PSA Kinetics and Risk of Death From Prostate Cancer: Title and subTitle BreakIn Search of the Holy Grail of Surrogate End Points

Adenocarcinoma of the prostate is the most commonly diagnosed malignancy in men in the United States. It is estimated that, in 2005, 232 000 new cases will be diagnosed and 30 350 men will die from this disease.¹ Even though prostate-specific antigen (PSA) screening has not been conclusively demonstrated to reduce mortality from prostate cancer, this practice has been widely adopted in the United States. As a result, more patients are being diagnosed at a younger age and with disease localized to the prostate.^{1 - 2} For these patients, radical prostatectomy (RP) and radiation therapy (RT) remain the mainstays of curative therapy.³ Nevertheless, prostate cancer remains primarily a disease of older men, many of whom have other serious coexisting illnesses from which they may be far more likely to die. Thus, it remains a major challenge for the clinician to identify which patients need aggressive treatment and which do not.

The long natural history of localized prostate cancer makes it a particularly difficult disease in which to study the impact of any therapy.⁴ Most newly diagnosed patients with prostate cancer would live for many years without treatment,⁵ those that are treated often live more than a decade even if they are not cured by their therapy,⁶ and the majority of patients die of something other than prostate cancer whether or not they are treated.^{5 ,7 - 8} Since improved survival has traditionally been the criterion standard against which cancer treatments are measured, it may take 15 to 20 years from a study’s inception to demonstrate a survival advantage, or lack thereof, in localized prostate cancer.⁹ Contrast this situation to that of glioblastoma or pancreatic cancer, diseases which, if left untreated, have survival time measured in weeks: with these cancers, death usually follows closely on the heels of a relapse, and few patients die of competing causes. Moreover, new treatments for these malignancies can be definitively studied in one tenth of the time required for most prostate cancer trials,¹⁰ and improvement in survival should be the end point measured. If major advances in the treatment of prostate cancer are to occur in a more timely manner, end points other than overall survival must be identified and validated. Since the introduction of PSA screening into clinical practice, the quest to identify the best surrogate end point has been a major focus of prostate cancer research.

In this issue of JAMA, 2 provocative retrospective studies involving 2 very different patient populations make major contributions toward the identification of surrogate end points. Freedland et al¹¹ studied 379 patients who developed an increasing PSA level after RP, whereas D’Amico et al¹² evaluated 358 newly diagnosed men treated with external beam RT for localized prostate cancer. The surgical group was much younger (59 vs 71 years old), was followed for a much longer period of time (10 vs 4 years), and had more adverse clinical and pathologic features. Despite these differences, the methods of data analysis revealed important similarities.

First, both studies use death from prostate cancer (prostate cancer–specific mortality), rather than overall survival, as their primary end point. In theory, prostate cancer–specific mortality is an attractive end point, since death from causes other than prostate cancer are censored rather than counted as deaths. In practice, however, it can be difficult to determine exactly what an individual patient died from, especially in a retrospective analysis. Simply determining whether a patient is dead or alive is much more straightforward. The investigators in both studies adopted a reasonable solution to this conundrum, given the era in which the patients were treated. The investigators attributed death as due to prostate cancer in patients with either a documented history of hormone-refractory metastatic disease with evidence of an increasing PSA level at last follow-up (D’Amico et al) or evidence of metastatic disease with progression following hormone therapy (Freedland et al).

Second, both authors assessed whether changes in PSA levels over time (PSA kinetics) significantly correlated with prostate cancer–specific mortality. D’Amico et al evaluated PSA velocity (>2.0 ng/mL per year) in the year prior to diagnosis, whereas Freedland et al studied PSA doubling time (PSADT) after recurrence. In both studies, larger changes in PSA level over time significantly correlated with a higher rate of prostate cancer–specific mortality. In the study by D’Amico et al, this finding was particularly striking for low-risk patients (ie, those with a PSA level <10.0 ng/mL, Gleason score <7, and clinical tumor category <T2b). Patients with low-risk features but with a PSA velocity greater than 2.0 ng/mL per year had outcomes similar to those of patients with higher-risk disease, who might usually be offered androgen deprivation plus RT rather than RT alone. A similar finding has been previously reported for patients treated with RP.¹³ Taken together, these data strongly suggest that a greater than 2.0-ng/mL increase in the PSA level in the year preceding diagnosis predicts a poorer outcome with RP or RT than otherwise might be suspected.

Freedland et al evaluated PSADT, as well as time to recurrence after surgery (>3 years vs ≤3 years) and Gleason score (8-10 vs ≤7) and found all 3 parameters to be significant independent predictors of prostate cancer–specific mortality. They used this information to construct a series of tables estimating 5-, 10-, and 15-year prostate cancer–specific mortality in patients with an increasing PSA level after RP. These data support the findings of others, which indicate that in the setting of primary treatment failure after either surgery or radiotherapy, PSADT is the clinical factor most consistently correlated with death from prostate cancer.^{6 ,14 - 17} The data in Table 3 of the study by Freedland et al may be helpful for clinicians faced with recommending therapy for a patient with increasing PSA levels after prostatectomy; however, the wide confidence intervals around some estimates may limit the value of these data for prognostic purposes. The study, unfortunately, does not shed light on the important question of how best to treat these patients. Other data in the literature exist to help guide the clinician in this situation.¹⁸

As with most good retrospective studies, these 2 reports raise more questions than they answer. Given recent randomized trials reporting significant improvement in survival with chemotherapy for patients with hormone-refractory prostate cancer,^{19 - 20} a number of important questions persist: Is the definition of prostate cancer–specific mortality used in these studies still appropriate? In addition, which is the best index of change-in-PSA-with-time to use? Does it depend on the clinical situation; ie, is PSA velocity more appropriate for newly diagnosed patients and PSADT more appropriate for patients recurring after primary therapy? What about patients who present with an abnormal result on their first-ever screening PSA measurement? Should the clinician wait 6 months and repeat the PSA measurement to calculate the PSA velocity? Also, should this information be used clinically at the present time, or are other nomograms better able to predict treatment outcome?²¹

Good retrospective analyses serve to generate hypotheses for future clinical trials. The hypothesis that the rate of change in PSA with time predicts prostate cancer–specific mortality must be validated prospectively. If proven to be correct, progress in prostate cancer research will proceed at a much faster pace, since conclusive answers to clinical questions will be available in 5 to 10 years, rather than once per generation.