Although technically well executed, the decision analysis presented by Parikh and Scher in this issue of Cancer has significant limitations in guiding clinical decision making.1 From a clinician's view, decision modeling is useful as an academic exercise to form an initial estimate of how different therapies compare (perhaps, for example, to decide if 2 treatments proposed as competing arms in a randomized trial have clinical equipoise), but its actual clinical utility is limited by the abstractness of the technique, which divorces it from clinical reality in several ways: 1) patient-derived utilities are usually not measured in the heat of battle, when patients are in the emotional throes of dealing with a new cancer diagnosis without really understanding whether they have life-threatening disease or actually knowing through personal experience how side effects will affect their daily routine; 2) the predicted cancer-related outcomes are based on prior reports that cannot take into account technical and experiential advances since their initial publication—in the Parikh and Scher model, the radiotherapy + hormone (RT + H) results were based on a trial started in 1992 (reference 9 in Parikh and Scher1), early in the prostate-specific antigen (PSA) era before PSA-induced stage migration and before the use of high-dose conformal techniques, and the RP + RT trial was based on a Southwest Oncology Group trial started in 1987 (reference 13 in Parikh and Scher1) and is subject to similar limitations; and 3) the advantage measured in quality-adjusted life-years is usually modest and only describes what is likely to happen to an average patient, making it difficult to translate into a meaningful survival estimate for an individual patient. These issues are paramount for screen-detected prostate cancer (and most high-risk disease is detected by PSA screening in the current era), where estimates of cure are limited by the lack of randomized trials comparing different treatment modalities and that by various cohort analyses appear to be similar for surgery, external radiation therapy, and brachytherapy-based approaches, and where decrements in quality of life are domain-specific but similar in total burden across these therapies.2 The importance of treatment era and experience is further emphasized by data showing that at least for radical prostatectomy the chance for cure is influenced by the surgeon's prior volume of cases.3
In addition to these general concerns, the current analysis is limited by several specific issues: 1) true local recurrence after radical prostatectomy (RP) as documented by positive biopsy or radiographic findings is very rare, and its inclusion in the model does not reflect clinical reality in the PSA era; 2) the model does not take into account biochemical failure alone after RP, which is far more common than local recurrence, produces anxiety in patients and physicians, and frequently leads to inappropriate use of hormone therapy; 3) the model does not include the potential curative effect of salvage radiotherapy for biochemical recurrence after RP; and 4) the model does not include an estimation of the late adverse effects of external beam radiation therapy (EBRT) (“All toxicities from surgery or radiotherapy were assumed to occur within 3 years of therapy”), which can result in grade 3 or higher genitourinary and/or rectal toxicity and a slight increased risk of radiation-induced second malignancies.
Despite the limitations of this study, the authors have addressed an important and frequently asked clinical question: what is the best way to treat high-risk prostate cancer? Level 1 evidence supports combined androgen deprivation therapy (ADT) + EBRT over EBRT alone, and as a result most high-risk patients have received this type of therapy, with reported 5-year and 10-year cancer-specific mortality of 3.2%4 and 11.3%,5 respectively. The major downside to this approach is the acute and chronic toxicity of ADT, which has been shown in prospective clinical trials to be associated with significant adverse effects, including hot flashes, obesity, osteopenia and osteoporosis resulting in increased fracture risk, sarcopenia, metabolic syndrome, and diabetes.6 Surveillance, Epidemiology, and End Results data also demonstrate that ADT is significantly associated with an increased incidence of cardiovascular disease, including coronary heart disease, myocardial infarction, and life-threatening ventricular arrhythmias.7, 8 Even ADT of limited duration used in patients undergoing EBRT with curative intent has been observed to cause long-lasting adverse effects on sexuality and vitality.2 A therapy that provides similar rates of cancer control in men with high-risk disease but avoids ADT would clearly be advantageous. In this regard, several retrospective analyses report excellent long-term outcomes of men with high-risk prostate cancer treated initially with RP, with 10-year cancer-specific mortality rates of 8% to 15% and metastasis rates in the range of 15% to 27%.9-12 Two series of RP used as initial therapy for men with high-risk disease demonstrated similar high rates of disease control, while noting that 70% of patients had avoided ADT at 10 years after surgery.12, 13
Furthermore, in contrast to the conclusions of the current decision model, 2 large comparative series have suggested that RP is associated with a lower risk of metastases and cancer-specific mortality than EBRT, with the differences in outcomes driven mostly by those with high-risk disease. In a large academic single-institution series, Zelefsky et al reported the 8-year probability of freedom from metastases was 97% for RP and 93% for EBRT, with a hazard ratio (HR) of 0.35 in favor of surgery after adjusting for case mix.14 The results were similar for prostate cancer-specific mortality, with an HR of 0.32. In the CaPSURE registry, an amalgam of academic and community practices, Cooperberg et al reported an adjusted HR of 2.21 for cancer-specific mortality in patients treated by EBRT relative to RP.15 Both series are limited in their interpretation because of their nonrandomized nature and relatively short follow-up, with a limited number of prostate cancer-related deaths.
In light of all these data, one is still left wondering what is the best way to treat men with high-risk prostate cancer. A decision model cannot replace a randomized trial, and in the absence of data from the latter, the relative advantages of surgery versus EBRT and hormones in this situation remain debatable. The good news is that even men with high-risk cancers can look forward to reasonable long-term disease-free survival whether treated by surgery or radiation; recent data in a multi-institutional pooled analyses of >24,000 RPs showed that 50% of men with Gleason 8 disease or seminal vesicle invasion were still alive 15 years after surgery.16 The bad news is that not all men with high-risk disease are cured by any modality, and all therapies come with a physical and emotional price. The development of new systemic agents directed at the androgen receptor and other molecular targets holds promise for a new era in therapy for biologically aggressive cancers of any stage, and impels us to design and rigorously evaluate new approaches of combined modality therapy rather than continue to argue over the relative merits and demerits of 2 therapies, surgery and radiation, that as currently practiced have clearly defined limits of benefit and harm.