Although multiple sequential recurrences are one of the most important characteristics of superficial transitional cell carcinoma (TCC) of the bladder, few studies have examined multiple sequential recurrence patterns and the clinicopathologic and biochemical factors associated with these patterns.
Two hundred seventy superficial TCC bladder carcinoma patients were followed. Clinical, pathologic, and tumor marker (p53, MIB-1, bcl-2, c-erb B-2, and epidermal growth factor receptor) data were collected at baseline and during followup. The Kaplan-Meier (KM) method was used to describe multiple recurrences. The Wei, Lin, and Weissfeld (WLW) marginal proportional hazards model was used to assess the effects of clinicopathologic and immunohistochemic factors on multiple recurrences.
Among the 270 patients, 126 (46.7%) had one or more recurrences, 38 (14.1%) had two or more recurrences, and 14 (5.2%) had three or more recurrences during the followup. The median times for the first, the second, and the third recurrences were 23 months, 15 months, and 13 months, respectively. In KM analysis, Stage T1, higher grades, and Ki-67 stain positivity were associated with the first recurrence, and p53 stain positivity was marginally significant. Other markers were not significant. For the second recurrence, only p53 stain positivity was significant. In multivariate analysis (WLW method), stage was significantly associated with the first recurrence (risk ratio [RR] = 2.03), and Ki-67 was marginally significant (RR = 1.49). For the second recurrence, only p53 positivity was statistically significant (RR = 2.73).
Bladder carcinoma is the second most common malignancy encountered by urologists. It is estimated that in 2001, 54,300 new cases of bladder carcinoma will be diagnosed in the United States, and approximately 12,400 deaths will be attributed to the cancer.1 Fortunately, approximately 80% of patients with newly diagnosed bladder carcinoma present with superficial disease (Ta/T1), which can be managed with transurethral resection alone or with intravesical therapy.2 However, more than 50% of the patients will have recurrences, and more than 20% of them will eventually die of the disease.3 If the prognosis of an individual patient could be predicted earlier in the course of the disease, patients with a high likelihood of recurrence could be treated more aggressively, and patients who could be clinically cured with local therapy would be spared from unnecessary radical procedures.
Many investigations have been done to identify prognostic factors for the first recurrence of superficial transitional cell carcinoma (TCC) of the bladder after initial treatment. These prognostic factors include the number, size, grade, and stage of the primary tumor, as well as various biologic markers.4 Undoubtedly, these prognostic factors provide invaluable information for clinicians to manage their patients. However, while multiple sequential recurrences are one of the characteristics of bladder carcinoma during its course, very few studies have systematically investigated multiple recurrences of the cancer. To date, the patterns of multiple sequential recurrences of bladder carcinoma have not been well documented in the literature, and factors associated with the multiple sequential recurrence patterns have not been defined. A salient description of the patterns of multiple sequential recurrences can help us better understand the disease course, and, more importantly, the identification of clinicopathologic and biochemical factors for multiple sequential recurrences can help us better select patients for more aggressive treatments. The current article describes multiple sequential recurrence patterns among superficial bladder carcinoma patients and identifies clinicopathologic and biochemical factors associated with the patterns.
Patients for the current study came from the Washington University Division of Urological Surgery and Division of Surgical Pathology. Between January 1994 and April 1999, 270 consecutive patients with superficial TCC of the bladder without carcinoma in situ (CIS) were diagnosed and treated in the two divisions. These patients were followed prospectively with cystoscopy and urinary cytology every three months for the first two years, then every six months for another two years and annually thereafter. For patients with tumor recurrence, the followup scheme restarted. Clinicopathologic and biochemical data were collected at the time of diagnosis and initial treatment for all 270 patients. Initial treatment of superficial TCC of the bladder included transurethral resection alone or with intravesical Bacille Calmette-Guerin (BCG) for patients with more aggressive tumors. During the followup, clinicopathologic and biochemical marker data were obtained for patients who had cancer recurrence. Informed consent was obtained from all study patients, and the study procedures were approved by the Washington University Human Studies Committee.
Clinicopathologic data included tumor stage, grade, and the presence of upper tract tumors. For those with multiple tumors, only the tumor with the greatest level of invasion was used. Patients were pathologically staged according to the American Joint Committee on Cancer system.5 Papillary epithelial confined tumors were classified as Stage Ta, and tumors that had invaded the lamina propria or submucosa were classified as Stage T1. All tumors were graded by one of the authors (P.A.H.) according to the World Health Organization scheme of Grades 1-3, corresponding to well differentiated, moderately well differentiated, and poorly differentiated carcinomas, respectively.6 Biochemical data included five commonly utilized tumor markers: p53, Ki-67, bcl-2, epidermal growth factor receptor (EGFR), and c-erb B-2. The procedures for immunohistochemic staining and the definition of positivity for these markers have been reported previously.7 Briefly, p53 was positive if ≥ 20% of the tumor nuclei reacted for p53, and Ki-67 was positive if ≥ 25% of tumor nuclei reacted for MIB-1. Epidermal growth factor receptor, bcl-2, and c-erb B2 were positive if there was any strong staining of cytoplasmic or membranous areas of the tumor cells. Forty three patients in the current study did not have enough tissue to perform all immunohistochemic analyses. When the effects of clinicopathologic factors on tumor recurrences were investigated, these 43 patients were included in the analyses; when effects of markers on the recurrences were examined, the patients with missing marker values were excluded from the analyses.
Eligibility of patients in the study was checked at the time of tumor recurrences. A patient was excluded from the study from the time when we found his/her tumor stage was more advanced than T1, but the patient's information up to and including that time was used in the current data analysis. The outcome of interest was the time to each of multiple sequential tumor recurrences. The three examples below will help clarify the dynamic followup process with multiple sequential recurrences as the outcome of interest and the data structure used in the current study.
Example 1. Patient A was diagnosed with a Stage Ta tumor and treated on January 1, 1994. Then, on September 1, 1994, he had his first tumor recurrence, and the recurrent tumor was staged as Ta. Twelve months later, he had his second tumor recurrence, with Stage T1. Nine months after that, he had his third recurrence, with Stage T2. If the first three tumor recurrence times were of interest, then all the information from the patient would be included in the data analysis.
Example 2. If Patient B had only one recurrence with tumor stage lower than T2, then the first recurrence time was used. The time from the first recurrence up to the time of data analysis was treated as a censoring observation for the second recurrence.
Example 3. If Patient C had a first tumor recurrence with Stage T2, then the first tumor recurrence time was included in the data analysis. After that, Patient C was excluded from the study, since by definition his tumor was not superficial TCC anymore.
Statistical data analyses included simple descriptive statistics and an estimation of the tumor recurrence free survival functions using traditional Kaplan-Meier product limit methods for each of multiple sequential recurrences. The recurrence free survival functions were compared for statistically significant differences by the log-rank test. The Wei-Lin-Weissfeld (WLW) marginal proportional hazards model was used to assess independent effects of clinicopathologic-biochemical factors on multiple sequential tumor recurrences.8 This statistical method models each of the sequential recurrences by a Cox proportional hazards model, taking into consideration correlation among patient multiple recurrence times without imposing any specific structure on the correlation. The effect of an individual factor in the presence of other factors on the recurrences was determined by G statistics (−2 log likelihood ratio with and without the factor in the model) and the corresponding degree of freedom. We tested correctness of fit of the final model by comparing the likelihood from the null model and the likelihood from the final model. Proportional hazards assumption was examined by plotting log (−log tumor free survival function) versus time for different levels of each variable in the final model. All data analyses were conducted using SAS software and a publicly-distributed SAS Macros (PHRWLW) modified for the current study.9
Table 1 presents the characteristics of 270 superficial bladder carcinoma patients at diagnosis and initial treatment. Among the 270 patients, most were male (73%) and white (71%). The mean patient age was 71 years, with a standard deviation of 11.9, and the median followup time was 19 months, with a range of 1-54 months. At initial diagnosis, about 80% of patients had Stage Ta tumors and 68% had Grade 2 tumors. Upper tract tumors were present in about 12% of the patients, and intravesical therapy (BCG) was used in about a quarter of the patients. Among the five tumor markers, EGFR had the highest proportion of positivity (73%), followed by c-erb B-2 (65.2%), while bcl–2 had the lowest proportion of positivity (16.7%).
Table 1. Characteristics of 270 Patients with Superficial TCC of the Bladder
A dynamic process of followup and multiple sequential tumor recurrence is depicted in Figure 1. Among the 270 patients, 126 had at least one tumor recurrence during followup. Ninety one out of the 126 patients were verified as Stage Ta or T1 tumors at first recurrence. Among the 91 patients, 38 had at least two tumor recurrences, with 35 patients being confirmed as Ta or T1 tumors at the second recurrence. Among these 35 patients, 14 had at least three tumor recurrences.
Figure 2 depicts the multiple sequential tumor recurrence process using Kaplan-Meier product limit tumor recurrence free survival functions. Tumor free survival functions A, B, and C describe the first, second, and third tumor recurrences, respectively. Time 0 for the three curves represents different points during followup: for the first recurrence, Time 0 indicates the time at initial diagnosis and treatment of primary tumor; for the second recurrence, Time 0 is the time at diagnosis and treatment of the first recurrent tumor; for the third recurrence, the time starts from diagnosis and treatment of the second recurrent tumor. Curve C goes down faster than Curve B, which in turn goes down faster than Curve A, indicating that tumor recurrence occurs faster as the number of recurrences increases. In fact, the median time for the first tumor recurrence was 23 months, and for the second and third recurrences, the times were 15 months and 13 months, respectively.
In order to get valid and reliable estimates of effect of clinicopathologic and biochemical factors on tumor recurrences, we only used the first two recurrences in our following analysis, since the number of third recurrence was too small.
Table 2 summarizes the Kaplan-Meier analyses for the first two recurrences by presenting the median time for tumor recurrences and corresponding P values from the log-rank test. After the initial treatment of primary tumors, patients with Stage Ta were less likely to have tumor recurrence than patients with Stage T1. In fact, the median time to first recurrence for patients with a Ta tumor was more than three times longer than that for patients with a T1 tumor. Patients with a higher grade tumor were likelier to have a first recurrence than patients with a low grade tumor. The median time to recurrence for patients with Grade 1 tumors was about three times longer than that for patients with Grade 3 tumors (37 months vs. 13 months). Ki-67 stain positivity increased the risk for a first tumor recurrence, as did p53 stain positivity, even though the statistical significance was marginal. The other three tumor markers, c-erb B-2, bcl-2, and EGFR, were not significantly associated with a first tumor recurrence.
Table 2. Median Times of Tumor Recurrences by Levels of Clinicopathologic and Biochemical Factors for 270 Patients with Superficial TCC of the Bladder
For second tumor recurrence, the effects of tumor stage, tumor grade, and Ki-67 on tumor recurrence were much smaller than their effects on first tumor recurrence. The median time for Ta patients was 15 months, compared to 13 months for T1 patients, and the median time for G2 patients was 14 months, compared to 16 months for G3 patients. The tumor stage and grade effects were not statistically significant. Patients with and without upper track tumor(s) had almost the same median time. Among all the variables, only the effect of p53 was statistically significant. Patients with a positive p53 stain were likelier to have a second tumor recurrence, compared to patients with a negative stain, with the median times to the second recurrence being 6 months and 18 months, respectively.
Table 3 presents risk ratios and the corresponding 95% confidence intervals for important factors from the final WLW marginal model. The criterion for a variable to be included in the final model was that the variable must have a statistically significant effect on the first or second tumor recurrence. The last column of the table is the common risk ratio, which describes the effect of a variable on the first and second tumor occurrences. A common risk ratio was estimated only for a variable with effects on the first and second recurrences that were not significantly different from each other. The 95% confidence intervals for the effects of all variables were adjusted for the correlation between the first and second recurrence times by the WLW method. Three variables (tumor stage, p53 and Ki-67) were included in the final model; among these three variables, common risk ratios were estimated for tumor stage and Ki-67. No meaningful interaction was found among the three variables in the final model, and reasonably parallel lines of log (−log tumor recurrence survival function) of the three variables suggested no evidence for a violation of the proportional hazards assumption.
Table 3. Relative Risk and 95% CI of Tumor Recurrences from the Final WLW Model
Compared to patients with Stage Ta tumors, patients with Stage T1 tumors were over twice as likely to experience a first tumor recurrence and were 1.74 times likelier to have a second tumor recurrence, with a common risk ratio of about 2 (P = 0.00). p53 stain positivity did not increase the risk for a first recurrence but significantly increased the risk for a second recurrence by 173% (P = 0.03). Ki-67 stain positivity increased the risk of tumor recurrence by 49% for the first recurrence and by 33% for the second recurrence, with a common effect of 45% (P = 0.05).
To our knowledge, there are very few studies systematically examining multiple sequential recurrences of superficial bladder carcinoma in the medical literature. The current results suggest that multiple sequential recurrences are common phenomena among patients with superficial bladder TCC. The times to tumor recurrences tend to be faster and faster as the number of recurrences increases. Furthermore, the current results suggest two tumor markers -Ki-67 and p53- have independent effects on tumor recurrences, in addition to the traditional clinical and pathologic factors.
Tumor stage and grade are the two most frequently cited clinical and pathologic predictors for tumor recurrence and tumor progression in the medical literature. Stage T1 tumors should be considered potentially aggressive, particularly if they are high grade.10 Furthermore, Stage T1 tumors are sometimes incompletely resected, even when the endoscopist thinks that complete resection has been accomplished. In one study from Germany, more than 40% of T1 tumors believed to be completely resected were found on resection at six weeks to have residual disease.11 Both potential aggressiveness and possibility of incomplete resection of Stage T1 tumors may in part explain why patients with Stage T1 tumors were likelier to have tumor recurrences than patients with Stage Ta tumors. Tumor grade was shown to be an important predictor for tumor recurrence in Kaplan-Meier analysis, but not in the final WLW multiple regression model. The current results do not imply that tumor grade is not an important factor for tumor recurrences. In patients in the current study, the effect of tumor grade was not as strong as tumor stage, and tumor grade was highly associated with tumor stage. In fact, among patients with Stage T1 tumor, none had well differentiated tumors and 40% had poorly differentiated tumors, while among patients with Stage Ta tumors, 26.5% had well differentiated tumors and 3.7 % had poorly differentiated tumors (P < 0.001). In the multivariate analyses, therefore, the effects of tumor grade on tumor recurrences were greatly compromised by the presence of tumor stage in the model, resulting in no statistically significant effects of tumor grade.
Patients with a positive Ki-67 stain were shown to have a higher risk of tumor recurrence than patients with the negative stain in the current study, consistent with other studies.12–16 Ki-67 is a murine monoclonal antibody reacting with a nuclear antigen expressed in proliferating cells and is thought to recognize a nuclear protein involved in the DNA replicase complex. In the current literature, Ki-67 is reported as one of the most promising cellular proliferation markers. Increased expression of this antigen indicates a higher level of proliferative activity in tumor cells and is associated with more aggressive tumors with increased propensities for recurrence, progression, and metastasis.
Patients with a positive p53 stain were shown to have a higher risk of a second tumor recurrence, compared to patients with a negative stain, but not for the first tumor recurrence. Findings about the relationship between p53 overexpression and subsequent tumor recurrence or progression in superficial bladder carcinoma have been contradictory. Although some authors have shown an association of p53 overexpression with tumor progression,17 others have reported no association.18, 19 Mutations in the p53 gene are the most common genetic defect in human tumors. The p53 gene functions as a tumor suppressor gene and has a vital role in the regulation of the cell cycle.20 Our finding that overexpression of p53 is linked to an increased risk of a second tumor recurrence but not for a first tumor recurrence suggests possible differences in tumor biology between the initial tumors and the first recurrent tumors.
In contrast to Ki-67 and p53, the effects of overexpression of bcl-2, c-erb B-2, and EGFR on tumor recurrence were not important in the current study. Although some studies have reported a prognostic value for c-erb B-2 expression in human bladder carcinoma,21, 22 others have suggested otherwise.23–25 Conflicting results about the relationship between bcl-2 and tumor recurrence and progression also exist in the literature.26, 27 Overexpression of EGFR correlates with an increase in cancer specific death, significantly short intervals to recurrence, higher rates of recurrence, and increased rates of progression in patients with superficial bladder tumors.28–30 However, the current study suggests that the effect of overexpression of EGFR on tumor recurrence is smaller than the effects of Ki-67 and p53.
Many studies have shown that BCG therapy reduces the risk of recurrence and progression of superficial TCC.31–35 The observational nature of the current study precluded us from investigating the effects of BCG on tumor recurrences. BCG is routinely indicated for patients with more aggressive tumors to prevent tumor recurrence, not randomly assigned to superficial bladder carcinoma patients. In our institute, BCG is given to all patients with CIS, and other relative indications include multifocality, T1 lesions, and high grade carcinoma. Such selective treatment of high risk patients means the high risk patients have recurrence experiences similar to the low risk patients. In the current study cohort, the median times for the first recurrence were 14 and 23 months for patients with and without BCG treatment, respectively (P = 0.33), the median time for the second recurrence was 16 months for both groups. Thus, the current data does not provide any evidence against the effectiveness of BCG treatment in delaying tumor recurrence. An interesting finding from the current study is that the use of BCG may affect expression of tumor markers. The marker positivity in subsequent recurrences is likelier to change among patients with BCG treatment than among patients without BCG treatment. For example, among patients with BCG treatment, 87% of those initially positive for p53 changed to negative and 33% of those initially positive for Ki-67 changed to negative. Conversely, among patients without BCG treatment, only 44% of those positive for p53 changed to negative and 28% of those positive for Ki-67 changed to negative. The differences in changes in marker positivity between patients with and without BCG treatment suggest possible differences in mechanisms of recurrence after BCG treatment. Therefore, the marker usefulness for predicting recurrence may change with subsequent recurrences.
During followup, there were only 15 (5.6%) patients who developed invasive tumors. Thirteen of them were detected at the first recurrence, and three were found at the second recurrence. The small number of patients with invasive tumors precluded a formal statistical analysis of the relationship between the markers and tumor progression. However, a rough comparison suggests that marker positivity is associated with the development of an invasive tumor. For example, 11.1% of the patients initially positive for p53 developed invasive tumors, compared to only 4.2% of the patients initially negative (P = 0.06, Fisher exact test). Similarly, about 14% of the patients who were initially positive for Ki-67 developed invasive tumors, compared to only 2% of those who were initially negative (P < 0.01, Fisher exact test). In the current study, the tumor markers did not seem to be associated with the ability of BCG to prevent progression. That is, the effect of BCG treatment on tumor progression was independent of marker positivity. For example, among patients initially negative for p53, 2.7% of those with BCG treatment developed invasive tumors, compared to 0% of those initially positive for p53. Among patients initially negative for Ki-67, none developed invasive tumors, compared to 5% of those initially positive for Ki-67. Because of the very small number of invasive tumors, these results must be interpreted with caution. The fact that the tumor progression rate is much lower than tumor recurrence calls for multicenter collaborative efforts to adequately address the tumor progression issue.
The current study has several limitations, one of which is the possibility that we missed other important factors: the number and size of tumor at initial diagnosis and treatment. In one study, tumors > 5 cm had a 35% progression rate, as compared with a 9% rate for smaller tumors.36 The number of tumors is important for predicting recurrence. Recurrence rates in patients with single tumors range from 18% to 60%, as opposed to rates of 40–90% in those with multiple tumors.37 However, the pathologic characteristics of superficial bladder tumors that have the greatest prognostic significance are tumor grade and tumor stage.2 For patients with multiple tumors in the current study, the tumor with the greatest level of invasion was used for staging and grading. In addition to tumor stage and grade, the presence of upper tract tumors was included in the current data analyses. Thus, possible effects of the number and size of tumors may be partially if not completely explained by tumor stage, tumor grade, and the presence of upper tract tumors in the current data. Therefore, it is unlikely that missing these variables distorted our results.
The current results have important implications for clinical practice. First, the time to recurrence becomes shorter and shorter as the number of recurrences increases, implying that the frequency of followup visits may need to be adjusted for the progressively decreasing time to tumor recurrences so that clinicians are able to manage patients more effectively. Second, we found that, in addition to the traditional prognostic factors of stage and grade, Ki-67 also provides prognostic value, mainly for the first tumor recurrence. This information may help clinicians to identify patients at a higher risk of a first recurrence. Third, the finding that p53 does not have an independent effect on the first recurrence but has an independent effect on the second recurrence is particularly interesting. It suggests that continuous monitoring of tumor markers in patients with recurrent tumors might be necessary, as some markers play independent roles not in the first recurrence but in subsequent recurrences. Furthermore, this information may help clinicians identify patients at a higher risk for a second recurrence among those who have had a first recurrence. Since the patterns of multiple sequential recurrences and related prognostic factors among superficial bladder carcinoma patients have not been well described in the literature, the current findings shed some light in this area and provide important information for clinicians relevant to their daily practice.
The authors thank Dr. M'Liss Hudson for initiating the bladder carcinoma research project during her time at the Washington University School of Medicine and Drs. Nelson Diaz and Bing Li for database work.