In this population-based study, we aimed to report real-world cancer-specific mortality of different NMIBC stages and develop a prognostic nomogram for CSS in NMIBC patients.
Methods
The National Cancer Institute Surveillance, Epidemiology, and End Results (SEER) database, based on 17 different registries, was used for this research. A systematic search was performed to identify patients with NMIBC according to the 6th Derived American Joint Committee on Cancer (AJCC) T edition (ICD codes from C67.0 to C67.9 with the exclusion of C67.7). In the subsequent selection, only patients treated with TURBT between 2004 and 2015 were included in the analysis. Exclusion criteria were as follows: any evidence of lymph node involvement or distant metastases, previous history of MIBC, lack of histopathological confirmation of BC, discrepancy in reporting tumor stage and grade, previous radiotherapy for BC, missing tumor grading or T category, unknown survival status.
The available data included information on the patient’s demographics, tumor histopathological (e.g., T category, grading) and clinical characteristics (e.g., tumor size, location, history of previous recurrence), therapy used (e.g., surgery type, chemotherapy), and survival outcomes (cancer-specific and all-cause mortality during the follow-up).
Ethics
The institutional review board waived the need for study approval. The study was performed in accordance with the Declaration of Helsinki and its later amendments.
Statistical Analysis
The dataset was divided into development and validation cohorts in the proportion of 70 to 30. Group characteristics are presented as the number of patients and percentage for categorical variables. Median survival follow-up was computed using the reverse Kaplan-Meier method.
Cox proportional hazard regression was performed to identify factors prognostic for CSS and overall survival (OS). Independent risk factors used in the multivariable analysis were used to create a prognostic nomogram for 5-year CSS. A hazard ratio (HR) supplemented with a 95% confidence interval (95% CI) was derived. For all analyses, we considered a two-sided p value < 0.05 as statistically significant. Calibration plots were generated to illustrate the nomogram-predicted and actual outcomes. The concordance index (C-index) was calculated for development and validation cohorts to demonstrate the discrimination of the nomogram. The area under the receiver operating characteristic curve (AUC) was calculated to measure the accuracy. Furthermore, decision curve analysis (DCA) was performed to illustrate the net benefit for the nomogram. Statistical analyses were performed using SAS software version 9.4 (Cary, North Carolina, U.S.) and R studio using R programming language version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria) with ‘rms’, ‘gtsummary,’ and ‘dcurves’ packages.
Discussion
In this population-based study, we compared the prognosis of different NMIBC stages and developed a prognostic nomogram for CSS of NMIBC patients. Firstly, we identified tumor T category, histological grading, and patient’s age as the strongest predictors of CSS. We confirmed that among NMIBC, T1HG and Tis are characterized by the most unfavorable disease-specific survival (up to 19% of patients die due to BC). Rarely diagnosed T1LG and TaHG are also associated with a non-negligible risk of CSM (approximately 10%), whereas TaLG is associated with favorable survival. Secondly, we constructed a prognostic nomogram for 5-year CSS, which included tumor T category and grade, age, tumor size, histology type, tumor location, the primary character of the tumor, patient’s race, marital status, and socioeconomic status. Thirdly, internal validation of the nomogram confirmed its acceptably high accuracy, discrimination abilities, and good calibration. Fourthly, the clinical net benefit seems to be higher for our nomogram compared with conventionally used tumor T stage and grade in predicting 5-year CSS. Fifthly, the significant burden of other-cause-specific mortality among all stages of NMIBC emphasizes the importance of considering comorbidities during clinical decision-making.
Our data underscore the need for aggressive treatment for T1HG and Tis, but also careful management of “forgotten” and relatively rare TaHG and T1LG which are also deadly tumors (up to 10% CSM). T1HG is widely considered the most aggressive type of NMIBC with high progression rates and relatively poor survival prognosis (approximately 10% 5-year CSM). In our population-based study, T1HG was characterized by 81.5% CSS at the median follow-up of 124 months. This is rather consistent with the CSS reported in other studies summarized in the available systematic review (from 78% to 93% CSS).
9 Because of such unfavorable outcomes, the debate on the indication for radical cystectomy (RC) in T1HG should be continued. In the treatment of NMIBC, RC has its established role in the BCG-unresponsive disease and tumors not amenable to complete transurethral resection. A recent systematic review of 10 studies demonstrated better 5- and 10-year CSS results for T1G3 patients who underwent early compared with deferred cystectomy.
10 On the other hand, there is still no convincing evidence of the survival benefit from RC compared with conservative management with TURBT and adjuvant BCG in T1HG patients.
9 The randomized controlled BRAVO-feasibility trial included a small cohort of patients who received either RC or a full course of BCG. None of the patients (0/20 patients) who underwent RC had developed metastatic progression at the study closure, compared with two patients with metastatic BC in the conservative management arm (2/25 patients).
11 Results of large cohort randomized controlled trials are awaited, but accrual for such trials is poor. The prognosis of T1HG seems to improve over the years. In the historic, classical study, Shahin et al. suggested the rule of 30%; namely, 30% of patients never have a recurrence, 30% undergo deferred cystectomy and 30% die because of metastatic progression.
12,13 On the other hand, competing risk of other-cause-specific mortality in patients with T1HG raises a concern about overtreatment when offering RC. As we demonstrated, T1HG patients have worse non-BC-specific survival than patients with TaLG. Perhaps, higher exposure to carcinogenic factors (e.g., smoking) leads to the development of more aggressive cancer, and simultaneously increases the risk of cardiovascular diseases and secondary malignancies contributing to non-BC mortality. The presence of more aggressive cancer in more comorbid patients poses a clinical dilemma as to the optimal therapy. Consequently, improved risk-stratification and taking comorbidities into consideration is obligatory to avoid over- and undertreatment in T1HG BC.
Our study indicates that TaHG is associated with nearly twice lower, but still substantial disease-specific mortality when compared to T1HG BC. A study by Quhal et al. showed that TaHG is associated with a substantially lower risk of progression than T1HG.
14 Nevertheless, Bree et al. concluded that all TaHG tumors should be regarded as high-risk as they confer a significantly higher progression risk (approximately 6% TaHG progress to ≥ T2) than intermediate-risk low-grade Ta when treated with BCG.
15 We believe that in light of this real-world data, TaHG should be always regarded as high-risk due to the substantial 9.2% CSM in the long-term observation.
Furthermore, T1LG is another rarely reported tumor category, as submucosa invasive tumors are suspected to consist of histologically undifferentiated cells able to infiltrate, which are classified as grade 3 according to WHO 1973 or high-grade according to WHO 2004 grading systems, respectively. A study by van de Putte et al., in which experienced pathologists re-revised T1 grading in 601 patients, yielded no grade 1 tumors but reported 188 (31%) grade 2 tumors and 413 (69%) grade 3 lesions according to the WHO 1973 classification.
6 In the same study, according to the new WHO 2004 classification, 47 (6%) low-grade and 554 (94%) high-grade T1 tumors were diagnosed. In the aforementioned study, grade 3, but not high-grade constituted an independent prognostic factor for progression and CSS.
6 Noteworthy, Beijert et al. performed a large multicenter retrospective study in which T1G1 was identified as rare (1.9% of all NMIBC), but with PFS similar to T1G2 and worse than TaG1.
7 Available evidence and our real-world data suggest that despite their invasive character, T1 tumors might be low-grade, and this is independently associated with better prognosis. Nevertheless, T1LG should always be stratified into the high-risk group due to the significant risk of CSM.
Our nomogram underlines the crucial prognostic role of T category, grading, and age which are cardinal factors determining CSS. Moreover, we report the additional prognostic role of other tumor features and sociodemographic factors. So far, the EORTC 2016 nomogram has been the most commonly used for survival assessment of NMIBC. EORTC 2016 included T category and grade as risk factors for disease-specific survival; age and grade for OS.
3 However, the above risk stratifications were validated only for intermediate- and high-risk patients treated with maintenance BCG therapy.
3 Conversely, our nomogram is constructed for all NMIBCs and includes several more risk factors. Internal validation, decision-curve, and calibration analyses confirm its applicability in that setting. DCA confirms the advantage of our nomogram compared with conventional tumor T category and grade, which were proposed by Cambier et al. as risk factors in the EORTC 2016 nomogram.
We showed the impact of age not only on OS but also CSS. Age has been consistently reported as an adverse prognostic factor in other nomograms such as CUETO and EORTC 2016.
2,3 Our study confirms the substantial role of age in BC prognosis, which might be not only related to a less aggressive therapeutic approach in elderlies, but also the biology of the tumor and impaired mechanisms of self-defense (e.g., immune surveillance) in older patients. Our previous study underscored the very high BC mortality and progression burden and suboptimal management in elderlies with T1HG BC.
16
Factors of secondary relevance which, however, should also be considered, are: tumor size, tumor character (primary vs recurrent), histology type, and tumor location in the bladder. In general, histology types such as squamous cell, neuroendocrine, and micropapillary cancer are more aggressive than urothelial cancer and confer substantial progression risk factors.
17 Moreover, variant histology tumors are more often diagnosed at an advanced stage.
17 Recent multicenter reports suggest that immediate cystectomy should be preferred in the case of T1HG with pure squamous or micropapillary differentiation.
18 Multiple tumor locations and large size are regarded as risk factors for recurrence and progression according to several available nomograms.
1‐3,19 Our data demonstrated that the location of the tumor also matters—multiple site lesions and tumors of overlapping sites have worse prognoses. The prognostic value of tumor location has not been well evidenced yet. Recently, Fukushima et al. showed that among intermediate-risk tumors, bladder neck involvement sub-stratifies patients according to the risk of progression.
20
Sociodemographic factors such as race, income, and marital status are also associated with CSS, but their relevance is less pronounced than that of clinical and histopathological tumor features and patient’s age. Conversely, another population-based study showed that OS and RFS do not differ between White and Black patients in the setting of equal access to healthcare resources.
21 Income and financial status undoubtedly affect access to novel therapies and perhaps to high-quality healthcare. Marital status underlines the significance of relatives’ support in the patient’s adherence to the urologist’s recommendations and struggle against cancer. A study by Heyes et al.
22 demonstrated the value of a supportive partner in dealing with the effects of bladder cancer daily and underlined the importance of partner adaptation to their disease, surveillance, and treatment protocols.
Prediction of CSS, which is the most significant ultimate end-point in oncological studies, is of high relevance in NMIBC in order to identify the group of highest-risk patients, who should be qualified for intensified treatment and very stringent follow-up. A novel therapeutic scheme is still awaited for high-risk NMIBC. An aggressive therapeutic approach might include RC or, in the near future, a combined adjuvant therapy after complete TURBT accompanied by intensive cystoscopic surveillance.
23 Awaited results of ongoing trials with a BCG and immune checkpoint inhibitors combination might soon change the current gold standard therapy for high- and very-high-risk subgroups, making risk-stratification even more important.
Despite its strengths, our study has certain limitations which must be acknowledged. T1HG diagnosis might have potentially included a proportion of understaged tumors (we lacked data about detrusor muscle in the specimen and about repeated TURBT performance). Data about tumor multiplicity was not precise and based solely on ICD-10 C67 coding. Primary tumor status was defined accordingly, with the SEER coding manual, and the cohort included only 8% of patients with recurrent tumors. The data about comorbidities and smoking status which constitute confounders in the survival analysis were not available.
24 We did not analyze the subsequent therapies after TURBT and the data about recurrence and progression were not available. On the other hand, we managed to analyze a very large cohort of patients with NMIBC showing real-world long-term survival outcomes. Although this population-based study has several unavoidable limitations, we believe that it provides important CSM data and a clinically useful nomogram.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.