Int. J. Radiation Oncology Biol. Phys., Vol. 81, No. 4, pp.

e361–e367, 2011
Copyright Ó 2011 Elsevier Inc.
Printed in the USA. All rights reserved
0360-3016/$ - see front matter

doi:10.1016/j.ijrobp.2011.04.048

CLINICAL INVESTIGATION Prostate

PERINEURAL INVASION PREDICTS INCREASED RECURRENCE, METASTASIS,

AND DEATH FROM PROSTATE CANCER FOLLOWING TREATMENT WITH
DOSE-ESCALATED RADIATION THERAPY

FELIX Y. FENG, M.D.,*y YUSHEN QIAN, B.S.,* MATTHEW H. STENMARK, M.D.,*

SCHUYLER HALVERSON, B.S.,* KEVIN BLAS, B.S.,* SEAN VANCE, B.S.,*
HOWARD M. SANDLER, M.D., M.S.,z AND DANIEL A. HAMSTRA, M.D., PH.D.*
*The University of Michigan Medical Center, Ann Arbor, MI; yThe Ann Arbor Veteran Affairs Medical System, Ann Arbor, MI; and
z
The Cedars Sinai Medical System, Los Angeles, CA

Purpose: To assess the prognostic value of perineural invasion (PNI) for patients treated with dose-escalated
external-beam radiation therapy for prostate cancer.
Methods and Materials: Outcomes were analyzed for 651 men treated for prostate cancer with EBRT to
a minimum dose $75 Gy. We assessed the impact of PNI as well as pretreatment and treatment-related factors
on freedom from biochemical failure (FFBF), freedom from metastasis (FFM), cause-specific survival (CSS),
and overall survival.
Results: PNI was present in 34% of specimens at biopsy and was significantly associated with higher Gleason score
(GS), T stage, and prostate-specific antigen level. On univariate and multivariate analysis, the presence of PNI was
associated with worse FFBF (hazard ratio = 1.7, p <0.006), FFM (hazard ratio = 1.8, p <0.03), and CSS (HR = 1.4,
p <0.05) compared with absence of PNI; there was no difference in overall survival. Seven-year rates of FFBF,
FFM, and CCS were 64% vs. 80%, 84% vs. 92%, and 91% vs. 95% for those patients with and without PNI,
respectively. On recursive partitioning analysis, PNI predicted for worse FFM and CSS in patients with GS
8–10, with FFM of 67% vs. 89% (p <0.02), and CSS of 69% vs. 91%, (p <0.04) at 7 years for those with and without
PNI, respectively.
Conclusions: The presence of PNI in the prostate biopsy predicts worse clinical outcome for patients treated with
dose-escalated external-beam radiation therapy. Particularly in patients with GS 8–10 disease, the presence of PNI
suggests an increased risk of metastasis and prostate cancer death. Ó 2011 Elsevier Inc.

Risk factors, Clinical outcome, PSA, Cause-specific survival.

INTRODUCTION sive biology (3). PNI is frequently assessed on pathologic

evaluation of biopsy specimens; its assessment is reproduc-
The National Comprehensive Cancer Network (NCCN)
ible and requires no additional cost or technology.
risk-group stratification uses clinical T stage, biopsy Glea-
The presence of PNI on needle biopsy has previously been
son score (GS), and prostate-specific antigen (PSA) level
demonstrated to predict an increased likelihood of high-risk
to predict the risk of prostate cancer recurrence (1). How-
pathologic features at the time of radical prostatectomy (4).
ever, because biochemical failure (BF) is not highly corre-
Further, although some studies report the presence of PNI as
lated with clinically symptomatic disease or death (2),
an independent risk factor for BF (5) or prostate cancer death
there is a need to develop prognostic systems more closely
(6) following radiation therapy (RT), other studies have
associated with clinical outcome. One factor that may
failed to corroborate its prognostic value (7, 8). However,
improve prognostic systems is perineural invasion (PNI),
all these studies were primarily conducted before the era
defined as the presence of cancer tracking along or around
of radiation therapy (RT) dose escalation. Growing
a nerve within the perineural space. PNI is thought to be
evidence suggests that higher doses of RT improve local
a mechanism of prostate cancer extension from the prostatic
control, BF, and metastasis in patients treated with
to periprostatic soft tissue and also a marker of more aggres-

Reprint requests to: Daniel A. Hamstra, M.D., Ph.D., University Supplemental material cited in the article is available online.
of Michigan Health System, 1500 East Medical Center Drive, Floor Conflicts of interest: none.
B2, Room C490, Ann Arbor, MI 48109-5010. Tel: (734) 936-4302; Received Jan 26, 2011, and in revised form April 15, 2011.
Fax: (734) 763-7370; E-mail: dhamm@med.umich.edu Accepted for publication April 19, 2011.
Authors FYF and YQ contributed equally to this work and should
be considered as co–first authors.
e361
e362 I. J. Radiation Oncology d Biology d Physics Volume 81, Number 4, 2011

Table 1. Pretreatment and treatment-related features

Characteristic All PNI (–) PNI (+) p value

N 651 431 220

Age (years)
Median (IQR) 69.4 (63–74) 69.3 (63–74) 69.4 (61–74) 0.9*
Race >0.1y
Caucasian 88% 88% 89%
African American 9% 9% 9%
Other 3% 3% 2%
T stage <0.0001y
T1–T2a 73% 79% 61%
T2b–T2c 17% 14% 25%
T3–T4 10% 7% 14%
PSA (ng/mL) <0.03*
Median (IQR) 8.3 (5.6–13.7) 8.0 (5.5–12.7) 8.9 (5.9–19.4)
Range 0.8–240 0.8–240 1.3–219
Biopsy cores, median (IQR)
No. taken 7 (6–12) 7 (6–12) 7 (6–12) >0.9*
Percent Positive 38% (17%–64%) 33% (17%–50%) 58% (33%–83%) <0.01*
Gleason score <0.0001y
2–6 34% 44% 15%
7 46% 43% 52%
8–10 20% 13% 33%
NCCN risk group <0.0001y
Low 23% 29% 10%
Intermediate 44% 46% 42%
High 33% 25% 48%
RT dose (Gy) <0.005*
Median (IQR) 77.0 (75.8–77.8) 77.0 (75.8–77.8) 77.7 (77.0–77.8)
Pelvic RT <0.0001y
Yes 36% 27% 52%
No 64% 73% 48%
Androgen deprivation therapy <0.0001y
No 60% 74% 54%
Yes 40% 26% 46%
Duration (months), Median (IQR) 7.0 (6.0–24.4) 6.3 (5.8–22.3) 9.1 (6.1–25.0) <0.001*

Abbreviations: IQR = interquartile range; NCCN = National Comprehensive Cancer Network; PNI = perineural invasion; RT = radiation
therapy.
* t test.
y
Chi-square.

external-beam RT (EBRT) (9). The impact of PNI in the era
of dose-escalated EBRT is unclear, so this study assessed the
prognostic value of PNI on BF, freedom from metastasis
(FFM), and cause-specific survival (CSS) in patients with
prostate cancer treated with dose-escalated EBRT.
METHODS AND MATERIALS
Patient selection
As part of an institutional review board–approved analysis, 718
men with clinically localized prostate cancer treated with definitive
EBRT to at least 75 Gy, with or without androgen deprivation ther-
apy (ADT), between 1998 and 2008 were identified. Information
regarding the presence or absence of perineural invasion was
available on biopsy reports from 91% (651/718) of patients, who
form the basis for this evaluation.
high-risk disease, and those with evidence of metastatic disease
by imaging were excluded.
Treatment
All treatments used CT planning with either three-dimensional
conformal RT or intensity-modulated RT (IMRT). The prescribed
dose levels ranged from 75.0 to 79.2 Gy to the planning target
volume using conventional fractions of 1.8–2.0 Gy delivered 5
days weekly. Per standard institutional practice, low-risk patients
were treated to the prostate only, intermediate-risk patients to the
prostate and seminal vesicles, and high-risk patients to the pelvic
lymph nodes followed by a boost to the prostate and seminal
vesicles. The frequency and duration of ADT use correlated with
NCCN risk group, with utilization in 11%, 26%, and 80% of
low-, intermediate-, and high-risk patients for a median duration
of 4.1, 6.0, and 21.3 months, respectively.

Risk stratification Endpoints

Using clinical T stage, serum PSA, and biopsy GS, patients were Freedom from biochemical failure (FFBF) was defined based on
stratified by NCCN risk criteria into low-, intermediate-, and high- the Radiation Therapy Oncology Group Phoenix definition (nadir +
risk groups. Staging was routinely performed for patients with 2 ng/mL) (10). Freedom from metastases was defined as the
 Perineural invasion and prostate cancer d F. Y. FENG et al. e363

absence of any clinical, radiographic, or pathologic evidence of
metastatic disease. Overall survival (OS) was defined as death
due to any cause. Cause-specific survival was (1) death attributed
to prostate cancer or death in any patient with either (2)
castration-resistant prostate cancer or (3) metastatic disease before
death. The time of follow-up was based on the last day of RT.
Statistical analysis
Student’s t test was used for the comparison between
PNI-positive and -negative groups and continuous variables. The
chi-square test was used for univariate analysis between PNI groups
and categorical variables. Univariate survival analyses were con-
ducted using the log-rank test and Kaplan-Meier methods, and mul-
tivariable analyses were conducted using Cox proportional hazards
models. Recursive partitioning analysis was performed to correlate
outcome based on variables that were significant on multivariate
analysis. All statistical analysis was performed using MedCalc
(v11.4.4.0, MedCalc Software, Mariakerke, Belgium)
and the median follow-up was 62.2 months (interquartile
range, 36–87). The presence of PNI was correlated with in-
creased risk features (Table 1). For instance, cT3–4 disease
was present in 7% and 14% of patients without or with
PNI, respectively (p <0.0001). Similarly, GS 8–10 was
present in 13% of those without and 33% of those with
PNI (p <0.0001). As a result, NCCN high-risk features
were present in 25% and 48% of patients without or with
PNI (p <0.0001). Conversely, when stratified by NCCN
risk groups, PNI was present in 15% of low-, 32% of inter-
mediate-, 49% of high-risk patients. RT dose was higher in
those with PNI, but the difference was <1 Gy (median
dose 77.0 vs. 77.7 Gy, p <0.005). Pelvic RT was also more
frequently used in those with PNI (52%) than in those with-
out PNI (27%, p <0.0001). Finally, those with PNI were both
more likely to have received ADT (p <0.0001) and to have
ADT delivered for a longer duration (p <0.001).

RESULTS
The presence of PNI correlates with worse clinical
Clinical characteristics and treatment outcome
PNI was present in 34% (220/651) of patients. The On univariate analysis, the presence of PNI was associ-
median age was 69.4 years (interquartile range, 63–74), ated with decreased FFBF (hazards ratio [HR] = 2.2,

Freedom From Biochemical Failure as A Function of Perineural Invasion Freedom From Metastasis As A Function of Perineural Invasion
100 100
Survival Probability (%)

Survival Probability (%)

80 80

60 60

40 40

20 p<0.0001 20 p<0.0001
No No
Yes Yes
0 0
0 12 24 36 48 60 72 84 96 108 120 0 12 24 36 48 60 72 84 96 108 120
Time (months) Time (months)

Cause Specific Survival As A Function of Perineural Invasion Overall Survival As A Function of Perineural Invasion
100 100
Survival Probability (%)

Survival Probability (%)

80 80

60 60

40 40

20 p<0.04 20 p<0.09
No No
Yes Yes
0 0
0 12 24 36 48 60 72 84 96 108 120 0 12 24 36 48 60 72 84 96 108 120
Time (months) Time (months)
Fig. 1. The presence of perineural invasion predicts for worse clinical outcome. Kaplan-Meier analyses are presented for
freedom from biochemical failure (A), freedom from metastasis (B), cause-specific survival (C), and overall survival (D)
for patients as a function of the presence or absence of perineural invasion.
e364 I. J. Radiation Oncology d Biology d Physics
p <0.0001) and FFM (HR = 2.4, p <0.0004) and also with
decreased CSS (HR = 2.1, p <0.04; Fig. 1A–1C). However,
there was no significant difference in OS when analyzed by
PNI (p >0.1; Fig. 1D). The corresponding 7-year endpoints
for patients with and without PNI are presented in Table 2.
Univariate analysis was also conducted to assess the correla-
tion between other clinical and treatment-related features
with outcome (Table E1). As expected, PSA, T stage, GS,
and percentage of positive biopsy cores were all associated
with FFBF, FFM, and CSS (all p values <0.005). Radiation
dose, within the narrow range of doses evaluated, did not
predict for improvement in any clinical outcome, although
there was a trend to worse CSS with higher radiation dose,
likely because slightly higher radiation doses were used in
higher-risk patients. Similarly, the use of a pelvic RT field
correlated with worse clinical outcome for all endpoints
(all p values <0.001), again likely secondary to the use of
pelvic RT in higher-risk disease. Specifically, pelvic RT
was used in 3%, 13%, and 91% of low-, intermediate-, and
high-risk patients respectively (p <0.0001). Finally, ADT
use was also closely associated with higher clinical risk
groups, and as a result, ADT use correlated with worse
FFBF (p <0.002) and demonstrated a trend toward decreased
FFM and CSS.
Multivariate analysis confirms adverse impact of PNI
Multivariate Cox proportional hazards models were con-
ducted to test further the association between PNI and clin-
ical endpoints. Variables included in the models were PSA
(#20 vs. >20 ng/mL), clinical T stage (T1–T2a, T2b–T2c,
T3–T4), GS (2–6, 3+4, 4+3, 8–10), percent positive biopsy
cores (#50% vs. >50%), perineural invasion (no vs. yes),
pelvic RT (no vs. yes), and ADT use (no vs. yes). Percent
positive biopsy cores, T stage, and the use pelvic RT were
not prognostic in any model and were, therefore, excluded
from the final models (Table 3). PSA >20 ng/mL was asso-
ciated with worse FFBF (HR = 3.0, p <0.0001), FFM
(HR = 4.0, p <0.0001), and CSS (HR = 3.2, p <0.003). Glea-
son patterns of 3+4 and worse were correlated with worse
FFBF (HR $1.8, p #0.03), FFM (HR $2.4, p #0.04),
and CSS (HR $7.4, p #0.02) when compared with GS
Table 2. 7-year freedom from clinical endpoints (plus or
minus the standard error of the mean) as well as hazard ratios
(and 95% CI) from log-rank test for those without or with
perineural invasion
All patients
p value HR (95% CI) No PNI Yes PNI
FFBF <0.0001 2.2 (1.5–3.1) 80% (77–83) 64% (60–68)
FFM <0.0004 2.4 (1.4–4.0) 92% (90–94) 84% (81–87)
CSS <0.04 2.1 (1.0–4.4) 95% (93–97) 91% (88–94)
OS >0.1 1.4 (0.9–2.0) 84% (81–87) 78% (74–82)
Abbreviations: CI = confidence interval; CSS = cause-specific
survival; FFBH = freedom from biochemical failure; FFM = free-
dom from metastasis; HR = hazard ratio; OS = overall survival;
PNI = perineural invasion.
Volume 81, Number 4, 2011
2–6. For each endpoint, the presence of Gleason 8–10 dis-
ease carried an increased hazard compared with lower Glea-
son scores. The use of ADT was prognostic for FFM (HR =
0.32, p <0.0008) and CSS (HR = 0.42, p <0.05) but not for
BF. After adjusting for other pathologic and treatment
related features, the presence of PNI was associated with
worse FFBF (HR = 1.7, p <0.006), FFM (HR = 1.8,
p <0.03), and CSS (HR = 1.4, p <0.05) compared with the
absence of PNI.
PNI predicts worse metastasis and prostate cancer death
for Gleason 8–10 prostate cancer
Given the correlation between PNI and clinical endpoints,
a recursive partitioning model (RPA) was constructed to in-
tegrate PNI with clinical prognostic factors in predicting the
risk of metastasis (Fig. 2) and CSS (Fig. 3). In these models,
GS, PSA, PNI, and the use of ADT were the only variables
that retained prognostic significance for both FFM (Fig. 2)
and CSS (Fig. 3). For metastasis, nine terminal nodes were
identified, with the 7-year FFM ranging from 25% (for those
in node IX who had GS 8–10, PNI, and no ADT use) to 98%
(for those in node IV with GS 2–7 and a PSA >20 ng/mL
with or without PNI treated with ADT). Gleason score 2–7
vs. 8–10 was the primary decision point, with a 3.6-fold in-
creased risk of metastasis with GS 8–10 compared with
those with GS 2–7 disease (p <0.0001). In this higher-risk
group with GS 8–10, the additional prognostic variables
were the presence of PNI (HR = 3.3, p <0.02) and the use
of ADT, which predicted for decreased metastasis both in
those with and without PNI (HR = 0.13–0.17, p <0.008).
Within the GS 8–10 group, the 7-year FFM ranged from
25% in node IX (GS 8–10, PNI, no ADT) to 95% in node
VI (GS 8–10, no PNI, with ADT use).
Because Gleason score was the predominate stratification
variable, we next assessed outcome in those with a lower risk
of recurrence based on Gleason score. In those with GS 2–7,
92% were free of metastasis at 7 years; within this group, the
7-year FFM was 94% in those patients with PSA #20 and
84% in those with PSA >20 (HR = 3.1, p <0.0005). For
the GS 2–7 group, ADT appeared to provide a benefit in re-
ducing metastasis only in those with PSA >20 (HR = 0.26,
p <0.009). In those with a PSA #20, the only predictor of
FFM was GS, with 7-year FFM of 96%, 95%, and 86% for
GS 2–6, 3+4, and 4+3, respectively. Although ADT duration
and the use of pelvic RT were not prognostic in this model,
their use was tightly correlated with one another and with the
different RPA nodes, which may have confounded the ability
to ascertain their unique importance. Therefore, the overall
rate and duration of ADT as well as the rate of pelvic RT
for each of the nine nodes are provided in Table E2.
The same RPA was subsequently applied to CSS (Fig. 3),
yielding a 7-year CSS ranging from 25% to 100%. In this
model, only 7 terminal nodes were identified (there was no
statistical difference based on ADT use between nodes IV
and V and nodes VI and VII). As in the metastasis model,
Gleason score was the strongest prognostic factor with worse
outcome for those with Gleason 8–10 compared with GS 2–7
 Perineural invasion and prostate cancer d F. Y. FENG et al. e365

Table 3. Multivariate Cox proportional analysis of outcome for FFBF, FFM, and CSS

FFBF FFM CSS

Covariate p value HR (95% CI) p value HR (95% CI) p value HR (95% CI)

PSA
#20 Reference Reference Reference
>20 <0.0001 2.9 (2.0–4.4) <0.0001 4.0 (2.3–6.9) <0.002 3.2 (1.5–6.8)
Gleason score
2–6 Reference Reference Reference
3+4 <0.03 1.8 (1.1–3.1) <0.04 2.4 (1.1–5.7) >0.3 2.2 (0.4–12.3)
4+3 <0.002 2.5 (1.4–4.4) <0.007 3.5 (1.4–8.6) <0.02 7.4 (1.5–36.5)
8–10 <0.0001 4.3 (2.4–7.8) <0.0001 10.5 (4.2–26.3) <0.0001 23.9 (5.1–112)
PNI
No Reference Reference Reference
Yes <0.006 1.7 (1.2–2.4) <0.03 1.8 (1.1–3.1) <0.05 1.4 (1.1–3.5)
ADT use
No Reference Reference Reference
Yes >0.1 0.69 (0.45–1.1) <0.0008 0.32 (0.16–0.62) <0.05 0.42 (0.17–0.99)

Abbreviations: ADT = androgen deprivation therapy; CI = confidence interval; CSS = cause-specific survival; FFBH = freedom from bio-
chemical failure; FFM = freedom from metastasis; HR = hazard ratio; OS = overall survival; PNI = perineural invasion; PSA = prostate-
specific antigen; RT = radiation therapy.
Variables not prognostic in any model were T stage, percent positive cores, and the use of a pelvic RT fields and were therefore excluded.

(HR = 7.0, p <0.0001), whereas the presence of PNI was DISCUSSION

only prognostic in men with GS 8–10 (HR = 3.1, p <0.04).
The purpose of our study was to assess the prognostic value
Similar to the FFM model, the use of ADT only provided
of PNI in prostate cancer patients treated with dose-escalated
benefit in those with either GS 8–10 disease or with GS
EBRT. Previous studies on PNI were conducted before dose-
2–7 and a PSA >20 ng/mL. With small patient numbers
escalated RT became the standard of care and included a broad
and few events, the impact of ADT was of borderline signif-
range of RT doses (5–7). However, because dose-escalated
icance for the comparison across nodes IV and V as well as
EBRT improves local and biochemical control of prostate can-
VI and VII (see Fig. 3), but all nodes were presented for
cer (9), the impact of PNI might be different in this setting.
comparison.

GS 2-7 (n=522) GS 8-10 (n=128)

92% (90-94) Gleason Score 77% (72-82)
p<0.0001
3.6 (1.8-7.2)
<20
– (n=455) >20 (n=67)
94% (92-96) PSA 84% (74-94)
p<0.0005
3.1 (1.3-7.3)

GS 2-6 GS 7 (n=249) YES NO NO (n=55) YES (n=73)

(n=203) Gleason Score 92% (89-95) (n=44) ADT (n=23) 89% (84-94) PNI 67% (60-74)
p<0.01 p<0.009 p<0.02
3.3 (1.5-7.3) 0.26 (0.09-0.80) 3.3 (1.5-4.4)

I IV V
96% (94-98) 98% (96-99) 59% (48-60)

GS 3+4 GS 4+3 YES NO YES NO

(n=156) Gleason Score (n=93) (n=46) ADT (n=9) (n=69) ADT (n=4)
p<0.03 p<0.008 p<0.002
2.7 (1.1-6.9) 0.13 (0.01-1.4) 0.17 (0.01-2.2)

II III VI VII VIII IX

95% (92-98) 86% (81-91) 95% (92-98) 65% (48-82) 69% (62-76) 25% (4-46)

Fig. 2. Recursive partitioning model for freedom from metastasis. Values outside each node represent the number of
patients and the 7-year freedom from metastasis along with the standard error of the mean. Values inside each node
are the p values from log-rank test for the comparison at each node with corresponding hazard ratios and 95% confidence
intervals. p value for overall model <0.0001. ADT = androgen deprivation therapy; GS = Gleason score; PNI = perineural
invasion; PSA = prostate-specific antigen.
e366 I. J. Radiation Oncology d Biology d Physics Volume 81, Number 4, 2011

GS 2-7 (n=522) GS 8-10 (n=128)

97% (96-98) Gleason Score 79% (74-84)
p<0.0001
7.0 (2.6-19.0)
<20
– (n=455) >20 (n=67)
98% (97-99) PSA 90% (85-95)
p<0.002
4.7 (1.2-18.6)

GS 2-6 GS 7 (n=249) YES NO NO (n=55) YES (n=73)

(n=203) Gleason Score 97% (95-99) (n=44) ADT* (n=23) 91% (86-96) PNI 69% (61-77)
p<0.05 p<0.06 p<0.04
4.4 (1.0-22) 0.23 (0.05-1.1) 3.1 (1.2-8.0)

I IV V
99% (98-99.5) 100% (95-100) 76% (65-87)

GS 3+4 GS 4+3 YES NO YES NO

(n=156) Gleason Score (n=93) (n=46) ADT* (n=9) (n=69) ADT (n=4)
p<0.03 p<0.4 p<0.004
7.7 (1.2-50) 0.5 (0.07-4.2) 0.19 (0.02-2.1)

II III VI VII VIII IX

100% (95-100) 93% (89-97) 93% (88-98) 83% (68-98) 74% (66-82) 25% (3-47)

Fig. 3. Recursive partitioning model for cause-specific survival (CSS). Values outside each node represent the number of
patients and the 7-year CSS along with the standard error of the mean. Values inside each node are the p values are from
log-rank test for the comparison at each node with corresponding hazard ratios and 95% confidence intervals. p value for
overall model <0.0001. *ADT use did not predict for improvement in CSS between nodes IVand Vas well as nodes VI and
VII as it did for freedom from metastasis, but these nodes were retained for comparison. ADT = androgen deprivation
therapy; GS = Gleason score; PNI = perineural invasion; PSA = prostate-specific antigen.

The prevalence of PNI increased with increasing adverse
pathologic features and NCCN risk group. On univariate
analysis the presence of PNI was significantly associated
with worse FFBF, FFM, and CSS, and after controlling for
clinical- and treatment-related factors on multivariate analy-
sis, PNI also independently predicted for worse FFBF, FFM,
and CSS. Recursive partitioning analysis demonstrated that
PNI predicted worse FFM and CSS in patients with GS 8–10.
In this group, the presence of PNI was strongly associated
with worse FFM at 7 years (67% with PNI vs. 89% without
PNI, p <0.02) as well as worse CSS (69% with PNI vs. 91%
without PNI, p <0.04). The negative impact of PNI was ob-
served in GS 8–10 patients that were treated both with and
without ADT, suggesting that ADT did not completely over-
come the negative prognostic association of PNI. By RPA,
the presence of PNI did not influence either FFM or CSS
in those with Gleason 2–7 disease. This could in part be be-
cause of both the lower frequency of PNI in this group (28%
vs. 57% in those with Gleason 8–10) and also to the lower
rate of prostate cancer death (occurring in 3% of those
with Gleason 2–7 compared with 13% in those with Gleason
8–10). In fact, the impact of PNI on FFM and CSS by Cox-
proportional hazards regression including all Gleason scores
would indicate that PNI is prognostic in these patients with
Gleason 2–7 (although the absolute rates of clinical failure
were clearly lower, and as a result the absolute impact of
PNI would also be lower), which would be in accordance
with a previous study using conventional dose EBRT in
which PNI predicted worse CSS in men with low-
intermediate risk disease (6).
The strengths of this study include relatively uniform
treatment delivered over a 10-year period with a median
follow-up >5 years. The limitations include the retrospective
nature of the report and the small numbers of patients in
some of the RPA subgroups. This single-institutional experi-
ence would need to be replicated in a larger cohort before the
RPA model could be considered a legitimate prognostic tool.
In addition, recent reports have suggested that nomogram
analysis using a multivariate model might provide an alter-
native and perhaps superior means to evaluate the impact
of PNI on clinical outcome (11). Nevertheless, the RPA
analysis provides a unique manner to evaluate the interaction
between prognostic variables that is lost in the Cox propor-
tional hazards regression. In addition, the concordance of the
results using two methods of multivariate analysis (Cox
proportional hazards regression and RPA) lends further
credence to the prognostic significance of PNI.
Therefore, PNI may represent a biomarker indicating the
need for further treatment intensification beyond dose-
escalated EBRT and conventional ADT. Such intensification
could be achieved with combined EBRT, ADT, and brachy-
therapy (12, 13) or through the addition of chemotherapy to
EBRT and ADT (14). It is possible that PNI has a more ad-
verse impact for patients treated with conventional-dose RT,
which might explain its association with CSS in a previous
study of low- to intermediate-risk patients, an association
 Perineural invasion and prostate cancer d F. Y. FENG et al.
that we did not observe here (6). In addition, at least when
measured in regard to BF, there does not appear to be
a negative impact for lower-risk patients treated with brachy-
therapy (13). Chemotherapy, although clearly providing
a benefit in men with castrate-resistant prostate cancer, has
not at present been demonstrated to provide a benefit when
used adjuvantly after EBRT and ADT (14), although results
from randomized studies with docetaxel are still pending.
Interestingly, both the multivariate models and the RPA
identified worse outcomes for both FFM and CSS in patients
with Gleason score 4+3 compared with Gleason score 3+4.
The influence of primary Gleason pattern 4 is an area that
has been extensively studied previously in patients treated
with radical prostatectomy (15, 16), external-beam radiation
(17), and prostate brachytherapy (18, 19). Given the close
association between Gleason 4+3 on biopsy and later
upstaging to higher Gleason score and/or pathologic stage at
time of RP, it is possible that patients with intermediate-risk
prostate cancer with primary Gleason pattern 4 represent
a more aggressive clinical phenotype and should be treated
REFERENCES
1. Bahnson RR, Hanks GE, Huben RP, et al. NCCN Practice
Guidelines for Prostate Cancer. Oncology (Williston Park)
2000;14:111–119.
2. Kwan W, Pickles T, Duncan G, et al. PSA failure and the risk of
death in prostate cancer patients treated with radiotherapy. Int J
Radiat Oncol Biol Phys 2004;60:1040–1046.
3. Liebig C, Ayala G, Wilks JA, et al. Perineural invasion in can-
cer. Cancer 2009;115:3379–3391.
4. Lee IH, Roberts R, Shah RB, et al. Perineural invasion is
a marker for pathologically advanced disease in localized pros-
tate cancer. Int J Radiat Oncol Biol Phys 2007;68:1059–1064.
5. Bonin SR, Hanlon AL, Lee WR, et al. Evidence of increased
failure in the treatment of prostate carcinoma patients who
have perineural invasion treated with three-dimensional con-
formal radiation therapy. Cancer 1997;79:75–80.
6. Beard C, Schultz D, Loffredo M, et al. Perineural invasion asso-
ciated with increased cancer-specific mortality after external
beam radiation therapy for men with low- and intermediate-risk
prostate cancer. Int J Radiat Oncol Biol Phys 2006;66:403–407.
7. Beard CJ, Chen MH, Cote K, et al. Perineural invasion is asso-
ciated with increased relapse after external beam radiotherapy
for men with low-risk prostate cancer and may be a marker for
occult, high-grade cancer. Int J Radiat Oncol Biol Phys 2004;
58:19–24.
8. Wong WW, Schild SE, Vora SA, et al. Association of percent pos-
itive prostate biopsies and perineural invasion with biochemical
outcome after external beam radiotherapy for localized prostate
cancer. Int J Radiat Oncol Biol Phys 2004;60:24–29.
9. Kuban DA, Levy LB, Cheung MR, et al. Long-term failure pat-
terns and survival in a randomized dose-escalation trial for
prostate cancer. Who dies of disease? Int J Radiat Oncol Biol
Phys 2011;79:1310–1317.
10. Roach M 3rd, Hanks G, Thames H Jr., et al. Defining biochem-
ical failure following radiotherapy with or without hormonal
therapy in men with clinically localized prostate cancer:
Recommendations of the RTOG-ASTRO Phoenix Consensus
Conference. Int J Radiat Oncol Biol Phys 2006;65:965–974.
e367
as such (15, 16). We treated those Gleason 7 patients with
primary or secondary Gleason pattern 4 similarly in regard
to RT dose, ADT use, ADT duration, and pelvic RT (Table 1
and Table E2). However, the patients with Gleason score
4+3 disease and PSA <20 ng/mL (RPA node III) had worse
outcome for both FFM and CSS compared with those with
Gleason 3+4 and a PSA <20 ng/mL (RPA node II).
Therefore, it is possible that outcomes in these subgroups
would be improved through more uniform use of ADT and/
or pelvic RT.
CONCLUSION
The presence of PNI in the prostate biopsy holds indepen-
dent prognostic utility in predicting clinical outcome for pa-
tients treated with dose-escalated EBRT with or without
ADT. Particularly in patients with GS 8–10 prostate cancer,
the presence of PNI suggests an increased risk of metastasis
and prostate cancer death, which may warrant more aggres-
sive treatment.
11. Shariat SF, Karakiewicz PI, Suardi N, et al. Comparison of
nomograms with other methods for predicting outcomes in
prostate cancer: A critical analysis of the literature. Clin
Cancer Res 2008;14:4400–4407.
12. Stone NN, Potters L, Davis BJ, et al. Multicenter analysis of
effect of high biologic effective dose on biochemical failure
and survival outcomes in patients with Gleason score 7–10
prostate cancer treated with permanent prostate brachytherapy.
Int J Radiat Oncol Biol Phys 2009;73:341–346.
13. Merrick GS, Butler WM, Galbreath RW, et al. Perineural inva-
sion is not predictive of biochemical outcome following pros-
tate brachytherapy. Cancer J 2002;8:79–80.
14. Sandler HM, Hunt D, Sartor AO, et al. A Phase III protocol of
androgen suppression (AS) and radiation therapy (RT) versus
AS and RT followed by chemotherapy with paclitaxel, estra-
mustine, and etoposide (TEE) for localized, high-risk, prostate
cancer, RTOG 9902. The 2010 American Society of Clinical
Oncology Annual Meeting Chicago, IL, 2010.
15. Chan TY, Partin AW, Walsh PC, et al. Prognostic significance
of Gleason score 3+4 versus Gleason score 4+3 tumor at radical
prostatectomy. Urology 2000;56:823–827.
16. Sakr WA, Tefilli MV, Grignon DJ, et al. Gleason score 7 pros-
tate cancer: A heterogeneous entity? Correlation with patho-
logic parameters and disease-free survival. Urology 2000;56:
730–734.
17. Anderson PR, Hanlon AL, Horwitz E, et al. Outcome and
predictive factors for patients with Gleason score 7 prostate
carcinoma treated with three-dimensional conformal external
beam radiation therapy. Cancer 2000;89:2565–2569.
18. Potters L, Purrazzella R, Brustein S, et al. The prognostic sig-
nificance of Gleason Grade in patients treated with permanent
prostate brachytherapy. Int J Radiat Oncol Biol Phys 2003;56:
749–754.
19. Merrick GS, Butler WM, Wallner KE, et al. The impact of
primary Gleason grade on biochemical outcome following
brachytherapy for hormone-naive Gleason score 7 prostate
cancer. Cancer J 2005;11:234–240.