Introduction

Bladder cancer is the second most common genitourinary malignancy in the United States [1]. It is estimated that 69,250 people will be diagnosed and 14,990 will die of bladder cancer in 2011 [2]. While only approximately one third of patients have muscle-invasive disease, these patients have a significantly greater risk of metastasis, progression, and death [3].

Radical cystectomy (RC) with pelvic lymphadenectomy is currently the standard of care for treating muscle-invasive disease. However, with persistent failure rates of 30-45 % from surgery alone, interest in neoadjuvant chemotherapy (NAC) arose [4]. The poor survival in muscle-invasive disease may be attributed to micrometastasis already present at the time of diagnosis, which NAC could possibly treat. Level 1 evidence for NAC is based on methotrexate, vincristine, doxorubicin, and cisplatin (MVAC), with cisplatin described as the single most effective chemotherapeutic agent against urothelial carcinoma of the bladder [5, 6].

A range of studies have analyzed prognostic factors in bladder cancer. Canter et al. [7] illustrated that increasing age, advanced pathological classification, and positive lymph nodes (LN) were adversely associated with cancer-specific survival. The Southwest Oncology Group Intergroup trial concluded that increased survival was associated with cisplatin based NAC, completion of RC, and removing a minimum number of LN [3]. However, much of the data are from large, multi-institutional clinical trials over the span of many years while others cohorts include patients with both muscle-invasive and non-muscle invasive disease [5–7]. The study is unique in that we analyzed a recent, large, single institution RC cohort restricted to pathologically proven muscle-invasive disease at biopsy to assess survival outcomes in current day practice.

Materials and methods

Data were collected from all patients who had a RC at this institution from 2007 to 2010. Pathology reports from RC and bladder biopsy/transurethral resections (TUR) were reviewed. Parameters obtained included gender, age, cT classification prior to RC, NAC usage, regimen, and number of cycles received, pT classification at TUR and RC, grade at RC and TUR, number of nodes resected, nodal status, presence of lymphovascular invasion (LVI) and/or carcinoma in situ (CIS) at TUR and RC, surgical margin status, and date of death. LVI information was available for all but three patients due to inaccessible TUR reports. Patients that received NAC regardless of number of cycles or regimen were designated as having received NAC. Exclusion criteria for NAC included renal insufficiency precluding treatment with cisplatin and Eastern Cooperative Oncology Group performance status ≥3. NAC was given at our hospital or an outside facility. RC was offered to almost all patients with pT2 disease (tumor invasive into the muscularis propria) with very few determined to be non-surgical candidates.

Pathologic classification was based upon AJCC seventh edition definitions [8]. The cohort for analyses was restricted to patients with TUR of pT2 to ensure consistency. A central review of all 180 RC cases was completed by one pathologist (D.L.Z.). Stage-based response (SBR) was defined as a decrease in pT between TUR (pT2) and RC (<pT2). Nodal involvement was recorded as pN0 if all nodes were negative or pN1+ if any nodes were positive. Patients with no nodes removed (pNX) were excluded from survival analysis using nodal involvement as a factor.

Overall survival (OS) was defined as time from RC to date of death due to any cause or date of last chart review. Surviving subjects were censored at the time of last chart review in time-to-event analyses. Kaplan–Meier survival curves were used to evaluate differences in OS between groups. Cox regression models were used to model OS based on RC pT classification, nodal involvement, NAC use, and SBR. Survival analyses included univariate and multivariable modeling. Due to limited numbers in some subgroups, 95 % confidence intervals (CI) for survival estimates at given time points were used to assess variability and differences between groups in lieu of formal comparisons. Prognostic utility of collapsed subgroups based on RC pT classification were explored and analyzed in a similar manner. Comparisons between groups were evaluated using chi-square tests or Fisher exact tests if fewer than five subjects in a cell. Differences in continuous measures between groups were evaluated using two-sample t tests or Kruskal–Wallis tests if comparing across more than two subgroups. Logistic regression was used to measure significance of prediction factors against binary response variables. Given the hypothesis-generating nature of these analyses for potential prospective studies, multiple comparison corrections were not used, and statistical significance was declared at P < 0.05.

Results

The study reviewed all patients who had RC for primary bladder carcinoma at our institution between 2007–2010 (n = 305). At TUR, 180 patients were pT2. All statistics were performed on these 180 patients.

Table 1 outlines the clinicopathological characteristics of the cohort. In 2007, 7.1 % (2/28) of patients received NAC, which increased to 40.8 % (20/49) in 2008, 53.1 % (26/49) in 2009, and 59.3 % (32/54) in 2010 as in recent years use of NAC is strongly encouraged for every candidate patient.

Table 1 Cohort characteristics of patients receiving RC between 2007 and 2010 for pT2 bladder cancer
Full size table

Figure 1a depicts OS by RC stage (P < 0.00001). pT4a (invasion of prostatic stroma, uterus, or vagina) was comparable to pT4b (invasion of pelvic or abdominal wall) and pT0 (no evidence of tumor) similar to pTis (carcinoma in situ), therefore these groups were combined (pT4a/pT4b and pT0/is) and are each reflected by only one curve. OS rate (1.5 years) (95 % CI) based on stage was pT0/pTis, 80 % (65–98 %); pT1 (subepithelial invasion), 77 % (54–100 %); pT2a, 80 % (62–100 %); pT2b, 61 % (43–87 %); pT3a (microscopic perivesicular fat invasion), 58 % (42–81 %); pT3b (macroscopic perivesicular fat invasion), 43 % (25–72 %); pT4, 12 % (4–35 %). Estimates are not given for pTa (noninvasive papillary carcinoma) given that there was only one pTa patient. Based on the World Health Organization/International Society of Urologic Pathology grading system included in the seventh edition of AJCC, 139 patients (77.2 %) were high grade at RC. Five patients (2.8 %) were low grade at RC and 36 patients (20 %) did not have a grade given due to no residual tumor or predominance of glandular/squamous differentiation.

Fig. 1
figure 1

Survival analysis showing survival time in months for the proportion of patients remaining alive. a OS for RC pT classification (P < 0.00001). b OS by collapsed RC pT classification (P < 0.0001). c OS by SBR (P = 0.000426). d OS by RC Nodal status (three patients were pNX and were not included) (P < 0.00001). e OS by LVI and node status (P < 0.0001) status. f OS by NAC status (P = 0.746)

Full size image

Subclassifying pT2 revealed no difference in OS (P = 0.18). Subclassifying pT3 showed a non-statistically significant separation in curves (P = 0.09). Restricting subclassification to the pN0 population did not yield a significant difference among pT2 and pT3 subcategories (P = 0.8 and P = 0.57, respectively). OS (1.5 years) (95 % CI) based on substage for pN0 patients was pT2aN0, 84 % (65–100 %); pT2bN0, 74 % (55–100 %); pT3aN0, 59 % (40–86 %); pT3bN0, 63 % (41–96 %). Median OS (95 % CI) in months was pT3a, 28.3 (15.3—not applicable (NA)); pT3b, 13.0 (7.9—NA); and pT4, 8.4 (4.0–13.9). Sufficient deaths did not occur in the remaining stages to calculate median OS with CI.

Graphically, there were noted similarities in survival distribution, with certain pT classifications having comparable survival rates across different time points. pT was collapsed into three groups for additional analysis, with pT0/a/is/1/2a surviving longer than pT2b/3a/3b and pT4 having the worst survival (P < 0.00001) (Fig. 1b). OS (1.5 years) (95 % CI) were pT0/a/is/1/2a, 79 % (68–92 %); pT2b/3a/3b, 53 % (41–67 %); pT4, 12 % (4–35 %).

SBR was shown to have improved OS (P = 0.00043) with a 1.5-year OS (95 % CI) of 79 % (67–95 %) vs. 45 % (37–56 %) without SBR (Fig. 1c).

Nodal involvement was strongly associated with poor outcome (P < 0.00001) (Fig. 1d). OS (1.5 years) (95 % CI) were 66 % (56–76 %) for pN0 and 30 % (19–48 %) for pN1+. Median OS (95 % CI) was 11 months (9.2–17.8) in pN1+ and could not be calculated for pN0 due to insufficient deaths. pN1+ was not restricted to high pT, as 4.5 % of the cohort and 14.8 % of the pN1+ group were pT0 (n = 1), pTis (n = 1), pT1 (n = 2), and pT2a (n = 4).

Of the patients, 44.1 % had LVI. Figure 1e demonstrates OS based on presence of LVI (P < 0.0001). OS (1.5 years) (95 % CI) was 70 % (60–82 %) without LVI versus 33 % (23–48 %) with LVI. When analyzing LVI and nodal status together, 50.0 % (87/174) of patients had no LVI/pN0, 6.3 % (11/174) no LVI/pN1+, 19.0 % (33/174) LVI/pN0, and 24.7 % (43/174) LVI/pN1+ with 1.5-year OS (95 % CI) of no LVI/pN0, 78 % (68-89 %); LVI/pN0, 38 % (23-61 %); no LVI/pN1+, 28 % (9-82 %); and LVI/pN1+, 30 % (17-51 %). Distribution of number of LN for the subgroups showed no statistical significant difference (P = 0.28) (mean number of LN: no LVI/pN0, 31.8; LVI/pN0, 29.9; no LVI/pN1+, 37.3; and LVI/pN1+, 28.3). Number of LN removed was not a significant predictor of nodal status (P = 0.65) or LVI (P = 0.14).

Of the patients, 38.2 % (68/178) had CIS on TUR or RC. OS (95 % CI) in months was 22.7 (14.9—NA) without CIS verses 19.9 (13.9—NA) with CIS (P = 0.889), demonstrating no statistically significant difference.

Tables 2 and 3 reveal characteristics of patients that received NAC verses immediate RC. More NAC patients were cT3/4 (42/81, 51.9 % vs. 45/99, 45.5 %). More frequent decrease in T classification comparing cT prior to RC versus pT at RC in the NAC group was observed (45.7 % vs. 24.2 %). There was a higher incidence of NAC patients with pT0 (23.5 % vs. 8.1 %) and the low pT group pT0/a/is/1/2a (45.7 % vs. 28.3 %). However, in each cT group, there were more pN1+ patients with a history of NAC, although this difference was not statistically significant (P = 0.63 for cT2; P = 0.21 for cT3; P = 0.24 for cT4). Of note, all eight patients that had low RC pT (pT0/a/is/1/2a) with pN1+ disease were in the NAC group (representing 9.9 % of the NAC group). Of these eight patients, three have died with a median follow-up in the remaining five living patients of 14 months. There was no significant difference in OS (P = 0.746) based on use of NAC (Fig. 1f) with median OS (95 % CI) in months as 24.0 (15.6—NA) with NAC and 22.2 (14.0—NA) for RC alone. OS (1.5 years) (95 % CI) was 58 % (46–72 %) in the NAC group and 50 % (40–63 %) with no NAC and the 3-year OS (95 % CI) was 31 % (15–63 %) with NAC compared to 42 % (31–56 %) without NAC. An additional consideration when interpreting OS is the time interval from TUR to RC which was on average 56.6 days longer for patients receiving NAC. We also attempted to see if there was a survival benefit with NAC for certain stage groups, with no benefit identified.

Table 2 NAC versus immediate RC
Full size table
Table 3 cT within NAC versus immediate RC group
Full size table

On multivariable analysis, pT4 followed by high RC pT group and positive nodes were negative prognostic factors (Table 4). SBR was not significant in the multivariable setting. Although SBR was a significant prognostic univariate factor, it was actually the RC pT0/a/is/1/2a group that conferred an improved OS. SBR was a surrogate of this pT group, without the inclusion of pT2a. When LVI was combined with RC pT, it was a significant prognostic factor, but nodal status was a stronger predictor of OS. This was confirmed looking at all subsets models. NAC was not a significant factor of OS across all pT classifications, collapsed pT groups, nodal status, SBR, and did not impact the other prognostic indicators. Thus, only RC pT, pT groupings, and nodal status remained significant in multivariable analysis.

Table 4 Multivariate analysis of prognostic factors
Full size table

Discussion

Muscularis propria-invasive bladder cancer has been associated with significant risk of progression, metastasis, and death [3]. Standard of care for ≥ pT2 disease is RC with lymphadenectomy, with increasing usage of NAC [4, 5, 9]. Past studies have evaluated different prognostic markers in order to stratify survival. Factors such as use of NAC, completion of RC, and extended LN dissection have been shown to improve survival, whereas advanced pT and LN metastases were linked to poor outcomes [3, 7]. Our aim was to analyze prognostic factors of bladder cancer in RC. This study is relevant in that it reflects a contemporary cohort at a single institution with similar surgical practices. Unlike previous reports, analysis was restricted to patients with TUR proven pT2 disease to ensure consistency within the cohort. The data provide reference OS values for patients with pathologically diagnosed muscle-invasive bladder cancer based on current medical guidelines outside the context of a clinical trial.

Over 50 % of patients at our institution had pT3/4 disease at RC. Not surprisingly, the OS of 57.8 % for the cohort was low. Incidence of pT3/4 disease from other single institutions ranged from 30 % to 53 % [10–13]. Of note, these studies had widely variable accrual lengths and included all RC patients, compared to only pT2 disease at TUR in our cohort [10–13]. We demonstrated a statistically significant relationship between pT classification at RC and OS. Similar to previous publications, pT4 had the worst OS (12 % at 1.5 years), with T4a and T4b disease behaving comparably [14, 15]. There was a non-statistical separation in survival curves between pT3a and pT3b and no difference in OS for pT2a and pT2b. Limiting analysis to the pN0 population did not show a difference in survival. Prior investigations have found no difference in recurrence free survival (RFS) for pT3 substages except within pN0 in which pT3aN0 had improved RFS compared to pT3bN0 [16, 17]. The literature reports conflicting results regarding RFS for pT2 subcategories [18, 19]. A limitation of subclassifying pT3 tumors is that it is based on gross impression of fat invasion, making standardization difficult. Similarly, subclassifying pT2 is somewhat subjective because it reflects an estimation of the depth of muscularis propria invasion. Demonstrating a survival benefit within these substages was also limited by the number of patients and number of deaths that occurred, especially within the pT2 population. We found OS between pT0, pTa, pTis, and pT1 to be comparable, although analysis was limited by small numbers and few deaths. There is contradictory published data comparing pT0 to pTa/Tis [20, 21]. Nonetheless, pT0 has improved survival compared to ≥pT1 at RC [20]. Based on similarities in survival distribution, pT were collapsed into pT0/a/is/1/2a, pT2b/3a/3b, and pT4 groups, representing a novel simplified model for bladder cancer OS. There was a statistically significant difference between these three groups, with pT0/a/is/1/2a patients surviving longer than pT2b/3a/3b, and pT4 disease having the worst survival. Moreover, on multivariable analysis, RC pT group pT2b/3a/3b remained a strong negative predictor of OS.

Our investigation indicates that patients with SBR (pT2 at TUR and <pT2 at RC) have improved OS, with a 1.5-year survival rate of 79 % vs. 45 % without SBR. There is limited data regarding the prognostic implications of SBR as most studies focus on complete pT response (pT0). When comparing SBR as a potential prognostic factor versus a three-tiered RC pT group, the collapsed pT group was a stronger indicator of OS. This illustrates that SBR is a surrogate to analyze lower RC pT, without the inclusion of pT2a. Thus, our data indicate that a stage of ≤pT2a is a stronger prognostic factor and may be more relevant to describe the effectiveness of a treatment than SBR and/or complete pT response. It is also interesting to note that presence of SBR is not specific to patients with NAC: 18.2 % of the RC cohort also had a decrease from pT stage at TUR to pT stage at RC. This illustrates that use of BCG therapy as well as physically debulking the tumor during TUR may play a role in downstaging tumors. While previous studies have used downstaging or a complete pT response as an indicator of successful use of NAC, our study demonstrates that downstaging is not limited to this population of patients and may be relevant in the immediate RC cohort.

Of our cohort, 30 % was pN1+ at RC and represented a significant negative predictor of survival. Nodal status impacting prognosis has been well illustrated in the literature [3, 7, 10, 15, 22, 23]. At 1.5 years after RC, 66 % of our patients with pN0 were alive versus 30 % with pN1+, consistent with published outcomes [3, 7, 10, 15, 22, 23]. Abdel-Latif et al. [23] demonstrated a 3-year RFS of 78 % with pN0 versus 38 % with positive nodes. Moreover, Stein et al. [10] found a 69 % 5-year OS with pN0 verses 31 % with pN1+. One limitation in the ability to compare our results with other institutions is that as a tertiary care center, we treat more patients with high stage and positive nodal disease. Our institution also routinely uses extended lymph node dissection, which is not practiced in all hospitals. While pN1+ was a predictor of poor outcome, an important finding of our analysis was that positive nodes were not limited to high pT disease, with 4.5 % of patients having pN1+ with ≤pT2a disease.

Almost half of the cohort had LVI. LVI did not impact survival in node positive patients; however, LVI was a negative predictor of survival in the pN0 group. There is data illustrating LVI at RC as an independent predictor of survival and recurrence in bladder cancer [24, 25]. Shariat et al. [26] studied pN0 RC patients and demonstrated that LVI was strongly associated with clinical outcome. The finding of pN0 patients with LVI and poor survival raises the possibility that the nodal resection was less substantial in this group. However, our data showed that LN number was comparable. A second possibility is that the LN metastases may not have been present for histologic assessment, in that either the tumor was cut through and discarded during processing or that the metastatic deposit is still embedded in paraffin and was not available for microscopic examination. A third possibility is that a diagnostic error occurred in which the pathologist did not identify the positive LN. A past article compared histologic impression versus keratin immunohistochemistry to detect micrometastasis in node negative bladder cancer, with no identified missed metastases [27]. We hypothesize that these patients may have a different path of tumor spread, for example, directly to non-regional lymph nodes or bone marrow. Future research is required to understand the mechanism by which patients come to have pathologically identified LVI and negative nodes and the biologic relevance of these findings.

Between the NAC vs. non-NAC groups, age and surgical parameters such as number of LN and margin status were comparable. However, there was extensive heterogeneity in type of NAC given, number of cycles, medical setting (tertiary care center vs. community practice) and time between NAC and RC, reflecting the current variable practice patterns. With NAC, we observed more frequent decrease of T classification comparing cT prior to RC and pT at RC (45.7 % vs. 24.2 %), increased pT0 (23.5 % vs. 8.1 %) and a higher rate of SBR (33.3 % vs. 18.2 %). High pT (≥pT2b) at RC occurred less often with NAC (34.6 % vs. 56.6 %). Similarly, previous authors have demonstrated that NAC is associated with more frequent complete pathologic response than RC alone (33–38 % vs. 5–15 %) [3, 12, 20, 28, 29]. However, we did not identify a statistically significant increase in OS in NAC patients (NAC OS 22 vs. 24 months without NAC). Literature regarding the survival advantage of NAC is controversial, but evidence from clinical trials suggests a benefit of up to 5.5 % [3, 28–34]. Perhaps to achieve this benefit, NAC must be better standardized as is done in a clinical trial and therefore our data does not discredit the use of neoadjuvant chemotherapy in patients with muscle-invasive bladder cancer. Surprisingly, pN1+ was more common in NAC patients (39.5 % vs. 22.2 %). For each cT classification, more node positivity was seen in the NAC cohort, although this difference was not statistically significant. Moreover, we identified a group of pN1+ patients with a low pT (≤pT2a). All patients in this group received NAC, with deaths observed. These findings suggest that focusing on pT classification or the frequency of decreased cT to pT at RC without the corresponding node status in NAC patients may not be fully indicative of a tumor's potential lethality, a concept not discussed in the literature. As NAC has consistently been shown to decrease pT at RC, systemic chemotherapy could have utility in the treatment of non-muscle-invasive, BCG refractory patients that are not ideal surgical candidates or have limited life expectancy.

Limitations of this investigation included its retrospective nature, short follow-up, limited number of patients, and multiple surgeons. Another consideration was potential bias between NAC and non-NAC groups. Poor renal function and performance status excluded patients from NAC. Age was slightly higher in the non-NAC group, which in conjunction with the NAC exclusionary criteria should help eliminate a bias of unhealthier patients receiving NAC. NAC was heterogeneous, as previously discussed. Other histopathological factors which may have prognostic significance, such as tumor multifocality and grade, were not analyzed. Lastly, being a tertiary care center with potentially higher cT/pT tumor, more positive nodes, routine use of extended LN dissection, and frequent use of NAC, our data may not be representative of other patient cohorts.

Conclusions

We performed a retrospective single institutional study of muscle-invasive bladder cancer at RC. Negative prognostic factors in bladder cancer included pT4, pT2b/3a/3b disease, pN1+, and LVI. Collapsing pT into pT0/a/is/1/2a, pT2b/3a/3b, and pT4 was a stronger predictor of survival than SBR. pN0 patients with LVI had decreased OS. The presence of positive nodes at RC was not limited to high pT disease with pN1+ seen in NAC patients with low RC pT. pT0 and SBR were more common in patients with a history of NAC, although use of NAC was not associated with increased OS. Our findings may be more indicative of outcomes outside the context of a clinical trial and highlight the variability of multidisciplinary management in bladder cancer in its current usage today.