F.Y.F. and D.A.H. contributed equally to this article.
Interval to biochemical failure as a biomarker for cause-specific and overall survival after dose-escalated external beam radiation therapy for prostate cancer†
Version of Record online: 26 AUG 2011
Copyright © 2011 American Cancer Society
Volume 118, Issue 8, pages 2059–2068, 15 April 2012
How to Cite
Kapadia, N. S., Olson, K., Sandler, H. M., Feng, F. Y. and Hamstra, D. A. (2012), Interval to biochemical failure as a biomarker for cause-specific and overall survival after dose-escalated external beam radiation therapy for prostate cancer. Cancer, 118: 2059–2068. doi: 10.1002/cncr.26498
- Issue online: 6 APR 2012
- Version of Record online: 26 AUG 2011
- Manuscript Accepted: 18 JUL 2011
- Manuscript Revised: 29 JUN 2011
- Manuscript Received: 11 MAY 2011
- risk factors;
- prostate-specific antigen;
- cause-specific survival;
After external beam radiation therapy (EBRT) for prostate cancer, a short interval to biochemical failure of <18 months has been proposed as a surrogate for cause-specific survival. Because EBRT dose influences biochemical failure, the authors investigated the interval to biochemical failure in a cohort of patients treated with dose-escalated EBRT.
From 1998 to 2008, 710 patients were treated with EBRT (≥75 grays) ± androgen deprivation therapy (ADT) at the University of Michigan. Biochemical failure was defined using the Phoenix consensus definition (nadir + 2 ng/mL). A short interval to biochemical failure was defined as <18 months after completing radiotherapy and/or ADT. The associations between biochemical failure, the interval to biochemical failure, and clinical factors with cause-specific survival (CSS) and overall survival (OS) were evaluated.
There were 149 biochemical failures (21%), and short interval to biochemical failure accounted for 14% and 40% of biochemical failures in those with intermediate-risk or high-risk disease, respectively. Biochemical failure impacted CSS (P < .0001) but not OS (P = .36). However, a short interval to biochemical failure predicted decreased CSS (P < .0001; hazard ratio [HR], 5.6; 95% confidence interval [CI], 2.4-13.0) and OS (P < .0001; HR, 4.8; 95% CI, 2.3-10.3) when compared with a long interval to biochemical failure. The 8-year OS was 78% without biochemical failure, compared with 87% with a long interval to biochemical failure (P = .1; HR, 0.7; 95% CI, 0.4-1.1) and 38% with a short interval to biochemical failure (P < .0001; HR, 3.7; 95% CI, 2.3-5.9). On multivariate analysis, a short interval to biochemical failure increased the risk of prostate cancer death (P < .0001; HR, 18.1; 95% CI, 8.4-39) and all cause mortality (P = .0027; HR, 1.5; 95% CI, 1.2-2.1), whereas a long interval to biochemical failure did not.
The relation between the interval to biochemical failure, CSS, and OS was independently validated in patients treated with dose-escalated EBRT. Further evaluation of the interval to biochemical failure as a surrogate endpoint is warranted. Cancer 2012. © 2011 American Cancer Society.
The long natural history of prostate cancer (PCa) has led to the search for biomarkers to be used as surrogates for either cause-specific survival (CSS) or overall survival (OS). After prostatectomy, the presence of detectable prostate-specific antigen (PSA) is defined as biochemical failure, whereas classifying biochemical failure after external beam radiation therapy (EBRT) is more complex, because residual normal prostatic tissue remains, and the PSA remains detectable. The current consensus (Phoenix) definition of biochemical failure after EBRT or brachytherapy with or without androgen deprivation therapy (ADT) is when post-treatment PSA reaches at least 2 ng/mL greater than its lowest post-treatment (or nadir) value.1 However, biochemical failure by itself does not necessarily portend a grave prognosis, a fact recognized both by an American Society of Therapeutic Radiation Oncology consensus panel and the National Comprehensive Cancer Network (NCCN) guidelines.2, 3 In fact, given the lack of clearly proven benefit of early salvage ADT, these bodies recommend that clinical failure—not biochemical failure—indicates the need for salvage ADT.
Because of the poor specificity of biochemical failure for later clinical failure, several other metrics have been evaluated as potential surrogates for clinical trial design or by which to plan the use of salvage therapy. The level and timing of PSA nadir have both been correlated with biochemical failure, distant metastasis (DM), and CSS in radiotherapy (RT) series4-9; both PSA velocity and/or doubling time10-12 have also been evaluated as potential surrogates. However, each of these has been demonstrated as an incomplete surrogate for CSS, with even worse correlation with OS. The interval to biochemical failure has also been evaluated as a surrogate for DM and CSS in both surgical and RT series.13, 14 Investigators at Fox Chase Cancer Center evaluated the time from the end of all therapy (EBRT and ADT) to biochemical failure, and identified 18 months as a cutpoint that was optimal as a potential surrogate for CSS.14 This cutpoint was subsequently evaluated in a multi-institutional series and found to be superior to both PSA nadir and PSA doubling time as a surrogate for CSS.15 Analysis of the TROG 96.01 randomized trial also established an interval to biochemical failure of <24 months as a potential surrogate for CSS among patients randomized to low-dose (66 grays [Gy]) RT alone or with either 3 or 6 months of ADT.16 These studies all used either uniform low-dose RT16 or a broad range of RT doses,14, 15 but because the level and timing of the PSA nadir as well as biochemical failure are functions of RT dose, this might influence the prognostic significance of the interval to biochemical failure. Therefore, we sought to validate the interval to biochemical failure in a cohort of patients uniformly treated with dose-escalated RT.
MATERIALS AND METHODS
Between 1998 and 2008, 718 PCa patients were treated with dose-escalated EBRT ± ADT at the University of Michigan Medical Center (Ann Arbor, Mich). Information about the interval to biochemical failure was available for 99% (710 of 718) of patients, who form the cohort for this analysis.
This was a retrospective review that was institutional review board-approved and not deemed to require informed consent. However, from 2002 to the present, at the time of radiation treatment patients have signed informed consent to have their clinical information included in the prostate database at the University of Michigan Medical Center.
All patients were clinically staged per the American Joint Committee on Cancer sixth edition staging criteria into low-risk, intermediate-risk, and high-risk groups. Patients with high-risk disease routinely underwent staging with computed tomography (CT) and bone scan to rule out metastatic disease. Per NCCN guidelines, staging was not part of standard practice for those with low-risk or intermediate-risk disease. In addition, neither endorectal coil or pelvic magnetic resonance imaging were routinely used. The Charlson Comorbidity Index was used to evaluate comorbid illness and was evaluable in 97% (691 of 710) of patients.17
Patients were treated based on CT planning with either 3-dimensional conformal EBRT or intensity-modulated RT prescribed such that the 95% isodose surface encompassed the planning target volume (PTV). Low-risk patients were treated at the prostate alone, and intermediate-risk patients were routinely treated at the prostate and seminal vesicles. High-risk patients routinely (87%) received treatment to the pelvic lymph nodes followed by a boost to the prostate and seminal vesicles. ADT was used in 39% of all patients: 11% (median, 4 months), 27% (median, 6 months), and 79% (median, 21 months) for low-risk, intermediate-risk, and high-risk patients, respectively. For purposes of analysis, short-term ADT was defined as <12 months, whereas long-term ADT was defined as ≥12 months.
Follow-up and Endpoints
Follow-up and PSA were routinely obtained at 3- to 4-month intervals for the first 2 years, every 6 months for 5 years, and every 6 to 12 months thereafter. The Phoenix definition (nadir + 2 ng/mL) was used to define biochemical failure1; DM was defined as any clinical, radiographic, or pathologic evidence of DM. OS was defined as death from any cause, whereas CSS was defined as 1) death attributed to PCa or death in any patient with either 2) castration-resistant PCa or 3) documented evidence of metastatic disease before death. The time of follow-up was based upon the last day of RT; however, the interval to biochemical failure was defined based upon the time from the end of all therapy (RT or ADT) to biochemical failure. Those who progressed while still on ADT were scored as having an interval to biochemical failure of zero.
Comparison between continuous variables was performed with 1-way analysis of variance, whereas comparisons between categorical variables used the chi-square test. The log-rank test and Kaplan-Meier methods compared the influence of single variables on survival endpoints, whereas multivariate analyses were conducted using stepwise Cox proportional hazard models.
Patient Characteristics and Follow-up
Median age at the time of treatment was 69 years (interquartile range [IQR], 63-74), with no difference in age by biochemical failure status (none, a short interval to biochemical failure [<18 months from the end of all therapy], or long interval to biochemical failure, Table 1). Median follow-up was 64 months (IQR, 36-89): 57 months (IQR, 34-82) in those without biochemical failure, 69 months (IQR, 35-91) in those with a short interval to biochemical failure, and 98 months (IQR, 71-131) in those with long interval to biochemical failure. Not surprisingly, patients with a short interval to biochemical failure were more likely to have higher-risk disease as measured by PSA, TNM classification, Gleason score, and risk group in comparison to those without biochemical failure or with a long interval to biochemical failure (Table 1).
|Variable||No Biochemical Failure||Biochemical Failure||P|
|Short IBF||Long IBF|
|Number of patients||563||41||106|
|Age, median y (IQR)||69.5 (63-74)||66.4 (59-75)||69.8 (64-73)||.5a|
|PSA (IQR)||7.8 (5.3-12.1)||26.1 (12.4-48)||11.3 (6.0-24)||<.001a|
|Clinical TNM classification||<.001a|
|Dose, median Gy (IQR)||78 (76-78)||77 (77-78)||77 (76-78)||.7a|
|Duration, mo||6.8 (6.0-24.3)||24.9 (6.2-31)||6.4 (6.0-20.4)||<.001a|
|NCCN risk group||<.0001b|
Biochemical Failure and Clinical Outcome
Biochemical failure was observed in 147 patients (21% of all) at a median of 42 months (IQR, 22-67) from the end of EBRT. A short interval to biochemical failure was seen in 41 patients, occurring a median 13 months (IQR, 8-22) from the end of RT, whereas for those with a long interval to biochemical failure this occurred a median 56 months (IQR, 40-80) after completing RT. Among the 160 low-risk patients, there were 13 biochemical failures (8%) and no patients with a short interval to biochemical failure. Among 319 patients with intermediate-risk disease, there were 49 biochemical failures (15%), and 7 of these were with a short interval to biochemical failure (2% of all intermediate-risk patients and 15% of all biochemical failures in this group). Among the 231 high-risk patients, there were 85 biochemical failures (36%), and 34 of these had a short interval to biochemical failure (15% of all high-risk patients and 40% of biochemical failures in this group). The rates of PCa death (±standard error of the mean) were greater with increasing risk: 1% (±1%), 3% (±1%), and 15% (±5%) at 8 years for low-risk, intermediate-risk, and high-risk patients, respectively (P < .0001). No patient died of PCa without first experiencing biochemical failure. As a result, cause-specific survival was worse in those with biochemical failure for the whole population (P < .0001) and when broken down by NCCN risk groups (Fig. 1). However, the positive predictive value of biochemical failure was low, as to date only 24% (36 of 147) of patients with biochemical failure have died of PCa. In addition, there was no correlation between biochemical failure and OS (P = .38) even for patients with high-risk disease (P = .8, Fig. 2).
Interval to Biochemical Failure and Clinical Outcome
Because biochemical failure was poorly correlated with CSS and OS, the impact of a short interval to biochemical failure for subsequent clinical events was evaluated. At 8 years, patients with a long interval to biochemical failure were less likely to experience metastasis, as compared with those with a short interval to biochemical failure (metastasis-free rate, 79% ± 8% vs 33% ± 5%; P < .0001; hazard ratio [HR], 4.2; 95% confidence interval [CI], 2.2-8.0; Fig. 3A). Similarly, 7% ± 1% of patients died of PCa at 8 years, and short interval to biochemical failure predicted for worse CSS than long interval to biochemical failure or no biochemical failure. No patients without biochemical failure died of PCa; the 8-year freedom from PCa death was 87% ± 4% in patients with a long interval to biochemical failure, and 44% ± 9% in those with a short interval to biochemical failure (P < .0001; HR, 5.6; 95% CI, 2.4-12.9; Fig. 3B). Finally, a short interval to biochemical failure also predicted for worse OS, with an 8-year rate of OS of 38% ± 9% in those with a short interval to biochemical failure as compared with 87% ± 4% in those with a long interval to biochemical failure (P < .0001; HR, 4.9; 95% CI, 2.3-10.3). Furthermore, OS among patients with a long interval to biochemical failure did not differ significantly as compared with those with no biochemical failure (P > .06; HR, 0.7; 95% CI, 0.4-1.1; Fig. 3C).
Potential confounding factors that could limit the applicability of the interval to biochemical failure were next evaluated. The negative prognostic impact of a short interval to biochemical failure on CSS and OS was evident both in those treated with EBRT alone and those treated with EBRT plus ADT. In those treated with EBRT and no ADT, a short interval to biochemical failure carried an 8.8-fold increased risk for death from PCa as compared with a long interval to biochemical failure (P < .0001; 95% CI, 2.2-36; Fig. 4A) and a 7.3-fold increased risk of all cause mortality (P < .0001; 95% CI, 2.0-27; Fig. 4C). For patients who received both EBRT and ADT, a short interval to biochemical failure was associated with a 3.7-fold increased risk of PCa death (P = .003; 95% CI, 1.3-10.7; Fig. 4B) and a 3.3-fold increased risk of all cause mortality (P = .0016; 95% CI, 1.3-8.1; Fig. 4D) as compared with a long interval to biochemical failure. As above, patients with a long interval to biochemical failure did not have significantly different OS as compared with those without biochemical failure when treated with RT alone (Fig. 4B) or when treated with RT + ADT (Fig. 4D).
In addition, given the potential influence of age on OS, the impact of a short interval to biochemical failure was assessed as a function of age. When stratified by those <60, 60 to 70, or >70 years of age, there was a similar relation, such that in each age group a short interval to biochemical failure predicted for worse OS, whereas a long interval to biochemical failure did not (Table 2). We also considered the possibility that a short interval to biochemical failure resulted from less aggressive treatment, especially as compared with patients with a long interval to biochemical failure, and that this might explain a worse clinical outcome. However, we found no difference in RT dose based upon biochemical failure status (P = .7, Table 1) and that pelvic EBRT fields were most commonly used in patients who would later experience a short interval to biochemical failure (P < .0001, Table 1). Patients with a short interval to biochemical failure also received adjuvant ADT more frequently than those patients with a long interval to biochemical failure (61% vs 47%, P < .004) and were treated with a longer duration of ADT (median, 25 vs 6 months; P < .001). Finally, patients with a short interval to biochemical failure were not only more likely to receive salvage ADT after biochemical failure, but also as a group received it sooner than those with a long interval to biochemical failure (P < .001; HR, 2.7; 95% CI, 1.6-4.5; Fig. 5). Two years after biochemical failure, 85% ± 6% of those with a short interval to biochemical failure received salvage ADT, whereas only 53% ± 6% of those with a long interval to biochemical failure were accordingly treated.
|All, N=718||Age <60 Years, n=121||Age 60-70 Years, n=259||Age >70 Years, n=338|
|5 Years||8 Years||5 Years||8 Years||5 Years||8 Years||5 Years||8 Years|
|No BF||92% (91-93)||78% (75-81)||97% (95-99)||89% (83-95)||92% (90-94)||85% (81-89)||90% (88-92)||72% (67-77)|
|LIBF||95% (93-97)||87% (83-91)||92% (84-99)||92% (84-99)||97% (94-99)||95% (93-97)||92% (88-96)||79% (73-85)|
|SIBF||68% (60-76)||38% (29-47)||75% (62-78)||50% (33-67)||68% (60-76)||47% (33-61)||60% (46-74)||21% (8-34)|
Multivariate Regression Results
Finally, a step-wise Cox proportional hazards regression model was used to determine which factors were prognostic for prostate-cancer specific death and overall mortality (Table 3). Patients with Gleason scores >7 or a short interval to biochemical failure were more likely to die of PCa, whereas long-term ADT decreased the risk of PCa-related death. Similarly multivariate analysis of OS revealed that Gleason scores >7, advanced age, comorbid illness, and a short interval to biochemical failure increased the risk of all cause mortality. As above, long-term ADT decreased the risk of all cause mortality. Factors that were evaluated, but not prognostic for either CSS or OS, included PSA (as a continuous variable), Gleason score ≤7, clinical TNM classification, the use of short-term ADT, pelvic RT, and biochemical failure >18 months from the end of treatment.
|Covariate||Cause-Specific Survival||Overall Survival|
|p||HR||95% CI||p||HR||95% CI|
|Gleason score 2-6||Reference||Reference|
|Gleason score 8||<.04||2.5||1.1-5.4||.0010||2.4||1.4-4.2|
|Gleason score 9-10||<.0001||12.3||5.6-28||<.0001||17.9||9.6-33|
Given the long natural history of PCa, several biomarkers that may act as surrogates for CSS and OS have been evaluated.10, 14, 16 In the present study, we independently validated the interval to biochemical failure as a surrogate for metastasis-free survival and CSS in patients treated with dose-escalated EBRT (with or without ADT), using a previously determined cutpoint of 18 months. Multivariate analysis confirmed that a short interval to biochemical failure was the strongest prognostic factor for CSS, associated with an 18-fold increase in the risk of PCa death. More importantly, multivariate analysis demonstrated that patients with a short interval to biochemical failure were 1.5× more likely to die from any cause than those without biochemical failure. This was clearly influenced by the finding that 77% of the patients treated in this cohort had intermediate-risk or high-risk disease. Interestingly, those with a long interval to biochemical failure did not have worse OS when compared with those without biochemical failure, which was confirmed upon multivariate analysis. Not surprisingly, given that most deaths in the group were not because of PCa, patient age and the presence of comorbid illness more significantly impacted OS than did the interval to biochemical failure. From this retrospective review, were are unable to make comment as to the impact of metastatic failure versus local failure; however, given the benefit observed through the addition of RT to ADT and vice versa, it is clear that both local and systemic control are likely of importance.
The benefit of ADT in concert with EBRT has been demonstrated in multiple randomized trials, as shown by improved disease-specific (and in some studies, overall) survival in patients with locally advanced and/or high-grade PCa treated with conventional dose RT.18-20 However, the timing of ADT after biochemical failure is more controversial. The NCCN has suggested deferring salvage hormonal therapy until documented metastatic or symptomatic disease.2 Other authors, however, advocate early ADT use to prevent the morbidity of potentially symptomatic metastatic disease.21 Although randomized investigations have failed to show a survival benefit in upfront versus delayed use, there does appear to be a benefit to immediate (and indefinite) ADT use in at least the postoperative setting for lymph node-positive patients.22
A secondary analysis of RTOG 85-31 (which randomized patients with locally advanced PCa to EBRT alone vs EBRT with lifelong adjuvant ADT) evaluated patients allocated to the RT-alone arm and divided them into 2 groups: early salvage therapy versus late salvage therapy as defined by the PSA level at which salvage was initiated, using 10 ng/mL as a threshold. Multivariate analysis demonstrated a statistically significant improvement in OS among patients who were salvaged before their PSA level reached 10 ng/mL, although interestingly no significant differences were observed in either CSS or local control.23 This suggests a possible bias toward using salvage ADT in those who were overall in better health. A similar secondary analysis was conducted for the RTOG 86-10 trial (which randomized patients with locally advanced PCa to EBRT alone vs EBRT + 4 months neoadjuvant and concurrent ADT). For patients randomized to RT alone (without adjuvant ADT), the initiation of salvage ADT before metastasis improved both OS and CSS.24 The results of the present analysis would seem to reinforce the above findings, suggesting that patients with a short interval to biochemical failure harbor aggressive disease and may stand to benefit from early and aggressive salvage therapy.
The primary strengths of this study include the contemporary and consistent treatment of patients over a 10-year period. These patients were treated in the modern era of dose-escalated EBRT, and as such the minimum PTV treatment dose was 75 Gy. This is in contrast to the report by Buyyounouski and colleagues, where the median radiotherapy dose was 72 Gy and the minimum reported dose was 67 Gy,14 and the TROG 96.01 study, where the prescribed RT dose was 66.6 Gy.16 The dose of RT is relevant because increased dose is more likely to improve local and biochemical control and has been demonstrated to alter the timing of PSA response, which could potentially impact the relevance of a short interval to biochemical failure.25 Limitations in the current data include the retrospective analysis and the relatively few events. In addition, the duration of ADT can influence the timing of testosterone recovery,26 which in turn may delay the rise of PSA,27 thereby altering the prognostic significance of the interval to biochemical failure. Unfortunately, we were unable to control for testosterone recovery in the current analysis, as these data were not routinely obtained.
A short interval to biochemical failure, defined as within 18 months of completing therapy (EBRT and/or ADT), correlated with a significantly increased rate of distant metastasis, decreased CSS, and decreased OS, as compared with a long interval to biochemical failure or a lack of biochemical failure. We are encouraged that this easily determined PSA-derived endpoint predicts for OS after RT in PCa. Nevertheless, these data regarding the interval to biochemical failure are hypothesis generating and must be validated within other datasets or randomized trials before gaining widespread acceptance.
No specific funding was disclosed.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.
- 6Posttreatment prostate-specific antigen nadir highly predictive of distant failure and death from prostate cancer. Int J Radiat Oncol Biol Phys. 2002; 2002: 297-303., , .
- 8Radiotherapy for localized prostate carcinoma: the correlation of pretreatment prostate specific antigen and nadir prostate specific antigen with outcome as assessed by systematic biopsy and serum prostate specific antigen. Cancer. 1997; 79: 328-336., , , et al.