- Email: [email protected]
Is The Cancer Genome Atlas (TCGA) bladder cancer cohort representative of invasive bladder cancer?
Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
Original article
Is The Cancer Genome Atlas (TCGA) bladder cancer cohort representative of invasive bladder cancer? Roland Seiler, M.D.a,b,*, Peter C. Black, M.D.a, George Thalmann, M.D.b, Arnulf Stenzl, M.D.c, Tilman Todenhöfer, M.D.c a
Department of Urologic Sciences, University of British Columbia, Vancouver, British Columbia, Canada b Department of Urology, University of Bern, Bern, Switzerland c Department of Urology, University Hospital, Eberhard Karls University, Tübingen, Germany Received 4 October 2016; received in revised form 4 January 2017; accepted 30 January 2017
Abstract Purpose: The Cancer Genome Atlas (TCGA) Research Consortium has conducted a comprehensive molecular characterization of invasive bladder cancer (BCa). This open-access dataset has become the critical reference for studying biomarkers and mechanisms of disease in BCa. In order for this data to be considered representative, and to allow comparisons of markers between cohorts, clinicopathologic characteristics of this cohort need to conform to those established for this disease state. The aim of this study was to critically evaluate clinicopathologic characteristics and outcomes of the TCGA BCa cohort in comparison with published cystectomy series. Methods: Clinicopathologic parameters from the provisional TCGA BCa cohort were accessed. Descriptive statistics were performed in the background of widely cited cystectomy series. Results: The TCGA BCa cohort included a higher rate of patients with non–organ-confined disease (62%) and lymph node involvement (30%) compared to previous series. The 5-year overall survival was slightly lower when compared to standard cystectomy series (43%), but it was consistent with prior reports when stratified by tumor stage (69%, 48%, and 23% for pT r 2 pN0, pT 4 2 pN0, and any pT pNþ tumors, respectively). Importantly, established risk factors (pT 4 2 and pNþ) were confirmed as independent predictors of poor overall survival. Conclusions: The characteristics of the TCGA BCa cohort include a high proportion of advanced tumors, but outcomes in major subgroups show concordance with previous series. Therefore, molecular data from this cohort can be considered representative of invasive BCa and serve as a valuable resource to validate prognostic biomarkers. r 2017 Elsevier Inc. All rights reserved.
Keywords: TCGA; Bladder cancer; Outcome; Biomarkers; Prognostic factors
1. Introduction The comprehensive molecular characterization of urologic malignancies by The Cancer Genome Atlas (TCGA) and other groups is an important effort that has provided important insights into genomic and transcriptomic alterations of various tumor entities including invasive bladder cancer [1]. Moreover, these clinically annotated molecular * Corresponding author. Tel.: þ1-417-6438-1118; fax: þ1-413-16322181. E-mail address: [email protected] (R. Seiler).
http://dx.doi.org/10.1016/j.urolonc.2017.01.024 1078-1439/r 2017 Elsevier Inc. All rights reserved.
datasets are publically available and represent critical reference datasets for molecular research and the validation of biomarkers [2]. The use of the molecular data for both purposes, however, is facilitated if the datasets represent the target patient population. The cohort can be also used for validation of biomarkers that have been explored in cohorts with special features (e.g., metastatic tumors or patients treated with radiotherapy). However, cohorts with strongly differing characteristics require a meticulous matching of important variables such as age and tumor stage. The TCGA bladder data is only relevant to analyses of genomic
2
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
alterations or transcriptomic pathways in treatment-naïve invasive bladder cancer without distant metastasis if the TCGA patient cohort is representative of this patient population as reported in the literature. Similarly, many studies are now using the TCGA dataset to validate candidate biomarkers, but successful validation also depends on the clinicopathologic characteristics and outcomes of this validation cohort being similar to the wider bladder cancer population. The aim of this study was to analyze clinicopathologic characteristics and outcomes of the TCGA BCa cohort [1] in the context of previously published large single-center cystectomy series [3–5]. We hypothesize that these features of the TCGA BCa cohort are comparable to previous cystectomy cohorts and that the TCGA, therefore, represents a valuable resource for biomarker validation and the study of molecular pathways in bladder cancer. 2. Patients and Methods We downloaded the complete dataset of the bladder TCGA (n = 413) from cbioportal.org on November 11, 2016 and ported the file directly in the statistical platform (R Software Package, version 3.1.0). For 1 patient, no clinical data were available and 1 patient was recorded twice in the database. Patients with cM1 disease (n = 11) and those with neoadjuvant treatment before tissue harvesting (n = 10) were excluded from further analysis. Wilcoxon rank-sum test was used to compare continuous data between 2 groups. Kaplan-Meier plots were used to estimate disease-free survival (DFS) and overall survival (OS) from the date of surgery to the date of relapse or death, respectively. Patients free of recurrence or alive were censored at the date of last follow-up. Cox proportional hazards regression models were used to determine the effect of each variable on survival in both univariate and multivariate settings. For these survival analyses, only patients with available primary tumor and lymph node (LN) stage were included. Median follow-up was calculated using inverted Kaplan-Meier plots. All tests were conducted with type I error probability of 5%. From 3 previously published large single-center cystectomy series [3–5], variables such as age, gender, follow-up, OS, primary tumor, and LN stage were extracted and compared to the bladder TCGA in a descriptive manner.
3. Results 3.1. Baseline Patient Characteristics Median age of the 390 TCGA patients at surgery was 69 years (Table 1). Although the median follow-up was
Table 1 Characteristics, follow-up, and outcome of the provisional TCGA bladder cancer cohort (excluding 21 patients treated with neoadjuvant chemotherapy or evidence of distant metastasis). Patient data
(n ¼ 390)
Age, y (median, range) Female/male (n) Follow-up, median, mo Follow-up longer than 3 y (n) Disease-free survival (DFS) DFS status available 5-year DFS Overall survival (OS) OS status available 5-year OS Primary tumor stage (pT) Not available (NA) (n) pT r 2 (n) pT3 (n) pT4 (n) Lymphovascular invasion (no/yes) Lymph nodes Median no. of LN examined (range)
69 (34–90) 104/286 30 88/388 (22%)
Median no. of positive LN (range) Median no. of positive LN in LN-positive patients (range) pNx pN0 (%) pN1 (%) pN2–3 (%)
306/390 (78%) 42% 390/390 (100%) 43% 29 (7%) 120 (31%) 186 (48%) 55 (14%) 127/141 (122 NA) 18 (1–170) (101 NA) 0 (0–97) (106 NA) 2 (1–97) (13 NA) 38 (10%) 234 (60%) 45 (11%) 73 (19%)
30 months, the number of patients at risk at 5 years is low (Fig. 1 A and B). Outcome data for OS and DFS were available for 388 (99%) and 306 (78%) patients, respectively. Eighty-eight out of 388 patients (22%) were alive after 3 years and a follow-up longer than 3 years was available. Information on primary tumor and LN stage was available in 92% and 90% of patients, respectively. Complete data on both was available in 338 (87%) patients. Tumors with extravesical extension were found in 241/390 patients (62%): pT3 in 186 and pT4 in 55 patients. At least 1 positive LN was found in 118/390 (30%). 3.2. Survival Analysis The 5-year DFS and OS in the total cohort were 42% and 43%, respectively (Fig. 1 A and B). Survival analyses stratified by clinically important risk groups yielded expected results (Fig. 1 C and D, Table 2). Patients with organ-confined, LN-negative disease had the best outcome (5-DFS 74% and 5-OS 69%). Approximately half of patients with extravesical tumor extension but no LN metastasis survived for 5 years or longer (5-DFS 46% and 5-OS 48%). Patients with LN metastases had the worst outcome (5-DFS 27% and 5-OS 23%). When considering primary tumor and LN stage separately (Supplementary
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
3
Fig. 1. Kaplan-Meier plots for DFS (A) and OS (B) of the entire TCGA bladder cancer cohort. During a median follow-up of 30 months, 5-year DFS and OS probability was 42% and 43%, respectively. Kaplan-Meier plots stratified according to tumor stage for DFS (C) and OS (D). Patients without LN metastases and organ-confined primary tumors had the best outcome with 5-DFS and OS of 75% and 69%, respectively. Patients with LN involvement, irrespective of primary tumor stages had the worst outcome (5-DFS and OS probability was 27% and 23%, respectively). HR ¼ hazard ratio; RL ¼ reference level. (Color version of figure is available online.)
Figs. S1 and S2), both were significantly related to survival. Furthermore, patients with lymphovascular invasion had significantly shorter DFS and OS (Fig. 2 A and B, Table 2). Age was a risk factor for OS but not for DFS (Fig. 2 C and D, Table 2). Survival of female and male patients did not differ significantly. In multivariate analysis, extravesical tumor extension and LN metastasis were the most consistent risk factors for unfavorable outcome (Table 2). Patient age was independently related to OS but not to DFS. Presence of lymphovascular invasion failed to add independent prognostic information for DFS.
3.3. Effect of LN counts on outcome The number of LN examined pathologically at the time of radical cystectomy was available in 279 patients. A median of 18 LN were examined in LN-negative (range: 2–142) and LN-positive (range: 1–170) patients (Table 1). In LN-positive patients, a median of 2 (range: 1–97) LN were involved. Interestingly, the number of examined LN was related to patient outcomes (Fig. 3 A and B). Patients with a lower number of LN examined had shorter DFS and OS, whereas patients with the number of examined LN in the third quartile had the best outcome.
4
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
Table 2 Univariate and multivariate Cox proportional hazards models for risk factors known to be associated with disease-free survival (DFS) and overall survival (OS). Univariate survival analysis
Disease-free survival Sex Age pT3–4 pN1–3 LVI Overall survival Sex Age pT3–4 pN1–3 LVI
Disease-free survival Sex Age pT3–4 pN1–3 LVI Overall survival Sex Age pT3–4 pN1–3 LVI
RL
HR
P
Female rMedian pT r 2 pN0 Neg
0.84 1.13 2.81 2.43 2.35
0.4 0.53 o0.001 o0.001 o0.001
Female rMedian pT r 2 pN0 Neg Multivariate survival analysis RL
0.89 1.53 2.03 2.25 2.28
0.53 0.014 o0.001 o0.001 o0.001
HR
95% CI
P
Female rMedian pT r 2 pN0 Neg
0.77 1.16 2.36 1.78 1.54
0.51–1.16 0.79–1.7 1.44–3.89 1.16–2.74 0.93–2.56
0.2 0.46 o0.001 0.008 0.09
Female rMedian pT r 2 pN0 Neg
0.9 1.67 1.65 1.81 1.61
0.62–1.27 1.2–2.33 1.1–2.51 1.24–2.64 1.02–2.55
0.5 0.003 0.02 0.002 0.041
HR ¼ hazard ratio; LVI ¼ lymphovascular invasion; RL ¼ reference level. Bold text indicates a statistical significance (p o 0.05).
4. Discussion We analyzed clinical characteristics of the TCGA BCa cohort and observed an enrichment of patient with more advanced primary tumor stages; however, similar outcomes were seen in important subgroups when compared with prior large standard cystectomy series [3–5]. Importantly, previously identified risk factors for poor outcome are confirmed in the TCGA BCa cohort and strengthen the applicability of this cohort as a discovery and validation set for prognostic biomarkers. In the TCGA BCa cohort, the 5-year OS of the entire cohort (43%) is inferior when compared to standard cystectomy series (5-year OS 55%–60%) [3–5] (Table 3). This can be explained by the higher rate of extravesical extension in the TCGA BCa cohort compared to previous series (62% vs. 26%–52%) (Table 3). We can speculate that this enrichment may be due to the need to have adequate tumor bulk to allow successful biobanking. Consistent with this, the rate of patients with positive LN is also higher (30%) in the TCGA cohort when compared to standard cystectomy series (20%–24%) (Table 3). However, extended lymphadenectomy which is known to lead to higher rates of positive LN may have influenced these
results (Dhar et al. [6]). Unfortunately, information on surgical margins is not provided. Sex and median age at surgery were very similar to published series. The outcomes of clinically important subgroups of patients in the TCGA BCa cohort and large cystectomy series are virtually the same [3–5]. The 5-year OS of LN-negative patients with organ-confined disease and extravesical extension of the primary tumor was 470% and 50%, respectively. Both in the TCGA BCa cohort and previous datasets, the worst prognosis was observed in patients with LN involvement (5-year OS of 30% and lower). Taken together, the outcome of the entire TCGA BCa cohort is inferior to reference cystectomy series, but the outcome of specific subgroups is highly concordant with reference cystectomy series. The prognostic significance of parameters in univariate and multivariate analysis is in line with previous studies [3–5]. Extravesical extension and LN involvement stratified patient outcome independently. In addition, a high patient age was associated with short OS but not DSS, which corresponds with expectations. Requirements for the extent of pelvic lymph node dissection (PLND) appear not to have been specified for inclusion of patients in the TCGA cohort. Since more than
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
5
Fig. 2. Kaplan-Meier plots of TCGA bladder cancer cohort for DFS (left) and OS (right) stratified according to lymphovascular invasion (LVI) (A and B) and age dichotomized according to the median of the entire cohort (C and D). Presence of LVI was significantly associated with shorter DFS and OS. Age above the median was significantly associated with shorter OS but not DFS. (Color version of figure is available online.)
15 US centers were involved in sample collection, it cannot be expected that PLND was performed uniformly by all surgeons. This hypothesis is underlined by the significant association observed between the number of LN examined and patient outcomes. Previous studies showed that numbers of LN examined are not associated with survival, provided the PLND is performed uniformly [5,7,8]. However, the inhomogeneous PLND in the TCGA BCa cohort is likely of minor effect. A significant understaging of LNnegative patients would affect outcome in this patient subset, but the outcomes in the TCGA cohort were virtually the same as in series with a homogeneous, meticulous, and extended PLND [3–5].
An important limitation of the cohort is the absence of data on surgical resection margin status. Previous reports have established that this parameter has a significant negative effect on cancer-related outcome [9,10]. The follow-up of the TCGA BCa cohort is also relatively short, which affects survival analyses. Furthermore, an unknown proportion of patients might have received neoadjuvant chemotherapy before subsequent radical cystectomy but after sample acquisition from transurethral tumor resection. Although the molecular characteristics of the tissue are not affected by chemotherapy, the subsequent clinical course of the patients could be significantly affected. Administration of adjuvant chemotherapy is captured in the TCGA.
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
6
Fig. 3. Kaplan-Meier plots of TCGA bladder cancer cohort for DFS (A) and OS (OS) according to number of LN examined. Patients were categorized into quartiles of LN examined. Outcome of patients who had a lower number of LN examined (first and second quartile) was worse compared to patients with more LN removed (third and fourth quartile). (Color version of figure is available online.)
The fact, that the 3 previous cystectomy series show significant overlaps regarding their clinical features whereas the current series shows higher rates of advanced disease with worse outcome does not necessarily mean that the current cohort is less representative of invasive BC. We cannot exclude that the retrospective analysis of the 3 large cystectomy series introduced a bias that is not present in the TCGA cohort that has been collected prospectively.
5. Conclusion The TCGA provides an invaluable source for the assessment of molecular alterations and investigation of prognostic biomarkers in bladder cancer. The patients captured in the TCGA BCa cohort have higher risk disease than most reference cystectomy series, and therefore have lower rates of OS and disease-specific survival, but these outcomes are equivalent to prior series when adjusting for the risk of
Table 3 Characteristics of 3 large published cystectomy series.
No. of patients Median (range) age, y Female/male (n) Median follow-up, mo Overall survival (OS) 5-year OS Primary tumor stage (pT) rpT2 (n) pT3 (n) pT4 (n) Lymph node stage (pN) pN0 (n) pN1–3 (n) a
Stein et al.
Madersbacher et al.
Ghoneim et al.
TCGA
1,054 66 (22–93) 211/843 122
507 66 (35–89) 107/400 31
2,720 49 (n.a.) 630/2,090 43
390 69 (34–90) 104/286 30
60%
59%
56% (CSS)
43%
536 (51%) 381 (36%) 137 (13%)
245 (48%) 184 (36%) 78 (16%)
2,023 (74%) 453 (17%) 244 (9%)
120 (31%)a 186 (48%) 55 (14%)
808 (77%) 246 (23%)
383 (76%) 124 (24%)
2,165 (80%) 555 (20%)
234 (60%)b 118 (30%)
29/390 (7%) patients with not available pT stage. 38/390 (10%) patients with not available pN stage.
b
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
disease. Researchers can, therefore, be encouraged to use the TCGA BCa dataset for evaluation of prognostic biomarkers. Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j. urolonc.2017.01.024.
References [1] Cancer Genome Atlas Research Network: Comprehensive molecular characterization of urothelial bladder carcinoma. Nature 2014;507:315–22. [2] Kapur P, Pena-Llopis S, Christie A, et al. Effects on survival of BAP1 and PBRM1 mutations in sporadic clear-cell renal-cell carcinoma: a retrospective analysis with independent validation. Lancet Oncol 2013;14:159–67. [3] Ghoneim MA, Abdel-Latif M, el-Mekresh M, et al. Radical cystectomy for carcinoma of the bladder: 2,720 consecutive cases 5 years later. J Urol 2008;180:121–7.
7
[4] Madersbacher S, Hochreiter W, Burkhard F, et al. Radical cystectomy for bladder cancer today—a homogeneous series without neoadjuvant therapy. J Clin Oncol 2003;21:690–6. [5] Stein JP, Lieskovsky G, Cote R, et al. Radical cystectomy in the treatment of invasive bladder cancer: long-term results in 1,054 patients. J Clin Oncol 2001;19:666–75. [6] Dhar NB, Klein EA, Reuther AM, Thalmann GN, Madersbacher S, Studer UE. Outcome after radical cystectomy with limited or extended pelvic lymph node dissection. J Urol 2008;179:873–8. [7] Fleischmann A, Thalmann GN, Markwalder R, et al. Extracapsular extension of pelvic lymph node metastases from urothelial carcinoma of the bladder is an independent prognostic factor. J Clin Oncol 2005;23:2358–65. [8] Zehnder P, Studer UE, Skinner EC, et al. Super extended versus extended pelvic lymph node dissection in patients undergoing radical cystectomy for bladder cancer: a comparative study. J Urol 2011;186:1261–8. [9] Neuzillet Y, Soulie M, Larre S, et al. Positive surgical margins and their locations in specimens are adverse prognosis features after radical cystectomy in non-metastatic carcinoma invading bladder muscle: results from a nationwide case-control study. BJU Int 2013;111:1253–60. [10] Novara G, Svatek RS, Karakiewicz PI, et al. Soft tissue surgical margin status is a powerful predictor of outcomes after radical cystectomy: a multicenter study of more than 4,400 patients. J Urol 2010;183:2165–70.
Original article
Is The Cancer Genome Atlas (TCGA) bladder cancer cohort representative of invasive bladder cancer? Roland Seiler, M.D.a,b,*, Peter C. Black, M.D.a, George Thalmann, M.D.b, Arnulf Stenzl, M.D.c, Tilman Todenhöfer, M.D.c a
Department of Urologic Sciences, University of British Columbia, Vancouver, British Columbia, Canada b Department of Urology, University of Bern, Bern, Switzerland c Department of Urology, University Hospital, Eberhard Karls University, Tübingen, Germany Received 4 October 2016; received in revised form 4 January 2017; accepted 30 January 2017
Abstract Purpose: The Cancer Genome Atlas (TCGA) Research Consortium has conducted a comprehensive molecular characterization of invasive bladder cancer (BCa). This open-access dataset has become the critical reference for studying biomarkers and mechanisms of disease in BCa. In order for this data to be considered representative, and to allow comparisons of markers between cohorts, clinicopathologic characteristics of this cohort need to conform to those established for this disease state. The aim of this study was to critically evaluate clinicopathologic characteristics and outcomes of the TCGA BCa cohort in comparison with published cystectomy series. Methods: Clinicopathologic parameters from the provisional TCGA BCa cohort were accessed. Descriptive statistics were performed in the background of widely cited cystectomy series. Results: The TCGA BCa cohort included a higher rate of patients with non–organ-confined disease (62%) and lymph node involvement (30%) compared to previous series. The 5-year overall survival was slightly lower when compared to standard cystectomy series (43%), but it was consistent with prior reports when stratified by tumor stage (69%, 48%, and 23% for pT r 2 pN0, pT 4 2 pN0, and any pT pNþ tumors, respectively). Importantly, established risk factors (pT 4 2 and pNþ) were confirmed as independent predictors of poor overall survival. Conclusions: The characteristics of the TCGA BCa cohort include a high proportion of advanced tumors, but outcomes in major subgroups show concordance with previous series. Therefore, molecular data from this cohort can be considered representative of invasive BCa and serve as a valuable resource to validate prognostic biomarkers. r 2017 Elsevier Inc. All rights reserved.
Keywords: TCGA; Bladder cancer; Outcome; Biomarkers; Prognostic factors
1. Introduction The comprehensive molecular characterization of urologic malignancies by The Cancer Genome Atlas (TCGA) and other groups is an important effort that has provided important insights into genomic and transcriptomic alterations of various tumor entities including invasive bladder cancer [1]. Moreover, these clinically annotated molecular * Corresponding author. Tel.: þ1-417-6438-1118; fax: þ1-413-16322181. E-mail address: [email protected] (R. Seiler).
http://dx.doi.org/10.1016/j.urolonc.2017.01.024 1078-1439/r 2017 Elsevier Inc. All rights reserved.
datasets are publically available and represent critical reference datasets for molecular research and the validation of biomarkers [2]. The use of the molecular data for both purposes, however, is facilitated if the datasets represent the target patient population. The cohort can be also used for validation of biomarkers that have been explored in cohorts with special features (e.g., metastatic tumors or patients treated with radiotherapy). However, cohorts with strongly differing characteristics require a meticulous matching of important variables such as age and tumor stage. The TCGA bladder data is only relevant to analyses of genomic
2
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
alterations or transcriptomic pathways in treatment-naïve invasive bladder cancer without distant metastasis if the TCGA patient cohort is representative of this patient population as reported in the literature. Similarly, many studies are now using the TCGA dataset to validate candidate biomarkers, but successful validation also depends on the clinicopathologic characteristics and outcomes of this validation cohort being similar to the wider bladder cancer population. The aim of this study was to analyze clinicopathologic characteristics and outcomes of the TCGA BCa cohort [1] in the context of previously published large single-center cystectomy series [3–5]. We hypothesize that these features of the TCGA BCa cohort are comparable to previous cystectomy cohorts and that the TCGA, therefore, represents a valuable resource for biomarker validation and the study of molecular pathways in bladder cancer. 2. Patients and Methods We downloaded the complete dataset of the bladder TCGA (n = 413) from cbioportal.org on November 11, 2016 and ported the file directly in the statistical platform (R Software Package, version 3.1.0). For 1 patient, no clinical data were available and 1 patient was recorded twice in the database. Patients with cM1 disease (n = 11) and those with neoadjuvant treatment before tissue harvesting (n = 10) were excluded from further analysis. Wilcoxon rank-sum test was used to compare continuous data between 2 groups. Kaplan-Meier plots were used to estimate disease-free survival (DFS) and overall survival (OS) from the date of surgery to the date of relapse or death, respectively. Patients free of recurrence or alive were censored at the date of last follow-up. Cox proportional hazards regression models were used to determine the effect of each variable on survival in both univariate and multivariate settings. For these survival analyses, only patients with available primary tumor and lymph node (LN) stage were included. Median follow-up was calculated using inverted Kaplan-Meier plots. All tests were conducted with type I error probability of 5%. From 3 previously published large single-center cystectomy series [3–5], variables such as age, gender, follow-up, OS, primary tumor, and LN stage were extracted and compared to the bladder TCGA in a descriptive manner.
3. Results 3.1. Baseline Patient Characteristics Median age of the 390 TCGA patients at surgery was 69 years (Table 1). Although the median follow-up was
Table 1 Characteristics, follow-up, and outcome of the provisional TCGA bladder cancer cohort (excluding 21 patients treated with neoadjuvant chemotherapy or evidence of distant metastasis). Patient data
(n ¼ 390)
Age, y (median, range) Female/male (n) Follow-up, median, mo Follow-up longer than 3 y (n) Disease-free survival (DFS) DFS status available 5-year DFS Overall survival (OS) OS status available 5-year OS Primary tumor stage (pT) Not available (NA) (n) pT r 2 (n) pT3 (n) pT4 (n) Lymphovascular invasion (no/yes) Lymph nodes Median no. of LN examined (range)
69 (34–90) 104/286 30 88/388 (22%)
Median no. of positive LN (range) Median no. of positive LN in LN-positive patients (range) pNx pN0 (%) pN1 (%) pN2–3 (%)
306/390 (78%) 42% 390/390 (100%) 43% 29 (7%) 120 (31%) 186 (48%) 55 (14%) 127/141 (122 NA) 18 (1–170) (101 NA) 0 (0–97) (106 NA) 2 (1–97) (13 NA) 38 (10%) 234 (60%) 45 (11%) 73 (19%)
30 months, the number of patients at risk at 5 years is low (Fig. 1 A and B). Outcome data for OS and DFS were available for 388 (99%) and 306 (78%) patients, respectively. Eighty-eight out of 388 patients (22%) were alive after 3 years and a follow-up longer than 3 years was available. Information on primary tumor and LN stage was available in 92% and 90% of patients, respectively. Complete data on both was available in 338 (87%) patients. Tumors with extravesical extension were found in 241/390 patients (62%): pT3 in 186 and pT4 in 55 patients. At least 1 positive LN was found in 118/390 (30%). 3.2. Survival Analysis The 5-year DFS and OS in the total cohort were 42% and 43%, respectively (Fig. 1 A and B). Survival analyses stratified by clinically important risk groups yielded expected results (Fig. 1 C and D, Table 2). Patients with organ-confined, LN-negative disease had the best outcome (5-DFS 74% and 5-OS 69%). Approximately half of patients with extravesical tumor extension but no LN metastasis survived for 5 years or longer (5-DFS 46% and 5-OS 48%). Patients with LN metastases had the worst outcome (5-DFS 27% and 5-OS 23%). When considering primary tumor and LN stage separately (Supplementary
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
3
Fig. 1. Kaplan-Meier plots for DFS (A) and OS (B) of the entire TCGA bladder cancer cohort. During a median follow-up of 30 months, 5-year DFS and OS probability was 42% and 43%, respectively. Kaplan-Meier plots stratified according to tumor stage for DFS (C) and OS (D). Patients without LN metastases and organ-confined primary tumors had the best outcome with 5-DFS and OS of 75% and 69%, respectively. Patients with LN involvement, irrespective of primary tumor stages had the worst outcome (5-DFS and OS probability was 27% and 23%, respectively). HR ¼ hazard ratio; RL ¼ reference level. (Color version of figure is available online.)
Figs. S1 and S2), both were significantly related to survival. Furthermore, patients with lymphovascular invasion had significantly shorter DFS and OS (Fig. 2 A and B, Table 2). Age was a risk factor for OS but not for DFS (Fig. 2 C and D, Table 2). Survival of female and male patients did not differ significantly. In multivariate analysis, extravesical tumor extension and LN metastasis were the most consistent risk factors for unfavorable outcome (Table 2). Patient age was independently related to OS but not to DFS. Presence of lymphovascular invasion failed to add independent prognostic information for DFS.
3.3. Effect of LN counts on outcome The number of LN examined pathologically at the time of radical cystectomy was available in 279 patients. A median of 18 LN were examined in LN-negative (range: 2–142) and LN-positive (range: 1–170) patients (Table 1). In LN-positive patients, a median of 2 (range: 1–97) LN were involved. Interestingly, the number of examined LN was related to patient outcomes (Fig. 3 A and B). Patients with a lower number of LN examined had shorter DFS and OS, whereas patients with the number of examined LN in the third quartile had the best outcome.
4
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
Table 2 Univariate and multivariate Cox proportional hazards models for risk factors known to be associated with disease-free survival (DFS) and overall survival (OS). Univariate survival analysis
Disease-free survival Sex Age pT3–4 pN1–3 LVI Overall survival Sex Age pT3–4 pN1–3 LVI
Disease-free survival Sex Age pT3–4 pN1–3 LVI Overall survival Sex Age pT3–4 pN1–3 LVI
RL
HR
P
Female rMedian pT r 2 pN0 Neg
0.84 1.13 2.81 2.43 2.35
0.4 0.53 o0.001 o0.001 o0.001
Female rMedian pT r 2 pN0 Neg Multivariate survival analysis RL
0.89 1.53 2.03 2.25 2.28
0.53 0.014 o0.001 o0.001 o0.001
HR
95% CI
P
Female rMedian pT r 2 pN0 Neg
0.77 1.16 2.36 1.78 1.54
0.51–1.16 0.79–1.7 1.44–3.89 1.16–2.74 0.93–2.56
0.2 0.46 o0.001 0.008 0.09
Female rMedian pT r 2 pN0 Neg
0.9 1.67 1.65 1.81 1.61
0.62–1.27 1.2–2.33 1.1–2.51 1.24–2.64 1.02–2.55
0.5 0.003 0.02 0.002 0.041
HR ¼ hazard ratio; LVI ¼ lymphovascular invasion; RL ¼ reference level. Bold text indicates a statistical significance (p o 0.05).
4. Discussion We analyzed clinical characteristics of the TCGA BCa cohort and observed an enrichment of patient with more advanced primary tumor stages; however, similar outcomes were seen in important subgroups when compared with prior large standard cystectomy series [3–5]. Importantly, previously identified risk factors for poor outcome are confirmed in the TCGA BCa cohort and strengthen the applicability of this cohort as a discovery and validation set for prognostic biomarkers. In the TCGA BCa cohort, the 5-year OS of the entire cohort (43%) is inferior when compared to standard cystectomy series (5-year OS 55%–60%) [3–5] (Table 3). This can be explained by the higher rate of extravesical extension in the TCGA BCa cohort compared to previous series (62% vs. 26%–52%) (Table 3). We can speculate that this enrichment may be due to the need to have adequate tumor bulk to allow successful biobanking. Consistent with this, the rate of patients with positive LN is also higher (30%) in the TCGA cohort when compared to standard cystectomy series (20%–24%) (Table 3). However, extended lymphadenectomy which is known to lead to higher rates of positive LN may have influenced these
results (Dhar et al. [6]). Unfortunately, information on surgical margins is not provided. Sex and median age at surgery were very similar to published series. The outcomes of clinically important subgroups of patients in the TCGA BCa cohort and large cystectomy series are virtually the same [3–5]. The 5-year OS of LN-negative patients with organ-confined disease and extravesical extension of the primary tumor was 470% and 50%, respectively. Both in the TCGA BCa cohort and previous datasets, the worst prognosis was observed in patients with LN involvement (5-year OS of 30% and lower). Taken together, the outcome of the entire TCGA BCa cohort is inferior to reference cystectomy series, but the outcome of specific subgroups is highly concordant with reference cystectomy series. The prognostic significance of parameters in univariate and multivariate analysis is in line with previous studies [3–5]. Extravesical extension and LN involvement stratified patient outcome independently. In addition, a high patient age was associated with short OS but not DSS, which corresponds with expectations. Requirements for the extent of pelvic lymph node dissection (PLND) appear not to have been specified for inclusion of patients in the TCGA cohort. Since more than
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
5
Fig. 2. Kaplan-Meier plots of TCGA bladder cancer cohort for DFS (left) and OS (right) stratified according to lymphovascular invasion (LVI) (A and B) and age dichotomized according to the median of the entire cohort (C and D). Presence of LVI was significantly associated with shorter DFS and OS. Age above the median was significantly associated with shorter OS but not DFS. (Color version of figure is available online.)
15 US centers were involved in sample collection, it cannot be expected that PLND was performed uniformly by all surgeons. This hypothesis is underlined by the significant association observed between the number of LN examined and patient outcomes. Previous studies showed that numbers of LN examined are not associated with survival, provided the PLND is performed uniformly [5,7,8]. However, the inhomogeneous PLND in the TCGA BCa cohort is likely of minor effect. A significant understaging of LNnegative patients would affect outcome in this patient subset, but the outcomes in the TCGA cohort were virtually the same as in series with a homogeneous, meticulous, and extended PLND [3–5].
An important limitation of the cohort is the absence of data on surgical resection margin status. Previous reports have established that this parameter has a significant negative effect on cancer-related outcome [9,10]. The follow-up of the TCGA BCa cohort is also relatively short, which affects survival analyses. Furthermore, an unknown proportion of patients might have received neoadjuvant chemotherapy before subsequent radical cystectomy but after sample acquisition from transurethral tumor resection. Although the molecular characteristics of the tissue are not affected by chemotherapy, the subsequent clinical course of the patients could be significantly affected. Administration of adjuvant chemotherapy is captured in the TCGA.
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
6
Fig. 3. Kaplan-Meier plots of TCGA bladder cancer cohort for DFS (A) and OS (OS) according to number of LN examined. Patients were categorized into quartiles of LN examined. Outcome of patients who had a lower number of LN examined (first and second quartile) was worse compared to patients with more LN removed (third and fourth quartile). (Color version of figure is available online.)
The fact, that the 3 previous cystectomy series show significant overlaps regarding their clinical features whereas the current series shows higher rates of advanced disease with worse outcome does not necessarily mean that the current cohort is less representative of invasive BC. We cannot exclude that the retrospective analysis of the 3 large cystectomy series introduced a bias that is not present in the TCGA cohort that has been collected prospectively.
5. Conclusion The TCGA provides an invaluable source for the assessment of molecular alterations and investigation of prognostic biomarkers in bladder cancer. The patients captured in the TCGA BCa cohort have higher risk disease than most reference cystectomy series, and therefore have lower rates of OS and disease-specific survival, but these outcomes are equivalent to prior series when adjusting for the risk of
Table 3 Characteristics of 3 large published cystectomy series.
No. of patients Median (range) age, y Female/male (n) Median follow-up, mo Overall survival (OS) 5-year OS Primary tumor stage (pT) rpT2 (n) pT3 (n) pT4 (n) Lymph node stage (pN) pN0 (n) pN1–3 (n) a
Stein et al.
Madersbacher et al.
Ghoneim et al.
TCGA
1,054 66 (22–93) 211/843 122
507 66 (35–89) 107/400 31
2,720 49 (n.a.) 630/2,090 43
390 69 (34–90) 104/286 30
60%
59%
56% (CSS)
43%
536 (51%) 381 (36%) 137 (13%)
245 (48%) 184 (36%) 78 (16%)
2,023 (74%) 453 (17%) 244 (9%)
120 (31%)a 186 (48%) 55 (14%)
808 (77%) 246 (23%)
383 (76%) 124 (24%)
2,165 (80%) 555 (20%)
234 (60%)b 118 (30%)
29/390 (7%) patients with not available pT stage. 38/390 (10%) patients with not available pN stage.
b
R. Seiler et al. / Urologic Oncology: Seminars and Original Investigations ] (2017) ∎∎∎–∎∎∎
disease. Researchers can, therefore, be encouraged to use the TCGA BCa dataset for evaluation of prognostic biomarkers. Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j. urolonc.2017.01.024.
References [1] Cancer Genome Atlas Research Network: Comprehensive molecular characterization of urothelial bladder carcinoma. Nature 2014;507:315–22. [2] Kapur P, Pena-Llopis S, Christie A, et al. Effects on survival of BAP1 and PBRM1 mutations in sporadic clear-cell renal-cell carcinoma: a retrospective analysis with independent validation. Lancet Oncol 2013;14:159–67. [3] Ghoneim MA, Abdel-Latif M, el-Mekresh M, et al. Radical cystectomy for carcinoma of the bladder: 2,720 consecutive cases 5 years later. J Urol 2008;180:121–7.
7
[4] Madersbacher S, Hochreiter W, Burkhard F, et al. Radical cystectomy for bladder cancer today—a homogeneous series without neoadjuvant therapy. J Clin Oncol 2003;21:690–6. [5] Stein JP, Lieskovsky G, Cote R, et al. Radical cystectomy in the treatment of invasive bladder cancer: long-term results in 1,054 patients. J Clin Oncol 2001;19:666–75. [6] Dhar NB, Klein EA, Reuther AM, Thalmann GN, Madersbacher S, Studer UE. Outcome after radical cystectomy with limited or extended pelvic lymph node dissection. J Urol 2008;179:873–8. [7] Fleischmann A, Thalmann GN, Markwalder R, et al. Extracapsular extension of pelvic lymph node metastases from urothelial carcinoma of the bladder is an independent prognostic factor. J Clin Oncol 2005;23:2358–65. [8] Zehnder P, Studer UE, Skinner EC, et al. Super extended versus extended pelvic lymph node dissection in patients undergoing radical cystectomy for bladder cancer: a comparative study. J Urol 2011;186:1261–8. [9] Neuzillet Y, Soulie M, Larre S, et al. Positive surgical margins and their locations in specimens are adverse prognosis features after radical cystectomy in non-metastatic carcinoma invading bladder muscle: results from a nationwide case-control study. BJU Int 2013;111:1253–60. [10] Novara G, Svatek RS, Karakiewicz PI, et al. Soft tissue surgical margin status is a powerful predictor of outcomes after radical cystectomy: a multicenter study of more than 4,400 patients. J Urol 2010;183:2165–70.