A comparison of the outcomes of neoadjuvant and adjuvant chemotherapy for clinical T2-T4aN0-N2M0 bladder cancer

Authors


Abstract

BACKGROUND:

Despite evidence supporting perioperative chemotherapy, few randomized studies compare neoadjuvant and adjuvant chemotherapy for bladder cancer. Consequently, the standard of care regarding the timing of chemotherapy for locally advanced bladder cancer remains controversial. We compared patient outcomes following neoadjuvant or adjuvant systemic chemotherapy for cT2-T4aN0-N2M0 bladder cancer.

METHODS:

In a retrospective review of a single institutional database from 1988 through 2009, we identified patients receiving neoadjuvant or adjuvant multiagent platinum-based systemic chemotherapy for locally advanced bladder cancer. Survival analysis was performed comparing disease-specific survival (DSS) and overall survival (OS).

RESULTS:

A total of 146 patients received systemic perioperative chemotherapy (73 neoadjuvant, 73 adjuvant). Of these, 84% (122/146) received cisplatin-based chemotherapy compared with carboplatin-based chemotherapy (24/146, 16.4%). Most patients receiving cisplatin-based chemotherapy were treated with methotrexate/vinblastine/adriamycin/cisplatin (79/122, 64.8%), whereas the remaining patients received gemcitabine/cisplatin (GC) (43/122, 35.2%). In multivariable analysis, there was no significant difference in DSS (P = .46) or OS (P = .76) between neoadjuvant or adjuvant chemotherapy groups. There was statistically significant improvement in DSS when patients received neoadjuvant GC rather than adjuvant GC (P = .049, hazard ratio, 10.6; 95% confidence interval, 1.01-112.2).

CONCLUSION:

In this study, there was no statistically significant difference in OS and DSS between patients receiving neoadjuvant versus adjuvant systemic platinum-based chemotherapy for locally advanced bladder cancer. In addition, there was no significant difference between neoadjuvant and adjuvant cisplatin- or carboplatin-based chemotherapy. Chemotherapy sequence relative to surgery appeared less important than whether or not a patient actually received perioperative chemotherapy. Cancer 2011;. © 2011 American Cancer Society.

INTRODUCTION

In 2009, 70,980 new cases of bladder cancer were diagnosed, and 14,330 patients died secondary to bladder cancer in the United States.1 Although the majority of patients are diagnosed with non-muscle invasive bladder cancer (NMIBC) and experience a favorable outcome, many patients with locally advanced bladder cancer ultimately succumb to metastatic disease with a 5-year overall survival rate after radical cystectomy ranging from 36%-48%.2-5 Systemic chemotherapy either before (neoadjuvant) or after (adjuvant) local extirpative therapy can be used as an adjunct to surgery. The rationale for neoadjuvant chemotherapy is to potentially eradicate the primary lesion and to treat micrometastatic foci of disease before the patient is debilitated by a major surgical procedure. The disadvantage of neoadjuvant chemotherapy is the delay in definitive surgical therapy, especially in patients who are clinically understaged or may respond suboptimally to chemotherapy. The rationale for adjuvant chemotherapy is to reduce the incidence of local and metastatic recurrence in patients with locally advanced disease. Although prognostic stratification may be improved based on pathologic staging, one possible disadvantage of adjuvant chemotherapy is the potential delay in systemic treatment for patients recovering from cystectomy and possible postoperative complications.

Several studies have demonstrated that the integration of either neoadjuvant or adjuvant chemotherapy is associated with a survival advantage. Both the Nordic Cystectomy I trial and the United States Intergroup trial (SWOG 8710) demonstrated improved overall survival in patients who received neoadjuvant chemotherapy.6-8 Trials published by Stockle et al.9 and Skinner et al.10 have demonstrated a significant decrease in rates of progression among patients who received adjuvant chemotherapy. In addition, a recent multicenter retrospective study demonstrated that off-protocol adjuvant systemic chemotherapy was independently associated with improved survival, especially among higher-risk subgroups.11 Despite evidence supporting perioperative chemotherapy, there is a paucity of randomized controlled trials comparing neoadjuvant and adjuvant chemotherapy; therefore, the standard of care regarding the sequence of chemotherapy relative to surgery for the treatment of advanced bladder cancer remains controversial. Given the limited data available in the literature, we sought to compare the clinical outcomes of patients receiving either neoadjuvant or adjuvant systemic chemotherapy at the time of radical cystectomy at our institution.

MATERIAL AND METHODS

The institutional review board–approved bladder cancer outcomes database was reviewed to identify patients who received either cisplatin-based or carboplatin-based chemotherapy between January 1988 and July 2009. All patients had cT2-T4aN0-N2M0 transitional cell carcinoma according to the 2010 American Joint Committee on Cancer tumor–node–metastasis staging scheme.12 All patients with known metastatic disease at the initiation of induction or salvage chemotherapy were excluded from analysis. Patients with pathologically confirmed lymph node metastases at the time of cystectomy receiving adjuvant chemotherapy were included in the study. Each patient included in the final dataset had received multiagent systemic chemotherapy.13 Each patient also underwent extirpative therapy with radical cystectomy and pelvic lymph node dissection, except for 13 patients that refused cystectomy after clinical T0 response to neoadjuvant chemotherapy (data not shown). For the purposes of this study, adjuvant chemotherapy was defined as systemic chemotherapy received in the postoperative period within 6 months of surgery.

The primary endpoints analyzed were disease-specific survival (DSS) and overall survival (OS). Preoperatively, patients were evaluated via physical examination, endoscopic evaluation, laboratory studies, chest radiographs, and cross-sectional imaging of the abdomen and pelvis. Postoperatively, patients were followed with routine laboratory studies, physical examinations, and interval cross-sectional imaging. Additional radiographic and diagnostic tests were performed at the discretion of the attending physician. Length of follow-up and start of survival analysis was determined from date of cystectomy to time of death or last follow-up. Survival analysis was performed using the Kaplan-Meier method with patients stratified into neoadjuvant systemic chemotherapy and adjuvant systemic chemotherapy cohorts. Univariable and multivariable analysis were used to determine significant predictors of DSS and OS. Multivariable analysis was controlled for age, grade, stage, chemotherapy type, and intent, and additionally for race, sex, and smoking history in a second model. Furthermore, a subset analysis was performed for patients who received cisplatin-based chemotherapy based on their regimen. All analyses were performed using α = .05 to determine statistical significance. Statistical functions were performed using Stata 11 (College Station, Texas).

RESULTS

A total of 687 patients underwent radical cystectomy at our institution between January 1988 and July 2009 (Figure 1). After all patients with known metastatic disease who received either induction or salvage chemotherapy were excluded, our final dataset was comprised of 146 patients with cT2-T4aN0-2M0 bladder cancer who received either neoadjuvant chemotherapy (73/146, 50%) or adjuvant chemotherapy (73/146, 50%). Complete response rate after neoadjuvant systemic chemotherapy was 31.5% (26/73). This finding represents a systemic chemotherapy use rate of 146/687 (21%). Among the 687 patients undergoing cystectomy, 180 (26%) had recurrent NMIBC and 80 (12%) had non–transitional cell carcinoma (TCC) histology. One patient with adenocarcinoma and 1 patient with squamous cell carcinoma received systemic chemotherapy. After excluding all patients with NMIBC and non-TCC histology, the percentage of patients with muscle-invasive or more advanced bladder receiving systemic chemotherapy perioperatively was 21% (144/687).

Figure 1.

Consort diagram shows the study cohort of 146 patients receiving systemic chemotherapy for advanced bladder cancer.

Clinical characteristics of the cohort are presented in Table 1. Among the group receiving perioperative chemotherapy, 105 were men (52 neoadjuvant, 53 adjuvant) and 41 were women (21 neoadjuvant, 20 adjuvant). The median age was 65 years (range, 36-83 years). The majority of patients received cisplatin-based chemotherapy (122/146, 83.6%) compared with carboplatin-based chemotherapy (24/146, 16.4%) (Table 2). The majority of patients who received cisplatin-based chemotherapy were treated with methotrexate/vinblastine/adriamycin/cisplatin (MVAC) (79/122, 64.8%), whereas the remaining patients were treated with gemcitabine/cisplatin (GC) (43/122, 35.2%). The distribution between neoadjuvant and adjuvant chemotherapy was relatively equal in the cisplatin-based group, with 62/122 patients (50.8%) and 60/122 patients (49.2%), respectively. Among the patients treated with carboplatin-based therapy, 11/24 (45.8%) received neoadjuvant treatment and 13/24 (54.2%) received adjuvant treatment.

Table 1. Patient Characteristics
CharacteristicsNeoadjuvant (n = 73)Adjuvant (n = 73)P
  • Data are presented as no. (%) unless otherwise specified.

  • a

    Clinical stage.

  • b

    Pathologic stage.

Demographic   
 Age, y, mean ± SD64 ± 10.666 ± 12.3.33
 Men52 (71)53 (73)1.00
 Race (Caucasian)60 (82)56 (77).08
 Race (African American)3 (4)2 (3) 
 Race (Latin)8 (11)1 (1) 
 Race (Other)6 (8)10 (14) 
 Smoking history20 (27)19 (26).85
Pathologic   
 Tumor stage (>T2)18 (25)a40 (55)b≤.001
 Node stage (>N0)5 (7)a29 (40)b≤.001
 Follow-up, mo, median (range)12.81 (0-155)14 (0-159).64
 Death, any cause3850 
 Death, disease-specific2424 
Table 2. Distribution of Chemotherapy Regimens for Patients with cT2-T4aN0-2M0 Bladder Cancer (n = 146)
Chemotherapy RegimenNo. (%)
  1. Abbreviations: GC, gemcitabine/cisplatin; MVAC, methotrexate, vinblastine, adriamycin, cisplatin.

Cisplatin-based122 (83.6)
 MVAC79 (64.8)
 GC43 (35.2)
 Neoadjuvant62 (50.8)
 Adjuvant60 (49.2)
Carboplatin-based24 (16.4)
 Neoadjuvant11 (45.8)
 Adjuvant13 (54.2)

The median follow-up for the neoadjuvant group and adjuvant group was 12.8 months (mean, 27.6 months) and 14 months (mean, 32.8 months), respectively. When DSS was compared in a univariable manner using the Kaplan-Meier method, between patients receiving neoadjuvant versus adjuvant chemotherapy, no significant difference was detected between the 2 groups (P = .307) (Figure 2). There was also no significant difference in OS when patients were stratified by neoadjuvant versus adjuvant chemotherapy (P = .541) (Figure 3).

Figure 2.

Kaplan-Meier estimated disease-specific survival (DSS) is shown for patients treated with neoadjuvant chemotherapy versus adjuvant chemotherapy. No statistically significant difference was identified in DSS between the neoadjuvant and adjuvant systemic chemotherapy groups.

Figure 3.

Kaplan-Meier estimated overall survival (OS) is shown for patients treated with neoadjuvant chemotherapy versus adjuvant chemotherapy. No statistically significant difference was identified in OS between the neoadjuvant and adjuvant systemic chemotherapy groups.

Univariable and multivariable analysis data are presented in Table 3. For the entire cohort, multivariable analysis controlling for age, stage, grade, and chemotherapy type (OS univariable analysis P = .60; DSS univariable analysis P = .360) revealed no significant difference in OS (hazard ratio [HR], 1.08; 95% confidence interval [CI], 0.67-1.73; P = .76) or DSS (HR, 1.24; 95% CI, 0.70-2.18; P = .46) between treatment groups receiving neoadjuvant or adjuvant chemotherapy. A multivariate model including sex, race, and smoking history as well did not change these results (Table 3). In the cisplatin-based chemotherapy group, there was no significant difference in DSS or OS when patients who received neoadjuvant chemotherapy (median survival, 11 months) were compared with patients who received adjuvant chemotherapy (median survival, 12.5 months) on univariable analysis. Both univariable and multivariable analysis demonstrated no significant difference in DSS (P = .555) or OS (P = .573) when patients who received neoadjuvant MVAC (median survival, 16 months) were compared with patients who received adjuvant MVAC (median survival, 22 months).

Table 3. Univariable and Multivariable Analyses
 Overall SurvivalDisease-Specific Survival
UnivariableMultivariable: Age, Stage, Grade, Chemotherapy Type, IntentMultivariableUnivariableMultivariable: Age, Stage, Grade, Chemotherapy Type, IntentMultivariable
HR (95% CI)PHR (95% CI)PHR (95% CI)PHR (95% CI)PHR (95% CI)PHR (95% CI)P
  1. Abbreviations: CI, confidence interval; HR, hazard ratio.

Age0.99 (0.97-1.017).380.98 (0.97-1.00).110.99 (0.97-1.01).170.98 (0.96-1.00).0390.97 (0.94-0.99).0020.96 (0.94-0.99).006
Race (African American)2.19 (0.68-7.05).192.36 (0.72-7.74).163.14 (0.96-10.3).0592.95 (0.88-9.88).08
Race (Latin American)2.95 (1.32-6.57).011.91 (0.74-4.91).184.05 (1.77-9.26).0011.96 (0.68-5.60).21
Race (other)1.73 (0.82-3.68).151.05 (0.46-2.40).911.79 (0.75-4.27).191.09 (0.42-2.77).87
Sex (male referent)1.4 (0.88-2.21).161.31 (0.80-2.14).281.54 (0.89-2.66).121.43 (0.80-2.55).22
Smoking history1.22 (0.76-1.93).411.28 (0.79-2.07).311.44 (0.83-2.49).191.62 (0.91-2.88).1
Tumor stage (>T2)1.22 (0.79-1.07).371.08 (0.66-1.77).751.06 (0.63-1.77).831.3 (0.77-2.18).331.25 (0.69-2.25).461.21 (0.65-2.24).54
High grade1.48 (0.96-2.29).071.37 (0.87-2.17).181.42 (0.88-2.31).151.35 (0.80-2.27).271.16 (0.66-2.03).611.25 (0.69-2.26).46
Cisplatin0.36 (0.21-0.65).0010.33 (0.18-0.60)≤.0010.37 (0.19-0.74).0040.34 (0.18-0.65)≤0.0010.25 (0.13-0.51)≤.0010.29 (0.13-0.65).003
Adjuvant chemotherapy1.12 (0.73-1.73).601.08 (0.67-1.73).761.04 (0.62-1.72).891.28 (0.76-2.16).361.24 (0.70-2.18).461.17 (0.63-2.16).62

When patients who received neoadjuvant GC (median survival 11 months) were compared with patients who received adjuvant GC (median survival, 10.5 months), there was no significant difference in OS in both univariable and multivariable (P = .607) analyses. On multivariable analysis, there was a statistically significant improvement in DSS when patients received neoadjuvant GC rather than adjuvant GC (P = .049, HR 10.6, CI 1.01-112.2). Similar to the cisplatin-based chemotherapy group, there was no difference in DSS (P = .764) or OS (P = .388) using either univariable or multivariable analyses when comparing patients receiving neoadjuvant carboplatin-based chemotherapy (median survival, 8.9 months) with patients receiving adjuvant carboplatin-based chemotherapy (median survival, 10 months). Cisplatin-based chemotherapy regimen, however, was a significant independent predictor of improved OS (HR, 0.33; 95% CI, 0.18-0.60; P≤.001) and DSS (HR, 0.25; 95% CI, 0.13-0.51; P≤.001).

DISCUSSION

The integration of systemic chemotherapy with surgical treatment for locally advanced bladder cancer is not a novel concept. In 1988, Logothetis et al.14 reported that patients with pathologically proven extravesical disease had better survival when they received adjuvant chemotherapy after radical cystectomy. Although the literature contains studies that show no significant survival advantage associated with the administration of neoadjuvant chemotherapy, several studies have clearly demonstrated the benefit of neoadjuvant regimens. The Nordic Cystectomy I trial included 325 patients that were randomized to local radiation therapy and radical cystectomy with or without neoadjuvant cisplatin and adriamycin.6 The study did not demonstrate a significant difference but did report improved 5-year OS when the neoadjuvant chemotherapy group (59%) was compared with the group that did not receive chemotherapy (51%, P = .10), despite lack of methotrexate and vinblastine. In addition, there was a 15% improvement in OS (P = .03) when the survival analysis was limited to only patients with stage T3-T4a disease. The survival advantage associated with neoadjuvant chemotherapy was confirmed in a more recent follow-up study.15 More recently, the United States Intergroup trial (SWOG 8710) reported on 307 patients with clinical T2-T4N0M0 bladder cancer randomly assigned to radical cystectomy with or without 3 cycles of neoadjuvant MVAC. Despite a lack of statistical significance, the study demonstrated a dramatic improvement in overall median survival from 46 months in the surgery alone group to 77 months in the group that received neoadjuvant MVAC (P = .06) with a 14% improvement in overall 5-year survival.7 In addition, the pT0 rate at the time of cystectomy among those receiving neoadjuvant systemic chemotherapy was significantly higher (38%) than the cystectomy alone group (15%).

The literature also provides sufficient support for the use of adjuvant chemotherapy for patients with locally advanced disease and/or nodal involvement. Skinner et al.10 reported 87 patients who were randomized to radical cystectomy with or without adjuvant chemotherapy, mainly in the form of cisplatin/cyclophosphamide/adriamycin. Despite the lack of methotrexate and vinblastine, the study demonstrated significant improvement in median progression-free survival from 2.4 years in the surgery alone group to 4.3 years in the group that received adjuvant chemotherapy (P = .006). Although there was no statistically significant difference in the 5-year OS when patients received adjuvant chemotherapy, there was a nonstatistically significant improvement from 39% to 44%. Shortly after Skinner et al. reported their results, Stockle et al.9 published a series of 83 patients that were randomized to radical cystectomy with or without adjuvant MVAC or methotrexate/vinblastine/epirubicin/cisplatin. The trial was stopped before completion of patient accrual when an interim data analysis revealed a significant improvement in 5-year progression-free survival from 13% in the surgery alone group to 59% in the adjuvant chemotherapy group. The results of this study are confounded, however, because patients randomized to receive no adjuvant chemotherapy did not receive any chemotherapy even upon relapse. More recently, Paz-Ares et al.16 conducted a randomized phase III clinical trial (SOGUG 99/01) comparing adjuvant paclitaxel/gemcitabine/cisplatin (PGC) (68 patients) with observation (74 patients) following resection of muscle-invasive bladder cancer (pT3-4 and/or pN+). At a median follow-up of 30 months, OS was significantly prolonged in the PGC arm (median, not reported; 5-year OS, 60%) compared with observation (median, 26 months; 5-year OS, 31%) (P<.0009). Disease-free survival (P<.0001), time to progression (P<.0001), and DSS (P<.0002) were also superior in the PGC arm. Because the study was closed prematurely due to poor accrual, definitive conclusions are limited. In addition, a recent study of 932 patients receiving off-protocol adjuvant systemic chemotherapy identified adjuvant therapy to be independently associated with improved survival (HR, 0.83; 95% CI, 0.72-0.97; P = .017). The advantage of adjuvant chemotherapy was most pronounced in the highest-risk patient group with advanced pathologic stage and lymph node involvement.11

Although there is compelling evidence to support the use of either neoadjuvant or adjuvant chemotherapy for patients with locally advanced bladder cancer, the question of sequence of perioperative chemotherapy remains controversial. To our knowledge, no prospective study has directly compared a full course of neoadjuvant chemotherapy to a full course of adjuvant treatment. Despite the lack of direct comparison studies, the literature does provide some clues as to the optimal sequence of perioperative chemotherapy in relation to cystectomy. Millikan et al.17 evaluated 140 patients at the University of Texas MD Anderson Cancer Center who were randomized to radical cystectomy with either 5 cycles of adjuvant MVAC or 2 cycles of neoadjuvant MVAC with 3 cycles of adjuvant MVAC. With a median follow-up of 6.8 years, the study demonstrated no significant difference in DSS or OS. Based on these results, the importance of simply receiving perioperative chemotherapy appears to outweigh the specific sequence of the adjunct treatment.

The results from the current study confirm that there is no statistically significant difference detectable between neoadjuvant and adjuvant systemic chemotherapy in advanced bladder cancer in terms of OS and DSS in our cohort. In subset analysis, there was no significant difference between neoadjuvant and adjuvant cisplatin-based chemotherapy or neoadjuvant and adjuvant carboplatin-based chemotherapy. It should be noted that perioperative carboplatin-based regimens were used in a small number of patients, and this is not standard practice.18 In further analysis of the cisplatin group, there was no difference in OS or DSS when MVAC chemotherapy was given in the neoadjuvant or adjuvant settings. This lack of significant impact held true even when controlling for any pathologic differences in the groups receiving adjuvant and neoadjuvant therapy. With GC therapy, however, there was a significant difference with adjuvant systemic therapy associated with decreased DSS (HR, 10.6; 95% CI, 1.01-112.2; P = .049) compared with neoadjuvant therapy. This difference of DSS between neoadjuvant and adjuvant GC use may be confounded, because patients receiving GC chemotherapy likely had worse prechemotherapy characteristics (including decreased performance status) than those selected for MVAC therapy. This may have affected tolerability of surgery with adjuvant chemotherapy and subsequent survival. These findings must be confirmed with additional clinical trials accounting for preoperative clinical characteristics and chemotherapy selection.

Despite evidence to support the use of perioperative chemotherapy, a multimodal treatment plan is often underused. A review of 7161 patients from the National Cancer Database revealed that only 11.6% of patients with stage 3 bladder cancer were treated with perioperative chemotherapy from 1998 to 2003.19 The same study also demonstrated that adjuvant chemotherapy was more commonly used in 10.4% of patients, compared with 2.2% for neoadjuvant chemotherapy. Our rate of use was 21%, which likely reflects the tertiary academic nature of our medical center, and we had an equal distribution between neoadjuvant and adjuvant chemotherapy. Urologists and patients may be more inclined to pursue adjuvant chemotherapy in an effort to avoid a delay in surgical treatment, especially with the decreased morbidity of radical cystectomy and improvement in quality of life afforded by orthotopic bladder substitutions. Some urologists may also avoid neoadjuvant treatment due to a concern that preoperative chemotherapy may increase the technical difficulty of a radical cystectomy. However, the study by Millikan et al.17 reported no increase in perioperative morbidity after neoadjuvant chemotherapy. With the future advent of molecular markers that may predict chemotherapy response of bladder cancer patients, perhaps the nationwide rates of use for perioperative chemotherapy will increase.

Our study had several limitations. We were confined by the retrospective nature of our study and therefore are exposed to potential biases in the 2 treatment groups. However, in the absence of a prospective randomized trial, this level of evidence may be the only source of information to help guide clinician decision-making. In an effort to control for any differences in risk between the adjuvant and neoadjuvant groups, a multivariable logistic regression model was developed that did not indicate sequence of therapy (neoadjuvant or adjuvant) as a significant factor in outcome. Our observations were also limited to a single tertiary-care center experience. Based on the retrospective nature of the study, we were forced to draw conclusions from a heterogeneous patient population that received a variety of chemotherapeutic regimens that may have differing efficacy. In addition, comorbidity information and preoperative performance status were not available for our cohort.

Furthermore, disease recurrence may not be captured effectively without a prospective trial with a specified follow-up protocol. There is potential error in discussing DSS or disease-free survival rates in a nontrial population. We have attempted to counter this potential issue by obtaining our survival data from the Social Security Death Index in addition to our clinical records. We followed rigid criteria when determining death from bladder cancer, and there was no difference in the outcome determination between neoadjuvant or adjuvant treatment arms. Thus, any inherent inaccuracy secondary to the retrospective nature of this study would be equivalent in both study groups. In addition, median follow-up was shorter than potential follow-up, because more than 60% of patients were treated since 2002, and some patients were lost to follow-up.

Despite these limitations, we feel the present study provides valuable information for clinicians counseling patients with locally advanced bladder cancer in the perioperative setting. Ultimately, a prospective trial ideally designed as a noninferiority trial would be useful, but would require a large patient cohort to determine outcome differences between chemotherapy sequences. Furthermore, it would require a dedicated staff to uphold rigorous inclusion criteria of clinical staging for the neoadjuvant study group and pathologic staging for the adjuvant study group. In addition, the consequences of alternative chemotherapy regimens, including high-dose intensity MVAC, gemcitabine with taxane chemotherapy, and predictive molecular markers should be studied further within the context of chemotherapy sequence.

Conclusion

In this study, neoadjuvant chemotherapy did not demonstrate a significant difference when compared with adjuvant chemotherapy in terms of OS and DSS for both cisplatin-based and carboplatin-based regimens. Completed randomized clinical trials such as SWOG 8710 suggest that chemotherapy offers a survival advantage to patients in the perioperative setting. The current study suggests that the sequence of such chemotherapy is less important than whether or not a patient actually receives perioperative chemotherapy. The findings of this study should be considered when counseling patients regarding the sequence of perioperative chemotherapy and should be validated with a prospective randomized trial.

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

Ancillary