ADULT UROLOGY

IMPACT OF EXTENT OF LYMPHADENECTOMY ON

SURVIVAL AFTER RADICAL PROSTATECTOMY FOR
PROSTATE CANCER
SUE A. JOSLYN AND BADRINATH R. KONETY

ABSTRACT
Objectives. Controversy exists regarding the benefit of extended lymphadenectomy at radical prostatec-
tomy for prostate cancer. We sought to determine whether more extended lymphadenectomy, along with
radical prostatectomy, resulted in a decreased risk of prostate cancer-specific death at 10 years.
Methods. Data on all patients undergoing radical prostatectomy (with or without lymphadenectomy) for
prostate cancer obtained from the Surveillance, Epidemiology, and End Results Program (1988 to 1991)
were examined. All surviving patients had a minimal follow-up of 10 years. Multivariate Cox proportional
hazards analysis was used to determine the independent effect of lymphadenectomy on the risk of prostate
cancer-specific death.
Results. Patients undergoing excision of at least 4 lymph nodes (node-positive and node-negative patients)
or more than 10 nodes (only node-negative patients) had a lower risk of prostate cancer-specific death at 10
years than did those who did not undergo lymphadenectomy. The removal of a greater number of nodes was
associated with a greater likelihood of the presence of positive nodes. The presence of more than one
positive node was associated with a greater risk of prostate cancer-related death.
Conclusions. Performing more extensive pelvic lymphadenectomy in patients undergoing radical prosta-
tectomy could improve the accuracy of staging and reduce the risk of prostate cancer-specific death in the
long term. UROLOGY 68: 121–125, 2006. © 2006 Elsevier Inc.

S tudies of other cancers have demonstrated a

distinct therapeutic value of performing a thor-
ough lymphadenectomy.1,2 The possibility of ther-
greater risk of nodal metastases, more extensive
nodal dissection may be imperative.4 – 6 Most re-
cent studies addressing this issue have been single-
apeutic benefit for pelvic lymphadenectomy in institution series with modest numbers of patients
prostate cancer has been suggested by some stud- and relatively short follow-up times. We sought to
ies, but the results have been inconsistent.3– 6 The determine whether the extent of pelvic lymphade-
use of routine lymphadenectomy at radical prosta- nectomy at radical prostatectomy had any effect on
tectomy has declined considerably during the past subsequent survival.
decade for two major reasons: (a) a shift toward an
earlier stage in newly diagnosed cases of prostate MATERIAL AND METHODS
cancer7; and (b) the ability to better predict a low
likelihood of node-positive disease in many pa- SUBJECTS
tients with prostate cancer using staging tools such All men diagnosed with prostate cancer and treated with
as the Partin tables.8 However, in patients with a radical prostatectomy from 1988 to 1991 were identified from
the National Cancer Institute’s Surveillance, Epidemiology,
and End Results (SEER) Program (SEER November 2003
From the Division of Health Promotion and Education, University Public-Use Database) were included as subjects for this study.
of Northern Iowa, Cedar Falls, Iowa; and Departments of Urol- Detailed staging information has been recorded in the SEER
ogy and Epidemiology and Biostatistics, University of California database using the American Joint Committee on Cancer stag-
San Francisco, School of Medicine, San Francisco, California ing system9 only since 1988, thereby limiting the current
Reprint requests: Badrinath R. Konety, M.D., M.B.A., Depart- study to cases diagnosed beginning in 1988.
ment of Urology, University of California, San Francisco, School
of Medicine, 1600 Divisadero, A624, Box 1695, San Francisco,
CA 94143-1695. E-mail: [email protected] DEFINITIONS OF PROSTATE CANCER VARIABLES
Submitted: October 21, 2005, accepted (with revisions): Janu- The variables analyzed included the number of lymph nodes
ary 17, 2006 examined, number of positive lymph nodes, ratio of positive

© 2006 ELSEVIER INC. 0090-4295/06/$32.00

ALL RIGHTS RESERVED doi:10.1016/j.urology.2006.01.055 121
 TABLE I. Conversion table of categories for SEER grade code
SEER Code Gleason Score Gleason Pattern Histologic Grade Terminology
1 2, 3, 4 1, 2 I Well differentiated
2 5, 6 3 II Moderately differentiated
3 7, 8, 9, 10 4, 5 III Poorly differentiated
IV Undifferentiated/anaplastic
KEY: SEER ⫽ Surveillance, Epidemiology, and End Results.

nodes to number of nodes examined, use of radiotherapy, stage at decreased the sample, and the relevant sample size
diagnosis, age category, race, vital status (alive, dead), cause of analyzed for variables is included in the tables as
death, and histologic grade (modified Gleason score).
The number of lymph nodes removed, number of positive appropriate. The median survival time was 127
lymph nodes, and ratio of positive nodes to number of nodes months (range 0 to 167). The median age was 67
removed were grouped into discrete categories for the multi- years (range 36 to 90).
variate analyses. The categories for the number of lymph More than 85% of patients for whom radical
nodes removed were 0, 1 to 3, 4 to 6, 7 to 9, and 10⫹. The prostatectomy was the surgical treatment were
categories for the number of positive nodes included zero,
one, or more than one. The categories for the ratio of positive aged between 55 and 74 years (Table II), and more
nodes to the number of nodes removed included 0.01 to 0.25, than 90% were white. Most tumors were diagnosed
0.26 to 0.50, 0.51 to 0.75, and 0.76 to 1.00. at the localized stage and were SEER grade 2. More
The stage at diagnosis was categorized as localized, regional than 90% of patients did not receive any additional
by direct extension only, regional involving only lymph radiotherapy. Approximately 30% of patients did
nodes, regional by direct extension and lymph nodes, distant
sites/nodes involved, and unknown/unstaged/unspecified/death not have lymph nodes examined, and of those that
certificate only. did, most had no positive nodes. Almost 68% of the
The age at diagnosis was stratified by 10-year age categories. patients diagnosed between 1988 and 1991 were
Race was categorized as white and African American. Other still alive in 2001. Of those that had died, most had
race categories were not analyzed, because of the low numbers died of cancer and heart disease.
of subjects in some categories.
Histologic grade was coded using the following priority or- When controlling for age, race, stage, grade, and
der: (a) Gleason’s grade, (b) terminology of differentiation, radiotherapy, the risk of cancer-specific death was
and (c) histologic grade (I, II, III, IV). Table I presents a con- significantly lower for patients who had more than
version table of the categories for the SEER grade code using four lymph nodes removed (Table III). Although
the above categories. not statistically significant, the risk of death was
lower for all patients who had had any lymph
STATISTICAL ANALYSIS nodes removed than for those who had had no
The data were analyzed using Statistical Analysis Systems,
version 8.01, statistical software (SAS Institute, Cary, NC).
lymph nodes removed. The extent of lymphadenec-
A t test was used to determine whether an increased number of tomy did not affect all-cause mortality (Table III).
lymph nodes removed was associated with the discovery of The number of positive nodes and the ratio of pos-
lymph node metastasis. Multivariate Cox regression analyses itive to the total number of nodes removed were
were used to determine the factors significantly associated not significant predictors of the risk of death. More
with survival. Survival analyses included death from any can-
cer as the outcome of interest (failure), with all other causes of
extensive lymphadenectomy (10⫹ nodes re-
death and survivors categorized as censored. In survival anal- moved) was associated with a lower risk of prostate
yses with the Cox regression technique, it is preferable to use cancer death, even after restricting the analysis to
all competing causes of death (in this case, all cancer deaths) patients with negative lymph nodes (Table IV). Pa-
as the failure outcome for the most conservative explanation tients with lymph node involvement had a signifi-
of the results, because death from any cancer may not be an
independent event in a patient with previous prostate carci-
cantly greater number of nodes removed compared
noma. A survival time variable was calculated for each case by with those with negative nodes (13 nodes versus 9
determining the number of months between the date of diag- nodes per case, P ⬍0.0001).
nosis and the most recent follow-up visit. Subjects with un-
known survival status or who had died but for whom the
cancer status was undetermined were categorized as missing COMMENT
survival data.
The key findings of this analysis are that patients
with lymph node involvement had a significantly
RESULTS
greater number of nodes removed compared with
The SEER database included 57,764 men diag- those with no lymph node involvement and that
nosed with primary prostate cancer from 1988 to extensive lymphadenectomy reduces the long-
1991. Of the total number of patients, 13,020 term risk of prostate cancer-related death, even in
(22.5%) underwent radical prostatectomy as their patients with negative nodes compared with pa-
initial therapy. Missing data on certain variables tients without lymphadenectomy.

122 UROLOGY 68 (1), 2006

 TABLE II. Descriptive characteristics of TABLE II. Continued
prostate cancer cases diagnosed in SEER Variable n %
geographic regions between 1988 and 1991 †
Cause of death
and treated with radical prostatectomy Alive 8,840
Variable n % Cancer 1,846 44.16
Age category (yr) Prostate 884 21.15
⬍54 686 5.28 Bladder 47 1.12
55–64 4,111 31.57 Lung 303 7.25
65–74 7,240 55.61 Digestive 283 6.77
75–84 962 7.39 Heart disease 967 23.13
ⱖ85 21 0.16 Stroke 234 5.60
Race category Pneumonia/influenza 90 2.15
White 11,909 94.35 COPD 129 3.09
African American 713 5.65 Diabetes 63 1.51
Other or unknown 398 Accidents 72 1.72
Stage at diagnosis All other causes 779 18.64
Localized 8,854 68.02 KEY: SEER ⫽ Surveillance, Epidemiology, and End Results; COPD ⫽ chronic ob-
Regional by direct extension structive pulmonary disease.
* Differences in numbers due to missing values.
only 2,774 21.31 †
Percentage given as percentage of total deaths.
Regional involving only lymph
nodes 223 1.71
Regional by both direct
TABLE III. Cox regression analysis of effects
extension and lymph nodes 478 3.67
of number of nodes examined on risk of death
Distant site/node involved 80 0.61
Unknown/unstaged/unspecified/
(hazard ratio) in all patients (controlling for
death certificate only 611 4.67 age, race, stage, grade, and
Histologic grade (modified radiotherapy use)
Gleason score) Hazard Ratio
1 2,050 16.21 Mortality (95% CI) P Value
2 8,132 64.32 Cancer specific
3 2,393 18.93 Nodes examined (n)
4 69 0.55 None 1.00 (Ref.)
Missing 376 1–3 0.85 (0.68–1.06) 0.1580
Radiotherapy 4–6 0.77 (0.64–0.93) 0.0069*
No 11,740 90.67 7–9 0.82 (0.67–0.99) 0.0390*
Yes 1,208 9.33 ⱖ10 0.81 (0.70–0.94) 0.0047*
Missing 72 All causes
No. of lymph nodes examined Nodes examined (n)
None 2,666 29.04 None 1.00 (Ref.)
1–3 774 8.43 1–3 0.964 (0.804–1.155) 0.6879
4–6 1,273 13.86 4–6 0.854 (0.730–1.000) 0.0496
7–9 1,165 12.69 7–9 0.914 (0.781–1.070) 0.2657
ⱖ10 3,304 35.98 ⱖ10 0.892 (0.786–1.013) 0.0777
Missing 3,838
KEY: CI ⫽ confidence interval; Ref. ⫽ reference.
No. of positive nodes* * Statistically significant.
0 6,807 91.21
1 409 5.48
⬎1 247 3.31
No nodes examined 2,666
The second finding was not surprising and has
Missing 2,891 been observed in other genitourinary cancers by us
Ratio of positive nodes to nodes and others.1,8,10 One potential explanation for this
examined* observation is that thorough nodal resection elim-
0.01–0.25 431 4.26 inated micrometastases that were not detected by
0.26–0.50 86 1.32 routine histologic examination. Therefore, would a
0.51–0.75 22 0.34 more careful histologic examination have improved
0.76–1.00 5 0.08 the chances of detecting these micrometastases?
Missing, no positive nodes, no Wawroschek et al.11 performed serial sections and
nodes examined 12,476 step sections with immunohistochemistry on speci-
Vital status
mens from limited and extensive pelvic lymphad-
Alive 8,840 67.90
Dead 4,180 32.10
enectomies for prostate cancer. They observed a
14% increase in the number of metastases detected

UROLOGY 68 (1), 2006 123

 a standard lymphadenectomy under current prac-
TABLE IV. Cox regression analysis of effects
tice. The lack of preoperative PSA data (recorded
of number of nodes examined on risk of
only since 2001 in the SEER database) prevented
prostate cancer death only in patients with
us from verifying if this was indeed the case.
negative nodes (controlling for age, race,
Hence, on the basis of our and previous data, it
stage, grade, and radiotherapy use)
appears that patients with intermediate or high-
Lymph Nodes risk features (PSA greater than 10 ng/mL, Gleason
Examined (n) HR (95% CI) P Value
score greater than 7) would benefit the most from
None 1.00 (Ref.) extended lymphadenectomy.
1–3 0.96 (0.76–1.21) 0.7373 The actual definition of extended pelvic lymph-
4–6 0.86 (0.70–1.05) 0.1321
adenectomy for prostate cancer is variable. As de-
7–9 0.87 (0.71–1.07) 0.1957
ⱖ10 0.85 (0.72–0.99) 0.0382*
termined by cadaveric and human studies, the
maximal average number of nodes obtained by
KEY: HR ⫽ hazard ratio (relative risk of death); CI ⫽ confidence interval; Ref. ⫽
reference. extensive pelvic dissection is 22.14 Our results sug-
gest that any survival benefit of lymphadenectomy
in those with negative nodes can only be realized in
using the same step section technique in the lim- those undergoing extensive dissection (more than
ited dissections and a 27% increase in metastases 10 nodes).
detected with more extended lymphadenectomy. The only randomized study on extended versus
These data underscore the importance of an exten- limited pelvic lymphadenectomy in patients with
sive lymphadenectomy, regardless of the method prostate cancer was reported by Clark et al.,15 who
of histologic examination. found no difference in the yield of positive nodes
More extensive lymphadenectomy is more likely with more extended dissection. The two types of
to yield more accurate staging information. We ob- dissection were conducted in the same patient on
served a direct correlation between the number of contralateral sides and complications were as-
lymph nodes removed and the presence of lymph signed on the basis of the side of dissection. The
node metastases (Pearson’s r ⫽ 0.10; P ⬍0.0001). study design rendered it hard to compare these
Allaf et al.12 compared the results of limited and data with those from the present study or other
extended lymphadenectomy each performed by previous studies.
two separate surgeons and discovered a nodal pos- Several studies have indicated the probability of
itivity rate of 1.2% among those undergoing limited long-term survival even with the presence of lim-
dissections compared with a 3.3% nodal positivity ited lymph node metastases after radical prostatec-
rate among those undergoing extensive dissection. tomy with or without subsequent androgen de-
Although these data suggest the benefit of ex- privation therapy.16 –18 Our data have further
tended pelvic lymphadenectomy at prostatectomy, substantiated these earlier reports. We did observe
it would be hard to advocate that approach in all that those with more than one positive node had a
patients, particularly those with low risk of lymph significantly greater risk of cancer-related death
node involvement. Currently, it is common prac- (relative risk 6.232, 95% confidence interval 1.074
tice to entirely eliminate lymph node dissection in to 36.170, P ⫽ 0.04) compared with those with one
patients deemed at very low risk of lymph node positive node (relative risk 4.011, 95% confidence
metastases.8 However, the area under the curve of interval 0.670 to 24.029, P ⫽ 0.13) or no positive
the Partin tables for predicting lymph node metas- nodes (reference category). Comparisons among the
tases is 0.79 to 0.82, which suggests that such risk node-positive patients alone were precluded by the
assignment may be inaccurate in 18% to 21% of pa- small sample sizes (Table II). However, after con-
tients.8,13 Patients with a greater likelihood of lymph trolling for patient and disease-related factors, we
node metastases may be most likely to benefit from were unable to demonstrate any differences in sur-
extended lymphadenectomy. Heidenreich et al.6 de- vival on the basis of the proportion of resected
fined a high-risk group with a serum prostate-spe- nodes that were involved, probably because of the
cific antigen (PSA) greater than 10.5 ng/mL and overpowering effect of the disease stage.
Gleason score greater than 7 and found that 92% of Our results are in contrast to the single-institu-
patients in this group had positive nodes detected tional data analysis of DiMarco et al.,3 who found
by extensive lymphadenectomy and only 2.8% no benefit to extended pelvic lymphadenectomy in
without these criteria demonstrated positive patients with node-negative disease. Our results
lymph nodes. It is likely that the 15% risk reduc- indicated a significant risk reduction between
tion in the risk of death observed even in patients those with no nodes removed and those with 10 or
with negative nodes in our data set was mainly more nodes removed. In the study by DiMarco
attributable to those individuals with a high PSA et al.,3 no patients had no nodes removed, and
level at diagnosis who would have merited at least DiMarco et al. only examined the incremental

124 UROLOGY 68 (1), 2006

changes in the relative risk with each additional REFERENCES
node removed. This may partly explain the lack of 1. Mathiesen O, Carl J, Bonderup O, et al: Axillary sam-
significant differences between the groups. Be- pling and the risk of erroneous staging of breast cancer: an
analysis of 960 consecutive patients. Acta Oncol 29: 721–725,
cause our analyses and those of others have indi- 1990.
cated an increased likelihood of detecting positive 2. Konety BR, Joslyn SA, and O’Donnell MA: Extent of
nodes with more extensive dissection, a clear value pelvic lymphadenectomy and its impact on outcome in pa-
exists to more precise staging, particularly in those tients diagnosed with bladder cancer: analysis of data from the
with high-risk disease. The analysis by DiMarco Surveillance, Epidemiology, and End Results Program data-
base. J Urol 169: 946 –950, 2003.
et al.3 did not address this issue. On the basis of 3. DiMarco DS, Zincke H, Sebo TJ, et al: The extent of
previous studies, prostatectomy in the presence of lymphadenectomy for pTXN0 prostate cancer does not affect
limited microscopic nodal metastases can yield prostate cancer outcome in the prostate specific antigen era.
long-term survival in an appreciable segment of J Urol 173: 1121–1125, 2005.
patients. To more accurately determine the volume 4. Bader P, Burkhard FC, Markwalder R, et al: Is a limited
lymph node dissection an adequate staging procedure for
of nodal metastases, it may be necessary to perform prostate cancer? J Urol 168: 514 –518, 2002.
a more thorough dissection. 5. Bader P, Burkhard FC, Markwalder R, et al: Disease
One of the limitations posed by using the SEER progression and survival of patients with positive lymph
data was the lack of information regarding adju- nodes after radical prostatectomy: is there a chance of cure?
vant hormonal therapy. In addition, the SEER da- J Urol 169: 849 – 854, 2003.
6. Heidenreich A, Varga Z, and Von Knobloch R: Ex-
tabase does not have information regarding other tended pelvic lymphadenectomy in patients undergoing radi-
patient characteristics such as comorbidity and cal prostatectomy: high incidence of lymph node metastasis.
performance status, known to be a key variable J Urol 167: 1681–1686, 2002.
influencing survival and type of adjuvant therapy. 7. Miller DC, Hafez KS, Stewart A, et al: Prostate carci-
We were unable to control for margin status (al- noma presentation, diagnosis, and staging: an update from the
National Cancer Data Base. Cancer 98: 1169 –1178, 2003.
though it is incorporated into the staging system) 8. Partin AW, Kattan MW, Subong EN, et al: Combination
in this analysis. The lack of PSA data in the current of prostate-specific antigen, clinical stage, and Gleason score
data set could have been overcome by using more to predict pathological stage of localized prostate cancer: a
recent SEER data sets (later than 2001), which multi-institutional update. JAMA 277: 1445–1451, 1997.
have PSA information, but the follow-up time 9. American Joint Committee on Cancer: AJCC Cancer
Staging Handbook, 5th ed. Philadelphia, Lippincott-Raven,
would have been too short, precluding meaningful 1998, pp 205–206.
survival analysis. In our sample of 13,020 men 10. Herr HW, Bochner BH, Dalbagni G, et al: Impact of the
treated with radical prostatectomy, the data re- number of lymph nodes retrieved on outcome in patients with
garding the number of lymph nodes examined was muscle invasive bladder cancer. J Urol 167: 1295–1298, 2002.
missing in 3838 men (30%). However, the remain- 11. Wawroschek F, Wagner T, Hamm M, et al: The influence
of serial sections, immunohistochemistry, and extension of pel-
ing sample of more than 9000 was still of sufficient vic lymph node dissection on the lymph node status in clinically
power to detect significant effects for the number of localized prostate cancer. Eur Urol 43: 132–137, 2003.
nodes examined on survival, even while controlling 12. Allaf ME, Palapattu GS, Trock BJ, et al: Anatomical
for age, race, stage, grade, and radiotherapy use. extent of lymph node dissection: impact on men with clini-
However, one of the advantages of using the cally localized prostate cancer. J Urol 172: 1840 –1844, 2004.
13. Augustin H, Eggert T, Wenske S, et al: Comparison of
SEER database is that it is composed of accurate accuracy between the Partin tables of 1997 and 2001 to predict
and complete information on all prostate cancer final pathological stage in clinically localized prostate cancer.
cases and not just a sample. This allows increased J Urol 171: 177–181, 2004.
generalizability of results and decreases selection 14. Weingartner K, Ramaswamy A, Bittinger A, et al: Ana-
bias resulting from sampling. tomical basis for pelvic lymphadenectomy in prostate cancer:
results of an autopsy study and implications for the clinic.
J Urol 156: 1969 –1971, 1996.
CONCLUSIONS 15. Clark T, Parekh DJ, Cookson MS, et al: Randomized
prospective evaluation of extended versus limited lymph node
A greater number of lymph nodes removed and dissection in patients with clinically localized prostate cancer.
examined at prostatectomy for prostate cancer ap- J Urol 169: 145–148, 2003.
16. Daneshmand S, Quek ML, Stein JP, et al: Prognosis of
peared to increase the likelihood of finding lymph patients with lymph node positive prostate cancer following
node metastases and increase prostate cancer-spe- radical prostatectomy: long-term results. J Urol 172: 2252–
cific survival even in patients who have histologi- 2255, 2004.
cally uninvolved lymph nodes. Patients with more 17. Cheng L, Zincke H, Blute ML, et al: Risk of prostate
than one positive node had a significantly greater carcinoma death in patients with lymph node metastasis. Can-
cer 91: 66 –73, 2001.
risk of prostate cancer-related death. This survival 18. Pound CR, Partin AW, Eisenberger MA, et al: Natural
benefit may result from more accurate staging and history of progression after PSA elevation following radical
possible removal of occult metastases. prostatectomy. JAMA 281: 1591–1597, 1999.

UROLOGY 68 (1), 2006 125