Abstract

Background

Bladder cancer is among the most prevalent and expensive to treat cancers in the United States. In the absence of high-level evidence to guide the optimal management of bladder cancer, urologists may vary widely in how aggressively they treat early-stage disease. We examined associations between initial treatment intensity and subsequent outcomes.

Methods

We used the Surveillance, Epidemiology, and End Results–Medicare database to identify patients who were diagnosed with early-stage bladder cancer from January 1, 1992, through December 31, 2002 (n = 20 713), and the physician primarily responsible for providing care to each patient (n = 940). We ranked the providers according to the intensity of treatment they delivered to their patients (as measured by their average bladder cancer expenditures reported to Medicare in the first 2 years after a diagnosis) and then grouped them into quartiles that contained approximately equal numbers of patients. We assessed associations between treatment intensity and outcomes, including survival through December 31, 2005, and the need for subsequent major interventions by using Cox proportional hazards models. All statistical tests were two-sided.

Results

The average Medicare expenditure per patient for providers in the highest quartile of treatment intensity was more than twice that for providers in the lowest quartile of treatment intensity ($7131 vs $2830, respectively). High–treatment intensity providers more commonly performed endoscopic surveillance and used more intravesical therapy and imaging studies than low–treatment intensity providers. However, the intensity of initial treatment was not associated with a lower risk of mortality (adjusted hazard ratio of death from any cause for patients of low– vs high–treatment intensity providers = 1.03, 95% confidence interval 0.97 to 1.09). Initial intensive management did not obviate the need for later interventions. In fact, a higher proportion of patients treated by high–treatment intensity providers than by low–treatment intensity providers subsequently underwent a major medical intervention (11.0% vs 6.4%, P = .02).

Conclusions

Providers vary widely in how aggressively they manage early-stage bladder cancer. Patients treated by high–treatment intensity providers do not appear to benefit in terms of survival or in avoidance of subsequent major medical interventions.

CONTEXT AND CAVEATS
Prior knowledge

Little is known about how urologists vary in the aggressiveness with which they treat patients during the first 2 years after a diagnosis of early-stage bladder cancer.

Study design

Linked Surveillance, Epidemiology, and End Results (SEER)–Medicare data were used to identify patients who were diagnosed with early-stage bladder cancer and the physician primarily responsible for each patient's care and to examine associations between initial treatment intensity and subsequent outcomes, including survival.

Contribution

Urologists who provided the most aggressive treatment had, on average, more than double the Medicare expenditures per patient compared with those who provided the least aggressive treatment, but their patients did not appear to benefit in terms of survival or in avoidance of subsequent major medical interventions.

Implications

It may be possible to eliminate unnecessary procedures and thus reduce the costs of caring for patients with early-stage bladder cancer.

Limitations

The use of observational data did not allow the authors to account for unmeasured differences between patients in different treatment intensity groups. The use of SEER–Medicare data limits the generalizabity of the finding to patients older than 65 years.

From the Editors

Bladder cancer is the fifth most common new cancer diagnosis and among the most expensive cancers to treat in the United States ( 1 ). Nearly three-quarters of incident cases of bladder cancer are non–muscle-invasive (ie, early-stage) tumors ( 2 ), which are removed endoscopically. In up to half of these patients, the disease will progress to muscle-invasive cancer ( 3–6 ). Because mortality from muscle-invasive disease is common ( 7 , 8 ) and often requires a major medical intervention (radical cystectomy, systemic chemotherapy, and/or radiation therapy), effective strategies for the prevention and early detection of disease progression are of paramount importance.

How best to achieve this goal remains unclear. Common strategies for bladder cancer surveillance and treatment that may be useful include intensive intravesical therapy, repeat endoscopic resection after diagnosis, and frequent endoscopic surveillance ( 9 , 10 ). However, evidence from randomized clinical trials establishing the optimal approaches to bladder cancer surveillance and treatment is largely lacking. Rather, current guidelines for the surveillance and treatment of early-stage bladder cancer are most commonly based on the opinions of experts or on observational data and generally favor more intensive regimens ( 10 ). Consequently, urologists vary widely in how they approach early-stage bladder cancer.

In this context, we used linked Surveillance, Epidemiology, and End Results (SEER)–Medicare data to identify differences in the practice styles of US urologists during the first 2 years after an early-stage bladder cancer diagnosis. We were particularly interested in the extent to which the initial treatment intensity of urologists was associated with patients' outcomes.

Methods

Study Population

We used 1992–2005 SEER–Medicare linked data to identify a cohort of patients with early-stage bladder cancer. As detailed elsewhere ( 11 ), the files in this database provide a rich source of information on Medicare patients included in SEER, a nationally representative collection of population-based registries that collect information about all incident cancer patients from diverse geographic areas in the United States. By December 31, 2005, the SEER registries included approximately 26% of the US population ( 12 ). For each Medicare patient in SEER, the SEER-Medicare–linked files contain 100% of the Medicare claims from the inpatient, outpatient, and national claims history files.

We first identified all Medicare patients aged 65–99 years who had an incident bladder cancer detected before death between January 1, 1992, and December 31, 2002, as documented by bladder cancer codes 67.0–67.9 within the SEER–Medicare Patient Entitlement and Diagnosis Summary File. Next, we limited our study population to patients with early-stage bladder cancer [stage 0 or 1, defined according to the modified American Joint Commission on Cancer (AJCC) ( 13 )] by using specific codes provided by SEER.

To ascertain the physician who had the primary responsibility for each patient's bladder cancer care, we first identified all early-stage bladder cancer–related procedures [as listed in the appendix of Schrag et al. ( 14 )] that were performed within a 2-year period following the patient's diagnosis. Only claims for procedures that were performed for a primary diagnosis of bladder cancer were included. Next, we assigned each patient to the provider who had submitted the most of these Medicare procedure claims for the patient by using the Unique Physician Identifier Number. To ensure adequate reliability in our profiles of individual physician practice styles, we included only those providers who had treated at least 10 patients diagnosed with bladder cancer between January 1, 1992, and December 31, 2002. Using this method, our final study population consisted of 20 713 patients who were treated by 940 providers, 99.4% of whom were urologists.

Characterization of Treatment Intensity

Treatment intensity was defined in terms of early-stage bladder cancer expenditures, which were measured at the patient level and included inpatient and outpatient Medicare payments incurred within the first 2 years after bladder cancer diagnosis. We included only those expenditures that were associated with a primary diagnosis code for bladder cancer [ie, International Classification of Diseases, 9th Revision (ICD-9) ( 15 ) codes 188.x (bladder cancer), 233.7 (carcinoma in situ of the bladder), and V105.4 (personal history of bladder cancer)]. Payments were standardized to account for the regional variation in Medicare reimbursement ( 16 ). Expenditures related to systemic chemotherapy, radiation therapy, and those incurred after these interventions were not included. Because radical cystectomy is generally considered an effective treatment for patients with high-risk early-stage bladder cancer ( 17 , 18 ), expenditures related to cystectomy were included. All payments were price adjusted to 2005 dollars by using the Medicare Economic Index ( 16 ), and the sum of the price-adjusted payments was attributed to the primary bladder cancer care provider. The providers were first ranked according to their average expenditures for the 2-year period after the patients' bladder cancer diagnosis and then sorted into four treatment intensity groups (quartiles) that contained approximately the same number of patients per quartile.

To explore the practice patterns underlying treatment intensity, we characterized processes of care by using the ICD-9 and Healthcare Common Procedure Coding System (HCPCS) codes in the Medicare files. The HCPCS codes are composed primarily of Common Procedure Terminology codes ( 19 ) in addition to codes used exclusively by Medicare. For this study, we focused on processes of care that were plausibly relevant to surveillance and survival. As shown in the Appendix, we divided our process-of-care measures into three categories: surveillance related (including endoscopic examination of the bladder, upper urinary tract evaluation, urinary studies, and imaging studies), treatment related (including intravesical therapy and repeat endoscopic resection within 60 days of the initial resection), and medical services (including visits to the urologist and visits to other physicians).

Outcomes

For all outcome measures, we used the patient as the unit of analysis. The primary outcome was all-cause mortality, which avoided potential problems with misclassification of the cause of death ( 20–24 ) and was measured from January 1, 1992, through December 31, 2005, by using explicit vital status fields in SEER. However, because the vast majority of patients with early-stage bladder cancer are likely to die from competing causes ( 6 ), we also assessed bladder cancer–specific mortality as a secondary outcome by using the cause-of-death field available in SEER. Finally, we assessed the patient's need for a subsequent major medical intervention as evidenced by their treatment with radical cystectomy, systemic chemotherapy, and radiation therapy. These major interventions were identified using the appropriate ICD-9 and HCPCS codes within the inpatient, national claims history, and outpatient files.

Statistical Analysis

For all of the analyses, the exposure was provider treatment intensity, which was categorized as quartiles of patients. We first sought to understand differences in patient demographics and disease characteristics according to provider treatment intensity. Next, we characterized the practice styles of these providers by exploring associations between treatment intensity and processes of care. For all of these comparisons, statistical inference was made using chi-square and Kruskal–Wallis tests for categorical and continuous data, respectively.

To examine the association between treatment intensity and survival, we fit a Cox proportional hazards model by adjusting for patient and disease characteristics, including patient age (in 5-year age groups), sex, race (white, black, or other), the International Classification of Diseases for Oncology, 2nd edition (ICD-O-2) ( 25 ) tumor grade (low, high, or unknown), and the AJCC ( 13 ) tumor stage (Ta, Tis, T1, or Ta or T1 not otherwise specified). In addition, we adjusted for socioeconomic status by using a composite measure that was assessed at the level of the patient's ZIP code, as described by Diez Roux et al. ( 26 ). Patients were separated into three equally sized groups of socioeconomic status according to the summary score for this composite measure: low (score range = −12.23 to 1.19), medium (score range = 1.20 to 5.89), and high (score range = 5.90 to 20.76). Patient comorbidities were identified by using ICD-9 diagnosis codes ( 15 ) in Medicare inpatient and outpatient claims for health care encounters that had occurred during the 12-month period preceding the bladder cancer diagnosis. We used the Klabunde et al. ( 27 ) adaptation of the Charlson comorbidity index ( 28 ) to assess comorbidity. Patients were classified according to their comorbidity index score (0, 1, 2, or ≥3), which was treated as a categorical variable. Because patients who were treated by the same provider may have similar outcomes ( 29 ), we adjusted the models to account for this potential clustering by using more robust standard errors ( 30 ). Briefly, within-cluster correlations in mortality were used to derive variance–covariance estimators. These sandwich estimators were then incorporated into the Cox proportional hazards models that measured the associations between treatment intensity and outcomes. For all Cox models, we confirmed the assumption of proportionality by visual inspection of the hazard plots and by goodness-of-fit testing ( 31 ).

For the secondary outcomes (use of cystectomy, systemic chemotherapy, and/or radiation therapy), we fit logistic models to estimate the association between provider treatment intensity and patient-level outcome, by adjusting for patient age (5-year age groups), sex, race (white, black, or other), comorbidity (0, 1, 2, or ≥3), socioeconomic status (low, medium, or high), the ICD-O-2 tumor grade (low, high, or unknown), and the AJCC stage (Ta, Tis, T1, or Ta or T1 not otherwise specified). We computed adjusted percentages for each outcome by back-transforming the predicted use of the intervention from the logistic model. To examine the association between the use of a therapy and provider treatment intensity, Cox proportional hazards models were used to take into account the timing of the therapy while adjusting for the covariates.

To account for potential unmeasured confounding by disease severity (ie, patients who received more intensive treatment may have had more aggressive disease), we conducted a sensitivity analysis in which initial treatment intensity and patient survival were measured in separate populations. For this analysis, treatment intensity was assessed by profiling the providers' practice patterns using data from January 1, 1992, through December 31, 1998, and survival was then assessed among the same providers' patients who were diagnosed with early-stage bladder cancer between January 1, 1999, and December 31, 2002.

All analyses were carried out with SAS software (version 9.1; Cary, NC). All statistical tests were two-tailed, and the probability of a type I error was set at .05. The study protocol was approved by the institutional review board of the University of Michigan.

Results

Medicare expenditures for the initial management of early-stage bladder cancer varied by more than twofold among quartiles of provider treatment intensity and ranged from mean per-patient expenditures of $2830 for low–treatment intensity providers to $7131 for high–treatment intensity providers. Table 1 presents clinical and disease characteristics as the average percentage of patients treated by providers within each quartile of provider treatment intensity. Patient age at diagnosis, sex, comorbidity, and tumor grade did not vary according to the initial treatment intensity by the provider. Compared with providers in the lowest quartile of treatment intensity, those in the highest quartile of treatment intensity treated patients with slightly more severe bladder cancers, as evidenced by the higher proportion of their patients with high-grade (29.1% vs 28.5%, P  = .02) and stage T1 (28.6% vs 24.3%, P < .001) disease.

Table 1

Patient and disease characteristics by provider treatment intensity *

Characteristic  Quartiles of provider treatment intensity
 
P 
1 (low) 4 (high) 
No. of patients 5198 5154 5177 5184 — 
No. of providers 254 204 224 258 — 
Mean Medicare expenditures per   patient in 2005 US dollars 2830 3962 4956 7131 — 
Patient age at diagnosis (%), y     .32 
    65–69 17.5 17.9 17.0 18.6  
    70–74 25.4 25.9 26.7 26.5  
    75–79 25.6 25.9 25.3 25.3  
    80–84 17.9 17.4 18.5 17.0  
    ≥85 13.6 12.9 12.5 12.6  
Female sex (%) 25.2 24.9 26.2 24.9 .42 
Race (%)     <.001 
    White 92.9 94.4 91.8 92.4  
    Black 1.8 1.8 2.4 3.5  
    Other 5.3 3.8 5.8 4.1  
Socioeconomic status † (%)      <.001 
    Low 30.4 31.3 33.7 38.0  
    Medium 32.3 36.8 33.2 31.0  
    High 37.3 31.9 33.1 31.0  
Comorbidity index (%)     .19 
    0 45.1 44.0 43.3 42.5  
    1 29.6 30.1 29.3 30.5  
    2 14.1 14.7 15.5 14.8  
    ≥3 11.2 11.2 11.9 12.2  
Tumor grade ‡ (%)      .02 
    Low 65.1 64.1 63.8 62.6  
    High 28.5 28.7 28.9 29.1  
    Unknown 6.4 7.2 7.3 8.3  
Tumor stage § (%)      <.001 
    Ta 58.4 58.9 54.1 53.0  
    Tis 6.5 6.5 7.4 7.5  
    T1 24.3 23.8 26.1 28.6  
    Ta or T1, not otherwise specified 10.8 10.8 12.4 10.9  
Characteristic  Quartiles of provider treatment intensity
 
P 
1 (low) 4 (high) 
No. of patients 5198 5154 5177 5184 — 
No. of providers 254 204 224 258 — 
Mean Medicare expenditures per   patient in 2005 US dollars 2830 3962 4956 7131 — 
Patient age at diagnosis (%), y     .32 
    65–69 17.5 17.9 17.0 18.6  
    70–74 25.4 25.9 26.7 26.5  
    75–79 25.6 25.9 25.3 25.3  
    80–84 17.9 17.4 18.5 17.0  
    ≥85 13.6 12.9 12.5 12.6  
Female sex (%) 25.2 24.9 26.2 24.9 .42 
Race (%)     <.001 
    White 92.9 94.4 91.8 92.4  
    Black 1.8 1.8 2.4 3.5  
    Other 5.3 3.8 5.8 4.1  
Socioeconomic status † (%)      <.001 
    Low 30.4 31.3 33.7 38.0  
    Medium 32.3 36.8 33.2 31.0  
    High 37.3 31.9 33.1 31.0  
Comorbidity index (%)     .19 
    0 45.1 44.0 43.3 42.5  
    1 29.6 30.1 29.3 30.5  
    2 14.1 14.7 15.5 14.8  
    ≥3 11.2 11.2 11.9 12.2  
Tumor grade ‡ (%)      .02 
    Low 65.1 64.1 63.8 62.6  
    High 28.5 28.7 28.9 29.1  
    Unknown 6.4 7.2 7.3 8.3  
Tumor stage § (%)      <.001 
    Ta 58.4 58.9 54.1 53.0  
    Tis 6.5 6.5 7.4 7.5  
    T1 24.3 23.8 26.1 28.6  
    Ta or T1, not otherwise specified 10.8 10.8 12.4 10.9  
*

All P values are two-sided (chi-square test). — = not applicable.

Based on summary scores determined according to Diez Roux et al. ( 26 ); low (score range  = −12.23 to 1.19), medium (score range  = 1.20 to 5.89), and high (score range = 5.90 to 20.76).

International Classification of Diseases for Oncology, 2nd edition ( 25 ).

§

Modified American Joint Commission on Cancer ( 13 ).

As shown in Table 2 , high–treatment intensity providers (ie, those in the highest quartile of treatment intensity) had higher rates of all surveillance- and treatment-related processes of care during the initial management of patients with early-stage bladder cancer than low–treatment intensity providers (ie, those in the lowest quartile of treatment intensity). Compared with patients who were treated by low–treatment intensity providers, those treated by high–treatment intensity providers were, on average, followed up more rigorously with bladder endoscopy (8.3 vs 7.3 procedures, P < .001), urine cytology (2.3 vs 1.3 tests, P < .001), and radiographic imaging (6.0 vs 5.1 studies, P < .001). Treatment-related processes of care followed similar trends. Patients who were treated by high–treatment intensity providers received statistically significantly more instillations (5.0 vs 2.6, P < .001) and induction courses (0.6 vs 0.5, P < .001) of intravesical therapy than patients who were treated by low–treatment intensity providers.

Table 2

Provider practice styles according to average treatment intensity

Processes of care  Quartiles of provider treatment intensity
 
P
1 (low) 4 (high) 
Surveillance related      
    Endoscopic surveillance, mean number of procedures 7.3 7.7 7.9 8.3 <.001 
    Upper urinary tract evaluation, mean number of tests 0.9 1.0 1.0 1.2 <.001 
    Radiographic imaging, mean number of studies 5.1 5.1 5.4 6.0 <.001 
    Urinary cytology, mean number of tests 1.3 1.7 1.9 2.3 <.001 
    Urine cytology, mean % of patients 41.3 47.8 50.4 56.8  <.001 † 
    Urinalysis, mean number of tests 7.3 7.3 8.4 8.5 <.001 
Treatment related      
    Intravesical therapy, mean number of instillations 2.6 3.4 4.1 5.0 <.001 
    Induction courses ‡ of intravesical therapy, mean number  0.5 0.5 0.6 0.6 <.001 
    Induction intravesical therapy, mean % of patients 21.7 25.8 30.2 36.0  <.001 † 
    Repeat endoscopic resection, mean % of patients 4.7 5.6 6.4 9.6  <.001 † 
Medical services      
    Visits to the urologist, mean number 3.4 3.5 4.1 4.2 <.001 
    Visits to the other physicians, mean number 20.2 19.4 19.8 19.9 .12 
Processes of care  Quartiles of provider treatment intensity
 
P
1 (low) 4 (high) 
Surveillance related      
    Endoscopic surveillance, mean number of procedures 7.3 7.7 7.9 8.3 <.001 
    Upper urinary tract evaluation, mean number of tests 0.9 1.0 1.0 1.2 <.001 
    Radiographic imaging, mean number of studies 5.1 5.1 5.4 6.0 <.001 
    Urinary cytology, mean number of tests 1.3 1.7 1.9 2.3 <.001 
    Urine cytology, mean % of patients 41.3 47.8 50.4 56.8  <.001 † 
    Urinalysis, mean number of tests 7.3 7.3 8.4 8.5 <.001 
Treatment related      
    Intravesical therapy, mean number of instillations 2.6 3.4 4.1 5.0 <.001 
    Induction courses ‡ of intravesical therapy, mean number  0.5 0.5 0.6 0.6 <.001 
    Induction intravesical therapy, mean % of patients 21.7 25.8 30.2 36.0  <.001 † 
    Repeat endoscopic resection, mean % of patients 4.7 5.6 6.4 9.6  <.001 † 
Medical services      
    Visits to the urologist, mean number 3.4 3.5 4.1 4.2 <.001 
    Visits to the other physicians, mean number 20.2 19.4 19.8 19.9 .12 
*

P values based on two-sided Kruskal–Wallis test except where indicated.

Two-sided chi-square test.

Five or more treatments of intravesical therapy within a 45-day period.

Despite these differences in provider practice style, the median survival of patients was similar across all four quartiles of provider treatment intensity ( Table 3 ) and ranged from 76.5 months for those whose providers were in the second highest quartile to 79.8 months for those whose providers were in the second lowest quartile ( P = .50). Overall, 11 485 (55.4%) of the 20 713 patients died from any cause between January 1, 1992, and December 31, 2005. Patients treated by low–treatment intensity providers had a similar risk of death as those who were treated by high–treatment intensity providers (adjusted hazard ratio [HR] of death = 1.03, 95% confidence interval [CI] = 0.97 to 1.09) after adjusting for differences in demographics and cancer severity (ie, tumor grade and stage). When the patients were stratified by tumor grade and stage, we observed the anticipated effects of these markers of disease severity on survival, that is, patients with high-grade or T1 disease had generally lower survival than their counterparts with low-grade or Ta disease, respectively, at all levels of provider treatment intensity. However, as with the primary analysis, we observed no survival benefit associated with more intensive care. For example, among patients with T1 disease—a population with the highest risk of disease progression—those treated by low–treatment intensity providers had a similar risk of death as those treated by high–treatment intensity providers (adjusted HR of death = 0.98, 95% CI = 0.88 to 1.09).

Table 3

All-cause and bladder cancer–specific mortality risks according to average provider treatment intensity *

Patient stratification  Median survival in months by quartiles of provider treatment intensity
 
Adjusted HR of death (95% CI) for patients of low– vs high–treatment intensity providers
 
1 (low) 4 (high) All-cause mortality Bladder cancer–specific mortality 
All patients 76.8 79.8 76.5 78.0  1.03 (0.97 to 1.09) †  0.70 (0.59 to 0.83) † 
Stratified by grade       
    Low 85.1 86.6 83.4 85.9  1.03 (0.96 to 1.11) ‡  0.66 (0.52 to 0.85) ‡ 
    High 62.0 64.0 62.5 63.3  1.03 (0.92 to 1.15) ‡  0.81 (0.63 to 1.03) ‡ 
    Unknown 71.1 80.7 72.8 71.3  0.96 (0.78 to 1.19) ‡  0.39 (0.20 to 0.75) ‡ 
Stratified by stage       
    Ta 85.3 87.4 84.3 89.9  1.06 (0.98 to 1.14) §  0.74 (0.57 to 0.97) § 
    Tis 76.8 72.0 74.9 72.4  0.91 (0.74 to 1.11) §  0.49 (0.27 to 0.89) § 
    T1 75.8 76.6 76.7 78.9  0.98 (0.88 to 1.09) §  0.68 (0.54 to 0.86) § 
    Ta or T1 64.4 64.8 63.7 61.7  1.06 (0.89 to 1.27) §  0.72 (0.44 to 1.18) § 
Patient stratification  Median survival in months by quartiles of provider treatment intensity
 
Adjusted HR of death (95% CI) for patients of low– vs high–treatment intensity providers
 
1 (low) 4 (high) All-cause mortality Bladder cancer–specific mortality 
All patients 76.8 79.8 76.5 78.0  1.03 (0.97 to 1.09) †  0.70 (0.59 to 0.83) † 
Stratified by grade       
    Low 85.1 86.6 83.4 85.9  1.03 (0.96 to 1.11) ‡  0.66 (0.52 to 0.85) ‡ 
    High 62.0 64.0 62.5 63.3  1.03 (0.92 to 1.15) ‡  0.81 (0.63 to 1.03) ‡ 
    Unknown 71.1 80.7 72.8 71.3  0.96 (0.78 to 1.19) ‡  0.39 (0.20 to 0.75) ‡ 
Stratified by stage       
    Ta 85.3 87.4 84.3 89.9  1.06 (0.98 to 1.14) §  0.74 (0.57 to 0.97) § 
    Tis 76.8 72.0 74.9 72.4  0.91 (0.74 to 1.11) §  0.49 (0.27 to 0.89) § 
    T1 75.8 76.6 76.7 78.9  0.98 (0.88 to 1.09) §  0.68 (0.54 to 0.86) § 
    Ta or T1 64.4 64.8 63.7 61.7  1.06 (0.89 to 1.27) §  0.72 (0.44 to 1.18) § 
*

HR = hazard ratio; CI = confidence interval.

Adjusted for age, sex, race, socioeconomic status, comorbidity, tumor grade, and tumor stage.

Adjusted for age, sex, race, socioeconomic status, comorbidity, and tumor stage.

§

Adjusted for age, sex, race, socioeconomic status, comorbidity, and tumor grade.

Overall, 1613 (7.8%) patients died from bladder cancer between January 1, 1992, and December 31, 2005. However, as with the primary outcome, we observed no benefit of treatment intensity to bladder cancer–specific survival ( Table 3 ). In fact, patients who were treated by low–treatment intensity providers had a 30% lower risk of death compared with those treated by high–treatment intensity providers (adjusted HR of death from bladder cancer = 0.70, 95% CI = 0.59 to 0.83). Similar relationships between treatment intensity and bladder cancer–specific survival were evident after stratifying patients by tumor grade and stage.

As shown in Figure 1 , patients treated by high–treatment intensity providers were not less likely to require a subsequent major medical intervention than those treated by low–treatment intensity providers (11.0% vs 6.4%, P = .02). Indeed, patients treated by high–treatment intensity providers were more likely than patients treated by low–treatment intensity providers to undergo radical cystectomy, even after adjustment for differences between the two groups of patients (3.9% vs 1.6%, P < .001).

Figure 1

Use of major medical interventions among patients with early-stage bladder cancer by quartiles of provider treatment intensity, expressed as the percentage of patients adjusted for age, sex, race, comorbidity, socioeconomic status, and tumor grade and stage. The adjusted percentages were obtained by back-transforming data from the logistic models. Statistical inference was based on the Wald chi-square statistic obtained from the Cox proportional hazards models for comparisons between the highest and lowest treatment intensity quartiles within interventions, adjusting for the above covariates. All P values are two-sided.

Figure 1

Use of major medical interventions among patients with early-stage bladder cancer by quartiles of provider treatment intensity, expressed as the percentage of patients adjusted for age, sex, race, comorbidity, socioeconomic status, and tumor grade and stage. The adjusted percentages were obtained by back-transforming data from the logistic models. Statistical inference was based on the Wald chi-square statistic obtained from the Cox proportional hazards models for comparisons between the highest and lowest treatment intensity quartiles within interventions, adjusting for the above covariates. All P values are two-sided.

In a sensitivity analysis, we repeated the primary analysis by assessing treatment intensity and survival in separate patient populations. Briefly, treatment intensity was measured by use of providers' practice patterns for their patients diagnosed between January 1, 1992, and December 31, 1998. Overall mortality was then assessed for the same providers among their patients diagnosed with bladder cancer between January 1, 1999, and December 31, 2002. As with the primary analysis, we observed no differences in overall mortality according to treatment intensity (eg, adjusted HR of death for low vs high treatment intensity = 1.00, 95% CI = 0.89 to 1.13). A similar null relationship was observed when we used bladder cancer–specific survival as the outcome.

Discussion

We found that urologists vary widely in the intensity of treatment they provide during the first 2 years after a diagnosis of early-stage bladder cancer. On average, providers in the highest quartile of treatment intensity had more than double the Medicare expenditures per patient compared with those in the lowest quartile. The high–treatment intensity style of practice was characterized by a greater use of all measured health services, including intravesical therapy, endoscopy, urinary studies, and imaging. However, this aggressive early treatment approach did not improve survival or prevent patients from having to undergo major medical interventions in subsequent years. In fact, compared with patients treated by low–treatment intensity urologists, those treated by high–treatment intensity urologists were nearly two and one-half times more likely to undergo radical cystectomy and nearly twice as likely to receive any major medical intervention, even after accounting for patient differences.

These findings highlight the lack of clinical consensus in how best to manage patients with early-stage bladder cancer. Current guidelines for the management of non–muscle-invasive bladder cancer generally favor the more intensive regimens of endoscopy and intravesical therapy ( 9 ). However, neither of these regimens has convincingly demonstrated the ability to prevent disease progression or to prolong patient survival ( 32–34 ). Indeed, only one randomized trial ( 32 ) to our knowledge has explored the question of optimal endoscopic surveillance care. Because that study included only 97 patients, the findings were inconclusive. In light of the limited high-level evidence to guide clinical practice, the considerable variation in the early treatment of bladder cancer is not surprising.

One potential limitation of our analysis relates to unmeasured differences in patients among the physician treatment intensity groups. In particular, patients treated by high–treatment intensity urologists might have more aggressive disease than those treated by low–treatment intensity urologists, which could explain the apparent lack of benefit associated with treatment intensity. We addressed this well-described limitation of observational data ( 35 , 36 ) in several ways. First, we used a clinical registry to ascertain patients’ bladder cancer stage and grade, which are, arguably, the most important determinants of death in the bladder cancer patient population ( 7 , 37 ). Patients in the different treatment intensity groups were similar with respect to age, sex, and comorbidity. Second, we assessed treatment intensity at the level of the provider. Relative to a patient-level analysis, this approach is less susceptible to selection bias to the extent that it would require systematic variation in unmeasured risk factors across providers, which is probably less likely than variation in such risks across patients. Finally, our sensitivity analysis to assess treatment intensity and survival in separate patient populations also failed to demonstrate a survival advantage of more aggressive treatment.

Although more aggressive early treatment intensity was not associated with survival, it was associated with higher rates of major medical interventions, including radical cystectomy, systemic chemotherapy, and radiation therapy. There are several potential explanations for this finding. First, as discussed above, high–treatment intensity providers may have been treating sicker patients who, ultimately, required such interventions. However, unmeasured confounding seems unlikely given our approach of measuring treatment intensity at the level of the provider and the large differences in the rates of major intervention. Second, it is possible that more intensive therapy could, paradoxically, increase the risk of disease progression and thus the need for major medical interventions. However, we know of no biological mechanism to support this possibility. Third and perhaps most likely, provider practice styles with regard to the management of early-stage bladder cancer, as measured by their initial treatment intensity, may be consistent with those for more advanced disease. Simply put, urologists who treat aggressively early are likely to provide aggressive treatment in all aspects of bladder cancer care, and vice versa.

A second limitation of our findings relates to their applicability to the broader population of patients with early-stage bladder cancer. Because we relied on SEER–Medicare data, our findings may not be generalizable to patients younger than 65 years. However, it is important to note that nearly three-quarters of bladder cancer cases in the United States occur annually within the Medicare population ( 12 ). In early-stage bladder cancer, unlike in prostate cancer, treatment decisions generally are not made on the basis of the patient's age. Thus, extrapolation of our findings to the broader cohort (ie, all patients with early-stage bladder cancer) would appear to be reasonable. Although overall treatment intensity was not associated with better outcomes, it is possible that greater use of individual aspects of early-stage bladder cancer care (eg, endoscopic surveillance) could afford a benefit for some patients. Using observational data to identify such components of care may provide better and more efficient care in patients with early-stage bladder cancer. Finally, the lack of an association between treatment intensity and all-cause mortality among patients traditionally felt to be at high risk of disease progression (ie, those with stage T1 and/or high-grade tumors) does not preclude the possibility that some groups of patients may benefit from greater intensity of care; rather, it suggests that such patient populations are not readily identifiable by the grade and stage information captured in SEER.

Given the lack of association between treatment intensity and survival, our findings suggest the opportunity for reducing costs by eliminating unnecessary procedures and thus reducing wasteful spending for the care of patients with early-stage bladder cancer, which is already among the most expensive cancers in the United States ( 1 ). In light of the small but nontrivial risks associated with early-stage bladder cancer surveillance and treatment, the overuse of a high–treatment intensity practice style is worrisome given its lack of association with any benefit for the patients. Identifying best practices of care for patients diagnosed with early-stage bladder cancer must ultimately await the findings from future well-designed randomized clinical trials. In the meantime, urologists should not assume that more aggressive management of early-stage bladder cancer will translate into better outcomes for their patients.

Funding

American Cancer Society Pennsylvania Division—Dr. William and Rita Conrady Mentored Research Scholar Grant (MSRG-07-006-01-CPHPS to B.H.); American Urological Association Foundation (to B.H.); Astellas Pharma US, Inc. (to B.H.); National Cancer Institute (R01 CA098481-01A1, K05 CA115571-01 A2 to J.B.).

Appendix Table

Codes used to identify processes of care *

Category Process of care Variable specification ICD-9 diagnosis codes (MEDPAR) HCPCS codes (National Claims History and Outpatient files) Descriptor 
Surveillance related Endoscopic surveillance Discrete variable: use for each 1-month interval for 2 years after diagnosis (range = 0–24 months) 57.32, 57.39 52 000, 52 001, 52 005, 52 007 Cystoscopy: with irrigation and evacuation of blood clots; with ureteral catheter; with brush biopsy of the ureter 
   57.33 52 204, 52214 Cystoscopy with biopsy, fulguration 
    52 250 Cystoscopy with insertion of radioactive substance, with or without biopsy or fulguration 
    52 260, 52 265 Cystoscopy with bladder dilation 
   57.91 52 270, 52 275, 52 276, 52 277, 52 281, 52 282, 52 283, 52 285 Cystoscopy with urethral dilation or urethrotomy 
    52 290, 52 300, 52 301, 52 305 Cystoscopy with ureteral meatotomy 
   57.0 52 310, 52 315 Cystoscopy with removal of foreign body 
    52 320, 52 325, 52 327, 52 330, 52 332, 52 334 Cystoscopy for ureteral calculus 
    52 317, 52 318 Cystoscopy with lithalopaxy 
    52 341, 52 342, 52 343 Cystoscopy for ureteral stricture 
   57.92, 60.2 52 347, 52 400, 52 450, 52 500, 52 510, 52 601, 52 606, 52 612, 52 614, 52 620, 52 630, 52 640, 52 647, 52 648, 52 700 Cystoscopy with transurethral prostate surgery 
   56.31, 56.33 52 344, 52 345, 52 346, 52 351, 52 352, 52 353, 52 354, 52 355 Cystoscopy with ureteroscopy 
    52 224 Cystoscopy, with fulguration or treatment of minor (<0.5 cm) bladder lesions, with or without biopsy 
   57.49 52 234, 52 235, 52 240 Cystoscopy, with fulguration and/or resection of small (0.5 up to 2 cm), medium (2 up to 5 cm), large (≥5 cm) bladder tumors 
 Upper urinary tract evaluation Discrete variable: use for each 1-month interval for 2 years after diagnosis (range 0–24)  74 400, 74 410, 74 415 Intravenous urography 
    74 420 Retrograde urography 
    74 425 Antegrade urography 
    74 150, 74 160, 74 170 Abdominal CT 
    74 181, 74 182, 74 183, 74 185 Abdominal MRI 
 Urinary studies Discrete variables: use (uniquely for urinalysis and cytology) for each 1-month interval for 2 years after diagnosis (range 0–24)  81 000, 81 001, 81 002,  81 003, 81 005 Urinalysis 
 Dichotomous variable: use of urine cytology at least once during 2-year interval 
    88 104, 88 106, 88 107, 88 108, 88 112, 88 160, 88 161, 88 162, 88 271, 88 272, 88 273, 88 274, 88 275, 88 291 Urine cytology and cytogenetics 
 Imaging studies Continuous variable: number of unique claims over 2-year interval  71 010, 71 015, 71 020, 71 021, 71 022, 71 023, 71 030, 71 034, 71 035 Chest radiography 
    71 250, 71 260, 71 270, 71 275 Chest CT 
    71 550, 71 551, 71 552, 71 555 Chest MRI 
    74 150, 74 160, 74 170 Abdominal CT 
    74 181, 74 182, 74 183, 74 185 Abdominal MRI 
    72 191, 72 192, 72 193, 72 194 Pelvis CT 
    72 195, 72 196, 72 197 Pelvis MRI 
    76 497 Unlisted CT 
    76 498 Unlisted MRI 
    78 810 PET scan 
    74 400, 74 410, 74 415 Intravenous urography 
    74 420 Retrograde urography 
    74 425 Antegrade urography 
    76 700, 76 770, 76 705, 76 775, 76 778 Abdominal or renal ultrasound 
    78 700, 78 701, 78 704, 78 707, 78 708, 78 709, 78 715 Nuclear medicine renal scan 
    78 300, 78 305, 78 306, 78 315, 78 320 Nuclear medicine bone scan 
    78 800, 78 801, 78 802, 78 990 Radiopharmaceutical localization of tumor or distribution of radiopharmaceutical agents 
Treatment related Intravesical therapy  Continuous variables: number of unique claims over 2-year interval; and, measure number of induction courses † over 2-year interval   51 720 Bladder instillation of anticarcinogenic agent 
Dichotomous variable: use of induction chemotherapy at least once during 2-year interval 
    51 700 Bladder irrigation, with or without instillation 
    51 701 Insertion of a nonindwelling catheter 
    51 702, 51 703 Insertion of a temporary indwelling catheter 
    J8520, J8521 Capecitabine 
    J9000, J9001 Doxorubicin 
    J9031 Intravesical Bacille Calmette-Guérin 
    J9201 Gemcitabine 
    J9280, J9290, J9291 Mitomycin 
    J9340 Thiotepa 
 Repeat endoscopic resection Dichotomous variable: use of repeat endoscopic resection within 60 days of initial resection or biopsy 57.49 52 234, 52 235, 52 240 Cystoscopy, with fulguration and/or resection of small (0.5 up to 2 cm); medium (2 up to 5 cm); large (≥5 cm) tumors 
Medical services Visits to the urologist Continuous variable: number of unique claims over 2-year interval  99 201–99 205 New patient, outpatient 
    99 211–99 215 Established patient, outpatient 
    99 218–99 220, 99 234–99 236 Initial, established, observation unit 
    99 221–99 223 Initial, inpatient 
    99 231–99 233 Subsequent, inpatient 
    99 241–99 245 Consult, outpatient 
    99 251–99 255 Consult, inpatient, new 
    99 261–99 263 Consult, inpatient, established 
    99 271–99 275 Confirmatory consult 
 Visits to other physicians Continuous variable: number of unique claims over 2-year interval  99 201–99 205 New patient, outpatient 
    99 211–99 215 Established patient, outpatient 
    99 218–99 220, 99 234–99 236 Initial, established, observation unit 
    99 221–99 223 Initial, inpatient 
    99 231–99 233 Subsequent, inpatient 
    99 241–99 245 Consult, outpatient 
    99 251–99 255 Consult, inpatient, new 
    99 261–99 263 Consult, inpatient, established 
    99 271–99 275 Confirmatory consult 
Category Process of care Variable specification ICD-9 diagnosis codes (MEDPAR) HCPCS codes (National Claims History and Outpatient files) Descriptor 
Surveillance related Endoscopic surveillance Discrete variable: use for each 1-month interval for 2 years after diagnosis (range = 0–24 months) 57.32, 57.39 52 000, 52 001, 52 005, 52 007 Cystoscopy: with irrigation and evacuation of blood clots; with ureteral catheter; with brush biopsy of the ureter 
   57.33 52 204, 52214 Cystoscopy with biopsy, fulguration 
    52 250 Cystoscopy with insertion of radioactive substance, with or without biopsy or fulguration 
    52 260, 52 265 Cystoscopy with bladder dilation 
   57.91 52 270, 52 275, 52 276, 52 277, 52 281, 52 282, 52 283, 52 285 Cystoscopy with urethral dilation or urethrotomy 
    52 290, 52 300, 52 301, 52 305 Cystoscopy with ureteral meatotomy 
   57.0 52 310, 52 315 Cystoscopy with removal of foreign body 
    52 320, 52 325, 52 327, 52 330, 52 332, 52 334 Cystoscopy for ureteral calculus 
    52 317, 52 318 Cystoscopy with lithalopaxy 
    52 341, 52 342, 52 343 Cystoscopy for ureteral stricture 
   57.92, 60.2 52 347, 52 400, 52 450, 52 500, 52 510, 52 601, 52 606, 52 612, 52 614, 52 620, 52 630, 52 640, 52 647, 52 648, 52 700 Cystoscopy with transurethral prostate surgery 
   56.31, 56.33 52 344, 52 345, 52 346, 52 351, 52 352, 52 353, 52 354, 52 355 Cystoscopy with ureteroscopy 
    52 224 Cystoscopy, with fulguration or treatment of minor (<0.5 cm) bladder lesions, with or without biopsy 
   57.49 52 234, 52 235, 52 240 Cystoscopy, with fulguration and/or resection of small (0.5 up to 2 cm), medium (2 up to 5 cm), large (≥5 cm) bladder tumors 
 Upper urinary tract evaluation Discrete variable: use for each 1-month interval for 2 years after diagnosis (range 0–24)  74 400, 74 410, 74 415 Intravenous urography 
    74 420 Retrograde urography 
    74 425 Antegrade urography 
    74 150, 74 160, 74 170 Abdominal CT 
    74 181, 74 182, 74 183, 74 185 Abdominal MRI 
 Urinary studies Discrete variables: use (uniquely for urinalysis and cytology) for each 1-month interval for 2 years after diagnosis (range 0–24)  81 000, 81 001, 81 002,  81 003, 81 005 Urinalysis 
 Dichotomous variable: use of urine cytology at least once during 2-year interval 
    88 104, 88 106, 88 107, 88 108, 88 112, 88 160, 88 161, 88 162, 88 271, 88 272, 88 273, 88 274, 88 275, 88 291 Urine cytology and cytogenetics 
 Imaging studies Continuous variable: number of unique claims over 2-year interval  71 010, 71 015, 71 020, 71 021, 71 022, 71 023, 71 030, 71 034, 71 035 Chest radiography 
    71 250, 71 260, 71 270, 71 275 Chest CT 
    71 550, 71 551, 71 552, 71 555 Chest MRI 
    74 150, 74 160, 74 170 Abdominal CT 
    74 181, 74 182, 74 183, 74 185 Abdominal MRI 
    72 191, 72 192, 72 193, 72 194 Pelvis CT 
    72 195, 72 196, 72 197 Pelvis MRI 
    76 497 Unlisted CT 
    76 498 Unlisted MRI 
    78 810 PET scan 
    74 400, 74 410, 74 415 Intravenous urography 
    74 420 Retrograde urography 
    74 425 Antegrade urography 
    76 700, 76 770, 76 705, 76 775, 76 778 Abdominal or renal ultrasound 
    78 700, 78 701, 78 704, 78 707, 78 708, 78 709, 78 715 Nuclear medicine renal scan 
    78 300, 78 305, 78 306, 78 315, 78 320 Nuclear medicine bone scan 
    78 800, 78 801, 78 802, 78 990 Radiopharmaceutical localization of tumor or distribution of radiopharmaceutical agents 
Treatment related Intravesical therapy  Continuous variables: number of unique claims over 2-year interval; and, measure number of induction courses † over 2-year interval   51 720 Bladder instillation of anticarcinogenic agent 
Dichotomous variable: use of induction chemotherapy at least once during 2-year interval 
    51 700 Bladder irrigation, with or without instillation 
    51 701 Insertion of a nonindwelling catheter 
    51 702, 51 703 Insertion of a temporary indwelling catheter 
    J8520, J8521 Capecitabine 
    J9000, J9001 Doxorubicin 
    J9031 Intravesical Bacille Calmette-Guérin 
    J9201 Gemcitabine 
    J9280, J9290, J9291 Mitomycin 
    J9340 Thiotepa 
 Repeat endoscopic resection Dichotomous variable: use of repeat endoscopic resection within 60 days of initial resection or biopsy 57.49 52 234, 52 235, 52 240 Cystoscopy, with fulguration and/or resection of small (0.5 up to 2 cm); medium (2 up to 5 cm); large (≥5 cm) tumors 
Medical services Visits to the urologist Continuous variable: number of unique claims over 2-year interval  99 201–99 205 New patient, outpatient 
    99 211–99 215 Established patient, outpatient 
    99 218–99 220, 99 234–99 236 Initial, established, observation unit 
    99 221–99 223 Initial, inpatient 
    99 231–99 233 Subsequent, inpatient 
    99 241–99 245 Consult, outpatient 
    99 251–99 255 Consult, inpatient, new 
    99 261–99 263 Consult, inpatient, established 
    99 271–99 275 Confirmatory consult 
 Visits to other physicians Continuous variable: number of unique claims over 2-year interval  99 201–99 205 New patient, outpatient 
    99 211–99 215 Established patient, outpatient 
    99 218–99 220, 99 234–99 236 Initial, established, observation unit 
    99 221–99 223 Initial, inpatient 
    99 231–99 233 Subsequent, inpatient 
    99 241–99 245 Consult, outpatient 
    99 251–99 255 Consult, inpatient, new 
    99 261–99 263 Consult, inpatient, established 
    99 271–99 275 Confirmatory consult 
*

ICD-9 = International Classification of Diseases, 9th Revision; MEDPAR = Medicare Provider Analysis and Review file; HCPCS = Healthcare Common Procedure Coding System; CT = computed tomography; MRI = magnetic resonance imaging; PET = positron emission tomography.

Induction intravesical therapy: at least five unique claims for instillation of any anticarcinogenic agent within a 45-day period.

References

1.
Riley
GF
Potosky
AL
Lubitz
JD
Kessler
LG
Medicare payments from diagnosis to death for elderly cancer patients by stage at diagnosis
Med Care
 , 
1995
, vol. 
33
 
8
(pg. 
828
-
841
)
2.
Snyder
C
Harlan
L
Knopf
K
Potosky
A
Kaplan
R
Patterns of care for the treatment of bladder cancer
J Urol
 , 
2003
, vol. 
169
 
5
(pg. 
1697
-
1701
)
3.
Haukaas
S
Daehlin
L
Maartmann-Moe
H
Ulvik
NM
The long-term outcome in patients with superficial transitional cell carcinoma of the bladder: a single-institutional experience
BJU Int
 , 
1999
, vol. 
83
 
9
(pg. 
957
-
963
)
4.
Holmang
S
Hedelin
H
Anderstrom
C
Holmberg
E
Busch
C
Johansson
SL
Recurrence and progression in low grade papillary urothelial tumors
J Urol
 , 
1999
, vol. 
162
 
3 pt 1
(pg. 
702
-
707
)
5.
Holmang
S
Hedelin
H
Anderstrom
C
Johansson
SL
The relationship among multiple recurrences, progression and prognosis of patients with stages Ta and T1 transitional cell cancer of the bladder followed for at least 20 years
J Urol
 , 
1995
, vol. 
153
 
6
(pg. 
1823
-
1826
discussion 1826–1827
6.
Herr
HW
Tumor progression and survival of patients with high grade, noninvasive papillary (TaG3) bladder tumors: 15-year outcome
J Urol
 , 
2000
, vol. 
163
 
1
(pg. 
60
-
61
)
7.
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
, vol. 
19
 
3
(pg. 
666
-
675
)
8.
Dimopoulos
MA
Moulopoulos
LA
Role of adjuvant chemotherapy in the treatment of invasive carcinoma of the urinary bladder
J Clin Oncol
 , 
1998
, vol. 
16
 
4
(pg. 
1601
-
1612
)
9.
Hall
MC
Chang
SS
Dalbagni
G
, et al.  . 
Guideline for the management of nonmuscle invasive bladder cancer (stages Ta, T1, and Tis): 2007 update
J Urol
 , 
2007
, vol. 
178
 
6
(pg. 
2314
-
2330
)
10.
 
The National Comprehensive Cancer Network. Bladder Cancer V.2.2008. http://www.nccn.org/professionals/physician_gls/PDF/bladder.pdf . Accessed June 4, 2008
11.
Warren
JL
Klabunde
CN
Schrag
D
Bach
PB
Riley
GF
Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population
Med Care
 , 
2002
, vol. 
40
 
(8 suppl)
 
IV-3-18
12.
 
Surveillance, Epidemiology, and End Results (SEER) Program. http://seer.cancer.gov/ . Accessed December 12, 2006
13.
Greene
FL
Page
EL
Fleming
ID
AJCC Cancer Staging Manual
 , 
2002
New York
Springer-Verlag
14.
Schrag
D
Hsieh
LJ
Rabbani
F
Bach
PB
Herr
H
Begg
CB
Adherence to surveillance among patients with superficial bladder cancer
J Natl Cancer Inst
 , 
2003
, vol. 
95
 
8
(pg. 
588
-
597
)
15.
US Public Health Service
International Classification of Diseases, 9th revision. (PHS) 92-1260
 , 
1992
Washington, DC
US GPO
16.
Brown
ML
Riley
GF
Schussler
N
Etzioni
R
Estimating health care costs related to cancer treatment from SEER-Medicare data
Med Care
 , 
2002
, vol. 
40
 
(8 suppl)
 
IV-104-17
17.
Herr
HW
Sogani
PC
Does early cystectomy improve the survival of patients with high risk superficial bladder tumors?
J Urol
 , 
2001
, vol. 
166
 
4
(pg. 
1296
-
1299
)
18.
Stein
JP
Indications for early cystectomy
Semin Urol Oncol
 , 
2000
, vol. 
18
 
4
(pg. 
289
-
295
)
19.
American Medical Association
Current Procedure Terminology. CPT 2005, Standard Edition
 , 
2004
Chicago, IL
AMA Press
20.
Bach
PB
Guadagnoli
E
Schrag
D
Schussler
N
Warren
JL
Patient demographic and socioeconomic characteristics in the SEER-Medicare database applications and limitations
Med Care
 , 
2002
, vol. 
40
 
(8 suppl)
 
IV-19-25
21.
Hoel
DG
Ron
E
Carter
R
Mabuchi
K
Influence of death certificate errors on cancer mortality trends
J Natl Cancer Inst
 , 
1063
, vol. 
85
 
13
(pg. 
1063
-
1068
)
22.
Feuer
EJ
Merrill
RM
Hankey
BF
Cancer surveillance series: interpreting trends in prostate cancer—part II: cause of death misclassification and the recent rise and fall in prostate cancer mortality
J Natl Cancer Inst
 , 
1999
, vol. 
91
 
12
(pg. 
1025
-
1032
)
23.
Penson
DF
Albertsen
PC
Nelson
PS
Barry
M
Stanford
JL
Determining cause of death in prostate cancer: are death certificates valid?
J Natl Cancer Inst
 , 
2001
, vol. 
93
 
23
(pg. 
1822
-
1823
)
24.
Weinstock
MA
Reynes
JF
Validation of cause-of-death certification for outpatient cancers: the contrasting cases of melanoma and mycosis fungoides
Am J Epidemiol
 , 
1998
, vol. 
148
 
12
(pg. 
1184
-
1186
)
25.
World Health Organization
International Classification of Diseases for Oncology, Second Edition
 , 
1990
Geneva
World Health Organization
26.
Diez Roux
AV
Merkin
SS
Arnett
D
, et al.  . 
Neighborhood of residence and incidence of coronary heart disease [see comment]
N Engl J Med
 , 
2001
, vol. 
345
 
2
(pg. 
99
-
106
)
27.
Klabunde
CN
Potosky
AL
Legler
JM
Warren
JL
Development of a comorbidity index using physician claims data
J Clin Epidemiol
 , 
2000
, vol. 
53
 
12
(pg. 
1258
-
1267
)
28.
Charlson
ME
Pompei
P
Ales
KL
MacKenzie
CR
A new method of classifying prognostic comorbidity in longitudinal studies: development and validation
J Chronic Dis
 , 
1987
, vol. 
40
 
5
(pg. 
373
-
383
)
29.
Panageas
KS
Schrag
D
Riedel
E
Bach
PB
Begg
CB
The effect of clustering of outcomes on the association of procedure volume and surgical outcomes [see comment]
Ann Intern Med
 , 
2003
, vol. 
139
 
8
(pg. 
658
-
665
)
30.
Lin
DY
Cox regression analysis of multivariate failure time data: the marginal approach
Stat Med
 , 
1994
, vol. 
13
 
(21)
(pg. 
2233
-
2247
)
31.
Kleinbaum
D
Evaluating the proportional hazards assumption
Survival Analysis
 , 
1996
New York
Springer
(pg. 
129
-
166
)
32.
Olsen
LH
Genster
HG
Prolonging follow-up intervals for non-invasive bladder tumors: a randomized controlled trial
Scand J Urol Nephrol Suppl
 , 
1995
, vol. 
172
 (pg. 
33
-
36
)
33.
Soloway
MS
Sofer
M
Vaidya
A
Contemporary management of stage T1 transitional cell carcinoma of the bladder
J Urol
 , 
2002
, vol. 
167
 
4
(pg. 
1573
-
1583
)
34.
Shelley
MD
Court
JB
Kynaston
H
Wilt
TJ
Coles
B
Mason
M
Intravesical bacillus Calmette-Guerin versus mitomycin C for Ta and T1 bladder cancer
Cochrane Database Syst Rev.
 , 
2003
3
 
CD003231
35.
Stukel
TA
Fisher
ES
Wennberg
DE
Alter
DA
Gottlieb
DJ
Vermeulen
MJ
Analysis of observational studies in the presence of treatment selection bias: effects of invasive cardiac management on AMI survival using propensity score and instrumental variable methods
JAMA
 , 
2007
, vol. 
97
 
3
(pg. 
278
-
285
)
36.
Radford
MJ
Foody
JM
How do observational studies expand the evidence base for therapy? [comment]
JAMA
 , 
2001
, vol. 
286
 
10
(pg. 
1228
-
1230
)
37.
Heney
NM
Nocks
BN
Daly
JJ
, et al.  . 
Ta and T1 bladder cancer: location, recurrence and progression
Br J Urol
 , 
1982
, vol. 
54
 
2
(pg. 
152
-
157
)
The study sponsors had no role in the design of the study; the collection, analysis, or interpretation of the data; the writing of the manuscript; or the decision to submit the manuscript for publication.