A population-based competing-risks analysis of the survival of patients treated with radical cystectomy for bladder cancer

Authors

  • Giovanni Lughezzani MD,

    1. Cancer Prognostics and Health Outcomes Unit, University of Montreal Health Center, Montreal, Quebec, Canada
    2. Department of Urology, Vita-Salute San Raffaele University, Milan, Italy
    Search for more papers by this author
    • The first 2 authors contributed equally to this article.

  • Maxine Sun BSc,

    1. Cancer Prognostics and Health Outcomes Unit, University of Montreal Health Center, Montreal, Quebec, Canada
    Search for more papers by this author
    • The first 2 authors contributed equally to this article.

  • Shahrokh F. Shariat MD,

    1. Department of Urology, Weill Medical College of Cornell University, Ithaca, New York
    Search for more papers by this author
  • Lars Budäus MD,

    1. Cancer Prognostics and Health Outcomes Unit, University of Montreal Health Center, Montreal, Quebec, Canada
    2. Martiniclinic, Prostate Cancer Center Hamburg-Eppendorf, Hamburg, Germany
    Search for more papers by this author
  • Rodolphe Thuret MD,

    1. Cancer Prognostics and Health Outcomes Unit, University of Montreal Health Center, Montreal, Quebec, Canada
    2. Department of Urology, University of Montpellier Health Center, Montpellier, France
    Search for more papers by this author
  • Claudio Jeldres MD,

    1. Cancer Prognostics and Health Outcomes Unit, University of Montreal Health Center, Montreal, Quebec, Canada
    2. Department of Urology, University of Montreal, Montreal, Quebec, Canada
    Search for more papers by this author
  • Daniel Liberman MD,

    1. Cancer Prognostics and Health Outcomes Unit, University of Montreal Health Center, Montreal, Quebec, Canada
    2. Department of Urology, University of Montreal, Montreal, Quebec, Canada
    Search for more papers by this author
  • Francesco Montorsi MD,

    1. Department of Urology, Vita-Salute San Raffaele University, Milan, Italy
    Search for more papers by this author
  • Paul Perrotte MD,

    1. Department of Urology, University of Montreal, Montreal, Quebec, Canada
    Search for more papers by this author
  • Pierre I. Karakiewicz MD

    Corresponding author
    1. Cancer Prognostics and Health Outcomes Unit, University of Montreal Health Center, Montreal, Quebec, Canada
    2. Department of Urology, University of Montreal, Montreal, Quebec, Canada
    • Cancer Prognostics and Health Outcomes Unit, University of Montreal Health Center (CHUM), 1058, Rue St-Denis, Montreal, Quebec, Canada H2X 3J4
    Search for more papers by this author
    • Fax: (514) 227-5103


Abstract

BACKGROUND.

Patients treated with radical cystectomy represent a very heterogeneous group with respect to cancer-specific and other-cause mortality. Comorbidities and comorbidity-associated events represent very important causes of mortality in those individuals. The authors examined the rates of cancer-specific and other-cause mortality in a population-based radical cystectomy cohort.

METHODS.

The authors identified 11,260 patients treated with radical cystectomy for urothelial carcinoma of the urinary bladder between 1988 and 2006 within 17 Surveillance, Epidemiology, and End Results registries. Patients were stratified into 20 strata according to patient age and tumor stage at radical cystectomy. Smoothed Poisson regression models were fitted to obtain estimates of cancer-specific and other-cause mortality rates at specific time points after radical cystectomy.

RESULTS.

After stratification according to disease stage and patient age, cancer-specific mortality emerged as the main cause of mortality in all patient strata. Nonetheless, at 5 years after radical cystectomy, between 8.5% and 27.1% of deaths were attributable to other-cause mortality. The 3 most common causes of other-cause mortality were other malignancies, heart disease, and chronic obstructive pulmonary disease. The most prominent effect on cancer-specific mortality was exerted by locally advanced bladder cancer stages. Conversely, age was the main determinant of other-cause mortality. Interestingly, even after adjusting for bladder cancer pathologic stage, cancer-specific mortality was higher in older individuals than their younger counterparts.

CONCLUSIONS.

The current study provides a valuable graphical aid for prediction of cancer-specific and other-cause mortality according to disease stage and patient age. It can help clinicians to better stratify the risk-benefit ratio of radical cystectomy. Hopefully, these findings will be considered in treatment decision making and during informed consent before radical cystectomy. Cancer 2011. © 2010 American Cancer Society.

Bladder cancer (BC) is the second commonest genitourinary malignancy.1 The standard of care for recurrent non-muscle-invasive or muscle invasive BC consists of radical cystectomy.2 The latter offers the best cancer control outcomes and exceeds the benefits of chemotherapy, radiotherapy, or the combination of both.2, 3 Despite its efficacy, radical cystectomy represents a high-risk surgery. The 90-day mortality after radical cystectomy ranges from 0.5% to 25%, depending on patient age, tumor stage, and grade.4

Independently of BC stage, age represents an important determinant of other-cause mortality. Older patients may be sicker and less capable of tolerating cancer manifestations and related therapies. Therefore, older individuals may benefit from radical cystectomy to a lesser extent than younger individuals. For example, some older patients may die from various causes related to perioperative radical cystectomy mortality, even after fully curative radical cystectomy.

Despite the established role of patient age on other-cause mortality after radical cystectomy, there are no available graphical tools capable of illustrating the likelihood of cancer-specific mortality relative to other-cause mortality, according to age and disease characteristics. Specifically, such graphical aids could indicate the probability of cancer-specific mortality relative to other-cause mortality at specific time points after radical cystectomy. This information can help in discriminating between individuals who will acquire the maximum benefit from radical cystectomy relative to those who may only marginally benefit from this surgery. To address this goal, we relied on Poisson regression models to graphically illustrate the probability of cancer-specific and other-cause mortality at different time points after radical cystectomy, according to age and disease stage, in a manner similar to that used by Albertsen et al for prostate cancer and by Hollingsworth et al for kidney cancer.5, 6

MATERIALS AND METHODS

Patient Population

Patients treated with radical cystectomy and pelvic lymphadenectomy for urothelial carcinoma of the urinary bladder between 1988 and 2006 were identified within 17 Surveillance, Epidemiology, and End Results (SEER) registries. This cohort represents a 26% sample of the US population and is considered representative of the entire population in terms of demographics, cancer incidence, and mortality rates.7 The registries include the Atlanta, Detroit, San Francisco-Oakland, Seattle-Puget Sound, Los Angeles, and San Jose-Monterey metropolitan areas, as well as the states of Connecticut, Hawaii, Iowa, New Mexico, Utah, Alaska, Georgia, California, Kentucky, Louisiana, and New Jersey.

The BC diagnostic code (International Classification of Disease [ICD] for Oncology, Second edition [C67.0-C67.9]) and the radical cystectomy and pelvic lymph node dissection treatment code (2-digit procedure code: 50) were used as main selection criteria. Only individuals aged 18 years or older were considered. Malignant histological subtypes other than urothelial carcinoma (ICD-O-3 codes 8120 and 8130) were excluded from the current analysis. Further exclusions consisted of patients with unavailable TNM stage and patients with established distant metastases (M1).

The TNM classification was used according to the specifications of the American Joint Committee on Cancer 2002 classification. Tumor grades I (well differentiated), II (moderately differentiated), III (poorly differentiated), and IV (undifferentiated) were used according to the most recent update of the SEER database.7 The cause of death was defined according to SEER-specific cause of death code (code 29020). For the purpose of this analysis, deaths from BC were coded as cancer-specific mortality. All other deaths were considered as other-cause mortality.

Statistical Analyses

Patients were stratified according to age in decades (≤59, 60-69, 70-79, and ≥80 years). The analysis of variance and the chi-square tests were used for comparisons of means and proportions, respectively.

Further stratification was performed according to pT and pN stages: pT1N0 versus pT2N0 versus pT3N0 versus pT4N0 versus pTanyN1-3. Subsequently, we tabulated the number of patients with each of the 3 outcomes of interest (dead from BC, dead from other cause, and alive) for each of the 20 age, pT stage, and pN stage combinations at 5 years after radical cystectomy, because this represented the furthest time with an adequate number of individuals at risk of either other-cause mortality or cancer-specific mortality within each age-stage stratum.

We used separate Poisson regression models to obtain estimates of the cancer-specific mortality rates after accounting for other-cause mortality and vice versa, for each of the 20 age-tumor stage combinations. We applied the fitted (Poisson model derived) cancer-specific and other-cause mortality rates to the proportion of patients still alive at the beginning of each successive 1-year follow-up interval. This provided us with the estimates of the proportions of patients who died from BC or other causes versus patients who were alive at 1, 2, 3, 4, and 5 years after radical cystectomy. Finally, we plotted the smoothed cancer-specific and other-cause mortality estimates for each age-tumor stage stratum. We relied on the Lowess (locally weighted scatterplot smoothing) method for smoothing.

All statistical tests were performed using SPSS, version 16 (SPSS, Chicago, Ill) or S-PLUS Professional, version 1 (Mathsoft, Seattle, Wash). Moreover, all tests were 2-sided, with a significance level set at .05.

RESULTS

The study population consisted of 11,260 patients who underwent a radical cystectomy with pelvic lymphadenectomy for urothelial carcinoma of the urinary BC between 1988 and 2006 (Table 1). Mean and median age was 67.3 and 68.0 years. The majority were male (74.5%) and Caucasian (90.0%). Overall, 11.2, 40.7, 29.2, and 19.0% of patients had pathologic stages T1, T2, T3, and T4, respectively. Most patients had grades III (53.8%) or IV (39.0%) tumors. The proportion of patients with lymph node metastases (pN1-3) was 22.5%. Mean and median follow-up was 53.1 and 38.0 months. Overall, 31.5% of patients died of BC, and 20.8% of patients died of other competing causes of mortality. The 3 most common causes of other-cause mortality were other malignancies (7.6%), heart disease (4.6%), and chronic obstructive pulmonary disease (1.0%).

Table 1. Characteristics of the Study Population Diagnosed With Urothelial Carcinoma of the Urinary Bladder and Treated With Radical Cystectomy and Pelvic Lymphadenectomy Between 1988 and 2006 Within 17 Surveillance, Epidemiology, and End Results Registries
VariablesOverall Population≤59 Years60-69 Years70-79 Years≥80 YearsP
No. of patients11,260 (100%)2547 (22.6%)3517 (31.2%)3938 (35.0%)1258 (11.2%)
Age, y     
 Mean (median)67.3 (68.0)52.3 (54.0)64.9 (65.0)74.3 (74.0)82.9 (82.0) 
 Range26-10026-5960-6970-7980-100 
Sex     <.001
 Men8390 (74.5%)1992 (78.2%)2683 (76.3%)2869 (72.9%)846 (67.2%) 
 Women2870 (25.5%)555 (21.8%)834 (23.7%)1069 (27.1%)412 (32.8%) 
Race     <.001
 White10,134 (90.0%)2225 (87.4%)3169 (90.1%)3576 (90.8%)1164 (92.5%) 
 Black590 (5.2%)187 (7.3%)188 (5.3%)172 (4.4%)43 (3.4%) 
 Other536 (4.8%)135 (5.3%)160 (4.5%)190 (4.8%)51 (4.1%) 
Year of surgery     <.001
 1988-19921253 (11.1%)263 (10.3%)479 (13.6%)449 (11.4%)62 (11.1%) 
 1993-19971894 (16.8%)415 (16.3%)662 (18.8%)669 (17.0%)148 (11.8%) 
 1998-20023937 (35.0%)891 (35.0%)1142 (32.5%)1435 (36.4%)469 (37.3%) 
 2003-20064176 (37.1%)978 (38.4%)1234 (35.1%)1385 (35.2%)579 (46.0%) 
pT stage     <.001
 pT11257 (11.2%)327 (12.8%)419 (11.9%)404 (10.3%)107 (8.5%) 
 pT24581 (40.7%)1079 (42.4%)1470 (41.8%)1560 (39.6%)472 (37.5%) 
 pT33287 (29.2%)692 (27.2%)957 (27.2%)1205 (30.6%)433 (34.4%) 
 pT42135 (19.0%)449 (17.6%)671 (19.1%)769 (19.5%)246 (19.6%) 
pN stage     <.001
 pN08726 (77.5%)1892 (74.3%)2683 (76.3%)3144 (79.8%)1007 (80.0%) 
 pN1-32534 (22.5%)655 (25.7%)834 (23.7%)794 (20.2%)251 (20.0%) 
Tumor grade     .02
 Grade I60 (0.5%)18 (0.7%)22 (0.6%)14 (0.4%)6 (0.5%) 
 Grade II752 (6.7%)195 (7.7%)242 (6.9%)256 (6.5%)59 (4.7%) 
 Grade III6053 (53.8%)1356 (53.2%)1877 (53.4%)2154 (54.7%)666 (52.9%) 
 Grade IV4395 (39.0%)978 (38.4%)1376 (39.1%)1514 (38.4%)527 (41.9%) 
Follow-up of censored patients, mo     <.001
 Mean (median)53.1 (38.0)61.5 (47.0)57.8 (42.0)48.7 (35.0)32.6 (21.0) 
 Range0.1-2270.1-2270.1-2270.1-2260.1-202 
Vital status at last contact     <.001
 Alive5375 (47.7%)1546 (60.7%)1741 (49.5%)1637 (41.6%)451 (35.9%) 
 Deceased due to bladder cancer3542 (31.5%)701 (27.5%)1100 (31.3%)1270 (32.2%)471 (37.4%) 
 Deceased due to other causes2343 (20.8%)300 (11.8%)676 (19.2%)1031 (26.2%)336 (26.7%) 

When patients were stratified according to age in decades (≤59, 60-69, 70-79, and ≥80 years), statistically significant differences were recorded between the 4 age groups (Table 1). The proportion of females was higher among patients aged ≥80 years than patients aged ≤59 years (32.8% vs 21.8%) (P<.001) The majority of radical cystectomies in patients aged ≥80 years were performed in the most contemporary year quartile (2003-2006; 46.0%) (P<.001). A larger proportion of older individuals underwent a radical cystectomy for more advanced pT stage (P<.001). For example, the proportion of patients with extravesical disease (pT3-4) among individuals aged ≥80 years was 54.0% versus 44.8% in the subgroup of patients aged ≤59 years.

Table 2 shows the proportion of patients who died of BC relative to patients who died of other causes at 5 years after radical cystectomy. Data were stratified by age in decades and by pathologic T and N stages. The proportion of patients who died from BC 5 years after radical cystectomy increased from 18.4% in pT1N0 to 67.5% in pT1-4N1-3. Interestingly, a similar rate of BC deaths was observed in individuals with pT3N0(50.5%) and pT4N0(49.1%) disease.

Table 2. Five-Year Outcomes of Patients Treated With Radical Cystectomy for Urothelial Carcinoma of the Urinary Bladder Stratified According to Age Decades and Pathologic T and N Stages
Cause of DeathAge at SurgeryTotal (%)
≤59 Years60-69 Years70-79 Years≥80 Years
pT1N0 patients (n=1202)
 Bladder cancer27365019132 (18.4)
 Other cause19527123165 (23.0)
 No. of patients alive13415411813419 (58.5)
pT2N0 patients (n=4070)
 Bladder cancer123237267100727 (27.9)
 Other cause64146240111561 (21.5)
 No. of patients alive373449417631302 (50.0)
pT3N0 patients (n=2182)
 Bladder cancer123196307122748 (50.5)
 Other cause448315662345 (23.3)
 No. of patients alive9413114121387 (26.2)
pT4N0 patients (n=1273)
 Bladder cancer7215017867467 (49.1)
 Other cause305711549251 (26.4)
 No. of patients alive6975827233 (24.5)
pTanyN1-3 patients (n=2534)
 Bladder cancer3264034071561292 (67.5)
 Other cause5412515943381 (19.9)
 No. of patients alive9190518240 (12.6)

Table 3 shows the estimated 5-year cancer-specific and other-cause mortality rates in every age-tumor stage stratum. Figure 1 depicts the smoothed, model-derived, cumulative mortality estimates. The lowest estimated 5-year cancer-specific mortality rate was recorded in patients with pT1N0 disease, aged ≤59 years (21.6%). Conversely, the 5-year estimated cancer-specific mortality rate was the highest in patients with pT1-4N1-3 disease, aged ≥80 years (65.5%).

Figure 1.

Smoothed, model-derived, cumulative mortality estimates are shown. CSM indicates cancer-specific mortality; OCM, other-cause mortality; black, CSM; grey, OCM; white, survival.

Table 3. Estimated 5-Year CSM and OCM Rates in Patients With Urothelial Carcinoma of the Urinary Bladder Treated With Radical Cystectomy, Stratified According to Age and Pathologic T and N Stages
pT and pN StagesAge at Surgery
≤59 Years60-69 Years70-79 Years≥80 Years
% Alive% CSM% OCM% Alive% CSM% OCM% Alive% CSM% OCM% Alive% CSM% OCM
  1. CSM indicates cancer-specific mortality; OCM, other-cause mortality.

pT1N0 patients65.721.612.760.923.415.857.123.719.241.231.727.1
pT2N0 patients62.226.311.557.028.614.453.129.317.636.438.924.7
pT3N0 patients54.535.79.848.738.912.444.340.515.226.852.420.8
pT4N0 patients53.036.710.347.139.913.051.842.615.925.153.521.4
pTanyN1-3 patients45.645.98.539.249.910.934.552.313.217.265.517.3

DISCUSSION

BC may be cured with radical cystectomy. However, individuals treated with radical cystectomy may also die of other-cause mortality. In consequence, graphical aids are needed to discriminate between cancer-specific mortality and other-cause mortality after radical cystectomy. Age and disease stage are the predominant determinants of other-cause mortality and cancer-specific mortality, respectively.4, 8 In the current study, we relied on these 2 variables to model and graphically depict cancer-specific and other-cause mortality rates. We used a large population-based cohort to ensure adequate statistical power and to maximize the generalizability of our findings.

Our results indicate that cancer-specific mortality is clearly determined by pathologic T and N stages. Moreover, our findings show that age has an important effect on cancer-specific mortality rates in patients aged ≥80 years, irrespective of disease stage. For example, in patients with pT1N0 disease, the predicted 5-year cancer-specific mortality rates were 21.6%, 23.4%, 23.7%, and 31.7% for those aged ≤59, 60-69, 70-79, and ≥80 years, respectively (Table 3). Conversely, the 5-year other-cause mortality rates increased from 12.7% to 27.1% from the youngest age stratum to the oldest age stratum.

The effect on cancer-specific mortality according to age was most striking in patients with pT3-4N0 and in patients with pT1-4N1-3 BC. This observation is consistent with the established tendency to defer radical cystectomy in older individuals until all alternatives have been exhausted.9–13 Under this premise, older individuals are more likely to be treated with radical cystectomy for more advanced disease stages.9, 12 In consequence, cancer-specific mortality is higher in older patients, even after adjustment for disease stage. This trend was previously documented by several investigators10, 14–16 and was confirmed in the current study.

Unlike several previous analyses, the current study examined other-cause mortality after controlling for cancer-specific mortality. The other-cause mortality trends indicate that age is an important determinant of other-cause mortality. For example, in patients with pT1N0 disease, the 5-year other-cause mortality rates increased from 12.7% to 27.1%, in persons aged ≤59 and ≥80 years, respectively. The same observation was valid for pT2N0 individuals, where the 5-year other-cause mortality rates increased from 11.5% to 24.7% for the same age categories. Finally, the 5-year other-cause mortality rates increased from 9.8% to 20.8%, from 10.3% to 21.4%, and from 8.5% to 17.3% for pT3N0, pT4N0, and pT1-4N1-3 BC stage strata, respectively.

It is noteworthy that in patients with locally advanced pathological stages (pT3-4N0) or in patients with lymph node metastases (pT1-4N1-3), the effect of other-cause mortality is less important than in patients with localized disease (pT1-2N0). This observation is consistent with the role limiting effect of BC on all-cause mortality. For example, in individuals with pT4N0 disease ≥80 years old, the 5-year other-cause mortality rate was 21.4% versus 53.5% for 5-year cancer-specific mortality. Similarly, in the same age category, other-cause mortality and cancer-specific mortality rates are 17.3% and 65.5% in pT1-4N1-3 BC patients, respectively.

Taken together, these findings indicate that cancer-specific mortality drives the overall mortality rates in patients with pT3-4N0 or pT1-4N1-3 disease, irrespective of patients' age. Conversely, in patients with localized disease, other-cause mortality plays a substantially more important role. However, in both patient groups, neither other-cause mortality nor cancer-specific mortality can be ignored, and our graphical aid represents an important indicator of the observed rates of other-cause mortality and cancer-specific mortality across all tumor stage and patient age strata (Fig. 1).

Our report is the first to address and predict other-cause mortality and cancer-specific mortality using Poisson regression models in BC. The same methodology was previously used in prostate and renal cell cancers.5, 6 These analyses demonstrated very similar trends and also documented the importance of other-cause mortality and cancer-specific mortality.5, 6 Therefore, our findings are biologically consistent with other urologic malignancies, and our methodology was previously validated.5, 6 Despite methodological similarities with these 2 reports, our findings differ with respect to message. The previous studies were meant to demonstrate that an important proportion of individuals die of causes other than cancer. Our findings showed that even in patients with non–muscle-invasive urothelial carcinoma of the urinary bladder, cancer-specific mortality prevailed over other-cause mortality. In consequence, our findings are meant to sensitize the clinicians to the importance of cancer-specific mortality across all stages of BC, regardless of age.

Despite its validity and generalizability, our study is not devoid of limitations. One such limitation consists of the surgical selection bias. This phenomenon may affect other-cause mortality rates relative to patients with newly diagnosed BC or age-matched controls. For example, octogenarian radical cystectomy candidates may be in a better health condition than age-matched controls, who may have multiple other comorbidities.12, 17, 18

Lack of comorbidity data also limits the scope of the analysis. The graphical aid presented in the current study should ideally be stratified according to comorbidities and age combinations.12 This limitation affected other analyses that relied on the SEER database.6, 9 Although the current study does not contain comorbidities as a predictor of other-cause mortality, it is fully capable of indicating the rate of other-cause mortality, because the latter is accurately provided in the SEER database.19 Moreover, our findings allow accounting for the effect of age on cancer-specific mortality, and it is well accepted that age represents a proxy and a determinant of comorbidities.

In addition, the SEER database does not account for precystectomy transurethral resection pathology. In consequence, patients with T1 BC at radical cystectomy may have had T2 disease before radical cystectomy. This hypothesis may explain why individuals with T1 disease in our database had a significant effect on cancer-specific mortality, regardless of age. This limitation was shared with previous analyses that relied on the SEER database.6, 9

Lack of data on potential confounding factors such as smoking status, chemotherapy, or radiotherapy delivery represents additional limitations of our study. For example, neoadjuvant chemotherapy demonstrated a statistically significant survival benefit in several randomized controlled trials20, 21 and in a meta-analysis report.22 Unfortunately, the rate of chemotherapy use is low in the SEER database, as recently reported by Porter et al.23

Other variables may also be considered as predictors of cancer-specific mortality or other-cause mortality, and several such variables are not recorded in the SEER-limited database. Despite the extensive sample size of the SEER database, which significantly exceeds that of any other single-institutional or even multi-institutional database, the current study could have benefitted from an even larger sample size. The latter represents a rate-limiting factor, which prevented us from examining other potentially important variables other than age and stage. By using 4 age and 5 stage categories, we modeled 20 patient subgroups. Although the inclusion of a third assessable dichotomous variable, such as low versus high tumor grade, would have doubled the number of groups, it would also have resulted in an insufficient number of observations and events. It would also have been highly desirable to include other important predictors of cancer-specific and other-cause mortality, such as for example race and sex. However, additional stratification according to these variables would have resulted in even smaller patient subsets, with an insufficient number of individuals at risk and of events. Thus, under ideal conditions, larger population sample sizes could have been used.

Conclusions

Despite its limitations, the current study provides a valuable graphical aid for prediction of cancer-specific and other-cause mortality according to disease stage and patient age. This can help clinicians to better stratify the risk-benefit ratio of radical cystectomy. Hopefully, our findings will be considered in treatment decision making and during informed consent before radical cystectomy.

CONFLICT OF INTEREST DISCLOSURES

Pierre I. Karakiewicz is partially supported by the Fonds de la Recherche en Santé du Québec, University of Montreal Department of Surgery, and University of Montreal Health Center (CHUM) Foundation.

Ancillary