- Email: [email protected]
Overall survival after prostate-specific-antigen-detected recurrence following conformal radiation therapy
Int. J. Radiation Oncology Biol. Phys., Vol. 48, No. 3, pp. 629 – 633, 2000 Copyright © 2000 Elsevier Science Inc. Printed in the USA. All rights reserved 0360-3016/00/$–see front matter
PII S0360-3016(00)00717-3
CLINICAL INVESTIGATION
Prostate
OVERALL SURVIVAL AFTER PROSTATE-SPECIFIC-ANTIGEN-DETECTED RECURRENCE FOLLOWING CONFORMAL RADIATION THERAPY HOWARD M. SANDLER, M.D.,* RODNEY L. DUNN, M.S.,‡ P. WILLIAM MCLAUGHLIN, M.D.,* JAMES A. HAYMAN, M.D., M.B.A.,* MOLLY A. SULLIVAN, M.D.,* AND JEREMY M. G. TAYLOR, PH.D.*†‡ *Department of Radiation Oncology, University of Michigan Medical School; †Department of Biostatistics, University of Michigan School of Public Health; and ‡The University of Michigan Comprehensive Cancer Center, Ann Arbor, MI Purpose: To study the significance, in terms of overall and cause-specific survival, of biochemical failure after conformal external-beam radiation therapy (RT) for prostate cancer. Methods and Materials: Of the 1844 patients in the Radiation Oncology prostate cancer database, 718 were deemed eligible. Patients excluded were those with N1 or M1 disease, those treated after radical prostatectomy, those who received hormone therapy before radiation therapy, and those who died, failed clinically, or had no PSA response in the first 6 months after RT. Patients included were required to have a minimum of 2 post-RT PSAs separated by at least 1 week. Biochemical relapse was defined as 3 consecutive PSA rises. This resulted in 154 patients with biochemical failure. Survival was calculated from the third PSA elevation. The rate of rise of PSA was calculated by fitting a regression line to the four rising PSAs on a ln PSA vs. time plot. Results: There were 41 deaths among the 154 patients with failure in 23 of the 41 due to prostate cancer. The overall survival after failure was 58% at 5 years, while the cause-specific failure was 73% at 5 years. Among the 154 failures, several factors were evaluated for an association with overall survival: age at failure, pre-RT PSA, PSA at second rise, PSA nadir, time from RT to failure, time to nadir, Gleason score, T-stage, and rate of rise, both from the nadir and from the beginning of the rise. None of these factors were significantly associated with an increased risk of death. As expected, the group of patients with biochemical failure have significantly worse prognostic factors than those without biochemical failure: median pre-RT PSA 15.9 vs. 9.0 (p < 0.001), and Gleason score of 7 or greater for 48% of subjects vs. 40% (p ⴝ 0.1). Relative PSA rise and slope of ln PSA vs. time were associated with cause-specific mortality (p < 0.001 and p ⴝ 0.007, respectively). Conclusion: Overall survival after conformal radiotherapy for prostate cancer remains high 5 years after biochemical failure. This high survival rate occurs even though the group of patients with biochemical failure has worse than average adverse preradiation prognostic factors. Thus, although biochemical failure can identify patients who have recurrent disease after RT, the ultimate relationship between this endpoint and death remains to be better defined. © 2000 Elsevier Science Inc. Prostate cancer, Conformal radiation therapy, PSA.
INTRODUCTION Prostate cancer is a condition with a long natural history. This disease is one of a very small list of human malignancies for which the option of no immediate treatment, also known as watchful waiting, is a frequently discussed option. Nevertheless, a large number of patients undergo definitive treatment for clinically localized prostate cancer, with radiation therapy (1) and surgery (2) being the two most frequently used modalities. Prostate-specific antigen (PSA) is a prostate-specific tumor marker which has altered significantly how newly diagnosed prostate cancer is detected (3) and has become a useful tool for detecting treatment failure following therapy for local-
ized disease (4). In general, PSA-detected failures predate clinical failures by several years. Options for treatment upon biochemical relapse after external beam radiotherapy are limited. These choices include androgen ablation [either continuous or intermittent (5)], salvage prostatectomy (6), or other localized treatment modalities such as cryotherapy (7) or salvage brachytherapy (8), or watchful waiting without immediate intervention. Salvage treatments after radiation therapy failure are not without morbidity and these regimens may need to be used judiciously, if many men die with, but not from, prostate cancer. To assess the risk of prostate cancer mortality after biochemical failure following external beam radiation therapy
Reprint requests to: Dr. Howard M. Sandler, Department of Radiation Oncology, University of Michigan Medical Center, 1500 East Medical Center Drive, Ann Arbor, MI 48109-0010. E-mail [email protected]
Presented at the 41st Annual Meeting of the American Society for Therapeutic Radiology and Oncology, San Antonio, Texas, 1999. Accepted for publication 22 May 2000. 629
630
I. J. Radiation Oncology
●
Biology
●
Physics
for localized disease, we reviewed our large experience of patients treated for localized prostate cancer using conformal radiotherapy techniques. We assessed the overall survival and cause-specific survival of patients who have biochemical relapse following external beam radiotherapy. MATERIALS AND METHODS The study group consisted of patients treated with conformal radiation therapy beginning January 1986 and who underwent treatment at the University of Michigan Medical Center and at the University of Michigan Radiotherapy Affiliate, Providence Hospital, Southfield, Michigan. Patients were excluded from this analysis if they had pathologically-proven nodal metastases, underwent hormonal therapy prior to radiation therapy including neoadjuvant hormone therapy, or were treated in the postoperative/postprostatectomy setting. In addition, those who died, failed clinically, or had no reduction in PSA in the first 6 months after radiation therapy were excluded. The study group was intended to evaluate the risk of dying in a group of patients who respond to radiotherapy, but then experience biochemical relapse during follow-up. All patients were treated with conformal external beam radiotherapy, as previously described (9). Biochemical relapse was defined as 3 consecutive PSA rises, based on the ASTRO definition of failure (10). Survival, for the purposes of this study, was calculated from the third PSA rise rather than the ASTRO-defined date of PSA failure, i.e., midpoint between the last nonrising PSA and the first rise. The rationale for calculating the survival interval from the 3rd PSA rise until death was to assist in providing clinical data to patients, who would be interested in the probability of survival from the date that biochemical failure is confirmed, i.e., the 3rd PSA rise. Generally, patients were monitored with history/physical examination, and PSA determinations every 3– 6 months. More specifically, PSA was assessed with median frequency of 3.9 months with an interquartile range of 1.6 – 6.2 months. There were 3 patients who died and for whom no cause of death could be determined. All 3 of these patients were presumed to have died from prostate cancer. It was elected not to use distant metastasis as an endpoint, given that many patients were treated with androgen ablation at some point during their period of PSA-determined relapse, but prior to any clinical evidence for recurrent disease. Thus, patients who develop distant metastasis often did it in the context of hormone-refractory disease. All patients who died from prostate cancer had androgen ablation prior to their death. Various factors were assessed for their prognostic importance upon predicting survival and cause-specific survival. Variables examined included age at the time of failure, preradiation therapy PSA, the time interval between radiation therapy and failure, the PSA nadir, Gleason score, T-stage, the rate of lnPSA rise, and the relative PSA rise. The rate of PSA rise was determined by fitting a regression line to the 4 PSA values—the baseline PSA0 (the last
Volume 48, Number 3, 2000
non-rising PSA), and the 3 rising values, PSA1, PSA2, PSA3— on a lnPSA vs. time plot. The relative rise of PSA is defined as the natural log of the 3rd elevated PSA above baseline divided by the baseline PSA, ln(PSA3/PSA0). Rates of overall survival and cause-specific survival were assessed using the Kaplan–Meier method. Cox regression analysis was performed using relevant significant variables. RESULTS Seven hundred eighteen patients were eligible for analysis. Of this group, 154 developed 3 consecutive PSA rises and represent the group of patients who were evaluated for survival and cause-specific survival. Of note, the group of patients with failure represented a group with more-advanced disease than those who did not fail. Median pre-RT PSA for those with failure was 15.9 and median pre-RT PSA for those who were eligible but did not fail (n ⫽ 564) was 9.0 (p ⬍ 0.001, Wilcoxon rank–sum test). In addition, the 49% of those who failed had a Gleason score of 7 or greater while 40% of those without failure had a Gleason score of 7 or more (p ⫽ 0.099, Fisher’s exact test). The median age of the 154 patients at the time of biochemical failure was 74.5 years, and 10% were African-American; 16% had T1 tumors, and 67% had T2 tumors. There were a total of 41 deaths among the group of 154 patients with biochemical failure. Of these events, 23 were considered cause-specific and occurred as a direct result of advanced prostate cancer, and 18 were unrelated to prostate cancer, although biochemical failure had occurred. The overall survival of the entire group of 154 biochemical failure patients was 58% at 5 years, with a median of 5.9 years. This data are relatively mature at the 5-year time point, with 26 patients followed above and beyond 5 years, ensuring the reliability of the data as long as 5 years after biochemical failure. The median follow-up of the surviving patients is 3.1 years. The cause-specific survival at 5 years is 73% and the median survival of this group is 7.4 years (Figs. 1a and 1b). When overall survival was examined in relation to the prognostic factors listed above, which may have been expected to influence the outcome, no significant associations were detected. Similarly for cause-specific survival, a relationship was not observed between death due to prostate cancer and the pretreatment clinical variables. These analyses are summarized in Table 1. However, when the risk of cause-specific mortality was modeled using the pattern of the rising PSA profile, a statistical relationship was detected between cause-specific death and two constructs: the slope of the ln PSA vs. time plot and the relative PSA. When each of these variables was modeled independently, the relative PSA variable was a stronger predictor of cause-specific death (Figs. 2a and 2b). DISCUSSION Among this population of prostate cancer patients who were treated with conformal external beam treatment and
Overall survival after PSA-detected recurrence
●
H. M. SANDLER et al.
631
specific survival is 73% at 5 years and these data can provide radiotherapy patients with useful information for assessing management options if and when biochemical relapse occurs. Options for patients with biochemical relapse after radiotherapy vary in intensity from watchful waiting to salvage radical prostatectomy and it remains unclear what strategy should be employed for most patients. It is clear that patients who do relapse after RT represent a subset of patients who are worse than the average irradiated patient and many patients may have a component of metastatic disease, limiting the potential benefit of aggressive local salvage options. This analysis found that both the slope of the rising PSA profile, which corresponds to the PSA doubling time, and the relative PSA rise were statistically associated with an increased risk of cause-specific death. These variables are similar (correlation coefficient ⫽ 0.63), but can be distinguished. The slope of the rising lnPSA vs. time plot is a reflection of exponential growth and, when the relationship between lnPSA and time is linear, it indicates that the growth rate is constant. Thus, the slope of this line can be calculated by collecting PSA data over any time period. If the three rising PSA values are collected over a relatively short period of time, say a total of 3– 6 months, the survival time from the 3rd elevated PSA would be longer than if the data were collected over a longer period of time, say 3 years. So, one potential disadvantage of predicting survival after recurrence based upon PSA doubling time is that the survival from confirmation of failure depends on the frequency of PSA data collection. The relative PSA rise, ln(PSA3/PSA0) does not depend on the collection frequency of PSA, rather it is only dependent upon the magnitude of the PSA value that confirms 3 consecutive rises, i.e., PSA3. So, a patient with a rising PSA profile whose 3rd consecutive PSA elevation is, for example, 10.0 ng/mL would have a worse prognosis than someone whose 3rd PSA is only 5.0 ng/mL, regardless of the time interval
Fig. 1. Overall survival (a) and cause-specific survival (b) of all patients with biochemical failure from the date of the 3rd prostatespecific antigen (PSA) rise.
subsequently developed biochemical recurrence, both overall survival and cause-specific survival remain high. Cause-
Table 1. Results of univariate analysis (Cox regression model) Overall survival Measure
Relative risk*
Baseline PSA Age at 3rd Rise Value of 3rd Rise Value of Nadir Time from baseline to 3rd rise Time from baseline to nadir Time from nadir to 3rd rise Gleason Sum (6⫺ vs 7⫹) T-stage (1,2 vs. 3,4) Slope of Rise on Log Scale Log of Relative Rise
0.890 0.980 1.020 0.999 0.828 0.749 0.957 0.907 0.885 1.358 1.849
95% CI* 0.686 0.701 0.977 0.934 0.536 0.466 0.615 0.483 0.406 1.103 1.315
1.154 1.372 1.064 1.070 1.279 1.203 1.489 1.702 1.925 1.672 2.598
Cause-specific survival p value
Relative risk*
0.378 0.909 0.369 0.986 0.396 0.232 0.846 0.760 0.757 0.004 ⬍0.001
0.990 1.042 1.040 1.021 0.605 0.518 0.782 1.557 0.453 1.693 2.491
95% CI* 0.741 0.675 0.997 0.951 0.311 0.239 0.397 0.612 0.183 1.331 1.640
1.321 1.610 1.086 1.097 1.175 1.120 1.538 3.961 1.123 2.153 3.784
p value 0.943 0.851 0.069 0.568 0.138 0.095 0.475 0.353 0.087 ⬍0.001 ⬍0.001
* Relative risk and confidence intervals are for change from 25th to 75th percentile for continuous covariates and for a 1 unit change in categorical covariates.
632
I. J. Radiation Oncology
●
Biology
●
Physics
Fig. 2. a. Cause-specific survival of patients based upon the relative rise of prostate-specific antigen (PSA) at the time of failure. The median ln relative rise is 1.07, corresponding to a ratio of PSA3/PSA0 of 2.9 (p ⬍ 0.001). b. Cause-specific survival of patients based upon the slope of the ln PSA vs. time relationship. The median slope was 0.0019, which corresponds to a PSA doubling time of 1.05 years or 12.7 months (p ⫽ 0.007).
between the last nonrising PSA, and PSA3. Clearly, both the rate of rise of PSA and the magnitude of the PSA when failure is confirmed will be associated with cause-specific death; in this series, the latter is more strongly associated with mortality than the former. Others have examined the history of patients with biochemical relapse following radiotherapy. Pollack et al. (11) from M. D. Anderson Hospital calculated PSA doubling times (PSADT) in 100 patients with a rising PSA profile in the PSA era. Of this group, 44 patients had documented distant disease relapse. The median PSADT was 13.5 months and a short doubling time was correlated with both
Volume 48, Number 3, 2000
local and distant relapse. Given the relatively short followup, the authors were unable to assess the influence of PSADT on overall survival or cause-specific survival, but they were able to estimate the interval between progression to clinical disease relapse after biochemical relapse and found this interval to be approximately 40 months. These data were updated by Zagars and Pollack (12) who found that doubling times of 8 months or less were associated with a 7-year risk of distant metastasis of 54% and doubling times of longer than 8 months had a low 7-year risk of distant metastasis of only 7%. In addition, the authors found that the doubling time was related to the location of failure. That is long doubling times seemed associated with local recurrence and short doubling times with distant metastasis. This observation was also made by Sartor et al. (13) Lee and others (14) studied rising PSA profiles following radiation therapy. They analyzed the PSA pattern among 151 men with a rising PSA after definitive radiotherapy and, with shorter median follow-up period, nevertheless, detected similar 5-year overall and cause-specific survival rates to the analysis currently presented. Lee et al.’s observed 5-year overall and cause-specific survival was 65% and 76%, respectively. In addition, Lee et al. noted a median PSA doubling time of 13 months, which compares closely to the value of 12.7 months observed by us (Fig. 2b). Recently, the natural history of PSA elevation after radical prostatectomy has been described in a large series of patients from The Johns Hopkins Medical Institutions (15). This experience is unusual in part because it truly represents a natural history experience. That is, patients were followed without intervention when their PSA began to rise following surgery, and no androgen ablation therapy was instituted until metastatic disease was discovered. In this series, the median time until distant metastases were discovered was 8 years from the time of PSA elevation. Time to distant progression was influenced by the interval between surgery and PSA elevation, the Gleason score and the PSA doubling time. Although this experience is quite different from the experience of patients who undergo radiation therapy, in part because the baseline characteristics of this select set of surgical patients is quite different, the authors note that there is a tremendous heterogeneity in the time-to-distant failure following PSA recurrence, which was also observed in the current series. In summary, the time between PSA relapse and causespecific death can be quite prolonged. In this series, the median time between relapse and cause-specific death was greater than 7 years. Given the long natural history of prostate cancer after relapse, it seems appropriate to be cautious when considering potentially toxic salvage treatments, especially in those with a longer PSA doubling time (flatter slope of ln PSA) or with a smaller relative rise. Conversely, those with more aggressive patterns of recurrence might be candidates for more intensive salvage treatments. In particular, because many will have subclinical distant disease, therapies that address the presence of distant disease, such as chemotherapy or androgen ablation or combinations of the two, might be attractive.
Overall survival after PSA-detected recurrence
●
H. M. SANDLER et al.
633
REFERENCES 1. Shipley WU, Thames HD, Sandler HM, et al. Radiation therapy for clinically localized prostate cancer: a multi-institutional pooled analysis. JAMA 1999;281:1598 –1604. 2. Walsh PC. Anatomic radical prostatectomy: Evolution of the surgical technique. J Urol 1998;160:2418 –2424. 3. Partin AW, Stutzman RE. Elevated prostate-specific antigen, abnormal prostate evaluation on digital rectal examination, and transrectal ultrasound and prostate biopsy. Urol Clin North Am 1998;25:581–589. 4. Hanlon AL, Moore DF, Hanks GE. Modeling postradiation prostate specific antigen level kinetics: Predictors of rising postnadir slope suggest cure in men who remain biochemically free of prostate carcinoma. Cancer 1998;83:130 –134. 5. Gleave M, Bruchovsky N, Goldenberg SL, et al. Intermittent androgen suppression for prostate cancer: rationale and clinical experience. Eur Urol 1998;34(Suppl. 3):37– 41. 6. Tefilli MV, Gheiler EL, Tiguert R, et al. Salvage surgery or salvage radiotherapy for locally recurrent prostate cancer. Urology 1998;52:224 –229. 7. Perrotte P, Litwin MS, McGuire EJ, et al. Quality of life after salvage cryotherapy: The impact of treatment parameters. J Urol 1999;162:398 – 402. 8. Grado GL, Collins JM, Kriegshauser JS, et al. Salvage brachytherapy for localized prostate cancer after radiotherapy failure. Urology 1999;53:2–10.
9. Fukunaga-Johnson N, Sandler HM, McLaughlin PW, et al. Results of 3D conformal radiotherapy in the treatment of localized prostate cancer. Int J Radiat Oncol Biol Phys 1997; 38:311–317. 10. Anonymous. Consensus statement: Guidelines for PSA following radiation therapy. Int J Radiat Oncol Biol Phys 1997; 37:1035–1041. 11. Pollack A, Zagars GK, Kavadi VS. Prostate specific antigen doubling time and disease relapse after radiotherapy for prostate cancer. Cancer 1994;74:670 – 678. 12. Zagars GK, Pollack A. Kinetics of serum prostate-specific antigen after external beam radiation for clinically localized prostate cancer. Radiother Oncol 1997;44:213–221. 13. Sartor CI, Strawderman MH, Lin XH, et al. Rate of PSA rise predicts metastatic versus local recurrence after definitive radiotherapy. Int J Radiat Oncol Biol Phys 1997;38: 941–947. 14. Lee WR, Hanks GE, Hanlon A. Increasing prostate-specific antigen profile following definitive radiation therapy for localized prostate cancer: Clinical observations. J Clin Oncol 1997;15:230 –238. 15. Pound CR, Partin AW, Eisenberger MA, et al. Natural history of progression after PSA elevation following radical prostatectomy. JAMA 1999;281:1591–1597.
PII S0360-3016(00)00717-3
CLINICAL INVESTIGATION
Prostate
OVERALL SURVIVAL AFTER PROSTATE-SPECIFIC-ANTIGEN-DETECTED RECURRENCE FOLLOWING CONFORMAL RADIATION THERAPY HOWARD M. SANDLER, M.D.,* RODNEY L. DUNN, M.S.,‡ P. WILLIAM MCLAUGHLIN, M.D.,* JAMES A. HAYMAN, M.D., M.B.A.,* MOLLY A. SULLIVAN, M.D.,* AND JEREMY M. G. TAYLOR, PH.D.*†‡ *Department of Radiation Oncology, University of Michigan Medical School; †Department of Biostatistics, University of Michigan School of Public Health; and ‡The University of Michigan Comprehensive Cancer Center, Ann Arbor, MI Purpose: To study the significance, in terms of overall and cause-specific survival, of biochemical failure after conformal external-beam radiation therapy (RT) for prostate cancer. Methods and Materials: Of the 1844 patients in the Radiation Oncology prostate cancer database, 718 were deemed eligible. Patients excluded were those with N1 or M1 disease, those treated after radical prostatectomy, those who received hormone therapy before radiation therapy, and those who died, failed clinically, or had no PSA response in the first 6 months after RT. Patients included were required to have a minimum of 2 post-RT PSAs separated by at least 1 week. Biochemical relapse was defined as 3 consecutive PSA rises. This resulted in 154 patients with biochemical failure. Survival was calculated from the third PSA elevation. The rate of rise of PSA was calculated by fitting a regression line to the four rising PSAs on a ln PSA vs. time plot. Results: There were 41 deaths among the 154 patients with failure in 23 of the 41 due to prostate cancer. The overall survival after failure was 58% at 5 years, while the cause-specific failure was 73% at 5 years. Among the 154 failures, several factors were evaluated for an association with overall survival: age at failure, pre-RT PSA, PSA at second rise, PSA nadir, time from RT to failure, time to nadir, Gleason score, T-stage, and rate of rise, both from the nadir and from the beginning of the rise. None of these factors were significantly associated with an increased risk of death. As expected, the group of patients with biochemical failure have significantly worse prognostic factors than those without biochemical failure: median pre-RT PSA 15.9 vs. 9.0 (p < 0.001), and Gleason score of 7 or greater for 48% of subjects vs. 40% (p ⴝ 0.1). Relative PSA rise and slope of ln PSA vs. time were associated with cause-specific mortality (p < 0.001 and p ⴝ 0.007, respectively). Conclusion: Overall survival after conformal radiotherapy for prostate cancer remains high 5 years after biochemical failure. This high survival rate occurs even though the group of patients with biochemical failure has worse than average adverse preradiation prognostic factors. Thus, although biochemical failure can identify patients who have recurrent disease after RT, the ultimate relationship between this endpoint and death remains to be better defined. © 2000 Elsevier Science Inc. Prostate cancer, Conformal radiation therapy, PSA.
INTRODUCTION Prostate cancer is a condition with a long natural history. This disease is one of a very small list of human malignancies for which the option of no immediate treatment, also known as watchful waiting, is a frequently discussed option. Nevertheless, a large number of patients undergo definitive treatment for clinically localized prostate cancer, with radiation therapy (1) and surgery (2) being the two most frequently used modalities. Prostate-specific antigen (PSA) is a prostate-specific tumor marker which has altered significantly how newly diagnosed prostate cancer is detected (3) and has become a useful tool for detecting treatment failure following therapy for local-
ized disease (4). In general, PSA-detected failures predate clinical failures by several years. Options for treatment upon biochemical relapse after external beam radiotherapy are limited. These choices include androgen ablation [either continuous or intermittent (5)], salvage prostatectomy (6), or other localized treatment modalities such as cryotherapy (7) or salvage brachytherapy (8), or watchful waiting without immediate intervention. Salvage treatments after radiation therapy failure are not without morbidity and these regimens may need to be used judiciously, if many men die with, but not from, prostate cancer. To assess the risk of prostate cancer mortality after biochemical failure following external beam radiation therapy
Reprint requests to: Dr. Howard M. Sandler, Department of Radiation Oncology, University of Michigan Medical Center, 1500 East Medical Center Drive, Ann Arbor, MI 48109-0010. E-mail [email protected]
Presented at the 41st Annual Meeting of the American Society for Therapeutic Radiology and Oncology, San Antonio, Texas, 1999. Accepted for publication 22 May 2000. 629
630
I. J. Radiation Oncology
●
Biology
●
Physics
for localized disease, we reviewed our large experience of patients treated for localized prostate cancer using conformal radiotherapy techniques. We assessed the overall survival and cause-specific survival of patients who have biochemical relapse following external beam radiotherapy. MATERIALS AND METHODS The study group consisted of patients treated with conformal radiation therapy beginning January 1986 and who underwent treatment at the University of Michigan Medical Center and at the University of Michigan Radiotherapy Affiliate, Providence Hospital, Southfield, Michigan. Patients were excluded from this analysis if they had pathologically-proven nodal metastases, underwent hormonal therapy prior to radiation therapy including neoadjuvant hormone therapy, or were treated in the postoperative/postprostatectomy setting. In addition, those who died, failed clinically, or had no reduction in PSA in the first 6 months after radiation therapy were excluded. The study group was intended to evaluate the risk of dying in a group of patients who respond to radiotherapy, but then experience biochemical relapse during follow-up. All patients were treated with conformal external beam radiotherapy, as previously described (9). Biochemical relapse was defined as 3 consecutive PSA rises, based on the ASTRO definition of failure (10). Survival, for the purposes of this study, was calculated from the third PSA rise rather than the ASTRO-defined date of PSA failure, i.e., midpoint between the last nonrising PSA and the first rise. The rationale for calculating the survival interval from the 3rd PSA rise until death was to assist in providing clinical data to patients, who would be interested in the probability of survival from the date that biochemical failure is confirmed, i.e., the 3rd PSA rise. Generally, patients were monitored with history/physical examination, and PSA determinations every 3– 6 months. More specifically, PSA was assessed with median frequency of 3.9 months with an interquartile range of 1.6 – 6.2 months. There were 3 patients who died and for whom no cause of death could be determined. All 3 of these patients were presumed to have died from prostate cancer. It was elected not to use distant metastasis as an endpoint, given that many patients were treated with androgen ablation at some point during their period of PSA-determined relapse, but prior to any clinical evidence for recurrent disease. Thus, patients who develop distant metastasis often did it in the context of hormone-refractory disease. All patients who died from prostate cancer had androgen ablation prior to their death. Various factors were assessed for their prognostic importance upon predicting survival and cause-specific survival. Variables examined included age at the time of failure, preradiation therapy PSA, the time interval between radiation therapy and failure, the PSA nadir, Gleason score, T-stage, the rate of lnPSA rise, and the relative PSA rise. The rate of PSA rise was determined by fitting a regression line to the 4 PSA values—the baseline PSA0 (the last
Volume 48, Number 3, 2000
non-rising PSA), and the 3 rising values, PSA1, PSA2, PSA3— on a lnPSA vs. time plot. The relative rise of PSA is defined as the natural log of the 3rd elevated PSA above baseline divided by the baseline PSA, ln(PSA3/PSA0). Rates of overall survival and cause-specific survival were assessed using the Kaplan–Meier method. Cox regression analysis was performed using relevant significant variables. RESULTS Seven hundred eighteen patients were eligible for analysis. Of this group, 154 developed 3 consecutive PSA rises and represent the group of patients who were evaluated for survival and cause-specific survival. Of note, the group of patients with failure represented a group with more-advanced disease than those who did not fail. Median pre-RT PSA for those with failure was 15.9 and median pre-RT PSA for those who were eligible but did not fail (n ⫽ 564) was 9.0 (p ⬍ 0.001, Wilcoxon rank–sum test). In addition, the 49% of those who failed had a Gleason score of 7 or greater while 40% of those without failure had a Gleason score of 7 or more (p ⫽ 0.099, Fisher’s exact test). The median age of the 154 patients at the time of biochemical failure was 74.5 years, and 10% were African-American; 16% had T1 tumors, and 67% had T2 tumors. There were a total of 41 deaths among the group of 154 patients with biochemical failure. Of these events, 23 were considered cause-specific and occurred as a direct result of advanced prostate cancer, and 18 were unrelated to prostate cancer, although biochemical failure had occurred. The overall survival of the entire group of 154 biochemical failure patients was 58% at 5 years, with a median of 5.9 years. This data are relatively mature at the 5-year time point, with 26 patients followed above and beyond 5 years, ensuring the reliability of the data as long as 5 years after biochemical failure. The median follow-up of the surviving patients is 3.1 years. The cause-specific survival at 5 years is 73% and the median survival of this group is 7.4 years (Figs. 1a and 1b). When overall survival was examined in relation to the prognostic factors listed above, which may have been expected to influence the outcome, no significant associations were detected. Similarly for cause-specific survival, a relationship was not observed between death due to prostate cancer and the pretreatment clinical variables. These analyses are summarized in Table 1. However, when the risk of cause-specific mortality was modeled using the pattern of the rising PSA profile, a statistical relationship was detected between cause-specific death and two constructs: the slope of the ln PSA vs. time plot and the relative PSA. When each of these variables was modeled independently, the relative PSA variable was a stronger predictor of cause-specific death (Figs. 2a and 2b). DISCUSSION Among this population of prostate cancer patients who were treated with conformal external beam treatment and
Overall survival after PSA-detected recurrence
●
H. M. SANDLER et al.
631
specific survival is 73% at 5 years and these data can provide radiotherapy patients with useful information for assessing management options if and when biochemical relapse occurs. Options for patients with biochemical relapse after radiotherapy vary in intensity from watchful waiting to salvage radical prostatectomy and it remains unclear what strategy should be employed for most patients. It is clear that patients who do relapse after RT represent a subset of patients who are worse than the average irradiated patient and many patients may have a component of metastatic disease, limiting the potential benefit of aggressive local salvage options. This analysis found that both the slope of the rising PSA profile, which corresponds to the PSA doubling time, and the relative PSA rise were statistically associated with an increased risk of cause-specific death. These variables are similar (correlation coefficient ⫽ 0.63), but can be distinguished. The slope of the rising lnPSA vs. time plot is a reflection of exponential growth and, when the relationship between lnPSA and time is linear, it indicates that the growth rate is constant. Thus, the slope of this line can be calculated by collecting PSA data over any time period. If the three rising PSA values are collected over a relatively short period of time, say a total of 3– 6 months, the survival time from the 3rd elevated PSA would be longer than if the data were collected over a longer period of time, say 3 years. So, one potential disadvantage of predicting survival after recurrence based upon PSA doubling time is that the survival from confirmation of failure depends on the frequency of PSA data collection. The relative PSA rise, ln(PSA3/PSA0) does not depend on the collection frequency of PSA, rather it is only dependent upon the magnitude of the PSA value that confirms 3 consecutive rises, i.e., PSA3. So, a patient with a rising PSA profile whose 3rd consecutive PSA elevation is, for example, 10.0 ng/mL would have a worse prognosis than someone whose 3rd PSA is only 5.0 ng/mL, regardless of the time interval
Fig. 1. Overall survival (a) and cause-specific survival (b) of all patients with biochemical failure from the date of the 3rd prostatespecific antigen (PSA) rise.
subsequently developed biochemical recurrence, both overall survival and cause-specific survival remain high. Cause-
Table 1. Results of univariate analysis (Cox regression model) Overall survival Measure
Relative risk*
Baseline PSA Age at 3rd Rise Value of 3rd Rise Value of Nadir Time from baseline to 3rd rise Time from baseline to nadir Time from nadir to 3rd rise Gleason Sum (6⫺ vs 7⫹) T-stage (1,2 vs. 3,4) Slope of Rise on Log Scale Log of Relative Rise
0.890 0.980 1.020 0.999 0.828 0.749 0.957 0.907 0.885 1.358 1.849
95% CI* 0.686 0.701 0.977 0.934 0.536 0.466 0.615 0.483 0.406 1.103 1.315
1.154 1.372 1.064 1.070 1.279 1.203 1.489 1.702 1.925 1.672 2.598
Cause-specific survival p value
Relative risk*
0.378 0.909 0.369 0.986 0.396 0.232 0.846 0.760 0.757 0.004 ⬍0.001
0.990 1.042 1.040 1.021 0.605 0.518 0.782 1.557 0.453 1.693 2.491
95% CI* 0.741 0.675 0.997 0.951 0.311 0.239 0.397 0.612 0.183 1.331 1.640
1.321 1.610 1.086 1.097 1.175 1.120 1.538 3.961 1.123 2.153 3.784
p value 0.943 0.851 0.069 0.568 0.138 0.095 0.475 0.353 0.087 ⬍0.001 ⬍0.001
* Relative risk and confidence intervals are for change from 25th to 75th percentile for continuous covariates and for a 1 unit change in categorical covariates.
632
I. J. Radiation Oncology
●
Biology
●
Physics
Fig. 2. a. Cause-specific survival of patients based upon the relative rise of prostate-specific antigen (PSA) at the time of failure. The median ln relative rise is 1.07, corresponding to a ratio of PSA3/PSA0 of 2.9 (p ⬍ 0.001). b. Cause-specific survival of patients based upon the slope of the ln PSA vs. time relationship. The median slope was 0.0019, which corresponds to a PSA doubling time of 1.05 years or 12.7 months (p ⫽ 0.007).
between the last nonrising PSA, and PSA3. Clearly, both the rate of rise of PSA and the magnitude of the PSA when failure is confirmed will be associated with cause-specific death; in this series, the latter is more strongly associated with mortality than the former. Others have examined the history of patients with biochemical relapse following radiotherapy. Pollack et al. (11) from M. D. Anderson Hospital calculated PSA doubling times (PSADT) in 100 patients with a rising PSA profile in the PSA era. Of this group, 44 patients had documented distant disease relapse. The median PSADT was 13.5 months and a short doubling time was correlated with both
Volume 48, Number 3, 2000
local and distant relapse. Given the relatively short followup, the authors were unable to assess the influence of PSADT on overall survival or cause-specific survival, but they were able to estimate the interval between progression to clinical disease relapse after biochemical relapse and found this interval to be approximately 40 months. These data were updated by Zagars and Pollack (12) who found that doubling times of 8 months or less were associated with a 7-year risk of distant metastasis of 54% and doubling times of longer than 8 months had a low 7-year risk of distant metastasis of only 7%. In addition, the authors found that the doubling time was related to the location of failure. That is long doubling times seemed associated with local recurrence and short doubling times with distant metastasis. This observation was also made by Sartor et al. (13) Lee and others (14) studied rising PSA profiles following radiation therapy. They analyzed the PSA pattern among 151 men with a rising PSA after definitive radiotherapy and, with shorter median follow-up period, nevertheless, detected similar 5-year overall and cause-specific survival rates to the analysis currently presented. Lee et al.’s observed 5-year overall and cause-specific survival was 65% and 76%, respectively. In addition, Lee et al. noted a median PSA doubling time of 13 months, which compares closely to the value of 12.7 months observed by us (Fig. 2b). Recently, the natural history of PSA elevation after radical prostatectomy has been described in a large series of patients from The Johns Hopkins Medical Institutions (15). This experience is unusual in part because it truly represents a natural history experience. That is, patients were followed without intervention when their PSA began to rise following surgery, and no androgen ablation therapy was instituted until metastatic disease was discovered. In this series, the median time until distant metastases were discovered was 8 years from the time of PSA elevation. Time to distant progression was influenced by the interval between surgery and PSA elevation, the Gleason score and the PSA doubling time. Although this experience is quite different from the experience of patients who undergo radiation therapy, in part because the baseline characteristics of this select set of surgical patients is quite different, the authors note that there is a tremendous heterogeneity in the time-to-distant failure following PSA recurrence, which was also observed in the current series. In summary, the time between PSA relapse and causespecific death can be quite prolonged. In this series, the median time between relapse and cause-specific death was greater than 7 years. Given the long natural history of prostate cancer after relapse, it seems appropriate to be cautious when considering potentially toxic salvage treatments, especially in those with a longer PSA doubling time (flatter slope of ln PSA) or with a smaller relative rise. Conversely, those with more aggressive patterns of recurrence might be candidates for more intensive salvage treatments. In particular, because many will have subclinical distant disease, therapies that address the presence of distant disease, such as chemotherapy or androgen ablation or combinations of the two, might be attractive.
Overall survival after PSA-detected recurrence
●
H. M. SANDLER et al.
633
REFERENCES 1. Shipley WU, Thames HD, Sandler HM, et al. Radiation therapy for clinically localized prostate cancer: a multi-institutional pooled analysis. JAMA 1999;281:1598 –1604. 2. Walsh PC. Anatomic radical prostatectomy: Evolution of the surgical technique. J Urol 1998;160:2418 –2424. 3. Partin AW, Stutzman RE. Elevated prostate-specific antigen, abnormal prostate evaluation on digital rectal examination, and transrectal ultrasound and prostate biopsy. Urol Clin North Am 1998;25:581–589. 4. Hanlon AL, Moore DF, Hanks GE. Modeling postradiation prostate specific antigen level kinetics: Predictors of rising postnadir slope suggest cure in men who remain biochemically free of prostate carcinoma. Cancer 1998;83:130 –134. 5. Gleave M, Bruchovsky N, Goldenberg SL, et al. Intermittent androgen suppression for prostate cancer: rationale and clinical experience. Eur Urol 1998;34(Suppl. 3):37– 41. 6. Tefilli MV, Gheiler EL, Tiguert R, et al. Salvage surgery or salvage radiotherapy for locally recurrent prostate cancer. Urology 1998;52:224 –229. 7. Perrotte P, Litwin MS, McGuire EJ, et al. Quality of life after salvage cryotherapy: The impact of treatment parameters. J Urol 1999;162:398 – 402. 8. Grado GL, Collins JM, Kriegshauser JS, et al. Salvage brachytherapy for localized prostate cancer after radiotherapy failure. Urology 1999;53:2–10.
9. Fukunaga-Johnson N, Sandler HM, McLaughlin PW, et al. Results of 3D conformal radiotherapy in the treatment of localized prostate cancer. Int J Radiat Oncol Biol Phys 1997; 38:311–317. 10. Anonymous. Consensus statement: Guidelines for PSA following radiation therapy. Int J Radiat Oncol Biol Phys 1997; 37:1035–1041. 11. Pollack A, Zagars GK, Kavadi VS. Prostate specific antigen doubling time and disease relapse after radiotherapy for prostate cancer. Cancer 1994;74:670 – 678. 12. Zagars GK, Pollack A. Kinetics of serum prostate-specific antigen after external beam radiation for clinically localized prostate cancer. Radiother Oncol 1997;44:213–221. 13. Sartor CI, Strawderman MH, Lin XH, et al. Rate of PSA rise predicts metastatic versus local recurrence after definitive radiotherapy. Int J Radiat Oncol Biol Phys 1997;38: 941–947. 14. Lee WR, Hanks GE, Hanlon A. Increasing prostate-specific antigen profile following definitive radiation therapy for localized prostate cancer: Clinical observations. J Clin Oncol 1997;15:230 –238. 15. Pound CR, Partin AW, Eisenberger MA, et al. Natural history of progression after PSA elevation following radical prostatectomy. JAMA 1999;281:1591–1597.