Multiple biomarkers improve prediction of bladder cancer recurrence and mortality in patients undergoing cystectomy

Abstract

BACKGROUND.

Tested was whether the assessment of 5 established bladder cancer biomarkers (p53, pRB, p21, p27, and cyclin E1) could improve the ability to predict disease recurrence and cancer-specific survival after radical cystectomy in patients with pTa-3N0M0 urothelial carcinoma of the bladder (UCB).

METHODS.

The study comprised 191 patients with pTa-3N0M0 UCB treated with radical cystectomy and bilateral lymphadenectomy (median follow-up, 3.1 years). Biomarker expression was assayed on serial tissue microarray slides using quantitative immunohistochemistry using advanced cell imaging and color detection software. Predictive accuracy was quantified using the concordance index and 200-bootstrap resamples were used to reduce overfit bias. Bootstrap-adjusted predictive accuracy estimates were compared using the Mantel-Haenszel test.

RESULTS.

UCB recurred in 36 (18.8%) patients and 30 (15.7%) died of bladder cancer; 157 (82.2%) patients had altered expression of at least 1 biomarker. In univariate analyses the number of altered biomarkers had the highest predictive accuracy for both disease recurrence (76.8%, P < .001) and cancer-specific mortality (78.3%, P < .001). Addition of the number of altered biomarkers increased the predictive accuracy of nomograms based on the TNM staging system for disease recurrence and cancer-specific mortality by 10.9% (83.4% vs 72.5%, P < .001) and 8.6% (86.9% vs 78.3, P < .001), respectively.

CONCLUSIONS.

Assessment of the number of altered biomarkers in the cystectomy specimen improves the prediction of bladder cancer recurrence and survival in patients with pTa-3N0M0 disease. Prospective evaluation of alteration in these biomarkers can help identify patients who would benefit from adjuvant treatment after radical cystectomy. Cancer 2008. © 2007 American Cancer Society.

Bladder cancer is the fourth most common cancer in males and the ninth most common cancer in females, with 61,420 new cases and 13,060 deaths estimated for 2006 in the US.1 For patients with muscle-invasive urothelial carcinoma of the bladder (UCB) and those patients with high-risk nonmuscle invasive cancers who recur despite intravesical therapy, radical cystectomy with bilateral pelvic and iliac lymphadenectomy provides accurate staging and excellent local and regional control. Unfortunately, the 5-year all-cause survival rate in patients with pathologic muscle-invasive organ-confined UCB is only 60% to 75%.2–7 Failure to cure these patients is often due to the presence of occult metastases at the time of primary local therapy.

Whereas adjuvant chemotherapy is effective for patients with locally advanced UCB,8–10 the use of such treatment is not standard care for patients with pTa-3N0M0, as only a proportion of them are at risk for disease progression. We and others have previously shown that nomogram-based prediction models based on current histopathologic criteria can improve accuracy over AJCC-based stage groupings.11–14 However, these models were not perfect and there is still a need to improve the predictive accuracy for the individual patient based on biologic information. Accurate prediction of response to therapy, disease recurrence, and survival may aid individual patient counseling, frequency and extent of monitoring, and adjuvant therapy planning. To date, results of multiple adjuvant chemotherapy trials in bladder cancer patients who are at high risk for recurrence after surgery based on pathologic staging alone have been variable and inconsistent.

Over the past 2 decades the molecular dissection of cancer has increased our understanding of the pathways that are altered in neoplastic cells. Protein expression profiling of UCB offers an alternative means to distinguish aggressive tumor biology and may improve the accuracy of outcome prediction for patients treated with radical cystectomy. As key gatekeepers of the cell cycle, p53, pRB, p21, p27, and cyclin E1 are indispensable for urothelial genome stability and growth balance. Structural and functional defects of these biomarkers are common in human UCB and have been associated with poor oncologic outcomes.15–27

Several studies have shown that accumulation of altered expression of several of these immunohistochemical markers adds important prognostic information in patients treated with radical cystectomy.15–28 A clinical trial is under way testing whether alterations of p53 nuclear expression can be used prospectively to identify high-risk patients and whether there is a benefit to administering adjuvant methotrexate, vinblastine, doxorubicin (Adriamycin), and cisplatin (M-VAC) chemotherapy in p53-altered patients.25, 26, 29 In the present study we hypothesized that the expression of a panel of established cell cycle regulators (ie, p53, pRB, p21, p27, and cyclin E1) could help identify patients with pTa-3 N0M0 UCB who are at increased risk for disease progression and therefore would be candidates for systemic adjuvant chemotherapy. Therefore, we determined whether the immunohistochemical expression signature of these biomarkers could improve the accuracy of a nomogram that includes standard histopathologic features for prediction of disease recurrence and bladder cancer-specific survival in patients with pTa-3 N0M0 UCB at radical cystectomy.

MATERIALS AND METHODS

Patient Population

All studies were undertaken with the approval and oversight of the Institutional Review Board for the Protection of Human Subjects. The initial study population comprised 305 patients treated with radical cystectomy and bilateral pelvic lymphadenectomy as definitive treatment for clinically localized bladder UCB between March 1984 and December 2002. This study targeted patients with pTa-T3, lymph node-negative bladder UCB who were chemotherapy-naive. Therefore, 77 patients with metastases to lymph nodes, 21 patients with T4 disease, and 16 patients who received adjuvant chemotherapy were excluded from analyses. This left 191 patients for analysis. None of the patients received neoadjuvant chemo- or radiation therapy.

Histology, tumor grade, tumor stage, and presence of carcinoma in situ were confirmed by blinded review of the original pathology slides. For each patient, comprehensive clinical and pathologic data elements were collected and entered into an Institutional Review Board-approved database. Multiple data reviews and quality checks were performed to assure the accuracy and completeness of data elements. The median number of lymph nodes removed was 19 (range, 10–53).

Staff pathologists with expertise in genitourinary pathology examined all specimens according to previously published protocols.3, 7 The 2002 TNM classification was used for pathologic staging and the 1973 WHO classification was used for pathologic grading. None of the patients had positive margins.

Follow-up protocol was described elsewhere.3, 7 Briefly, patients were seen postoperatively at least every 3 to 4 months for the first year, semiannually for the second year, and annually thereafter. Detection of cancers in the ureter and urethra was coded as second primaries and not as local or distant recurrence.

Immunohistochemistry and Scoring

We performed cyclin E1, p53, p21, pRB, and p27 immunohistochemical staining using serial sections from the same paraffin-embedded tissue microarray blocks on the Dako Autostainer (Carpinteria, Calif). Staining and scoring protocols for cyclin E1, p53, p21, pRB, and p27 were performed as described elsewhere.20, 24, 27, 30 Multiple positive and negative control sections were included in each run.

We used brightfield microscopy imaging coupled with advanced color detection software (Automated Cellular Imaging System, Clarient, Calif). We obtained the mean, maximum, range, and standard deviation of staining intensity and percentage of positive nuclei/area measurements by using 10 random hotspots within each specimen. The mean of the triplicate cores was calculated for data analysis. Nuclear p53 immunoreactivity was considered altered when samples demonstrated at least 10% nuclear reactivity.16–18, 20, 25–27 p21 immunoreactivity was considered altered when samples demonstrated less than 10% staining.19, 20 pRB immunoreactivity was assigned to 1 of 3 categories of nuclear staining in the tumor cells: no nuclear reactivity; normal heterogeneous; and strong homogeneous (>50%). Tumors with no pRB expression and those with a strong homogeneous staining pattern were categorized as having altered pRB status, because it has been shown that overexpression of pRB in bladder cancer may be also indicative of dysfunctional pRB status.15, 20, 30 Nuclear p27 and Cyclin E were considered altered when samples demonstrated less than 30% nuclear reactivity.23, 27

Statistical Analyses

Hospital charts and physician records were reviewed, abstracted using standardized forms and entered into 1 database. Outcomes were measured by time to disease recurrence or to bladder cancer-specific survival. The cause of death was determined by the treating physicians, by chart review corroborated by death certificates, or by death certificates alone. To reduce bias in attribution of cause of death, only subjects who had bladder cancer listed in part I of the death certificate were considered to have died of bladder cancer for this study. Perioperative mortality (any death within 30 days of surgery or before discharge) was censored at time of death for bladder cancer-specific survival analyses.

Univariate recurrence and survival probabilities after cystectomy were estimated using the Kaplan-Meier method. Univariate and multivariable Cox regression models addressed time to recurrence and cancer-specific mortality after cystectomy. The number of altered biomarkers was categorized as 0–2 versus 3 versus 4–5. Biomarker categories 0, 1, and 2 were grouped due to paucity of events in these 3 categories. Biomarker categories 4 and 5 were grouped due to small numbers of patients in these categories. The same considerations were used to group stages Ta, Tis, and T1 as pT1 or lower disease and grades 1 and 2 as grade 1 or 2.

In all models proportional hazards assumptions were systematically verified using the Grambsch-Therneau residual-based test.31 Because a proportion of patients treated with radical cystectomy for invasive bladder cancer die of other causes, competing risk regression was used to test the significance of all the above variables in competing risks regression models that predicted bladder cancer-specific mortality after accounting for other-cause mortality.32

Multivariate Cox regression coefficients were then used to generate prognostic nomograms. Nomograms are graphic representation of a statistical model that incorporate multiple continuous variables to predict a patient's risk of developing a specific endpoint. The ability to simultaneously evaluate many variables is a major advantage of nomograms over standard risk-grouping models (ie, TNM or pathologic groupings).33, 34 Predictive accuracy of these nomograms was quantified with receiver operating characteristics-derived area under the curve (AUC) estimates.35, 36 In Cox regression models the AUC is substituted with the Harrell concordance index.35, 36 A value of 1.0 indicates perfect predictions, whereas 0.5 is equivalent to a toss of a coin. Internal validation was performed using 200-bootstrap resamples.35, 36 Predictive accuracy estimates were expressed as proportions and compared with the Mantel-Haenszel test. Calibration plots were generated to explore nomogram performance. All reported P-values are 2-sided, and statistical significance was set at .05. All statistical tests were performed with S-Plus Professional (MathSoft, Seattle, Wash).

RESULTS

Descriptive Characteristics

Table 1 describes the study cohort. Altered biomarkers were identified in 157 patients (82.2%). Of these, 52 (27.2%), 36 (18.8%), 38 (19.9%), 27 (14.1%), and 4 (2.1%) had 1 to 5 altered biomarkers, respectively. Recurrence after cystectomy occurred in 36 (18.8%) patients and bladder cancer-specific mortality occurred in 30 (15.7%). Death due to other causes occurred in 34 (17.8%) patients. Follow-up ranged from 0.1 to 16.3 years (mean 4.2, median 3.1).

Table 1. Descriptive Characteristics of 191 Patients With pTa-3 N0M0 Urothelial Carcinoma of the Bladder Treated With Radical Cystectomy and Bilateral Lymphadenectomy

Variables	No. (%)
Total	191 (100.0)
Age, y
Mean [median]	65.3 [66.4]
Range	33.8–87.7
Sex
Women	32 (16.8)
Men	159 (83.2)
Pathologic grade
1	1 (0.5)
2	17 (8.9)
3	173 (90.6)
Pathologic stage
Tis	14 (7.3)
Ta	4 (2.1)
T1	41 (21.5)
T2	86 (45.0)
T3	46 (24.1)
Lymphovascular invasion	42 (22.0)
Concomitant carcinoma in situ	85 (44.5)
No. of altered biomarkers
0	34 (17.8)
1	52 (27.2)
2	36 (18.8)
3	38 (19.9)
4	27 (14.1)
5	4 (2.1)
Bladder cancer recurrence	36 (18.8)
Bladder cancer-specific mortality	30 (15.7)
Other cause mortality	34 (17.8)

Overall, 41.3% of patients with pT3 disease had 0–2 altered biomarkers and 29% of patients with pTa-2 had 3–5 altered biomarkers. In patients with pT3 disease (n = 46), those who had 3–5 altered biomarkers were at significantly increased risk of bladder cancer recurrence and death compared with those who had 0–2 altered biomarkers (P = .025 and P = .014, respectively). Similarly, in patients with pTa-2 disease (n = 145), those who had 3–5 altered biomarkers were at significantly increased risk of bladder cancer recurrence and death compared with those who had 0–2 altered biomarkers (P < .001 and P < .001, respectively).

Recurrence Models and Nomogram

Figure 1A displays the estimated recurrence-free survival probability. Figure 1B displays the recurrence-free probability stratified according to number of altered biomarkers categorized as 0–2 versus 3 versus 4–5 (all intergroup comparison log-rank P < .001). Figure 1C displays the recurrence-free probability stratified according to pathologic stage. The following intergroup comparisons were statistically significant: T1 versus T3 (P < .001) and T2 versus T3 (P < .001).

Figure 1. Kaplan-Meier plot estimates of (A) overall recurrence-free survival probability (dotted lines, 95% confidence intervals), (B) recurrence-free survival probability stratified according to the number of altered biomarkers, (C) recurrence-free survival probability stratified according to pathologic stage, (D) overall bladder cancer-specific survival probability (dotted lines, 95% confidence intervals), (E) bladder cancer-specific survival probability stratified according to the number of altered biomarkers, (F) the overall bladder cancer-specific survival probability stratified according to pathologic stage.

Table 2 shows the univariate and multivariate Cox regression models for prediction of bladder cancer recurrence after radical cystectomy. In univariate analyses, the number of altered biomarkers, age, pathologic T stage, and lymphovascular invasion were associated with disease recurrence (P < .001). Assessment of regression coefficients in the full multivariate model revealed that patients with 3 and 4–5 altered biomarkers were 3.8 and 11.2 times more likely to experience disease recurrence than patients with 0–2 altered biomarkers, respectively. By comparison, patients with pT3 disease were 1.8 times more likely to experience disease recurrence than patients with pT1 disease.

Table 2. Univariate and Multivariate Cox Regression Models and Competing Risk Models for Prediction of Bladder Cancer Recurrence in 191 Patients With pTa-3 N0M0 Urothelial Carcinoma of the Bladder Treated With Radical Cystectomy and Bilateral Lymphadenectomy

Predictors	Cox regression models					Competing risk regression models
	Univariate		Multivariate			Univariate	Multivariate
	RR; P	% Predictive accuracy	Model without biomarkers	Full model with biomarkers	Reduced model with biomarkers	P	P
			RR; P	RR; P	RR; P
RR indicates risk ratio.
Sex	0.6; .2	50.1	1.0; 1.0	1.6; .3	—	.3	.8
Age	1.1; .01	61.7	1.0; .1	1.0; .7	—	.01	.9
Pathologic stage	—; <.001	66.8	—; .004	—; .08	—; .04	—	—
T2 vs T1	0.9; 0.9	—	0.8; .6	0.7; .5	0.9; .9	.9	.6
T3 vs T1	5.0; <.001		2.9; .041	1.8; .3	2.4; .08	<.001	.3
Pathologic grade (3 vs 1 and 2)	2.3; .2	53.4	2.0; .4	1.9; .4		.3
Lymphovascular invasion	4.2; <.001	66.4	2.3; .035	2.3; .046	1.9; .1	<.001	.4
Concomitant carcinoma in situ	0.7; .4	53.0	0.7; .4	0.6; .2		.3	.3
No. of altered biomarkers	—; <.001	76.8	—	—; <.001	—; <.001	—	—
3 vs 0–2	5.4; <.001	—	—	3.8; .006	4.0; .004	<.001	.008
4–5 vs 0–2	15.2; <.001	—	—	11.2; <.001	10.7; <.001	<.001	<.001
% Predictive accuracy	—	—	72.5	81.7	83.4	—	—

Assessment of univariate predictive accuracy estimates revealed that the number of altered biomarkers was the most informative predictor (76.8%). Pathologic stage (66.8%) and lymphovascular invasion (66.4%) were the second and the third most informative predictors, respectively.

In multivariate recurrence analyses that adjusted for the effects of all variables, the number of altered biomarkers was independently associated with disease recurrence (P < .001). The predictive accuracy of a multivariate model that did not include the marker status was 72.5%. Addition of the number of altered biomarkers increased the predictive accuracy of the standard model to 81.7% and that of the most predictive and parsimonious model to 83.4%. The difference in predictive accuracy between the model with and without number of altered biomarkers was statistically significant (10.9%, 95% confidence interval [CI], 10.2–11.6%, P < .001).

Table 2 shows the univariate and the multivariate competing risk regression models for prediction of disease recurrence. These models take the risk of death secondary to noncancer causes into account. In multivariate competing risk regression models, only 3 (P = .008) or 4–5 (P < .001) altered biomarkers were independently associated with disease recurrence.

The nomogram predicting the probability of bladder cancer recurrence-free survival after radical cystectomy is shown in Figure 2A. Its calibration plot, which compares the nomogram predicted probabilities to the observed rate of recurrence-free survival at 1, 2, and 5 years, is shown in Figure 2B. Assessment of the nomogram axes indicates that the number of altered biomarkers contributes the highest number of risk points. For example, 3 altered biomarkers contributes 60 risk points and exceeds the contribution of pathologic stage T3 disease (40 risk points) or that of lymphovascular invasion (27 risk points). The presence of 4–5 altered biomarkers contributes 100 risk points.

Figure 2. (A) Postoperative nomogram predicting 1-, 2-, and 5-year risk of disease recurrence in patients with pTa-3 N0M0 urothelial carcinoma of the bladder treated with radical cystectomy and bilateral lymphadenectomy. Instructions for physicians: Locate the patient's T stage on the sex axis. Draw a straight line up to the points axis to determine how many points toward recurrence the patient should receive. Repeat this process for each of the remaining axes, drawing a straight line each time to the points axis. Sum the points received for each predictive variable and locate this number on the total points axis. Draw a straight line down from the total points to 1 of the recurrence-free prediction (RFS) axes for the patient's specific risk of remaining free from recurrence for 1, 2, and 5 years. (B) Calibration plot of the postoperative nomogram predicting risk of disease recurrence after radical cystectomy and bilateral lymphadenectomy. (C) Postoperative nomogram predicting 1-, 2-, and 5-year risk of cancer-specific survival in patients with pTa-3 N0 M0 urothelial carcinoma of the bladder treated with radical cystectomy and bilateral lymphadenectomy. Instructions for physicians: Locate the patient's T stage on the sex axis. Draw a straight line up to the points axis to determine how many points toward recurrence the patient should receive. Repeat this process for each of the remaining axes, drawing a straight line each time to the points axis. Sum the points received for each predictive variable and locate this number on the total points axis. Draw a straight line down from the total points to 1 of the RFS axes for the patient's specific risk of remaining free from recurrence for 1, 2, and 5 years. (D) Calibration plot of the postoperative nomogram predicting risk of cancer-specific survival after radical cystectomy and bilateral lymphadenectomy.

The performance characteristics of the recurrence nomogram indicate that 2- and 5-year predictions virtually parallel the observed recurrence-free rates.

Bladder Cancer-specific Survival Models and Nomogram

Figure 1D displays the estimated bladder cancer-specific survival probability. Figure 1E displays the bladder cancer-specific survival probability stratified according to number of altered biomarkers categorized as 0–2 versus 3 versus 4–5 (all intergroup comparison log-rank P < .001). Figure 1F displays the bladder cancer-specific survival probability stratified according to pathologic stage. The following intergroup comparisons were statistically significant: T1 versus T3 (P < .001) and T2 versus T3 (P < .001).

Table 3 shows the univariate and multivariate Cox regression models for prediction of bladder cancer-specific survival after radical cystectomy. In univariate analyses the number of altered biomarkers, age, pathologic T stage, and lymphovascular invasion were associated with the risk of dying of bladder cancer (all P ≤ .01).

Table 3. Univariate and Multivariate Cox Regression Models and Competing Risk Models for Prediction of Bladder Cancer-specific Survival in 191 Patients With pTa-3 N0M0 Urothelial Carcinoma of the Bladder Treated With Radical Cystectomy and Bilateral Lymphadenectomy

Predictors	Cox regression models					Competing risk regression models
	Univariate		Multivariate			Univariate	Multivariate
	RR; P	% Predictive accuracy	Model without biomarkers	Full model with biomarkers	Reduced model with biomarkers	P	P
			RR; P	RR; P	RR; P
RR indicates risk ratio.
Sex	0.7; .4	51.5	1.3; .6	2.6; .09	—	.5	.4
Age	1.1; .01	61.9	1.1; .03	1.1; .3	—	.02	.9
Pathologic stage	—; <.001	74.3	—; 0.002	—; .03	—; .03	—	—
T2 vs T1	2.1; 0.3	—	2.1; .3	2.0; .3	2.0; .3	.3	.3
T3 vs T1	11.7; <.001	—	7.3; .004	4.8; .02	4.8; .02	<.001	.02
Lymphovascular invasion	6.3; <.001	71.5	2.8; .01	2.9; .02	2.3; .046	<.001	.1
Concomitant carcinoma in situ	0.7; .4	52.2	0.9; .9	0.8; .7	—	.3	.6
No. of altered biomarkers	—; <.001	78.3	—	—; <.001	—; <.001	—	—
3 vs 0–2	5.8; .001	—	—	3.4; .03	3.9; .01	.002	.022
4–5 vs 0–2	17.9; <.001	—	—	12.5; <.001	11.2; <.001	<.001	<.001
% Predictive accuracy	—	—	78.3	85.4	86.9	—	—

In univariate analyses the number of altered biomarkers had the highest predictive accuracy for bladder cancer-specific survival (78.3%), followed by pathologic stage (74.3%) and lymphovascular invasion (71.5%).

In multivariate analyses the number of altered biomarkers (P < .001), pathologic stage (P = .03), and lymphovascular invasion (P = .02) were independently associated with bladder cancer-specific mortality.

The predictive accuracy of the multivariate model without the number of biomarkers was 78.3%. Addition of the number of altered biomarkers increased the predictive accuracy of the model to 85.4% and that of the most predictive and parsimonious model to 86.9%. The difference in predictive accuracy between the model with and without the number of altered biomarkers was 8.6% (95% CI, 7.6–8.8%, P < .001).

Table 3 also shows univariate and multivariate competing risks regression models for prediction of cancer-specific mortality. In multivariate competing risks regression models, only 3 (P = .008) or 4–5 (P < .001) altered biomarkers and pathologic stage T3 (P = .02) were independent predictors of bladder cancer-specific mortality.

The nomogram predicting the probability of bladder cancer-specific survival after radical cystectomy is shown in Figure 2C. The calibration plot of bladder cancer-specific mortality at 1, 2, and 5 years are shown in Figure 2D. Assessment of the nomogram axes indicates that the number of altered biomarkers contributes the highest number of risk points. For example, pathologic T3 contributes approximately 65 risk points versus 100 points for 4–5 altered biomarkers. The performance characteristics of the bladder cancer-specific mortality nomogram indicate that 1-, 2-, and 5-year predictions virtually parallel the observed bladder cancer-specific mortality rates.

DISCUSSION

Despite the increasing number of published studies that have added to the general knowledge about the best candidate for adjuvant chemotherapy after radical cystectomy, physicians and patients have had few tools to help them translate this body of general knowledge into individualized, evidence-based recommendations. Patients with lymph node involvement and/or metastatic disease are usually coun-seled in favor of chemotherapy. Yet the majority of patients who undergo radical cystectomy with bilateral pelvic lymphadenectomy have organ-confined disease (pTa-T3N0M0) and are infrequently provided with multimodal therapy. Such patients are a heterogeneous group encompassing a wide array of biologic behaviors. Despite apparent cure with extirpative surgery a significant number of these patients experience disease recurrence and ultimately succumb to disease progression. Whereas there is a suggestion that adjuvant chemotherapy may improve disease-free survival in this population,8–10 the accumulated data are underpowered and insufficient to base clinical decision-making.37, 38 This leaves the practitioner and patient in a quandary when discussing clinical outcomes after radical cystectomy as well as the potential benefit from adjuvant therapy. Molecular biomarkers may provide a better understanding of the biology of an individual's tumor and may help stratify the heterogeneous pTa-3N0M0 patient population into risk groups that can be used to guide clinical decision-making regarding observation versus adjuvant therapy. Moreover, accurate individual risk attribution via molecular staging may assist in the development of entry criteria for clinical trials.

We found that the addition of a panel of 5 well-established cell cycle regulatory biomarkers improved the predictive accuracy of competing-risk nomograms for prediction of bladder cancer recurrence and survival after cystectomy for patients with pTa-pT3 lymph node-negative tumors by a prognostically and clinically significant margin. Alterations in cell cycle regulators were common, with 82% of patients exhibiting at least 1 altered biomarker, 20% exhibiting 3 altered biomarkers, and 16% exhibiting 4 or 5 altered biomarkers. Patients with 3 or more altered biomarkers had a 4 to 10 times elevated risk of bladder cancer recurrence and mortality after radical cystectomy. Overall, less than 10% of patients with less than 3 altered biomarkers experienced disease recurrence within 5 years of surgery compared with 30% and 65% of patients with 3 and 4–5 altered biomarkers, respectively. Patients with 0–2 altered markers, regardless of pathologic stage, have excellent survival rates and potentially could be spared from unnecessary chemotherapy. Conversely, patients with 4–5 altered markers have a poor prognosis and therefore they may benefit the most from early aggressive therapy. Nomograms that incorporate pathologic and molecular information could form the basis for counseling patients regarding their risk of disease recurrence after surgery and for designing clinical trials to test adjuvant treatment strategies in high-risk patients.

Alterations in the expression of multiple biomarkers are thought to be necessary to adversely affect clinical outcomes after radical cystectomy. UCB is a multistep genetic process wherein individual alterations of molecular determinants may have only a restricted role. We and others have previously shown that altered expressions of cyclin E1, p53, pRB, p21, and p27 are independent determinants of prognosis that can act cooperatively/synergistically to promote tumor progression.19–21, 24, 27 In agreement with these findings, we found that patients with 3 or more altered biomarkers had a statistically and clinically significantly higher risk of disease recurrence or bladder cancer-specific mortality compared with patients with 0–2 altered biomarkers. The choice of the biomarkers was based on previous studies establishing their independent prognostic value in traditional multivariable analyses.15–28 Moreover, these biomarkers are commonly used in pathology departments and most pathologists are familiar with immunohistochemical staining and interpretation of these biomarkers. However, combinations of other biomarkers reflecting the functional status of different pathways/phenomena associated with bladder cancer progression may provide similar or even greater predictive accuracy.

This study has several potential limitations. First and foremost are the limitations inherent to any retrospective data collection. The sample size and relatively short follow-up may have limited our ability to detect small differences attributed to other variables. However, several large series have shown that bladder cancer recurrence and associated mortality is most likely to occur within the first 2 years of surgery. Hence, our median follow-up of 3.1 years should have been adequate to observe our intended endpoints.2–7

Another potential limitation is the questionable reliability of immunohistochemical techniques. Immunohistochemistry is semiquantitative and highly dependent on a range of variables such as choice of antibody, antibody concentration, fixation techniques, variability in the interpretation and stratification criteria, and inconsistency in specimen handling and technical procedures. To reduce the number of variables in immunohistochemistry analysis, we chose to use tissue microarrays with an automated autostainer and an automated scoring system based on brightfield microscopy imaging coupled with advanced color detection software. These techniques have been shown to result in experimental standardization and greater reproducibility.39–42

Finally, our prediction models provide additional guidance, but do not replace the decision-making process between the clinician and his or her patient with regard to administration and regimen of adjuvant chemotherapy. Our findings of the profound predictive capacity of cell cycle protein expression in individuals with bladder UCB treated with surgery need to be externally validated in prospective multicenter trials. At UT Southwestern Medical Center Dallas we have started prospectively performing immunohistochemical on all patients with high-grade or invasive disease (Fig. 3).

Figure 3. Possible pathology report for sample patient treated with radical cystectomy and bilateral lymphadenectomy for bladder cancer.

Conclusions

Patients and physicians desire reliable knowledge regarding the risks of cancer-associated death and disease progression after radical cystectomy. We have developed and internally validated highly accurate competing-nomograms that predict the risks of bladder cancer recurrence and survival after radical cystectomy in patients with pTa-3N0M0 UCB. These nomograms integrate the immunohistochemical status of 5 established cell cycle regulatory biomarkers with standard histopathologic variables. These nomograms may aid the decision-making process regarding the prospect of adjuvant chemotherapy in patients with pTa-3N0M0 disease after radical cystectomy.