christopher J.D. Wallis, MD, Ph.D., Assistant Professor in the Division of Urology at the University of Toronto.
Zachary Klaassen, MD, MSc, Urologic Oncologist, Assistant Professor Surgery/Urology at the Medical College of Georgia at Augusta University, Georgia Cancer Center
Biographies:
Christopher J.D. Wallis, MD, Ph.D., Assistant Professor in the Division of Urology at the University of Toronto.
Zachary Klaassen, MD, MSc, Urologic Oncologist, Assistant Professor Surgery/Urology at the Medical College of Georgia at Augusta University, Georgia Cancer Center
Read the Full Video Transcript
Christopher Wallis: Hello, and welcome to this UroToday Journal Club. Today, we are discussing a recent publication entitled, Different Responses to Neoadjuvant Chemotherapy in Urothelial Carcinoma Molecular Subtypes. I'm Chris Wallis, an Assistant Professor in the Division of Urology at the University of Toronto, and joining me today is Zach Klaassen, an Assistant Professor in the Division of Urology at the Medical College of Georgia. This is the citation for this recent publication in European Urology.
You can see, it's highlighted here in the NCCN clinical practice guidelines for bladder cancer, that the use of neoadjuvant chemotherapy, either one followed by radical cystectomy or followed by a partial cystectomy in the appropriate patients, is one of the few Level 1 recommendations in the treatment of patients with muscle-invasive bladder cancer. However, despite the strength of evidence, neoadjuvant chemotherapy is relatively underutilized, and the rationale for this ranges from toxicity considerations or the relatively modest absolute benefits. Additionally, there is morbidity associated with the use of neoadjuvant chemotherapy, and we lack biomarkers to predict those patients who will benefit.
In muscle-invasive bladder cancer, there has been a substantial amount of work undertaken in the past few decades looking at molecular subtypes, which represent discrete phenotypes within muscle-invasive bladder cancer. You can see here, there is a whole variety of different schema that have been proposed, but this is one that the authors of this paper have put forward. Previous work has suggested that each of these molecular subtypes may have a differential response to chemotherapy or systemic therapies of other types.
The goal of this paper was to investigate how molecular subtypes, as assessed based on TURBT specimens, impact the pathological response and survival for patients who receive chemotherapy prior to radical cystectomy. To assess this, the authors looked at patients undergoing preoperative cisplatin-based chemotherapy for muscle-invasive bladder cancer between 2004 and 2015 in two regions in Sweden. They included a chemotherapy-exposed cohort, which was predominantly patients treated with chemotherapy but included some patients with clinical node-positive or clinical T4b disease who received induction chemotherapy followed by consolidative surgery. A reference cohort who received a radical cystectomy alone was also accrued.
The authors assessed the primary outcome of pathological response rate, and they categorized this as a complete response, versus partial response, or no response. Partial response is defined as patients who were downstaged to non-muscle invasive disease. Pathologic downstaging was further considered as the difference between clinical and pathological stages and scored on an ordinal scale from 0 to 5. The difference of three steps or more was considered a major response to neoadjuvant therapy. The authors secondarily assessed cancer-specific survival and overall survival.
The authors reviewed TURBT specimens for all included patients using four expert uropathologists to determine T-stage, grade, the presence of lymphovascular invasion, keratinization, CIS, necrosis, or variant histology. Two 1 mm cores were taken from representative blocks for each tumor and arranged into tissue microarrays, which were subsequently stained using the Lund taxonomy IHC subtyping, as well as for anti-osteopontin. RNA was extracted from the formalin-embedded blocks, and the Affymetrix Gene ST 1.0 platform was used to undertake transcriptome-based analyses for these specimens.
In terms of statistical analysis, the authors use Chi-squared tests to compare the proportions of patients in each subtype who achieve pathologic complete response. They further used multivariable logistic regression models to adjust for the clinical T-stage to test the association between molecular subtype and pathologic complete response. Multivariable Cox proportional-hazards models were used to assess the association between these subtypes and survival endpoints, and ROC curves and AUCs were calculated for the ability of molecular subtypes to predict response to neoadjuvant chemotherapy.
At this point in time, I'm going to hand it over to Zach to walk us through the results.
Zachary Klaassen: Thanks, Chris. This is the flow chart showing the criteria for entry into the chemo cohort and the radical cystectomy cohort. And as you can see here on the left, there were 172 patients that had preoperative cisplatin-based chemotherapy. Ultimately, 124 of these were included in the neoadjuvant analysis, and 24 in the induction chemotherapy arm. Moving to the right, for radical cystectomy, there were 239 patients evaluated, and after exclusion criteria, 186 were included in the radical cystectomy analysis.
This is table 1 which looks at the clinical and pathological characteristics of the study cohorts. And as you can see here on the right, this is a busy table, I'll walk you through several key findings. On the far right is the RC cohort, to the left of that is the induction cohort, and to the left of that is the neoadjuvant cohort.
Looking at the ages of these patients stratified by these cohorts, the chemotherapy arm is slightly younger, 66 and 67 years of age, compared to 77 in the radical cystectomy cohort. Overall, more than three-quarters of patients were male. The most common T-stage for these patients was clinical T2. The most common node stage was N0 for the neoadjuvant cohort, and 99% of patients in the radical cystectomy cohort were clinical N0. For the induction cohort, the most common clinical node stage was clinical N1. Looking at variant histology, 17% in the neoadjuvant cohort, 12% in the induction cohort, and 9% in the radical cystectomy cohort. Going a little bit further down the table, we see the LundTax subtype based on RNA, and you can see a relatively balanced cohort amongst these different subtypes, which was also confirmed by the LundTax subtype IHC and the consensus subtype as well.
Looking at the type of chemotherapy received in terms of gemcitabine-cisplatin, 61% of patients in the neoadjuvant cohort and only 25% in the induction cohort. Dose-dense MVAC was more common in the induction cohort. In terms of chemotherapy, cycles received, in the neoadjuvant cohort, most commonly three cycles at 76%, and in the induction cohort, most commonly four cycles. Looking at the pathological T-stage, the most common pathological T-stage was pT0 in the neoadjuvant cohort, was pT2 in the induction cohort, and was pT3 in the radical cystectomy cohort.
This slide looks at the RNA-based molecular subtypes based on the LundTax classification, and you can see this is broken down by RNA subtype and IHC subtype. And you can see the subtypes here listed on the right, including UroA, B, and C, genomic unstable, basal/squamous, mesenchymal-like, as well as neuroendocrine. Below are just the representative images of the immunohistochemical staining, as well as the subtype-specific staining as well.
This figure looks at the pathologic responses stratified by molecular subtypes. We'll start on the left side. This is the pathologic response in the neoadjuvant cohort, and looking at the dark blue, this is the pathologic complete response, which was more common in the genomic unstable subtype and the second most common was in the urothelial-like subtype. Moving to the right, this is the pathologic response by subtype in the radical cystectomy cohort. As expected, less pathologic complete responses in this group, based on the fact that they did not have neoadjuvant chemo. And we can see that looking at the light blue, the pathologic MIBC or N-positive disease was most common in the basal/squamous subtype.
This is the figure looking at the number of steps of pathological downstaging by molecular subtypes. As Chris alluded to in the methods, a significant downstage was classified as 3 to 5, which is in the dark blue. And you can see that the genomic unstable, as well as the urothelial subtypes, had the most common or most frequent downstaging of 3 to 5 degrees.
This is the Kaplan-Meier curve for cancer-specific survival, stratified by the LundTax subtypes. And you can see here that, no surprise, the genomic unstable and the urothelial-like subtypes had the best cancer-specific survival, with the median not reached at over 5 years of follow-up. This is the Kaplan-Meier curves for recurrence-free survival stratified by the subtypes in the neoadjuvant and the radical cystectomy cohort. Again, we see the genomic unstable and the urothelial-like having the best RFS in the neoadjuvant cohort, and in the RC cohort, interestingly, we see that the squamous and neuroendocrine has the best RFS. However, this is a small number of patients with only 11 in this subtype.
This table looks at the multivariable Cox proportional hazards model for cancer-specific survival, with a reference of basal/squamous in the LundTax RNA subtype. You can see the significant improvement in cancer-specific survival for the urothelial C subtype, as well as the genomic unstable subtype. And no surprise here, with a reference to T2, patients with T3 and T4a disease had worse cancer-specific survival.
This table looks at the pathological responses stratified by cohorts. This is the neoadjuvant cohort in this study, as well as two outside cohorts, including the Seiler cohort and the Taber cohort. And I'll point your attention to the far right of the screen, looking at the composite response compared to the basal/squamous subtype. We see a significant benefit for urothelial A subtype, 54% composite response, which was statistically significant compared to the basal/squamous subtype, as well as the genomic unstable subtype, with a composite response of 53%, and again, statistically significant.
This slide looks at the boxplots of subtype-specific expression of SPP1 within the molecular subtypes, and you can see here that to summarize these figures, the expression of SPP1, which codes for osteopontin, is a biomarker for a response only within the non-urothelial subtype. So for instance, we see that in the basal/squamous in the chemo cohort, this predicted basal/squamous no pathological complete response, and we also see this signal in the GU or genomically unstable cohort, with no pathological complete response with a high SPP1 expression. And this was also seen in the statistical model here on the right.
So several discussion points from this study. This study showed an improved pathological response and survival outcomes for patients with GU LundTax molecular subtype, independent of tumor stage, following treatment with either neoadjuvant chemotherapy or radical cystectomy. And this was externally validated in other data sets as we described in the previous slides. Previous studies investigating the response to neoadjuvant chemotherapy using the different classification systems have not yielded consistent results, which has been somewhat confusing in the literature. And so the authors posed that this may be secondary to widely varied study designs, different survival endpoints, treatments, or selection endpoints, and unknown confounders that were unable to be controlled for. Finally, results in muscle-invasive bladder cancer may improve with second-generation biomarkers and those that operate only within a specific context defined by the molecular subtypes. And as we saw in this study, SPP1 in non-urothelial subtypes was able to be applied to the non-urothelial versions of the subtypes.
So in conclusion, this study shows that the existing molecular subtypes of muscle-invasive bladder cancer derived different clinical benefits from neoadjuvant chemotherapy. And ultimately, molecular classification and subtype-dependent second-generation biomarkers should be integrated into future prospective trials.
Thank you very much, and we hope you enjoyed this UroToday Journal Club discussion.
Christopher Wallis: Hello, and welcome to this UroToday Journal Club. Today, we are discussing a recent publication entitled, Different Responses to Neoadjuvant Chemotherapy in Urothelial Carcinoma Molecular Subtypes. I'm Chris Wallis, an Assistant Professor in the Division of Urology at the University of Toronto, and joining me today is Zach Klaassen, an Assistant Professor in the Division of Urology at the Medical College of Georgia. This is the citation for this recent publication in European Urology.
You can see, it's highlighted here in the NCCN clinical practice guidelines for bladder cancer, that the use of neoadjuvant chemotherapy, either one followed by radical cystectomy or followed by a partial cystectomy in the appropriate patients, is one of the few Level 1 recommendations in the treatment of patients with muscle-invasive bladder cancer. However, despite the strength of evidence, neoadjuvant chemotherapy is relatively underutilized, and the rationale for this ranges from toxicity considerations or the relatively modest absolute benefits. Additionally, there is morbidity associated with the use of neoadjuvant chemotherapy, and we lack biomarkers to predict those patients who will benefit.
In muscle-invasive bladder cancer, there has been a substantial amount of work undertaken in the past few decades looking at molecular subtypes, which represent discrete phenotypes within muscle-invasive bladder cancer. You can see here, there is a whole variety of different schema that have been proposed, but this is one that the authors of this paper have put forward. Previous work has suggested that each of these molecular subtypes may have a differential response to chemotherapy or systemic therapies of other types.
The goal of this paper was to investigate how molecular subtypes, as assessed based on TURBT specimens, impact the pathological response and survival for patients who receive chemotherapy prior to radical cystectomy. To assess this, the authors looked at patients undergoing preoperative cisplatin-based chemotherapy for muscle-invasive bladder cancer between 2004 and 2015 in two regions in Sweden. They included a chemotherapy-exposed cohort, which was predominantly patients treated with chemotherapy but included some patients with clinical node-positive or clinical T4b disease who received induction chemotherapy followed by consolidative surgery. A reference cohort who received a radical cystectomy alone was also accrued.
The authors assessed the primary outcome of pathological response rate, and they categorized this as a complete response, versus partial response, or no response. Partial response is defined as patients who were downstaged to non-muscle invasive disease. Pathologic downstaging was further considered as the difference between clinical and pathological stages and scored on an ordinal scale from 0 to 5. The difference of three steps or more was considered a major response to neoadjuvant therapy. The authors secondarily assessed cancer-specific survival and overall survival.
The authors reviewed TURBT specimens for all included patients using four expert uropathologists to determine T-stage, grade, the presence of lymphovascular invasion, keratinization, CIS, necrosis, or variant histology. Two 1 mm cores were taken from representative blocks for each tumor and arranged into tissue microarrays, which were subsequently stained using the Lund taxonomy IHC subtyping, as well as for anti-osteopontin. RNA was extracted from the formalin-embedded blocks, and the Affymetrix Gene ST 1.0 platform was used to undertake transcriptome-based analyses for these specimens.
In terms of statistical analysis, the authors use Chi-squared tests to compare the proportions of patients in each subtype who achieve pathologic complete response. They further used multivariable logistic regression models to adjust for the clinical T-stage to test the association between molecular subtype and pathologic complete response. Multivariable Cox proportional-hazards models were used to assess the association between these subtypes and survival endpoints, and ROC curves and AUCs were calculated for the ability of molecular subtypes to predict response to neoadjuvant chemotherapy.
At this point in time, I'm going to hand it over to Zach to walk us through the results.
Zachary Klaassen: Thanks, Chris. This is the flow chart showing the criteria for entry into the chemo cohort and the radical cystectomy cohort. And as you can see here on the left, there were 172 patients that had preoperative cisplatin-based chemotherapy. Ultimately, 124 of these were included in the neoadjuvant analysis, and 24 in the induction chemotherapy arm. Moving to the right, for radical cystectomy, there were 239 patients evaluated, and after exclusion criteria, 186 were included in the radical cystectomy analysis.
This is table 1 which looks at the clinical and pathological characteristics of the study cohorts. And as you can see here on the right, this is a busy table, I'll walk you through several key findings. On the far right is the RC cohort, to the left of that is the induction cohort, and to the left of that is the neoadjuvant cohort.
Looking at the ages of these patients stratified by these cohorts, the chemotherapy arm is slightly younger, 66 and 67 years of age, compared to 77 in the radical cystectomy cohort. Overall, more than three-quarters of patients were male. The most common T-stage for these patients was clinical T2. The most common node stage was N0 for the neoadjuvant cohort, and 99% of patients in the radical cystectomy cohort were clinical N0. For the induction cohort, the most common clinical node stage was clinical N1. Looking at variant histology, 17% in the neoadjuvant cohort, 12% in the induction cohort, and 9% in the radical cystectomy cohort. Going a little bit further down the table, we see the LundTax subtype based on RNA, and you can see a relatively balanced cohort amongst these different subtypes, which was also confirmed by the LundTax subtype IHC and the consensus subtype as well.
Looking at the type of chemotherapy received in terms of gemcitabine-cisplatin, 61% of patients in the neoadjuvant cohort and only 25% in the induction cohort. Dose-dense MVAC was more common in the induction cohort. In terms of chemotherapy, cycles received, in the neoadjuvant cohort, most commonly three cycles at 76%, and in the induction cohort, most commonly four cycles. Looking at the pathological T-stage, the most common pathological T-stage was pT0 in the neoadjuvant cohort, was pT2 in the induction cohort, and was pT3 in the radical cystectomy cohort.
This slide looks at the RNA-based molecular subtypes based on the LundTax classification, and you can see this is broken down by RNA subtype and IHC subtype. And you can see the subtypes here listed on the right, including UroA, B, and C, genomic unstable, basal/squamous, mesenchymal-like, as well as neuroendocrine. Below are just the representative images of the immunohistochemical staining, as well as the subtype-specific staining as well.
This figure looks at the pathologic responses stratified by molecular subtypes. We'll start on the left side. This is the pathologic response in the neoadjuvant cohort, and looking at the dark blue, this is the pathologic complete response, which was more common in the genomic unstable subtype and the second most common was in the urothelial-like subtype. Moving to the right, this is the pathologic response by subtype in the radical cystectomy cohort. As expected, less pathologic complete responses in this group, based on the fact that they did not have neoadjuvant chemo. And we can see that looking at the light blue, the pathologic MIBC or N-positive disease was most common in the basal/squamous subtype.
This is the figure looking at the number of steps of pathological downstaging by molecular subtypes. As Chris alluded to in the methods, a significant downstage was classified as 3 to 5, which is in the dark blue. And you can see that the genomic unstable, as well as the urothelial subtypes, had the most common or most frequent downstaging of 3 to 5 degrees.
This is the Kaplan-Meier curve for cancer-specific survival, stratified by the LundTax subtypes. And you can see here that, no surprise, the genomic unstable and the urothelial-like subtypes had the best cancer-specific survival, with the median not reached at over 5 years of follow-up. This is the Kaplan-Meier curves for recurrence-free survival stratified by the subtypes in the neoadjuvant and the radical cystectomy cohort. Again, we see the genomic unstable and the urothelial-like having the best RFS in the neoadjuvant cohort, and in the RC cohort, interestingly, we see that the squamous and neuroendocrine has the best RFS. However, this is a small number of patients with only 11 in this subtype.
This table looks at the multivariable Cox proportional hazards model for cancer-specific survival, with a reference of basal/squamous in the LundTax RNA subtype. You can see the significant improvement in cancer-specific survival for the urothelial C subtype, as well as the genomic unstable subtype. And no surprise here, with a reference to T2, patients with T3 and T4a disease had worse cancer-specific survival.
This table looks at the pathological responses stratified by cohorts. This is the neoadjuvant cohort in this study, as well as two outside cohorts, including the Seiler cohort and the Taber cohort. And I'll point your attention to the far right of the screen, looking at the composite response compared to the basal/squamous subtype. We see a significant benefit for urothelial A subtype, 54% composite response, which was statistically significant compared to the basal/squamous subtype, as well as the genomic unstable subtype, with a composite response of 53%, and again, statistically significant.
This slide looks at the boxplots of subtype-specific expression of SPP1 within the molecular subtypes, and you can see here that to summarize these figures, the expression of SPP1, which codes for osteopontin, is a biomarker for a response only within the non-urothelial subtype. So for instance, we see that in the basal/squamous in the chemo cohort, this predicted basal/squamous no pathological complete response, and we also see this signal in the GU or genomically unstable cohort, with no pathological complete response with a high SPP1 expression. And this was also seen in the statistical model here on the right.
So several discussion points from this study. This study showed an improved pathological response and survival outcomes for patients with GU LundTax molecular subtype, independent of tumor stage, following treatment with either neoadjuvant chemotherapy or radical cystectomy. And this was externally validated in other data sets as we described in the previous slides. Previous studies investigating the response to neoadjuvant chemotherapy using the different classification systems have not yielded consistent results, which has been somewhat confusing in the literature. And so the authors posed that this may be secondary to widely varied study designs, different survival endpoints, treatments, or selection endpoints, and unknown confounders that were unable to be controlled for. Finally, results in muscle-invasive bladder cancer may improve with second-generation biomarkers and those that operate only within a specific context defined by the molecular subtypes. And as we saw in this study, SPP1 in non-urothelial subtypes was able to be applied to the non-urothelial versions of the subtypes.
So in conclusion, this study shows that the existing molecular subtypes of muscle-invasive bladder cancer derived different clinical benefits from neoadjuvant chemotherapy. And ultimately, molecular classification and subtype-dependent second-generation biomarkers should be integrated into future prospective trials.
Thank you very much, and we hope you enjoyed this UroToday Journal Club discussion.