Systemic Therapy for Invasive Bladder Cancer: The Value Proposition.

The report from Yin et al. in this issue of is an important reminder of the utility of neoadjuvant cisplatin-based chemotherapy for invasive, clinically nonmetastatic bladder cancer [1]. In their updated analysis, these investigators suggest that the hazard ratio from this treatment strategy has improved to 0.87, from approximately 0.93; the analysis now includes more than 3,000 cases, some with longer follow-up. It is a pity that there has been a need to justify a treatment that my colleagues and I first tested more than 30 years ago [2, 3] and proved, with level 1 evidence, in two hallmark trials that were initiated more than 20 years ago [4-6] and were even confirmed by an early meta-analysis [7]. That said, updating of information is useful, in that Yin et al. have confirmed and extended the 10-year trends and added important case experience.

Why is this worthy of publication? First, there is an increasing focus on the value proposition in oncology, at a time when costs of treatment are burgeoning and patients are being expected to shoulder much higher proportions of the cost. Increasingly, health planners are looking at the ratio of outcome versus cost [8] and questioning the use of treatments that simply do not provide meaningful increments in survival when compared with physical or fiscal expenditure [9]. The American Society of Clinical Oncology and the European Society of Medical Oncology have issued remarkably similar documents that have attempted to measure “value” in oncology in a structured fashion [9, 10]; these have produced complex algorithms, but they do begin to address a thoughtful and critical appraisal of the return on investment for the expenditure of resources by the individual patient and the community at large. In that context, a hazard ratio for neoadjuvant treatment of bladder cancer of 0.87 is compelling, particularly with median survival figures measured in years rather than months. That said, the upper boundary of 0.95 in the confidence interval suggests that continued follow-up of this domain will be necessary.

The second reason that this study is important is that the reaffirmation of the important effect of neoadjuvant chemotherapy sets into stark contrast the proliferation of underpowered and inaccurate reports that have attempted to equate the usefulness of adjuvant chemotherapy in the same setting, as discussed in detail elsewhere [11]. Although the European Organisation for Research and Treatment of Cancer took on the tough challenge of a randomized trial of cystectomy versus cystectomy plus adjuvant chemotherapy, which was also open in North America, investigators in Europe and the U.S. chose not to support this seminal trial, which closed early because of lack of accrual [12].

Dr. Sternberg and her colleagues reported the available information, showing the presence of a progression-free survival benefit, which was not surprising (given the known activity of cisplatin-based chemotherapy for bladder cancer). However, once salvage chemotherapy had been given to patients relapsing after initial cystectomy-only treatment, no statistically significant difference in survival was observed, although there was a small trend in favor of adjuvant chemotherapy [12]. They also reported the unexpected finding that the real effect of adjuvant chemotherapy was seen in node-negative disease, which might indicate that chemotherapy was partially compensating for suboptimal surgery. Another explanation could also have been that four cycles of adjuvant chemotherapy were sufficient to improve survival for node-negative disease but were not adequate to affect node-positive tumors. Whichever is true, the bottom line is that overall survival was not significantly affected by the expenditure of resources required from the patient and the community when four cycles of adjuvant chemotherapy was delivered.

One of the frustrating aspects of this discussion is the range of attempts to convince our profession of the utility of this adjuvant approach despite the absence of real data. Thus, there have been several meta-analyses, all of which have included a seminal German study that compared cystectomy without salvage chemotherapy to cystectomy plus adjuvant chemotherapy [13, 14]; inclusion of this study prejudices the outcome clinically and statistically. In addition, the situation has been confused by post hoc large database studies (which have ignored randomization and have attempted to use statistical ploys and/or propensity matching to overcome the absence of real level 1 data) [15]. Bladder cancer is a heterogeneous disease, and the population of patients being treated is even more heterogeneous, so that propensity matching essentially reflects the “garbage in/garbage out” principle. In planning treatment of invasive disease, it is important to consider conventional predictors of outcome, such as grade and stage, solid growth pattern, size, aneuploidy and genomic heterogeneity, lympho-vascular invasion, and the presence of hydronephrosis. There is simply no way that propensity matching can account for this level of variation, particularly given the modest numbers of cases included in the recent published exercises. When one compounds this with a consideration of the huge potential costs [11] engendered in an unproven approach with, at best, marginal survival effect and significant morbidity, the argument simply becomes fatuous.

The other important consideration in the work of Yin et al. [1] is their attempted comparison of the methotrexate, vinblastine, doxorubicin, and cisplatin (MVAC) and gemcitabine-cisplatin (GC) regimens in the neoadjuvant context, and their conclusion that GC is inferior. Once again, the question is important because GC has become a standard approach since the demonstration that it is less toxic than MVAC in the metastatic setting [16], with a hazard ratio of 1.09 (which is not “noninferior”). Further confusion on this issue has been created by another propensity-matched retrospective analysis of pooled data from 28 centers (which included data dredging from several prior reports) [17]. Despite different median survival figures (MVAC, 35.5 months; GC, 26.8 months; p = .17) and absence of central pathology or surgical quality review, with only 212 heterogeneous cases, these authors concluded that MVAC and GC were not different, and this study has been used as the basis for routine introduction of GC in the neoadjuvant setting.

When I contemplate the respective levels of rhetoric and reason in the profusion of publications on invasive bladder cancer, I keep coming back to my five basic rules regarding the design of trials for invasive bladder cancer (Table 1) [18]:

Table 1.

Design rules for neoadjuvant bladder cancer trials

Understand the complex biology of invasive bladder cancer when designing trials.

Variable constants can impair outcomes (e.g., understand the importance of optimal surgery, radiotherapy, or chemotherapy in the design of these studies).

Repetition is not the best way to prove a concept.

Randomized trials prove more than historical comparisons (or retrospective analyses with so-called propensity matching).

If it doesn’t make sense, it’s probably wrong.

Fortunately, carefully designed and completed studies, such as the work of Yin et al. [1], despite some obvious methodological issues acknowledged by the authors, allow us to build structure and confidence around a decision that neoadjuvant MVAC chemotherapy followed by definitive local treatment should be the treatment of choice for otherwise healthy and robust patients with T2b–T4 invasive bladder cancer. Randomized clinical trials may not be easy or convenient, but they remain the mainstay of thoughtful, well-designed science that allows us to advance with confidence.