D'Amico AV, Whittington R, Malkowicz SB, Schultz D, Blank K, Broderick GA, Tomaszewski JE, Renshaw AA, Kaplan I, Beard CJ, Wein A. Biochemical Outcome After Radical Prostatectomy, External Beam Radiation Therapy, or Interstitial Radiation Therapy for Clinically Localized Prostate Cancer. JAMA. 1998;280(11):969–974. doi:10.1001/jama.280.11.969
From the Joint Center for Radiation Therapy, Harvard Medical School, Boston, Mass (Drs D'Amico, Kaplan, and Beard); Departments of Radiation Oncology (Drs Whittington and Blank), Urology (Drs Malkowicz, Broderick, and Wein), and Pathology (Dr Tomaszewski), Hospital of the University of Pennsylvania, Philadelphia; Department of Mathematics, University of Millersville (Dr Schultz), Millersville, Pa; and Department of Pathology, Brigham and Women's Hospital (Dr Renshaw), Boston, Mass.
Context.— Interstitial radiation (implant) therapy is used to treat clinically
localized adenocarcinoma of the prostate, but how it compares with other treatments
is not known.
Objective.— To estimate control of prostate-specific antigen (PSA) after radical
prostatectomy (RP), external beam radiation (RT), or implant with or without
neoadjuvant androgen deprivation therapy in patients with clinically localized
Design.— Retrospective cohort study of outcome data compared using Cox regression
Setting and Patients.— A total of 1872 men treated between January 1989 and October 1997 with
an RP (n=888) or implant with or without neoadjuvant androgen deprivation
therapy (n=218) at the Hospital of the University of Pennsylvania, Philadelphia,
or RT (n=766) at the Joint Center for Radiation Therapy, Boston, Mass, were
Main Outcome Measure.— Actuarial freedom from PSA failure (defined as PSA outcome).
Results.— The relative risk (RR) of PSA failure in low-risk patients (stage T1c,
T2a and PSA level ≤10 ng/mL and Gleason score ≤6) treated using RT,
implant plus androgen deprivation therapy, or implant therapy was 1.1 (95%
confidence interval [CI], 0.5-2.7), 0.5 (95% CI, 0.1-1.9), and 1.1 (95% CI,
0.3-3.6), respectively, compared with those patients treated with RP. The
RRs of PSA failure in the intermediate-risk patients (stage T2b or Gleason
score of 7 or PSA level >10 and ≤20 ng/mL) and high-risk patients (stage
T2c or PSA level >20 ng/mL or Gleason score ≥8) treated with implant compared
with RP were 3.1 (95% CI, 1.5-6.1) and 3.0 (95% CI, 1.8-5.0), respectively.
The addition of androgen deprivation to implant therapy did not improve PSA
outcome in high-risk patients but resulted in a PSA outcome that was not statistically
different compared with the results obtained using RP or RT in intermediate-risk
patients. These results were unchanged when patients were stratified using
the traditional rankings of biopsy Gleason scores of 2 through 4 vs 5 through
6 vs 7 vs 8 through 10.
Conclusions.— Low-risk patients had estimates of 5-year PSA outcome after treatment
with RP, RT, or implant with or without neoadjuvant androgen deprivation that
were not statistically different, whereas intermediate- and high-risk patients
treated with RP or RT did better then those treated by implant. Prospective
randomized trials are needed to verify these findings.
THERE ARE no completed prospective randomized trials, to our knowledge,
that compare definitive local treatment options for clinically localized adenocarcinoma
of the prostate. Retrospective comparisons1,2
stratified by the known prognostic factors and using actuarial analyses have
been published comparing radical prostatectomy (RP) with external beam radiation
therapy (RT). However, a direct comparison of the results of ultrasound-guided
interstitial prostate radiation (implant) therapy with RP or RT stratified
by the pretreatment prognostic factors has not been previously reported.
The utility of the pretreatment prostate-specific antigen (PSA),3 biopsy Gleason score,4
and American Joint Commission on Cancer Staging (AJCC) T stage5
in predicting postradiation6- 10
and postoperative11- 16
PSA outcome has been previously published by several investigators.
The purpose of this study is to provide PSA outcome data stratified
by the pretreatment PSA, biopsy Gleason score, and AJCC T stage in men treated
with RP, RT, or implant therapy with or without the addition of neoadjuvant
androgen deprivation for clinically localized prostate cancer.
Between January 1989 and October 1997, 1872 men with clinically localized
prostate cancer underwent definitive local therapy. Local therapy received
was RP (n=888) or implant with or without neoadjuvant androgen deprivation
therapy (n=218) at the Hospital of the University of Pennsylvania (HUP), Philadelphia,
or conformal RT (n=766) at the Joint Center for Radiation Therapy, Boston,
In all cases, staging evaluation included a history and physical examination
including a digital rectal examination, serum PSA, computed tomographic scan
of the pelvis or an endorectal and pelvic coil magnetic resonance imaging
scan of the prostate and pelvis, bone scan, and a transrectal ultrasound-guided
needle biopsy of the prostate with Gleason score histologic grading.4 A sextant biopsy was performed using a 18-gauge Tru-Cut
needle (Travenol Laboratories, Deerfield, Ill) via a transrectal approach.
The clinical stage was obtained from the digital rectal examination findings
using the 1992 AJCC staging system.5 Radiologic
and biopsy information was not used to determine clinical stage. The PSA level
was obtained on an ambulatory basis prior to radiologic studies and the biopsy
procedure. All PSA measurements3 were made
using the Hybritech (San Diego, Calif), Tosoh (Foster City, Calif), or Abbott
assays (Chicago, Ill).
A referee genitourinary pathologist reviewed the diagnostic biopsy specimens
for all patients undergoing surgery or implant at the HUP (J.E.T.) and RT
at the Joint Center for Radiation Therapy (A.A.R.). Surgical treatment consisted
of a radical retropubic prostatectomy and bilateral pelvic lymph node sampling.
All patients managed with definitive RT were treated using at least
10-MV photons and a conformal shaped 4-field technique. Those patients with
AJCC clinical stage T1c, T2a disease who also had a PSA level of 10 ng/mL
or less and biopsy Gleason score of 2 to 6 were treated to the prostate only
with a 1.5-cm margin. The median prescription dose was 66 Gy (66-70 Gy) and
was delivered using 2-Gy fractions. All other patients with clinically localized
disease received a median prescription dose of 45 Gy (45-50.4 Gy) in 1.8-Gy
fractions to the prostate and seminal vesicles plus a 1.5-cm margin. This
was followed by treatment to the prostate alone using a shrinking field technique
with a 1.5-cm margin to a median prescription dose of 22 Gy (18-22 Gy) in
1.8- to 2.0-Gy fractions. A 95% normalization was used.
Implant therapy was performed using palladium 103103 Pd)
seeds, a perineal template-guided, peripheral-loading technique, and a Bruel
& Kjaer 8551 transrectal ultrasound unit (Naerum, Denmark). The minimum
peripheral dose to the prostatic capsule was 115 Gy. A transrectal ultrasound
probe was used to image the prostate at 5-mm intervals preoperatively to ascertain
the optimal number and location of seeds needed to deliver the minimum peripheral
dose to the entire prostate gland volume. Individual seed strength ranged
from 58 to 61 MBq. The total amount implanted ranged from 1306 to 7189 MBq.
Postimplant dosimetry was performed on all patients based on films obtained
at 4 weeks after the implant. For the first 143 patients this consisted of
orthogonal films, and for the latter 75 patients, computed tomography was
used. Of the 218 patients who received implant therapy, 152 (70%) received
neoadjuvant androgen deprivation for a median of 3 months (2-10 months). Hormonal
therapy consisted of a luteinizing hormone-releasing hormone agonist that
was preceded by the use of a nonsteroidal antiandrogen for 7 to 10 days. Ninety-six
(63%) of 152 patients received 3 months of neoadjuvant androgen deprivation
therapy. The remaining 2 (1.33%), 15 (10%), 14 (9%), 20 (13%), 1 (1%), 2 (1.33%),
and 2 (1.33%) received 2, 4, 5, 6, 7, 9, or 10 months of neoadjuvant androgen
deprivation therapy, respectively.
The median follow-up of the 888 surgically managed patients at HUP was
38 months (8-100 months). The median follow-up for the 766 and 218 radiation-managed
patients at the Joint Center for Radiation Therapy and HUP was 38 months (8-75
months) and 41 months (3-72 months), respectively. The patients were seen
1 month postoperatively or after the end of radiation therapy, then at 3-month
intervals for 2 years, every 6 months for 5 years, and annually thereafter.
At each follow-up a serum PSA was obtained prior to performing the digital
rectal examination. All pretreatment PSA values were obtained within 1 month
of the date of surgery or start of radiation. No patient was lost to follow-up
and all patients were alive at the time of this analysis.
In order to have the multivariable analysis results of the Cox proportional
hazards regression model be applicable in the clinical setting for an individual
patient,3 risk groups were defined. These risk
groups were established from a review of the literature6- 19
and were based on the known prognostic factors: PSA level, biopsy Gleason
score, and 1992 AJCC T stage. Patients with AJCC clinical T stage T1c, T2a
and PSA level of 10 ng/mL or less and biopsy Gleason score of 6 or less have
been identified to be at low risk (<25% at 5 years) for posttherapy PSA
failure. Conversely, patients with AJCC stage T2c disease or a PSA level of
more than 20 ng/mL or a biopsy Gleason score of 8 or more have a risk higher
than 50% at 5 years of posttherapy PSA failure. The remaining patients with
PSA levels higher than 10 and 20 ng/mL or lower, a biopsy Gleason score of
7, or AJCC clinical stage T2b have been found to have an intermediate risk
(25%-50% at 5 years of posttherapy PSA failure). Patients with AJCC clinical
stage T1a, T1b were not managed using implant therapy because of the significant
rate or urinary incontinence noted17 using
this approach in patients with a history of a transurethral resection of the
prostate. Therefore, patients with AJCC clinical stage T1a, T1b disease managed
with RP or RT were excluded from the study to ensure statistically valid comparisons.
A Cox regression multivariable analysis20
was used to compare PSA outcome among the therapies within each risk group.
For each analysis the assumptions of the Cox model were tested and satisfied.
Coefficients from the Cox regression model were used to calculate the overall
relative risk of PSA failure for patients managed with RT or implant with
or without neoadjuvant androgen suppression as compared with patients managed
with RP. For the purposes of the multivariable analysis, the type of therapy
was treated as a categorical variable indicating RP at HUP, RT, implant, or
implant plus neoadjuvant androgen deprivation. Radical prostatectomy at HUP
was defined as the baseline group for the purposes of the multivariable analyses.
Patients were also stratified and analyzed with the traditional rankings of
a biopsy Gleason score of 2 through 4, 5 through 6, 7, and 8 through 10.
Prostate-specific antigen failure was defined according to the American
Society of Therapeutic Radiation and Oncology 1996 consensus statement21 for all study patients. The definition required that
a patient have 3 consecutive rising PSA values each obtained at least 3 months
apart before PSA failure was scored. The time of PSA failure was defined as
the midpoint between the time of the PSA nadir value and the time of the first
rising PSA value. Time zero was defined as the date of diagnosis for all study
Pairwise comparisons were made using the log-rank test. In the case
where a number of comparisons were made, the level of significance in order
to be called statistically significant was lowered from the convention of
.05 to .05 divided by the number of comparisons following the Bonferonni adjustment.22 For the purpose of illustration, estimates of PSA
outcome were calculated using the Kaplan-Meier23
actuarial method and graphically displayed. In the low-, intermediate-, and
high-risk patient groups the sample size and the number of events in this
study was sufficient to detect a 12%, 17%, and 15% difference in PSA survival,
respectively, with 80% power at a .05 level of significance. This was calculated
for a baseline PSA survival of 85%, 60%, and 30% at 5 years in the low-, intermediate-,
and high-risk patients, respectively.
The clinical pretreatment characteristics of the 1872 patients used
in the time to PSA-failure analyses are listed in Table 1 and are stratified by the type of treatment. Table 2 lists the clinical characteristics of the study patients
within each risk group. The pairwise P values from
the comparative analyses of the proportion of patients having a specific pretreatment
clinical characteristic between the treatment groups are shown in Table 3. After adjustment for the multiple
comparisons,22 no significant differences were
noted in low-risk and intermediate-risk patients. High-risk patients managed
with implant plus neoadjuvant androgen deprivation had an increased proportion
of patients with PSA levels lower than 10 ng/mL and decreased proportion of
patients with a PSA level of more than 20 ng/mL compared with patients managed
with RP (P=.003) or RT (P=.0002).
Both of these differences could bias the comparisons of PSA survival in favor
of the implant plus neoadjuvant androgen suppression patient cohort. The use
of multiple comparisons between treatment modalities (n=6) required that the
level of statistical significance as per Bonferonni adjustment22
be redefined as lower than .008.
Table 4 lists the P values from the Cox regression multivariable analyses
evaluating the effect of the treatment type on time to posttherapy PSA failure
stratified by risk group. The relative risks of PSA failure with a 95% confidence
interval are also listed. No significant difference (P≥.25)
in outcome was noted in low-risk patients (T1c, T2a and PSA level ≤10 ng/mL
and Gleason score ≤6) across all treatment modalities. The 95% confidence
intervals for the relative risk of PSA failure relative to RP for all patients
included 1.0. High-risk patients (T2c, PSA level >20 ng/mL, or Gleason score ≥8),
however, treated using a RP or RT did significantly better (P≤.01) then those managed with implant with or without neoadjuvant
androgen deprivation. Specifically, high-risk patients managed with implant
therapy had at least a 2.2-fold increased risk of PSA failure compared with
those treated with RP even if neoadjuvant androgen deprivation therapy was
used. Intermediate-risk patients (T2b, Gleason score of 7, or PSA level >10
and ≤20 ng/mL) did significantly worse (P≤.003)
if managed by implant alone, but fared equivalently (P=.18)
to those patients managed with RP if androgen deprivation was also administered.
Intermediate-risk patients managed with implant therapy alone had a 3.1-fold
increased risk of PSA failure compared with those patients managed with RP.
These results remained unchanged when patients were stratified using the traditional
groups of biopsy Gleason score. Specifically, patients with biopsy Gleason
score of 2 through 6 had no statistical difference in their estimates of PSA
failure-free survival across all the treatment modalities evaluated in this
study. However, patients with biopsy Gleason scores of 8 through 10 who were
managed with implant with or without neoadjuvant androgen deprivation therapy
had a lower PSA failure-free survival that approached statistical significance
(P≤.07) when compared with those patients managed
with RP or RT. Patients with biopsy Gleason scores of 7 did not have statistically
different PSA failure-free survival when managed with RP, RT, or implant plus
neoadjuvant androgen deprivation therapy (P≥.59).
However these patients did statistically worse (P≤.003)
if managed by implant alone. This analysis was repeated using the traditional
Gleason score groupings for patients with PSA levels lower than 20 ng/mL and
the results remained unchanged.
For the purpose of illustration, estimates of PSA outcome with pairwise P values evaluating the comparisons between treatment types
were calculated using the Kaplan-Meier23 actuarial
method and are graphically displayed by risk group in Figure 1, Figure 2, and Figure 3 and by biopsy Gleason score in Figure 4, Figure 5, Figure 6, and Figure 7.
Several studies from the urologic12- 16
and oncologic6- 11,17- 19
literature support that the combination of the AJCC clinical T stage, pretreatment
PSA, and biopsy Gleason score can predict the pathologic organ confinement
rate, biochemical failure rate, and subsequent metastatic rates in patients
managed with definitive local therapy for clinically localized prostate cancer.
Therefore, when attempting to compare PSA outcome across different treatment
modalities, it is important to control for the values of these 3 prognostic
factors. Using the results of the published literature,6- 19
the risk of postradiation and postoperative PSA failure was classified into
3 groups based on the pretreatment prognostic factors.
Using a multivariable time-to-PSA-failure analysis to compare PSA outcome
after RP, RT, or implant with or without neoadjuvant androgen deprivation
therapy for patients stratified by the defined pretreatment risk groups, several
observations were noted. First, the group of patients defined to be at low
risk for posttherapy PSA failure were estimated to derive equal benefit from
treatment with RP, RT, or implant (Figure
1) at 5 years. Moreover, the addition of neoadjuvant androgen deprivation
to implant therapy in low-risk patients provided no further benefit in the
estimated 5-year PSA outcome. Second, patients at high risk for posttherapy
PSA failure did significantly worse with implant therapy despite the addition
of neoadjuvant hormonal deprivation when compared with patients treated with
RP or RT (Figure 3). A statistically
significant increase in favorable prognostic factors was present in the high-risk
patients managed with implant plus neoadjuvant androgen suppression (ie, PSA
level <10 ng/mL) compared with patients managed with RP or RT (Table 2 and Table 3). Despite this potential bias in favor of the patients managed
with implant plus neoadjuvant androgen suppression, the PSA outcome of these
patients was still inferior to those patients managed with RP or RT. Finally,
patients in the intermediate category for posttherapy PSA failure did significantly
worse when managed with implant alone as compared with patients managed with
RP, RT, or implant plus neoadjuvant androgen deprivation (Figure 2). While a statistical difference may exist for intermediate-risk
patients managed with implant plus neoadjuvant androgen deprivation therapy
vs RP or RT, this study was not adequately powered to detect this difference.
Further follow-up is needed to ascertain if these results are maintained.
In particular, low-risk patients can sustain late PSA failures (ie, beyond
5 years). Moreover, men with low-grade or low-risk disease have a relatively
low rate of PSA progression requiring numbers of patients much larger than
presented in this study in order to prove a statistical difference. Therefore,
while small differences may exist, they are unlikely to reach statistical
significance. In addition, the intermediate-risk patients managed with a median
of 3 months of neoadjuvant androgen deprivation and implant therapy may be
experiencing a hormone-induced delay in PSA failure and not a true therapeutic
gain. With only 9 patients at risk after 2 years in the implant plus androgen
deprivation group compared with 116 and 77 in the RP and RT managed groups,
respectively, it is too soon to make conclusions regarding the relative efficacy
of these 3 treatments. Therefore, because of the small numbers and relatively
short follow-up, particularly in the patients receiving neoadjuvant androgen
deprivation, the results must be viewed as preliminary. However, these early
data suggest that in high-risk patients, who are in greater need of treatment
and who have the most to lose by ineffective therapy, implant therapy with
or without the addition of a median of 3 months of neoadjuvant androgen deprivation
was less effective than RP or RT at maintaining PSA-based survival. When examining
the PSA failure-free survival using the traditional groupings of biopsy Gleason
score, the exact results were found as those noted when the data were analyzed
according to the risk groups lending further support to this study's findings.
Several issues remain that are not addressed by the data in this study.
First, the comparison of PSA outcome for expectant management vs treatment
is lacking. This comparison would be particularly relevant in the low-risk
patients where 5-year PSA-progression rates numerically approximate the 10-year
clinical-progression rates noted from expectant management series.24,25 Second, the PSA outcomes of the now
widely practiced combination therapies of RT plus implant with or without
neoadjuvant androgen deprivation therapy need to be prospectively compared
with the PSA outcomes achieved after RP, RT, or implant. These comparisons
would be particularly relevant in the high-risk and intermediate-risk groups
where implant therapy alone may be insufficient. A final unanswered question
remains. That is whether the use of 103Pd as opposed to the conventional
iodine 125 125 I) affected the PSA outcome data reported in this
study. The physical characteristics of these 2 radionuclides differ in that
the half life and mean photon energy are 60 days, 27 keV and 17 days, 21 keV
for 125>I and 103Pd, respectively. These differences
result in an initial dose rate of 0.0772 Gy/h and 0.197 Gy/h for 125I and 103Pd, respectively. It is therefore conceivable that
the higher dose rate of palladium could have affected the results. Further
investigations of these issues are needed.
Nevertheless, considering the widespread increase in the use of implant
therapy throughout the United States, these data serve to heighten awareness
to the possibility that this form of prostate cancer therapy may only be clinically
efficacious in a select subgroup of patients and possibly inadequate in others.
While no definitive conclusions can be reached using nonrandomized retrospective
data, these analyses can provide the basis on which to design prospective
randomized clinical trials that could definitively compare PSA, cause-specific,
and overall survival outcomes among treatment modalities.