Introduction

The incidence of several types of cancers, such as bladder cancer, is increasing in the USA. Bladder cancer is the fourth most common cancer in men and accounts for 4.4 % of all primary malignancies in the USA, with approximately 54,610 men and 17,960 women newly diagnosed annually (Siegel et al. 2013). Muscle-invasive bladder cancer (MIBC) accounts for 25 % of newly diagnosed bladder cancers (Leow et al. 2014). Radical cystectomy alone can result in a 5-year survival rate in patients with organ-confined lymph node-negative MIBC of up to 80 %. This rate, however, decreases to 40–50 % in patients with extravesical disease, with a further reduction to 15–35 % if lymph node metastases are present (Stein et al. 2001).

Distant recurrences have been observed to occur more frequently than locoregional recurrences in 20–50 and 5–15 % of patients, respectively, suggesting that perioperative systemic therapy may improve outcomes (Quek et al. 2003; Shariat et al. 2006; Stein et al. 2001). The current National Comprehensive Cancer Network (NCCN) guidelines recommend radical cystectomy as the standard of care for all patients with nonmetastatic MIBC and sufficient performance status. Cisplatin-based neoadjuvant chemotherapy (NC) is strongly recommended based on level 1 evidence showing a survival benefit (Advanced Bladder Cancer Meta-analysis 2005b; Sternberg et al. 2013). Recent studies, however, demonstrate the underutilization of this treatment strategy. For example, a retrospective multi-institutional study of 4,541 patients across 14 academic centers in the USA from 2003 to 2008 found that only 12 % of patients received NC and 22 % received AC (Feifer et al. 2011).

AC has several putative advantages compared with NC, including its administration to selected patients, based on factors predictive of relapse, and its lack of delay of definitive treatment. Since many patients are unwilling to accept NC, clinicians must decide whether to recommend AC to moderate- to high-risk patients previously unexposed to NC. Although several randomized trials have investigated the efficacy and safety of AC after radical cystectomy (Lehmann et al. 2006; Resorlu et al. 2010), almost all these studies have provided insufficient evidence to support routine AC, due to small sample sizes, early stopping of patient entry, confusing analyses and terminology, and the reporting of questionable conclusions (Boccardo and Palmeri 2006; Sylvester and Sternberg 2000). Meta-analyses that included several trials each have been performed to evaluate AC for bladder cancer (Advanced Bladder Cancer Meta-analysis 2005a; Leow et al. 2014). However, these meta-analyses have limitations, such as significant heterogeneity among trials, lack of access to individual patient data, different chemotherapy regimens, and different definitions of outcomes.

In assessing the efficacy of AC, we rigorously adjusted for significant differences in patient characteristics by using inverse-probability-of-treatment weighting (IPTW), reducing the impact of treatment selection bias and potential confounding factors in an observational study. Using IPTW, we therefore evaluated the benefits of cisplatin-based AC plus surgery versus surgery alone in patients with bladder cancer.

Materials and methods

Study participants

This study was performed with the approval and oversight of the Institutional Review Board of the Asan Medical Center, which waived the requirement for informed consent due to the retrospective design of the study. The medical records of 756 consecutive patients who underwent radical cystectomy for bladder cancer between 1990 and 2012 were retrospectively reviewed. Patients receiving NC or radiotherapy were excluded. Finally, a total of 746 patients who received radical cystectomy for bladder cancer were included in this study. Patient demographic and clinical characteristics were evaluated. All patients underwent preoperative chest X-ray, computerized tomography of the abdomen and pelvis, and bone scan for disease staging. No patient showed evidence of metastatic disease on physical examination or staging.

Surgical procedures and pathological evaluation

Radical cystectomy and pelvic lymphadenectomy were performed by two senior surgeons (C.S.K. and H.A.) as previously described in detail (Jeong et al. 2011). Urinary diversions, including ileal conduit diversion, and orthotopic bladder substitution were performed following radical cystectomy and bilateral pelvic lymphadenectomy (Jeong et al. 2012). All surgical specimens were processed according to standard pathological procedures. Tumor grade was determined according to the 2004 WHO grading system (Montironi and Lopez-Beltran 2005), and pathological stage was reassigned according to the 2010 American Joint Committee on Cancer tumor node metastasis (TNM) staging system (Edge and Compton 2010). Positive soft tissue surgical margin status was defined as the presence of tumor in inked areas of soft tissue of cystectomy specimens; urethral or ureteral margin status was not considered positive in this analysis (Novara et al. 2010). Lymphovascular invasion was defined as the unequivocal presence of tumor cells in an endothelium-lined space without underlying muscular walls.

Adjuvant chemotherapy

AC at our institute is routinely recommended with full counseling for pathologic ≥T3 and/or node-positive patients, except for those who are medically intolerant or who refuse this treatment. Within 3 months of surgery, all patients started AC, consisting of three to six cycles of MVAC (methotrexate, vinblastine, doxorubicin, and cisplatin, n = 45) or GC (gemcitabine and cisplatin, n = 131). Patients treated with MVAC received methotrexate 30 mg/m2 on days 1, 15, and 22; vinblastine 3 mg/m2 on days 2, 15, and 22; doxorubicin 30 mg/m2 on day 2; and cisplatin 70 mg/m2 as a 1–8 h infusion on day 2. Patients treated with GC received gemcitabine 1,000 mg/m2 over 30–60 min on days 1, 8, and 15, plus cisplatin 70 mg/m2 on day 2. Cisplatin was administered with adequate pre- and post-hydration. Cycles were repeated every 28 days (von der Maase et al. 2000). From 1992 to 2000, only MVAC regimens were used. GC and MVAC regimens have similar survival benefits, but GC has a better safety profile and is better tolerated by patients with advanced or metastatic bladder cancer; therefore, only GC regimens were administered from 2001 to 2011 (von der Maase et al. 2000).

Follow-up

After radical cystectomy, patients were generally followed up every 3 months during the first year, every 6 months during years 2–6, and annually thereafter. Follow-up consisted of a medical history and a physical examination, as well as blood laboratory assays, urine sedimentation, culture, and cytology. All patients were assessed by imaging modalities, including chest radiography, computed tomography of the abdomen and pelvis, and bone scanning, at 6 and 12 months postoperatively, and annually thereafter. Recurrence was defined as local recurrence at or below the common iliac bifurcation or distant metastasis documented by imaging and biopsy, if indicated. Detection of cancer in the ureter or urethra was recorded as a second primary tumor rather than as local or distant recurrence. The median follow-up duration was 64.3 months (range, 1–231.4 months).

Statistical analysis

Clinicopathological features in the two groups were compared using Pearson’s chi-square test for categorical variables and Student’s t test for continuous variables. Quantitative data are expressed as the mean ± standard deviation. Recurrence-free survival (RFS) was calculated as the interval from radical cystectomy to the first documented clinical recurrence. Patients who died before any clinical recurrence were censored at death. Cancer-specific survival (CSS) was measured from the date of initiation of treatment to the date of death from bladder cancer. To evaluate the effect of AC according to pathologic status and to reduce the impact of treatment selection bias and potential confounding in an observational study, significant differences in patient characteristics were rigorously adjusted using multivariable model and IPTW method of propensity scores (Robins et al. 2000). For multivariate analyses, the Cox proportional hazards model was fitted with time to recurrence and death with and without variables selection procedure. As well as multivariable model, the propensity score was calculated to account for the confounding arising from differences in clinical characteristics. Propensity scores were estimated by multiple logistic-regression analysis. Significant differences of clinical/pathological characteristics are identified in Tables 1 and 2 and were used to calculate propensity scores of likelihood of AC. To evaluate the discrimination and calibration abilities of propensity scores, C statistics and the Hosmer–Lemeshow test were used. The model was well calibrated (Hosmer–Lemeshow test; p = 0.732) with reasonable discrimination (C statistic = 0.763). IPTW for AC use was applied to estimate hazard ratios of treatment effect in Cox’s proportional hazard model. All reported p values are two-sided, with p values <0.05 considered statistically significant. All statistical analyses were performed using SAS software, version 9.3 (SAS Institute, Cary, NC, USA).

Table 1 Clinical/pathological characteristics of the patient cohort
Full size table
Table 2 Effect of adjuvant chemotherapy on survival according to pathologic status
Full size table

Results

The descriptive characteristics of the 746 patients are shown in Table 1. Of these patients, 176 (23.6 %) received AC after radical cystectomy and 570 (76.4 %) received radical cystectomy alone. The mean age of the entire cohort was 62.4 ± 9.7 years (range: 29–85 years), with patients receiving AC being significantly younger at the time of radical cystectomy (p = 0.001). Patients in the cystectomy with AC group were also significantly more likely to have higher pathologic T stage (p = 0.001), lymph node metastasis (p = 0.001), and lymphovascular invasion (p = 0.001) than patients undergoing cystectomy alone. However, there were no between group differences in the number of positive lymph nodes and removed lymph nodes.

Of all, 253 patients had died. Tumor was the cause of death in 166 patients in the cystectomy alone group and in 87 patients in the cystectomy with AC group. The overall 5- and 10-year survival rates were 70.7 % and 65.1 %, respectively. However, the 5-year overall survival rate was significantly lower for patients with than without node metastasis (37.8 vs. 80.6 %, p < 0.001). Table 2 shows a Cox regression analysis of the effects of AC on RFS and CSS according to pathologic status. AC improved RFS (HR 0.50, p = 0.004) and CSS (HR 0.32, p = 0.001) only in patients with pathologic N2 stage or greater. In these patients, the 5-year RFS (21.9 vs. 16.4 %, p = 0.003, Fig. 1a) and CSS (32.5 vs. 12.6 %, p = 0.001, Fig. 1b) rates were significantly higher in patients who received AC than in those who underwent surgery alone.

Fig. 1
figure 1

a Recurrence-free survival and b cancer-specific survival in 126 patients with bladder cancer of pathologic stage N2 or greater

Full size image

Multivariable analysis showed that factors significantly predictive of RFS and CSS included age (RFS HR 1.01, p = 0.013; CSS HR 1.03, p = 0.001), body mass index (RFS HR 0.89, p = 0.001; CSS HR 0.86, p = 0.001), pathologic T stage (RFS HR 0.53, 95 % CI 0.39–0.71, p = 0.001; CSS HR 0.49, p = 0.001), lymph node metastasis (RFS HR 0.40, p = 0.001; CSS HR 0.28, p = 0.001), and lymphovascular invasion (RFS HR 0.52, p = 0.001; CSS HR 0.65, p = 0.004) (Tables 3, 4). In multivariable analysis, AC after radical cystectomy correlated significantly with CSS (HR 0.56, 95 % CI 0.39–0.80, p = 0.002). AC also correlated significantly with CSS when the significance of each variable affecting CSS was assessed by IPTW using propensity score (HR 0.83, 95 % CI 0.69–0.99, p = 0.043).

Table 3 Multivariable analysis of factors influencing recurrence-free survival
Full size table
Table 4 Multivariable analysis of factors influencing cancer-specific survival
Full size table

Discussion

Bladder cancer is a potentially lethal disease, with as many as one-quarter of patients who undergo cystectomy having micrometastatic disease at the time of surgery (International Bladder Cancer Nomogram et al. 2006). Bladder cancer can progress quickly and may justify some form of systemic treatment for tumor eradication and improved survival. This study evaluated the survival benefits of cisplatin-based AC in a large cohort of patients treated at a single institution with radical cystectomy and lymphadenectomy, with some also receiving AC. In addition, patients were compared after IPTW, thus controlling for potentially confounding covariables affecting survival, including age, sex, comorbidities, and pathologic status. Propensity scores were used to reduce any biases due to differences in the baseline characteristics of the two patient groups (Joffe and Rosenbaum 1999). Although matching was an option, weights based on propensity scores reduce the possibility that matching might exclude a substantial sample size of patients whose disease experience could make a valid contribution. Considering weights based on both the standardized morbidity ratio and the inverse probability of treatment (Robins et al. 2000; Sato and Matsuyama 2003), we chose the latter. These weights yield cohorts that are effectively from a common population, except for the difference in the propensity score response variable. Following IPTW, AC significantly improved CSS (HR 0.83, 95 % CI 0.69–0.99, p = 0.043).

Cisplatin-based combination NC has been shown to result in a 5 to 7 % absolute survival benefit in patients undergoing radical cystectomy for bladder cancer (Advanced Bladder Cancer Meta-analysis 2005b; Sternberg et al. 2013). Cisplatin-based NC is therefore an option for patients with localized cT2–4 urothelial carcinoma of the bladder (Sternberg et al. 2007); it should be ‘considered’ for patients with cT2N0M0 and ‘strongly considered’ for patients with cT3N0M0 disease (Montie et al. 2009). However, NC for all patients undergoing radical cystectomy has not been widely adopted, due to concerns about unnecessary treatment of patients who may not benefit from NC and disease progression that may result from delays in surgery caused by chemotherapy, especially if NC is ineffective. Furthermore, AC may be as effective as NC if the former is administered selectively based on adverse pathologic characteristics (Ruggeri et al. 2006). Nevertheless, the role of AC in patients with high-risk urothelial carcinoma of the bladder has not been well defined. Although a recent meta-analysis of individual patient data from available trials reported a 23 % relative reduction in the risk of death in patients receiving AC compared with those undergoing surgery alone (Leow et al. 2014), this analysis was limited by the methodological flaws of individual trials and their inclusion of insufficient numbers of patients and events. Thus, the routine use of AC for patients undergoing radical cystectomy has not gained widespread acceptance. Moreover, these trials were performed in highly selected patients and under favorable conditions. The results of these trials are therefore difficult to apply to patients in everyday practice, who may be older, have increased comorbidities, and may receive suboptimal chemotherapy regimens or doses (Markman 2007). This has increased the importance of rigorous adjustment for differences in patient characteristic by IPTW. In our study, using IPTW, AC after radical cystectomy improved survival in the entire cohort, with subgroup analysis showing that AC improved RFS and CSS in patients with a pathologic N2 or greater. Thus, the benefit of AC depended primarily on disease progression. Selective administration of AC to patients at higher risk of disease progression, such as those with pathologic N2 or greater, may optimize the therapeutic benefits of AC.

Patients with N ≥ 2 disease had poorer survival. In our study, all single positive lymph nodes were found below the aortic bifurcation, consistent with previous studies (Abol-Enein et al. 2004; Leissner et al. 2004). For example, one study reported that none of 290 patients who underwent extended lymphadenectomy had a single nodal metastasis located above the aortic bifurcation or in the presacral triangle (Leissner et al. 2004). A second study reported a single nodal metastasis in the endopelvic region, except for one in the common iliac chain (Abol-Enein et al. 2004). Therefore, patients with a single positive lymph node had less advanced disease than patients with N2 patients, and N2 disease is more likely to lead to systemic metastasis. In addition, the 5-year RFS and CSS rates in patients with N ≥ 2 disease were 18 and 26 %, respectively. The CSS rate was likely higher than the RFS rate because tumors were likely the cause of death in most of these patients.

This study had several limitations, including its retrospective nature, the relatively small patient cohort, and the significant differences in certain variables between patient groups. Residual confounding is anticipated as many patients undergoing radical cystectomy have significant age-related comorbidities that influence their ability to be selected for and/or tolerate AC and are casually related to outcomes (Koppie et al. 2008). Unfortunately, the comorbidity status of patients was not collected during data acquisition. The absence of this information would likely systematically bias the results in favor of AC because it is more likely administered to healthier patients. In addition, our estimates for the effect of AC in the low-risk subgroup were underpowered because of the lack of events in this population, enabling no conclusions to be drawn with respect to the advantages or disadvantages of AC in this population. Nevertheless, the clinical results obtained from this single group of patients over an extended period of time demonstrate that AC after radical cystectomy provides good survival results after controlling for potentially confounding covariables affecting survival, including age, sex, comorbidities, and pathologic status by IPTW.

Conclusions

Our results showed that, after adjusting for confounding variables using IPTW analysis, cisplatin-based AC showed survival benefit in patients with bladder cancer after radical cystectomy. Although AC was associated with a significant improvement in survival in the entire patient cohort, its benefits depended primary on disease progression. Selective administration of AC to patients at highest risk of disease progression, such as those with pathologic N2 or greater, may reinforce the therapeutic benefit of AC.