Adjuvant Chemotherapy for Bladder Cancer: Using Population-Based Data to Fill a Void of Prospective Evidence

In the article that accompanies this editorial, Galsky et al use the National Cancer Data Base (NCDB) to examine the benefit associated with postcystectomy adjuvant chemotherapy for bladder cancer. A great deal of effort has been invested in optimizing the paradigm for perioperative chemotherapy in bladder cancer. In the neoadjuvant setting, two phase III studies have suggested an overall survival (OS) advantage with the use of cisplatin-based combination chemotherapy before cystectomy. These results are bolstered by meta-analytic patient-level data derived from multiple prospective studies that suggest a 5% to 10% absolute improvement in 5-year OS using this approach. Consequently, guidelines such as those from the National Comprehensive Cancer Network and the European Association of Urology offer clear recommendations for use of cisplatin-based combination chemotherapy in the neoadjuvant setting in cisplatin-eligible patients. The same dogmatic stance cannot be taken in the adjuvant setting. A study reported by Skinner et al in 1991 randomly assigned 102 patients with either pT3-4 or node-positive disease to receive cisplatin-based chemotherapy or observation after radical cystectomy. The study suggested a significant improvement in OS with adjuvant chemotherapy (4.3 v 2.4 years; P5.0062), whichwas an encouraging initial signal for adjuvant treatment. A logical goal was to then increase the size and magnitude of subsequent trials to support these initial findings. Unfortunately, in the 25 years that have elapsed since this initial report, no single trial has definitively established the role of adjuvant therapy as a standard. Several studies reported more recently have had ambitious designs but have failed to accrue. Perhaps most notably, a phase III European Organisation for Research and Treatment of Cancer study had a planned enrollment of 1,344 patients. Ultimately, only 284 patients were accrued; although a 22% improvement in OS was noted with adjuvant cisplatin-based chemotherapy, this failed to reach statistical significance. Two other contemporary trials have faced a similar fate—in total, these studies accrued only 620 of 2,334 patients planned for enrollment. The cause of failure in these trials is likely multifactorial. A lack of a cohesive multidisciplinary approach in a urologic practice may limit referrals for adjuvant therapy trials. Furthermore, oncologists and urologists may be inclined to treat patients based on intuition; for instance, they may opt to offer adjuvant treatment to a patient with node-positive disease rather than subjecting him or her to random assignment. Whatever the issue, it is unlikely that more studies of adjuvant chemotherapy will overcome the obstacles faced thus far. In the absence of additional prospective trials, clinicians may be tempted to rely on meta-analytic data. Leow et al recently performed a pooled analysis across nine randomized studies including 945 patients. The study identified a cumulative OS benefit of 23% with adjuvant chemotherapy and a 34% improvement in disease-free survival, particularly in those patients with node-positive disease. The purist will be quick to acknowledge how unsatisfying it is to combine underaccruing prospective studies to form positive meta-analyses. Where can one turn beyond such efforts? Retrospective data sets are one possibility. Svatek et al assessed a cohort of 3,947 patients from 11 institutions, of whom 923 (23.6%) received adjuvant chemotherapy off protocol. The use of adjuvant chemotherapy was associated with improved OS (hazard ratio [HR], 0.75; 95% CI, 0.62 to 0.90; P 5 .002), with a more pronounced benefit across higher-risk subgroups. A challenge with use of such data is the potential for selection and referral biases. As eloquently stated by Booth and Tannock in a separate editorial, population-based observational research may serve as an excellent partner for randomized controlled trials and avoid the noted biases with retrospective institutional experiences. The study by Galsky et al evaluating NCDB data is therefore well poised to complement the beleaguered data sets for adjuvant therapy. As noted by the authors, the NCDB data set includes roughly 70% of all newly diagnosed cancer cases. Patients examined in their analysis had pT3-4 or node-positive disease after radical cystectomy and had not received any neoadjuvant chemotherapy. Among 5,653 patients identified from 2003 to 2006, 1,293 (23%) received adjuvant chemotherapy. Where their study differs from similar efforts is in the careful attention to confounding factors, principally achieved by use of propensity scores.