Preoperative Risk Assessment

Introduction

There was an estimated 233,000 cases of prostate cancer diagnosed in the United States in 2014. There are a variety of therapeutic options for men with clinically localized prostate cancer including radical prostatectomy, radiation therapy, and active surveillance. Multiple variables ultimately factor into a patient’s decision on which therapeutic avenue to pursue including overall health status and life expectancy, tolerance of side effects, and oncologic efficacy. Further complicating the decision-making process is the fact that a significant number of screen-detected prostate cancers have a low potential to ultimately result in patient mortality.

Pretreatment decision-making for any disease is a complex process and interpretation of a multitude of factors, patient and disease based, must be considered. Given the multitude of treatment options, heterogeneity of tumor biology, and quality of life implications associated with treatment, the decision to move forward with treatment and which treatment can be agonizing decisions for patients and their clinicians. There are a variety of predictive tools that have been developed to aid physicians and patients in prostate cancer decision-making. Pretreatment predictive tools have been reported in various forms including risk stratification groupings, regression tree analysis, artificial neural networks, and nomograms. Nomograms, defined as a graphic representation of a mathematical formula that incorporates several predictive variables into a model, have been increasingly popular in prostate cancer decision-making.

Incorporation of pretreatment decision tools is necessary as individual clinical judgment has been shown to be quite subjective and often inaccurate. Pretreatment tools have been used with a variety of different goals including predicting pathologic outcomes, non-prostate-cancer related mortality, and biochemical and prostate-cancer-specific survival. To this end, there now exists an extensive list of predictive tools that can be used as decision aids after an initial diagnosis of prostate cancer. The focus of this chapter will be the most commonly used and studied pretreatment decision tools.

Life expectancy

No discussion on prostate cancer treatment decision-making would be complete without framing it within the context of life expectancy. Outcomes for prostate cancer treatment are measured in years and decades and therefore a thoughtful analysis of a patient’s medical comorbidities and life expectancy is necessary for any prostate cancer treatment conversation. The national Comprehensive Cancer Network Guidelines do not recommend definite treatment for patients with low or intermediate risk prostate cancer whose life expectancy is <10 years. While on the surface this is a very rational approach, it is often difficult for physicians to accurately assess comorbidity and life expectancy. In fact, the accuracy of clinicians to predict 10-year life expectancy based on age and comorbidity has been reported to be as low as 69%.

The most basic life expectancy predictor is a life table, of which the most commonly used in clinical practice are the United States Life Tables prepared by the US Department of Health and Human Services. Using census data, life tables estimate life expectancy for a given age based on gender and race. A significant limitation of life tables is that they do not factor additional clinical information into the predictive model. For example, Walz et al. reported that life tables are only 60% accurate in predicting life expectancy in prostate cancer patients who were treated with radiation therapy. In an attempt to better individualize life expectancy, a multitude of predictive tools incorporating patient comorbidity have been reported. One of the most commonly used is the Charlson Comorbidity Index (CCI), which predicts the 10-year mortality for a patient based on 22 comorbid conditions. The CCI has also been applied to a number of malignancies to predict mortality. Specifically in prostate cancer, a CCI of > 2 has been shown to be a predictor of non-prostate-cancer mortality after radical prostatectomy.

Investigators have also incorporated comorbidity and cancer-specific variables to predict survival following prostate cancer treatment. Using four variables including age, CCI, biopsy Gleason score, and PSA, Tewari et al. retrospectively evaluated 1611 men with clinically localized prostate cancer and 4538 age-matched controls. The authors reported a validation C-index of 0.69 for overall survival using this four-variable model. Similarly, Walz et al. evaluated 9131 men treated with either radical prostatectomy or external beam radiation using age and CCI to predict 10-year life expectancy. Their nomogram for predicting 10-year life expectancy after either surgery or radiation was 84.3% accurate.

In summary, it is important for clinicians to take overall comorbidity and life expectancy into account when discussing various treatment alternatives with prostate cancer patients. While subjective physician assessment of life expectancy is inaccurate, several nomograms do exist that were developed specifically in prostate cancer patients, which can better discriminate life expectancy in men with prostate cancer.

Predictors of final pathology

A major handicap in pretreatment decision-making is the discordance between clinical and pathologic stage. Even in patients with clinical low-risk disease, upgrading and upstaging can be significant. For example, El Hajj et al. evaluated 626 patients who met active surveillance criteria based on the Prostate Cancer Research International: Active Surveillance (PRIAS) criteria and noted an upgrading to Gleason score 7 or higher in 45% and upstaging (>pT2) in 21% of patients. Pathologic upstaging for those with low-risk prostate cancer is currently a major limitation for patients seeking active surveillance protocols. Multiple predictive tools have been developed in an attempt to better predict final pathology based on preoperative clinical characteristics. Most of these tools have been developed based on large populations of radical prostatectomy cohorts. Each predictive tool is designed to predict various pathologic outcomes such as Gleason score upgrading, extra-capsular extension, seminal vesicle invasion, and lymph node metastasis.