The purpose of this study was to investigate the effect of neoadjuvant chemotherapy with gemcitabine and cisplatin (GC) on pathologic down-staging of patients with locally advanced urothelial cancer (UC) of the bladder.
The purpose of this study was to investigate the effect of neoadjuvant chemotherapy with gemcitabine and cisplatin (GC) on pathologic down-staging of patients with locally advanced urothelial cancer (UC) of the bladder.
This was a retrospective cohort study of patients treated with radical cystectomy (RC) for clinical stage cT2-T4, N any, M0 bladder UC at Strong Memorial Hospital from 1999 to 2009. The primary exposure variable was use of neoadjuvant chemotherapy (GC vs none). The primary outcome was stage pT0 at RC. Secondary outcomes included other down-staging end points in the bladder (<pT1, <pT2, <pT3), nodal status, and surgical margins. Linear probability models were used to estimate the effect of neoadjuvant GC on tumor down-staging with adjustment for clinical staging variables.
A total of 160 eligible patients were identified, of whom 25 were treated with neoadjuvant GC before RC (GC + RC) and 135 without neoadjuvant chemotherapy (RC only). Stage pT0 at cystectomy was found in 20% of patients in the GC + RC group and in 5% of patients in the RC group (adjusted risk difference [aRD] = 16%, P = .03). For other down-staging end points, the estimated treatment effect was as follows (all point estimates favoring chemotherapy): <pT1 aRD = 30% (P = .005); <pT2 aRD = 30% (P = .004); <pT3 aRD = 31% (P = .008); margins aRD = 8% (P = .41); nodes aRD = 4% (P = .74).
Neoadjuvant GC was found to be capable of down-staging UC in the bladder; however, no effect on disease in nodes was seen in this study. Cancer 2012;. © 2011 American Cancer Society.
Bladder cancer (BC) is the fifth most commonly diagnosed malignancy in the United States, with more than 70,000 new cases and more than 14,000 BC deaths reported in 2009.1 The overwhelming majority of deaths from BC occur among patients with muscle-invasive disease (stage categories T2-T4). Standard therapy for resectable (T2-T4a) muscle-invasive BC without known metastases includes radical cystectomy.2 Unfortunately, 30% to 50% of patients with apparently resectable muscle-invasive BC (“locally advanced” disease) in fact have undiagnosed micrometastases at the time of definitive surgery.2, 3
Early treatment of micrometastatic disease with neoadjuvant platinum-based combination chemotherapy (PBCC) administered before cystectomy has been compared with cystectomy alone in several randomized trials. A meta-analysis of these trials demonstrated that addition of a neoadjuvant PBCC regimen to local treatment improves the average 5-year survival by 5% on the additive scale (from 45% to 50%).4 Several trials also reported that the use of neoadjuvant PBCC may increase the probability of pathologic stage zero (pT0) at cystectomy from approximately 12% to 15% in the cystectomy-only arm to 33% to 38% in the PBCC-plus-cystectomy arm.5, 6
However, the PBCC regimens used in the trials of neoadjuvant chemotherapy for locally advanced BC were relatively toxic. In particular, a combination of methotrexate, vinblastine, doxorubicin, and cisplatin (MVAC), which appeared to be the most effective regimen in terms of pathologic down-staging and survival, had a relatively unfavorable toxicity profile, producing severe granulocytopenia (grade 4 or <500 cells/μL) in one-third of all patients treated.5
In clinical practice, a less toxic neoadjuvant chemotherapy regimen composed of gemcitabine and cisplatin (GC) is now commonly substituted for neoadjuvant MVAC.7 In the setting of metastatic disease, GC seems to result in similar clinical response rates and similar survival as MVAC but with a better toxicity profile, based on a report from a randomized trial.8
However, it has been argued that extrapolation of these results to resectable (nonmetastatic) BC is potentially invalid because clinical response of metastatic lesions may not be equivalent to pathologic down-staging or cure of locally advanced BC, and survival of patients with metastatic disease is usually very poor despite the use of combination chemotherapy (MVAC or other regimens).7, 8 Unlike metastatic BC, which is generally considered incurable, locally advanced disease is potentially curable in at least half of all cases. Patients with pathologically organ-confined disease at cystectomy (stage <pT3, pN0) generally have a favorable prognosis, and those achieving stage pT0 are almost always cured from BC.5, 9 However, the ability of GC to produce pathologic down-staging and improve survival of patients with locally advanced BC has not yet been clearly demonstrated.
This question has been investigated in 2 small single-institution series with somewhat conflicting results.7, 10 In the series published by Dash et al,10 42 patients received neoadjuvant GC and 54 patients received neoadjuvant MVAC prior to cystectomy. Chemotherapy allocation was primarily determined by calendar time (patients treated in more recent years received GC). The distribution of clinical stages favored MVAC: 55% of patients in the GC group and 41% of patients in the MVAC group had clinical stage T3/T4. Stage pT0 at the time of definitive surgery was documented in 26% (11/42) of patients treated with GC and in 28% (15/54) of patients treated with MVAC. The authors concluded that GC is equivalent to MVAC in its ability to produce pathologic down-staging of locally advanced BC.
In the second series published by Weight et al,7 20 patients treated with neoadjuvant GC followed by cystectomy were compared with 88 patients treated with cystectomy alone. The decision to administer neoadjuvant chemotherapy in this cohort was determined by patients' ability to tolerate chemotherapy as well as patients' and physicians' preferences. Among patients treated with chemotherapy, 21% had clinical stage T3/T4 cancers. The corresponding proportion of cT3/T4 cases for the cystectomy-only group was not reported. Stage pT0 at the time of definitive surgery was documented in 10% (2/20) of patients treated with neoadjuvant GC, and in 9% (8/88) of patients treated with cystectomy alone. The authors concluded that the use of neoadjuvant chemotherapy was not associated with pathologic down-staging in their series. In addition, delays in cystectomy due to chemotherapy were substantial. The median time from diagnosis of muscle invasion to cystectomy in patients treated with and without neoadjuvant chemotherapy was 7 months and 1.5 months, respectively.7
Because these 2 reports reached different conclusions, we decided to examine this question in our own institutional series. The purpose of the current study was to investigate the effect of neoadjuvant chemotherapy with GC on pathologic down-staging and survival of patients with locally advanced urothelial cancer (UC) of the bladder treated with radical cystectomy (RC).
This is a retrospective cohort study of patients treated with RC for clinical stage cT2-T4, N any, M0 bladder UC at Strong Memorial Hospital (Rochester, NY) from 1999 to 2009. Patients treated with other neoadjuvant therapy, with clinical stage <cT2, by partial cystectomy, with upper tract UCs or with non UCs of the bladder were not included. The primary exposure variable was use of neoadjuvant chemotherapy (GC vs none). The primary outcome was stage pT0 at RC. Secondary outcomes included other down-staging end points in the bladder (<pT1, <pT2, <pT3), nodal status, and surgical margins. The effect of chemotherapy on the hazard of death from all causes and from BC was examined in exploratory analyses.
The following additional covariates were analyzed: age, sex, race, clinical T stage, radiographic evidence of regional adenopathy, grade and histologic type (ie, pure UC or UC with non-UC component), major cardiovascular disease (heart surgery with or without myocardial infarction, stroke, congestive heart failure, and deep venous thrombosis not related to surgery), major respiratory disease (asthma, emphysema, chronic obstructive pulmonary disease), non-BC malignancies (all previously diagnosed malignancies other than nonmelanoma skin cancer), smoking status (current, former, never), creatinine concentration and clearance, white cell and platelet counts (documented prior to the initiation of definitive therapy for locally advanced BC), prior use of intravesical therapy, and use of adjuvant systemic chemotherapy. Postoperative chemotherapy was considered “adjuvant” as opposed to “salvage” if the decision to administer such therapy was made before known recurrence.
Clinical staging variables were determined based on transurethral resection pathology reports, radiology reports, and bimanual examinations. Central pathologic review was not performed on each specimen in this study. However, during the study period, consensus review was routinely used in the pathology department to determine clinical and pathologic staging variables in equivocal cases deemed appropriate for such review. The follow-up data included time from the initiation of definitive therapy (GC or cystectomy) for locally advanced BC until death from any cause or end of follow-up (December 31, 2009). The underlying cause of death was determined by death certificate. The follow-up data were obtained from the National Death Index and compared with information available in the chart for quality assurance.
Continuous variables were described and compared with medians and Wilcoxon tests, and categorical variables were compared with proportions and Fisher exact tests. Linear probability models (generalized linear models with identity link and binomial variance) fit by maximum likelihood were used to estimate the effect of neoadjuvant GC on pathologic down-staging end points, adjusted for clinical staging variables.11, 12 The effects of chemotherapy on the hazard of death from all causes and the hazard of death from BC were estimated using the stratified Cox model, adjusted for baseline covariates.13-15 Proportional hazards assumptions were tested via covariate-by-time interactions. Assumptions of linearity in the log-hazard were tested by adding quadratic terms for continuous covariates. Interactions were tested via product terms. Kaplan-Meier curves and log-rank tests were used to estimate and compare overall survival. All analyses were performed in SAS 9.2 (SAS Institute, Inc., Cary NC). All tests were 2-sided. This study was approved by the University of Rochester Ethics Committee.
During the study period, muscle-invading UC prior to RC was documented in 173 patients, of whom 38 received neoadjuvant chemotherapy and the remaining 135 were treated without neoadjuvant chemotherapy. Decision to administer neoadjuvant chemotherapy was determined by clinical extent of disease, patients' ability to tolerate combination chemotherapy with cystectomy, and patients' preferences. Ten of the 38 patients treated with neoadjuvant chemotherapy received drug combinations other than GC and were excluded from analysis. Three of the 28 patients treated with neoadjuvant GC were found to be ineligible after complete chart review. Hence, the study cohort included 160 eligible patients, of whom 25 received neoadjuvant GC and 135 did not receive neoadjuvant chemotherapy. Baseline characteristics of these patients documented before the initiation of definitive therapy for BC are shown in Table 1.
|Variable||GC+RC (n=25)||RC (n=135)||P|
|Age, y, mean (median)||64||(65)||70||(70)||.01a|
|Females, no. (%)||7||(28)||27||(20)||.42b|
|White, no. (%)||24||(96)||130||(96)||.99b|
|Clinical T3/T4, no. (%)||19||(76)||29||(21)||<.001b|
|Any regional adenopathy, no. (%)||9||(36)||15||(11)||.004b|
|Other malignancies, no. (%)||3||(12)||21||(16)||.77b|
|Serum creatinine, mg/dL, mean (median)||0.95||(0.9)||1.04||(1.0)||.44a|
|Creatinine clearance, mL/min, mean (median)||85.3||(83.9)||78.5||(76.7)||.21a|
|WBC, 1000 cells/μL, mean (median)||9.3||(7.5)||8.4||(7.7)||.69a|
|Platelets, cells/μL, mean (median)||274||(253)||264||(251)||.98a|
|Never smokers, no. (%)||7||(29)||21||(17)||.58bc|
|Former smokers, no. (%)||12||(50)||70||(56)|
|Current smokers, no. (%)||5||(21)||35||(28)|
|Intravesical therapy, no. (%)||6||(24)||26||(19)||.59b|
|Mixed histology, no. (%)||9||(36)||30||(22)||.20b|
|High grade, no. (%)||25||(100)||132||(98)||.99b|
Compared with patients treated without neoadjuvant chemotherapy, those who received neoadjuvant GC were on average younger (64 years vs 70 years) and had fewer comorbidities. In particular, the proportions of patients who had neither major cardiovascular disease (MCVD) nor major respiratory disease (MRD) were 76% in the GC + RC group and 53% in the RC group (Table 1). In addition, no one in the GC + RC group had a combination of MCVD and MRD, whereas in the RC group, 7% of patients had such combination. On the other hand, compared with patients in the RC group, patients in the GC + RC group were more likely to have direct extravesical extension of BC based on clinical staging (76% vs 21%) and were more likely to have regional adenopathy (36% vs 11%). The chemotherapy group also included a higher proportion of mixed histology tumors, although this difference was not statistically significant. The distribution of other covariates was fairly similar in the 2 treatment groups (Table 1).
Missing values were uncommon in this series. Serum creatinine, white counts, and platelet counts before the initiation of definitive therapy for BC were unknown for 7 patients. These patients were excluded from the calculation of the average serum creatinine and creatinine clearance and average cell counts in Table 1 but were included in other analyses. Smoking status was unknown for 10 patients. These patients were excluded from the calculation of proportions of current, former, and never smokers in Table 1 but were included in other analyses.
In this series, neoadjuvant GC was typically administered in 3-week cycles according to the following protocol: gemcitabine 2,000 mg/m2 per cycle (as a split dose) and cisplatin 70 mg/m2 per cycle (as a single dose or a split dose). During the course of treatment, doses and schedules were adjusted as deemed necessary by medical oncologists based on cell counts and other considerations. The numbers of cycles completed were as follows: 3 patients completed 2 cycles, 8 patients completed 3 cycles, and 14 patients completed 4 cycles. The mean and the median number of cycles completed were 3.44 and 4, respectively. Median time from diagnosis of muscle invasion to cystectomy was 4.8 months in patients treated with neoadjuvant chemotherapy and 1.5 months in patients treated without neoadjuvant chemotherapy.
There was evidence of tumor down-staging in the bladder after neoadjuvant chemotherapy. Stage pT0 at cystectomy was found in 5 of 25 patients (20%) in the GC + RC group and in 7 of 135 patients (5%) in the RC group (model-based risk difference adjusted for clinical stage = 16%, P = .03; Table 2). Compared with patients treated without neoadjuvant chemotherapy, those treated with neoadjuvant GC were also significantly less likely to have residual invasion of lamina propria, residual invasion of detrusor muscle, and direct extravesical extension at cystectomy (end points <pT1, <pT2, <pT3, respectively; Table 2).
|End Point||GC+RC (n=25)||RC (n=135)||cRD||95% CI||P||aRD||95% CI||P|
In contrast, there was no evidence of a significantly beneficial effect of chemotherapy on the risk of positive nodes at cystectomy, even after adjustment for clinical stage, radiographic regional adenopathy, and the number of nodes examined (adjusted risk difference 4%, P = .74; Table 2). The average number of nodes examined was slightly greater in the GC + RC group than in the RC group (21.4 vs 16.1, P = .02). The corresponding medians were 20 and 15, respectively. Among patients with positive nodes, the average node density (the ratio of the number of positive nodes to the number of nodes examined) was 0.24 in the GC + RC group and 0.30 in the RC group (P = .44). In both treatment arms, pathologically positive nodes were primarily, although not exclusively, associated with direct extravesical extension at RC (stage ≥pT3; Table 3).
|No. N+/n||(%)||No. N+/n||(%)|
In this series, pelvic adenopathy was not a highly accurate predictor of pathologic nodal status even in the absence of neoadjuvant chemotherapy. In the RC group, nodal metastases were found pathologically in 73% (11/15) of patients with baseline adenopathy and in 39% (47/120) of patients without adenopathy. Hence, both false-negatives and false-positives occurred, although the false-negatives were more common. In the GC + RC group, nodal metastases were found pathologically in 44% (4/9) of patients with adenopathy and in 44% (7/16) of patients without adenopathy. Four of 5 patients with pT0 after GC had adenopathy at baseline. Hence, radiographic adenopathy did not predict poor response to GC; however, it was not a highly accurate indicator of true nodal status. Although in patients with adenopathy the proportion of pN+ cases was greater in the RC group than in the GC + RC group (73% vs 44%), this difference was based on small numbers and was not statistically significant (P = .21). In patients without adenopathy, the proportion of pN+ cases was slightly greater in the GC + RC group than in the RC group (44% vs 39%), but this was not statistically significant (P = .79).
The proportions of patients with negative surgical margins at RC were similar in the 2 treatment groups (Table 2). After adjustment for clinical stage, the estimated risk difference was 8%, favoring chemotherapy; however, this was not statistically significant (P = .41; Table 2). The number of chemotherapy cycles completed when treated as an ordinal variable (2, 3, 4) was not significantly associated with tumor down-staging in this series (P > .1 for all down-staging end points in the crude analysis and after adjustment for clinical stage, adenopathy, and the number of nodes examined).
We also examined the association of mixed histology with tumor down-staging. The most common nonurothelial components were squamous (n = 18), glandular (n =8), and sarcomatoid (n = 3). Other components were present in 2 or fewer cases each and included nested cell, basal cell, clear cell, small cell, lymphoepithelial, plasmocytoid, signet ring, and carcinosarcoma. All patients with nonurothelial components also had malignant urothelial histology. Among the 25 patients in the GC + RC group, mixed histology, as determined from transurethral resection (TURBT) specimens, was found in 9 patients, of whom 1 (11%) had stage pT0 (UC with adenocarcinoma on TURBT), 2 (22%) had stage pTis (both were UC with squamous differentiation on TURBT), and 6 (67%) had muscle-invasive disease. Among the 135 patients in the RC group, mixed histology (based on TURBT specimens) was found in 29 patients, of whom none had stage pT0, 2 (7%) had stage pTis or pT1 (both were UC with carcinosarcoma on TURBT), and 27 (93%) had muscle-invasive or more advanced disease. Because of the small number of mixed histology cases in the GC + RC group, analysis of the interaction of treatment with histologic type was highly imprecise. However, there was no clear evidence that the down-staging effect was limited only to pure UCs or only to mixed histology cases. For the primary down-staging end point (stage pT0), the estimated risk difference for the treatment effect adjusted for clinical stage was 0.19 for pure UC and 0.12 for mixed tumors, interaction P = .64).
The median potential follow-up time in these series was 32 months. Approximately 55% of the patients were treated in the last 3 years (2007, 2008, 2009) and therefore had <3 full years of potential follow-up; the remaining 45% were treated in 1999 to 2006 and had at least 3 full years of potential follow-up. A total of 66 of 160 patients (41%) died during the study period. There were 8 deaths among 25 patients (32%) in the GC + RC group (all from BC) and 58 deaths among 135 patients (43%) in the RC group (76% from BC, 24% from other causes, most frequently cardiovascular disease). We did not observe any discrepancies between the underlying cause of death obtained from the National Death Index and information available from the medical record.
Crude Kaplan-Meier survival curves for all-cause mortality in each treatment group are shown in Figure 1. The survival function estimate for the GC + RC group was fairly imprecise, based on only 8 deaths, but risk appears lower for GC + RC over the first 15 months, though not thereafter, implying potentially nonproportional hazards. These survival curves may not accurately reflect the treatment effect because the 2 treatment groups were different with respect to certain baseline characteristics related to mortality (Table 1). Compared with patients in the RC group, those in the GC + RC group were younger and had fewer comorbidities but had more advanced tumors based on clinical staging (Table 1).
To estimate the treatment effect adjusted for baseline characteristics, patients who had both MCVD and MRD were excluded from analysis because no one in the GC + RC group had both MCVD and MRD, whereas 7% of patients in the RC group had this combination (Table 1). The restricted cohort included 25 patients (8 deaths) in the GC + RC group and 126 patients (55 deaths) in the RC group. The treatment effect on all-cause and BC-specific mortality was estimated using the Cox model, stratified on MCVD/MRD comorbidity indicator and clinical stage, and adjusted for regional adenopathy and age. No adjustment was made for pathologic staging variables because these variables are in the “causal pathway” from treatment to survival (chemotherapy may improve survival by shrinking or destroying the tumors, which would be manifested as pathologic down-staging).
The estimated treatment effect on all-cause mortality was hazard ratio (HR) = .61 (95% confidence interval [CI], 0.26-1.42; P = .25), in favor of GC + RC. For the BC-specific hazard, the estimated treatment effect was HR = 0.76 (95%CI, 0.31-1.83; P = .5). The interaction of treatment with histologic type was not statistically significant for either outcome (both P > .8). There was insufficient evidence of violations of the proportional hazards and linearity assumptions (all P > .1), but the power of these tests was very limited.
Creatinine clearance, white count, platelet count, history of other malignancies, smoking status, sex, and surgeon were not significantly associated with survival (all P > .1), after controlling for treatment, clinical stage, comorbidities, and age, and were thus not included in the final models.
Of note, pathologic variables documented at cystectomy such as the T stage, nodal status, lymphovascular invasion, and surgical margins were independently associated with survival in this study (Table 4; these are established prognostic factors for BC). The cutoffs for the nodal categories (1, 2-3, >3) were based on tertiles of positive nodes among patients with nodal metastases.
|Pathologic stage (≥T3 vs <T3)||2.79||(1.42-5.50)||.003|
|Lymphovascular invasion (present vs absent)||2.14||(1.19-3.85)||.01|
|Surgical margins (positive vs negative)||2.04||(1.03-6.07)||.04|
|No. of nodes examined (0, 1, 2, …)||0.96||(0.93-0.99)||.02|
|No. of positive nodes (1 vs 0)||1.13||(0.52-2.45)||.04a|
|No. of positive nodes (2-3 vs 0)||1.36||(0.64-2.89)|
|No. of positive nodes (>3 vs 0)||2.02||(1.04-3.90)|
Seven of 25 (28%) of patients in the neoadjuvant GC group and 42 of 135 (31%) of patients in the control group received adjuvant chemotherapy (AC) after cystectomy.
The most frequently used AC regimens were gemcitabine plus cisplatin (36% of all AC cases), gemcitabine plus Taxol (19% of all AC cases), and gemcitabine plus carboplatin (14% of all AC cases). The numbers of cycles completed were as follows: 4 patients received 1-2 cycles, 8 patients received 3 cycles, 25 patients received 4 cycles, 7 patients received >4 cycles, and for the remaining 5 patients the number of cycles was unknown.
The use of AC was strongly associated with pathologic stage. Among the 49 patients in the AC group, 46 (96%) had either direct extravesical extension or positive nodes at cystectomy. In contrast, among the 111 patients treated without AC, 64 (58%) had either direct extravesical extension or positive nodes at cystectomy. Among patients who had either direct extravesical extension or positive nodes at cystectomy, the use of AC was associated with improved survival (HR = 0.40; 95%CI, 0.21-0.78; P = .007; adjusted for pathologic variables in Table 4 as well as age and comorbidities). This comparison, however, was not based on a randomized treatment assignment and could be confounded by the patients' ability to tolerate chemotherapy after cystectomy, which may not be adequately reflected by comorbidities documented at baseline. Hence, the observed association of AC with improved survival may not represent the true treatment effect.
The purpose of this study was to investigate the effect of neoadjuvant chemotherapy with GC on pathologic down-staging and survival of patients with locally advanced BC. Our findings suggest that this chemotherapy regimen is capable of down-staging the tumors in the bladder. In particular, stage pT0 at cystectomy (the primary down-staging end point) was documented in 20% of patients in the GC + RC group and in 5% of patients in the RC group. Both of these proportions were noticeably lower than the corresponding proportions originally reported from the S8710 trial of MVAC, where stage pT0 was found in 38% and 15% of patients who received cystectomy in the MVAC + RC and RC arms, respectively.5 However, some patients in the S8710 trial did not receive cystectomy for various reasons, including disease progression/unresectable disease at laparotomy, personal choice, or unknown reasons. If one makes the conservative assumption that all patients who did not receive cystectomy in S8710 had residual disease, the proportion of pT0s in the MVAC + RC and RC arms of this trial would decrease to 30% and 11%, respectively.16, 17 Both of these proportions (computed under the most conservative assumptions) are still noticeably higher than what was observed in our current study of GC.
The difference between the proportion of pT0s after GC (20%) in our series and the corresponding proportion after MVAC (30%) in the Southwest Oncology Group (SWOG) study does not necessarily indicate that GC is inferior to MVAC. It must be recognized that in our series, one of the main factors that influenced the decision to administer neoadjuvant chemotherapy was clinical evidence of extravesical disease (clinical stage T3+ and/or adenopathy), while in the SWOG study treatment was randomized and all patients were clinically N0 at baseline. Therefore, patients treated with GC in our series on average had more advanced disease than patients treated with MVAC in the SWOG trial. Hence, if GC and MVAC were equally effective in terms of tumor down-staging, GC would be expected to produce a smaller proportion of pT0s in our setting.
The proportion of pT0s among patients treated without neoadjuvant chemotherapy in our series was only 5%, which is somewhat smaller than the corresponding conservatively estimated proportion in the SWOG trial (11%), although similar to what has been reported in other institutional series.3, 18 For example, Stein et al reviewed 1,054 cystectomies performed at the University of Southern California for biopsy-confirmed muscle-invasive and high-grade nonmuscle-invasive BC and reported that 6% of patients had stage pT0 at cystectomy.3 In our study, all patients had documented muscle invasion before the initiation of definitive therapy for BC, which may explain the relatively small proportion of pT0s.
Although in our series neoadjuvant GC appeared to shrink the tumors in the bladder, we did not observe a clear beneficial effect of GC on disease in the nodes. The proportions of patients with pathologically positive nodes were about the same in both treatment groups in the crude analysis (GC + RC = 44%; RC = 43%). Adjustment for adenopathy and clinical stage shifted this balance slightly in favor of GC, but the adjusted effect was small in magnitude (a risk difference of 4%) and not statistically significant.
One possible explanation for this is that GC is truly less effective in the nodes than in the bladder (eg, due to tissue-specific differences in drug concentrations or tumor-related factors). These possibilities, however, are purely hypothetical because no data are currently available to support or refute them. As an alternative explanation, one must also consider the possibility of residual confounding, which is a general limitation of observational studies of therapeutic interventions. It may be the case that the proportion of truly node-positive patients in the GC + RC group before the initiation of neoadjuvant chemotherapy was higher than the proportion of node-positive patients in the RC group. Although we made adjustment for adenopathy in the analysis, it is possible that compared with patients in the RC group, those in the GC + RC group had more extensive adenopathy that could not be completely accounted for in the analysis. In addition, adenopathy was not a highly accurate predictor of the true nodal status even without neoadjuvant chemotherapy. Both false-negatives and false-positives occurred, although the false-negatives were more common. These considerations explain why adjustment for adenopathy made relatively little difference in the analysis of tumor down-staging in the nodes. Nevertheless, it appears that if GC had any beneficial effect on disease in the nodes, it was not very large in magnitude because as many as 44% (11/25) of patients in our series had positive nodes after neoadjuvant chemotherapy.
The effect of MVAC and cisplatin, methotrexate, vinblastine on the risk of nodal metastases at cystectomy was not previously reported from randomized trials of these regimens because nodal status by itself was not an end point in these trials.5, 6 There is some evidence from observational studies that among patients who present with radiographic N+ or even microscopically confirmed N+ disease and achieve clinical response to neoadjuvant MVAC, some patients have pathologically negative nodes at cystectomy.19, 20 For GC, evidence of activity in the nodes is more limited. In Weight et al's series, positive nodes at cystectomy were found in 38 of 88 patients (43%) treated without neoadjuvant chemotherapy and in 11 of 29 patients (38%) treated with neoadjuvant chemotherapy (of which 20 patients received GC and 9 received other regimens).7 These proportions were similar to proportions observed in our series. Hence, although neoadjuvant GC may down-stage the primary tumors in a sizable proportion of patients, many patients have regionally metastatic disease at cystectomy despite neoadjuvant chemotherapy and have a substantial risk of death from BC. In our series, 8 of 25 patients (32%) in the GC + RC group died from BC.
The proportions of patients with positive surgical margins in our series were similar in the 2 treatment groups (GC + RC = 12%; RC = 11%). After adjustment for clinical stage and adenopathy, the risk difference was equal to 8% favoring GC, but this difference was not statistically significant. In the SWOG trial, positive margins were found in 7% of patients in the MVAC + RC arm and in 14% of patients in the RC arm (risk difference 7% favoring MVAC, P = .1).5 Because positive margins are relatively uncommon even without neoadjuvant chemotherapy, a definitive reduction in the risk of this outcome after chemotherapy is generally difficult to demonstrate. Overall survival in the RC group in our series was similar to that in the RC arm of the SWOG trial (3-year survival approximately 50%).
The strengths of our study include direct access to all medical records, which allowed for accurate documentation of all relevant information, and the use of the National Death Index, which eliminated the problem of loss to follow-up. The most important limitations of the current study are lack of randomization in the assignment of treatment regimens and relatively small sample size.
Lack of randomized treatment assignment is a general limitation of observational studies of therapeutic interventions because the decision to administer a particular treatment may often be influenced by factors that also affect the outcome. If all such factors are documented and accounted for in the analysis, the treatment effect can be estimated without bias. The feasibility of this may depend to some extent on the choice of the end point.
For tumor down-staging end points, adjustment for clinical T stage, adenopathy, and the number of nodes examined may be sufficient. Although the possibility of residual confounding cannot be entirely excluded here, its direction is predictable because in the absence of randomization, those who receive neoadjuvant chemotherapy on average had more extensive tumors (possibly even within each clinical stage) than those who are treated without neoadjuvant chemotherapy. In particular, it is possible that cT3/T4 cases in the GC + RC group had more extensive disease than the cT3/T4 cases in the RC-only group. However, if residual confounding occurred in our study in the analysis of tumor down-staging in the bladder (end points pT0, <pT1, <pT2, <pT3), it could not account for the observed treatment effect because any such residual confounding would produce bias in the opposite direction.
With nodal status at cystectomy, we observed no significantly beneficial treatment effect, and this could theoretically be attributed at least to some extent to residual confounding. Survival analysis could also theoretically be influenced by residual confounding because complete documentation by retrospective chart review of all factors that affect treatment decisions and survival may not be feasible. Nevertheless, we were able to document the clinical staging variables and the major comorbidities and incorporate them in the models. Hence, it is unlikely that confounding played a major role in survival analysis.
In addition to lack of randomization in treatment assignment, the current study was limited by a relatively small sample size in the GC + RC group. Despite this, there was significant evidence of tumor down-staging in the bladder by GC (end points pT0, <pT1, <pT2, <pT3; Table 2). Hence, because the null hypotheses were rejected, power is not technically an issue in the analysis of tumor down-staging in the bladder. Nevertheless, small sample size limits the precision of estimation irrespective of significance testing, and this must be acknowledged as a limitation. In survival analysis in particular, the treatment effect was estimated fairly imprecisely because there were only 8 events in the GC + RC group. Limited precision and power in survival analysis were recognized a priori at the design stage. For this reason (and because of concerns about residual confounding) survival analysis was considered mostly a descriptive technique, whereas the primary end point of the study was defined as tumor down-staging to pT0.
Another limitation of this study is lack of detailed information on drug delivery. In this series, neoadjuvant GC was typically administered in 3-week cycles according to the following protocol: gemcitabine 2,000 mg/m2 per cycle (as a split-dose) and cisplatin 70 mg/m2 per cycle (as a single dose or a split-dose). However, during the course of treatment, doses and schedules were adjusted as deemed necessary by medical oncologists based on cell counts and other considerations, and the details of this were often not available from the charts because chemotherapy was frequently administered in clinics not directly affiliated with Strong Hospital. Nevertheless, we were able to determine the total number of cycles completed for each patient and include this variable in the analysis.
Finally, it must be emphasized that pathologic response rates after GC in our study cannot be directly compared with response rates after MVAC in S8710 due to differences in study design (observational vs randomized) and inclusion criteria, as well as the high proportion (almost 20%) of cancelled or aborted cystectomies in S8710. Hence, we cannot make any definitive claims concerning GC's equivalence to MVAC in the neoadjuvant setting and acknowledge this as a limitation of the current study. Nevertheless, because published data on pathologic tumor response to GC are limited and somewhat contradictory, we believe that our study represents an important addition to the literature.7, 10
In summary, our findings suggest that neoadjuvant chemotherapy with GC is capable of down-staging the tumors in the bladder, although we did not observe a clearly beneficial effect of chemotherapy in the nodes. There was insufficient evidence that the use of neoadjuvant chemotherapy improved survival; however, this study was not adequately powered for survival analysis.
Funding for this work was provided by the Ashley Family Foundation.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.