The Role of PSA in the Radiotherapy of Prostate Cancer

Introduction

The diagnosis, staging, treatment, and follow-up of patients with prostate cancer have been dramatically altered by the availability of the serum marker known as prostate-specific antigen (PSA). Because of the sheer numbers of patients with this cancer (317,000 "new" cases were expected for 1996), no other single break- through in oncology has had an impact on so many people in such a short period of time. The explosion of published information related to this marker has been exponential, and there is no evidence yet that a plateau has been reached.

Obviously, a single brief review article cannot cover all of the interesting observations being made about this new tool. It is likely that as this manuscript is being prepared, new findings concerning the use of PSA are being discovered. With this disclaimer in mind, this review will attempt to highlight some of the practical uses of PSA for managing patients undergoing definitive local therapy.

What is PSA?

Prostate-specific antigen is a proteolytic enzyme that belongs to the serine protease family and is generally believed to have an approximate mass of 33 kd. Among its major functions, PSA causes liquefaction of semen and release of motile spermatozoa [1].

Several forms of PSA are found in the serum. Generally, 5% to 40% of the PSA is considered to be "free," and this portion may be inactive. From 60% to 95% of PSA may be complexed with alpha-1-antichymotrypsin (ACT). Prostate-specific antigen reacting with ACT is reportedly inactivated due to splitting of the reactive center [1]. Several studies have shown that the ratio of free to total PSA is significantly lower in patients with prostate cancer than in those with benign prostatic hypertrophy [2,3]. Based on this difference, several investigators have reported improvements in the specificity and sensitivity of PSA screening.

In order to avoid some of the confusion concerning the use and prognostic significance of a given PSA level, it is important to be aware of the differences between the two major types of assays (polyclonal vs monoclonal assays) used in the published literature. Several centers have reported comparisons of PSAs in patients undergoing both monoclonal and polyclonal assays [4,5]. Despite the fact that the normal range was 0 to 4.0 ng/mL for the monoclonal assay and 0 to 2.5 ng/mL for the polyclonal assay, the ratio observed was the opposite of what would have been expected based on these recommended "normal" ranges. This suggests that the manufacturers chose different sensitivities and specificities when their "normal" ranges were defined. A given PSA level determined by the monoclonal assay was approximately 1.5 to 1.85 times that when determined by the polyclonal assay. Furthermore, for most monoclonal assays, the "best" upper limit of normal is probably 3.3 ng/mL rather than 4.0 ng/ml [6]. Alternatively, some authorities have recommended an age-adjusted scale [7].

Staging With PSA

Prostate-specific antigen has been shown to be of value in the staging work-up and evaluation of patients with prostate cancer. Use of this serum marker can save millions of dollars by obviating the need for expensive, time-consuming studies. For example, bone scans are probably not justified if the PSA is 10 ng/mL and may not be warranted if the PSA is less than 50 ng/mL and serum alkaline phosphatase is normal [8,9]. In the assessment of lymph node status, neither abdominal/pelvic CT nor MRI appear to be warranted if the PSA is less than 20 ng/mL [10].

Problems With Use of PSA Alone

Numerous studies have demonstrated that there is a correlation between the level of PSA and the volume of both the normal prostate and prostatic carcinoma. Per gram of tissue, PSA levels are much higher for cancerous than for normal tissue. This may well be due to greater leakage of PSA into the circulation or disruption of normal prostate parenchyma associated with cancer, because higher levels and mRNA are expressed in benign than in malignant tissue [1]. As the tumor grade increases (Gleason score more than 7), there is a tendency toward a reduction in the amount of PSA produced per gram of tissue [11,12].

Anaplastic tumors may fail to produce an elevation of serum PSA or to stain positively for it. These phenomena complicate the use of PSA for accurately predicting tumor volume, because although higher-grade tumors tend to be larger, the volume tends to be underestimated for a given level of PSA. Because of this complex interaction between PSA and tumor grade, and the confounding effects of the presence of benign prostatic hypertrophy, PSA alone cannot be relied on to predict tumor volume.

PSA Density

Attempts to use the so-called PSA density to compensate for this problem have met with mixed results. Although many investigators agree that the use of PSA density may improve specificity and/or sensitivity for the purposes of screening, the impact on outcome following treatment is less clear [13-15]. In a recent report, M. D. Anderson investigators concluded that PSA density was "particularly useful" in subdividing patients with PSAs between 4 and 20 ng/mL. In contrast, investigators at Columbia concluded that use of PSA density did not add anything to their ability to predict biochemical failure. The authors admitted, however, that the small number of patients in their series may have limited the power of their analysis.

Free PSA and PSA Doubling Time

The application of free PSA to this problem may further improve our ability to predict tumor volume, but this possibility has only begun to be investigated. Also, some investigators have reported that the rate of rise of PSA, or PSA doubling time, is an independent prognostic factor for outcome following radiotherapy [16]. The optimal time interval for calculating this end point has not been well defined, however.

PSA, Tumor Grade, and Clinical Stage

Several studies have shown that combining PSA with tumor grade and clinical stage improves the prediction of pathologic stage. Partin et al published nomograms for defining the risk of extracapsular extension (ECE), seminal vesicle involvement (SV+), and lymph node involvement (N+) [11]. Based on these data, three simple equations have been described that help simplify the estimations of the risk of N+, SV+ and ECE [17-20]. These equations were verified initially for their predictive values using radical prostatectomy data from the University of California, San Francisco (UCSF), and have been since verified at other institutions as well [21,22]. The three equations are shown below:

Risk of N+ = 2/3 PSA + {(GS - 6) × 10}

Risk of SV+ = PSA + {(GS - 6) × 10}

Risk of ECE = 3/2 PSA + {(GS - 3) × 10}

In these equations, PSA is the highest value (not attributed to other causes, such as infection) prior to treatment using a monoclonal assay, which has a normal range of 0 to 4.0 ng/mL. When used in this equation, PSAs measured using a polyclonal assay (eg, Yang), with a normal range of ~ 0.0 to 2.5 ng/mL, may need to be divided by a conversion factor of approximately 1.5. For use in this equation, the nine possible Gleason score (GS) values are empirically grouped into five major scoring categories (ˆequal to or less than 4 = 4, 5 = 5, 6 = 6, 7 = 7, and equal to or more than 8 = 8), and the calculated risk values are constrained between 0% and 100%.

Table 1 compares the calculated risk of lymph node involvement using the above equation with the observed incidence based on data provided by Partin et al [11,17]. With the exception of the patients with high-grade tumors, there is excellent agreement between the calculated risk and observed incidence. In practice, patients at UCSF are arbitrarily considered to be at "high risk" for lymph node involvement if their calculated risk is 15%. Based on surgical data, patients at UCSF with a calculated risk of lymph node involvement of less than 15% have an observed incidence of ~ 6%, while patients with a calculated risk of 15% or more have an observed incidence of ~ 40%.

A shortcoming of the original "lymph node equation" is that it ignores clinical stage. Stage appears to independently affect the risk of lymph node involvement. To address this concern, the equation for predicting the risk of N+ was modified (data presented at the American College of Radiology, "Future Trends in Radiation Oncology," New Orleans, Louisianna, 1994) to include clinical stage information. The modified equation is shown below:

Risk of N+ = 2/3 PSA + {(GS - 6) + (TG - 1.5)} × 10

For this model, the T-groupings (TG) are as follows: TG 1 (stages T1c and T2a) is assigned a value of 1; TG 2 (T1b and T2b) is given a value of 2; and TG 3 (T2c and T3) is assigned a value of 3.

Using this equation, patients with clinical T2c-T3 lesions have a 15% higher risk of positive lymph nodes than was calculated with the previous equation and patients with clinical T1c or T2a lesions have a 5% lower risk. Figures 1A, 1B, and 1C are examples of how this model that includes T-stage more closely approximates the expected incidence of lymph node involvement, based on surgical data provided by Partin et al [11,23].

Unfortunately, to date there have been insufficient surgical data (too few T2c and T3 patients undergoing node dissections) to validate this equation. The recently opened Radiation Therapy Oncology Group (RTOG) 94-13 clinical trial may provide an opportunity to indirectly test the modified equation. Patients entered into the trial are being stratified by the T-groupings used in this equation (in addition to pretreatment PSA and Gleason score) and are being randomized to whole-pelvic irradiation or no whole pelvic radiotherapy. (The trial protocol is outlined in Table 2). If the patients with a higher calculated risk of lymph node involvement who do not receive whole-pelvic irradiation have a higher PSA failure rate than those receiving this therapy but an equal local control rate, this might support the value of the model (as well as the benefit of whole-pelvic irradiation).