BJUIB J U I N T E R N A T I O N A L

© 2 0 1 2 T H E A U T H O R S

B J U I N T E R N A T I O N A L © 2 0 1 2 B J U I N T E R N A T I O N A L | doi:10.1111/j.1464-410X.2012.11115.x 1

What ’ s known on the subject? and What does the study add? Prognostic factors such as serum PSA, tumor T stage, and Gleason grading are commonly used to predict disease progression and mortality in prostate cancer and to guide treatment decision-making. These markers are combined to defi ne risk strata that are commonly accepted in practice. Despite the assignment of patients to a specifi c risk stratum (e.g. intermediate-risk disease), however, within-stratum survival duration varies considerably, suggesting that many other factors, including lymphovascular invasion (LVI) may infl uence prognosis. LVI is currently a recognized prognostic factor in the management of some cancers (e.g. in early-stage breast cancer) and prostate cancer is known to spread via lymphatic channels. Furthermore, the reporting of microscopic lymphovascular invasion is now considered part of the standard pathologic report of prostatectomy specimens. Nevertheless, scientifi c studies in this area have produced confl icting conclusions regarding the utility of LVI as a prognostic indicator in prostate cancer. This paper provides a comprehensive review and synthesis of the recent literature.

Although a number of studies examining the role of LVI as an independent prognostic factor for biochemical recurrence in prostate cancer have been reported, the characteristics, quality and results of these studies vary considerably. The value of using LVI as a prognostic factor in prostate cancer remains unclear. This study provides a systematically-performed synthesis of the results of recent research including lymphovascular invasion (LVI) in the multivariate analyses of potential prostate cancer prognostic factors. Not only do we report on the results of these studies, we assess the heterogeneity of the study populations, disease characteristics, and quality of the studies. Ultimately, we determined that meta-analysis of the existing data is not possible, and thus, there is no ‘ best estimate ’ of the strength of association between LVI status and disease recurrence after prostatectomy. Most studies, but not all, reveal a weak or statistically insignifi cant association between LVI status and recurrence. We therefore conclude with a recommendation to clinicians that they should not overweight the importance of LVI status on clinical prognostication. The use of LVI status as a strong predictor of clinical outcomes is not recommended.

OBJECTIVES

• To synthesize the results of studies including lymphovascular invasion (LVI) in the multivariate analyses of potential prostate cancer prognostic factors. • To determine the role of LVI as an independent prognostic factor for biochemical recurrence in prostate cancer.

PATIENTS AND METHODS

• We performed a comprehensive systematic literature review of studies examining the association between LVI in prostatectomy specimens and prostate cancer recurrence. • Ovid MEDLINE, Embase, Web of Knowledge, Cochrane Database of Systematic Reviews, Database of Abstracts of Review of Effects (DARE) and Google Scholar were searched from January 2000 to February 2009. • The primary outcome of interest was biochemical recurrence measured by serum prostate specifi c antigen (PSA).

RESULTS

• One thousand two hundred and forty-eight papers met our search criteria. Of these, 19 articles meeting our selection criteria reported results of a multivariate analysis to evaluate LVI as an independent prognostic factor of biochemical recurrence. • Eleven (58%) of these studies concluded that LVI was an independent prognostic factor. • Signifi cant heterogeneity in the study population, disease characteristics and quality of the studies prevented meta-analysis of the results.

• In the nine studies in which the magnitude of independent association of LVI with recurrence was reported, it ranged from an odds ratio or relative risk of 1.37 to 4.39.

CONCLUSIONS

• The existing literature is confl icting and of insuffi cient homogeneity to defi nitively establish LVI as an important independent prognostic factor of biochemical recurrence in prostate cancer prostatectomy specimens.

• Additional adequately powered studies are required to determine the clinical value of reports of LVI involvement. • In the meantime, the use of LVI status as an independent prognostic factor for clinical prognostication and medical decision making is not recommended.

KEYWORDS

prognosis , lymphovascular invasion , prostatic neoplasm , multivariate analysis , prostatectomy

Study Type – Prognosis (systematic review) Level of Evidence 1a

Prognostic signifi cance of lymphovascular invasion in radical prostatectomy specimens Jonathan Ng , Aamer Mahmud * , Brenda Bass † and Michael Brundage ‡ Department of Family Medicine , * Department of Oncology , † Division of Cancer Care & Epidemiology and ‡ Departments of Oncology and Community Health & Epidemiology, Queen ’ s University, Kingston, Ontario, ON, Canada Accepted for publication 20 December 2011

N G E T A L .

© 2 0 1 2 T H E A U T H O R S

2 B J U I N T E R N A T I O N A L © 2 0 1 2 B J U I N T E R N A T I O N A L

INTRODUCTION

Serum PSA, tumour T stage and Gleason grading are commonly used prognostic factors that predict disease progression and mortality in prostate cancer and that guide treatment decision-making [ 1 ] . In the clinical context of early-stage prostate cancer, these markers are combined to defi ne risk strata that are commonly accepted in practice [ 2 ] .However, despite the assignment of patients to a specifi c risk stratum (e.g. intermediate-risk disease), within-stratum survival duration varies considerably, suggesting that other tumour, patient or treatment factors infl uence prognosis. Improved understanding of the factors that affect prognosis would potentially allow for better treatment selection and reduced variation in practice [ 1 ] .

Lymphovascular invasion (LVI) is a recognized prognostic factor in cancer management (e.g. in early-stage breast cancer) [ 3 ] . Prostate cancer is known to spread via lymphatic channels, and the reporting of microscopic lymphovascular invasion is now considered part of the standard pathologic report of prostatectomy specimens [ 4 ] . Nevertheless, the utility of LVI remains unproven as a prognostic indicator in prostate cancer, as scientifi c studies in this area to date have produced confl icting conclusions.

Systematic searches of the literature as it relates to the prognostic potential of LVI in prostate cancer are problematic. For example, literature searches specifi c to LVI as a keyword or subject heading tend to omit papers in which LVI has been included in a multivariate analysis that was intended to assess primarily other potential prognostic factors. Therefore, search terms that encompass a broad range of potential prognostic factors in prostate cancer are required to capture all relevant studies. In 2007, Harnden et al . [ 5 ] reported an exemplary systematic approach to investigating the prognostic value of a widely used histopathological factor [ perineural invasion (PNI) ] in prostatic cancer biopsies. As these authors point out, systematic searches may reveal heterogeneity in study quality, and sources of variability, bias and error that other approaches miss (e.g. consensus of expert opinion). It is this lack of homogeneity that limits synthesis of the available evidence [ 5 ] .

As with the literature addressing PNI, there is signifi cant heterogeneity and complexity in the available LVI literature relating to prostate cancer that makes comparing the conclusions of these studies diffi cult. Variability in study size, in the characteristics of the study populations, in defi nitions used for LVI status and for disease progression, as well as the repeated reporting of overlapping patient cohorts, all complicate the interpretation of the literature. The variability across studies is perhaps not surprising. Strict defi nitions for biochemical recurrence, for example, are critical for identifying patients at risk for progression and for conducting meaningful comparisons of treatment effi cacy. However, it was not until 2007 that the American Urological Association [ 6 ] published a recommendation for a standard defi nition of biochemical recurrence in patients treated for localized prostate cancer.

As noted by McShane et al . [ 7 ] in their paper addressing reporting recommendations for tumour marker prognostic studies, it is important to understand why multiple studies of the same marker lead to differing conclusions. The reporting recommendations developed by McShane et al . [ 7 ] facilitate evaluation of the appropriateness and quality of study design, methods and analysis, ultimately improving our ability to compare results across studies. Accordingly, we conducted a systematic review of the literature to summarize knowledge regarding LVI as a prognostic factor in early-stage prostate cancer with a view to resolving the current uncertainty around this issue, while remaining cognizant of the quality of the reporting of the studies included in the analysis.

We sought to focus the review on relatively recently reported publications (published in year 2000 or after) in order to take advantage of recent developments designed to improve consistency among studies. For example, in 1994 the Cancer Committee of the American College of Pathologists [ 8 ] proposed that prostatic microvascular invasion be reported by pathologists because of its potential predictive value for pathologic stage.Also, a recent conference of the International Society of Urological Pathology (ISUP) [ 4 ] on handling and staging of radical prostatectomy specimens, achieved consensus that LVI should be

routinely reported in radical prostatectomy specimens. To our knowledge, there is no current paper that provides a comprehensive review of the recent literature.

METHODS

A systematic review of the prostate cancer literature relating to the potential prognostic value of LVI was performed in March 2009. Online databases employed in the search included Ovid MEDLINE, Embase, Web of Knowledge, Cochrane Database of Systematic Reviews, Database of Abstracts of Review of Effects (DARE) and Google Scholar. Collaborative efforts between clinician specialists and research librarians resulted in several iterations and refi nements of the search. Ultimately, we used a series of search terms that captured papers relevant to theprediction of disease progression following radical prostatectomy. The specifi c search terms and permutations were:

1. prostate$.mp. 2. prognos$.mp. 3. exp Disease Progression/ 4. biochemical progression.mp. 5. biochemical relapse.mp. 6. time to relapse.mp. 7. time to progression.mp. 8. exp Recurrence/ or exp Neoplasm Recurrence, Local/or recurrence.mp. 9. PSA relapse.mp. 10. PSA recurrence.mp. 11. 3 or 4 or 5 or 6 or 7 or 8 or 8 or 9 or 10 12. exp “ predictive Value of Tests ” / or predict$.mp. 13. 2 or 12 14. 1 and 11 and 13 15. Limit 14 to year = “ 2000 – 2009 ”

This search was suffi ciently broad to include all papers that had examined LVI (also known as microvascular or lymphatic, angiolymphatic and vascular invasion) as a potential prognostic factor. Reference lists from relevant articles were also examined to identify potential additional studies.

We limited the search date range to January 2000 to February 2009, in order to avoid inclusion of studies with patient accrual taking place prior to 1980 (prior to which the many histopathological features of prostate cancer were under-reported by pathologists).

L V I A S A P R O G N O S T I C F A C T O R I N P R O S T A T E C A N C E R

© 2 0 1 2 T H E A U T H O R S

B J U I N T E R N A T I O N A L © 2 0 1 2 B J U I N T E R N A T I O N A L 3

Following the identifi cation of potentially relevant publications, the abstract of each identifi ed publication was reviewed by one of the co-author reviewers. Studies that were clearly not related to the prognostic signifi cance of LVI were removed. Of the remaining studies, if the abstract did not clearly indicate whether LVI was included as a study variable, the full publication was scanned for inclusion of LVI. When LVI was identifi ed as a study variable, the publication was reviewed regarding study size, study cohort, pathological fi ndings, outcomes, analyses, prognostic factors and overall study quality.

Study inclusion criteria were English language publications in the period of January 2000 – March 2009, study populations having prostatectomies with curative intent; biochemical progression was reported, LVI included as one of the potential prognostic indicators in the analyses and a multivariate analysis (with or without LVI, depending on univariate fi ndings) was reported. Articles were excluded if they did not relate LVI to recurrence rates. Review articles or articles looking at LVI in the context of treatment modalities other than radical prostatectomy were also excluded. In addition, we identifi ed and excluded those studies with suspected overlapping patient populations by examining the hospitals at which patients underwent prostatectomy, the range of dates during which the cohort was selected and the characteristics of the reported study cohorts.

Articles with LVI in their title were indentifi ed as ‘ LVI-focused reports ’ . Where one reviewer identifi ed study characteristics as uncertain, a second reviewer reviewed the publication.

Data abstraction was guided by the checklist for reporting recommendations for tumour marker prognostic studies (REMARK) developed by McShane et al . [ 7 ] in 2005. Data describing study and patient characteristics, study results and methodology were collected and tabulated.

RESULTS

The search strategy produced a total of 2432 potentially relevant papers. Forty papers were identifi ed for close reading.

Studies with suspected overlapping patient populations were identifi ed and only the most recent and largest cohort studies among the duplicates studies were retained, leaving 19 papers for inclusion in the fi nal analysis. Of the 19 papers included in this review [ 9 – 27 ] ; only one (Huang et al . [ 16 ] ) represented a prospective study.

STUDY POPULATIONS

Table 1 provides a summary of the study characteristics and outcomes, illustrating wide variation in the characteristics of the study populations.

Study sample sizes were those described by the study authors as the overall cohort, as listed in Table 1 . In some studies, the prognostic value of LVI was calculated on a study subset, defi ned as number of patients with both LVI status and known biochemical recurrence status (range 531 393).

The age of patients (median, mean or range) was reported in all but four studies. When reported, mean age ranged from 60.4 to 67.8 years. Median age, when reported, ranged from 60.9 to 67 years. Age range was reported in 12 studies and spanned 34 to 80 years.

Pre-operative PSA data were reported in the majority of studies although there were four studies that did not do so. The summary statistics used to describe the study cohort in terms of PSA levels varied considerably and included mean and/or median values as well as subgroups defi ned by PSA ranges (see Table 1 , which includes the available data reported in each publication). PSA ranges were reported in nine studies.

Gleason scores were recorded in all studies and were generally reported in a manner that allowed grouping into Gleason grade < 7 or Gleason grade ≥ 7.Some studies had predominantly high-grade disease, while others not ( Table 1 ).

Reporting of pathologic T-staging varied considerably and was explicitly stated in only 13 studies. Limited staging information (e.g. extra-prostatic extension rates) was reported in the remaining six studies.

The percentages of the patient populations that had LVI positive prostatectomy

specimens ranged from just over 5% to almost 50%. The range of recurrence rates also varied widely from 11 to 49%. Duration of follow-up, while not reported in two studies, ranged from 1 month to 172 months; mean or median follow-up ranged from 21.4 months to 91.2 months ( Table 1 ).

Table 1 includes the results of the multivariate analyses, which indicate that 11/19 studies (58%) support the conclusion of LVI as an independent predictor of recurrence [ 9 – 11,13 – 15,17,18,20,24,26 ] . When reported, the magnitude of independent association of LVI with recurrence ranged from an odds ratio or relative risk of 1.37 to 4.39 ( Fig. 1 ). Larger studies tended to have lower effect-size estimates of the prognostic signifi cance of LVI for biochemical recurrence.

Figure 2 illustrates the conclusions of the reports that were focused on LVI vs those that were not LVI-focused. LVI was the prognostic factor of interest in less than half of the studies analyzed (42%). These studies are identifi ed in the last row of Fig. 3 .

As shown in Fig. 2 , studies with a specifi c focus on LVI had a higher proportion of positive associations between LVI status and recurrence.

FIG. 1. Forest plot of studies providing statistical estimates of effect size as relative risk or odds ratio with 95% confi dence intervals.

1.5

1.37

1.671

1.92

1.6

4.39

2.49

3.51

4.39Ito (82)

Huang (111)

Whittemore (214)

May (412)

Cheng (459)

Lee (557)

Shariat (626)

Quinn (731)

Loeb (1393)

Relative Risk or Odds Ratio

Author (Study N)

0 1 2 3 4 5 6 7 8 9 10

N G E T A L .

© 2 0 1 2 T H E A U T H O R S

4 B J U I N T E R N A T I O N A L © 2 0 1 2 B J U I N T E R N A T I O N A L

TABL

E 1

Stud

y ch

arac

teris

tics

Auth

orSt

udy

N

Age:

Med

ian

year

s [ ra

nge ]

Pre-

oper

ativ

e PS

A M

ean

ng/m

L, [ ra

nge ]

or

oth

er s

umm

ary †

G

leas

on

> / = 7

Path

olog

ic s

tage

[ ECE

, SV

I rat

es ]

LVI p

ositi

ve

rate

: N (%

)Re

curr

ence

ra

te (%

)

Follo

w-u

p:

Med

ian

mon

ths

[ rang

e ]

LVI

inde

pend

ent

pred

icto

r?Ef

fect

siz

e Es

timat

e [ 9

5% C

I ]

Baba

ian

et a

l . [ 9

] 26

564

.2 [ 4

1 – 74

] < 1

0.0:

72.

2% 1

0 – 20

: 23

.0%

; > 20

: 4.7

%

86%

pT2:

69%

pT3

: 31%

N/A

16 > 4

8Ye

sH

R N

/A P

= 0

.036

Bayd

ar e

t al

. [ 10

] 71

63 [ 4

8 – 75

] 11

.5 [ 1

– 41 ]

75

%pN

1: 7

%11

(15.

5%)

3154

[ 4 – 1

45 ]

Yes

HR:

N/A

P =

0.0

23.

Chen

g et

al .

[ 11 ]

50

463

[ 34 –

80 ]

< 10:

61.

8% 1

0 – 20

: 38

.2%

64%

pT2:

69%

pT3

: 31%

pN

1: 4

%10

6 (2

1%)

3144

.3 *

[ 2 – 1

44 ]

Yes

HR

1.6.

[ 1.1

2, 2

.38 ]

P

= 0.

01

Enge

rs e

t al

. [ 12

] 60

67 [ 5

5 – 79

] 14

.4 [ 1

– 60 ]

82

%pT

2: 6

2% p

T3: 3

8%

pN1:

2%

6 (1

0%)

3286

[ 24 –

173 ]

N

oN

/A

Epst

ein

et a

l . [ 1

3 ]

60N

/AN

/A93

%pT

3: 1

00%

13 (2

1.7%

)N

/AN

/AYe

sN

/A

Ferr

ari e

t al

. [ 14

] 62

065

[ 42 –

78 ]

[ 7 – 1

0 ]

82%

pT2:

63%

pT3

– 4: 3

7%,

pN1:

6%

110

(18%

)28

90 [ 2

4 – 21

6 ]

Yes

P =

0.00

5

Her

man

et

al . [

15 ]

263

64N

/A73

%pT

3: 1

00%

91 (3

5%)

2436

[ 1 – 1

58 ]

Yes

P =

0.00

1

Hua

ng e

t al

. [ 16

] 13

167

.8 *

< 10:

39.

0% >

10:

61.0

%

N/A

pT1-

2: 6

1% p

T3: 4

4%

pT4:

2%

pN

1: 6

%17

(15.

3%)

2327

* (2

4.0)

No

HR

3.5

[ 0.7

9, 1

5.65

]

Ito e

t al

. [ 17

] 82

66.5

* [ 5

6 – 74

] 17

.2 [ 0

.6 – 1

10 ]

57%

pT1 –

2: 6

1%; p

T3: 3

9%38

(46.

3%)

4422

[ 9 – 8

4 ]

Yes

RR 4

.39

[ 1.4

0, 1

3.70

] , P

= 0.

019

Lee

et a

l . [ 1

8 ]

557

647.

772

– 83%

pT2:

64%

pT3

: 33%

pT

4: 2

% p

N1:

6%

90 (1

6.16

%)

2890

Yes

HR

1.92

[ 1.2

6, 2

.92 ]

P

≤ 0.

01

Loeb

et

al . [

19 ]

1709

N/A

N/A

32%

pT1 –

2: 7

4%, p

T3:2

6%11

8 (6

.9%

)11

74 *

[ 0 – 1

59 ]

No

RR =

1.5

[ 0.9

, 2.4

] P

= 0.

1

May

et

al . [

20 ]

412

63.7

* [ 4

4 – 79

] 12

.1 [ 0

.1 – 1

51 ]

41%

pT2:

73%

, pT3

: 27%

42 (1

0.2%

)17

52.5

* [ 1

0 – 11

6 ]

Yes

HR

4.39

[ 2.4

7, 7

.80 ]

P

< 0.

001

Mia

n et

al .

[ 21 ]

18

863

[ 43 –

73 ]

8.6

[ 1.6

– 42 ]

10

0%pT

2: 3

8%, p

T3: 5

6%

pN1:

6%

N/A

3261

[ 1 – 1

29 ]

No

N/A

Quin

n et

al .

[ 22 ]

73

263

[ 41 –

77 ]

9.0

[ 0.7

– 194

] 47

%pT

2: 5

6% p

T3: 3

9%

pT4:

3%

pN

1: 2

%38

(5.2

%)

2039

[ 1 – 1

68 ]

No

HR

1.37

[ 0.8

2, 2

.30 ]

P

= 0.

23

Shar

iat

et a

l . [ 2

3 ]

630

61 [ 4

0 – 75

] 6.

1 [ 0

.1 – 9

9 ]

59%

pT3A

: 29%

, pT3

B: 9

%32

(5.1

%)

1321

[ 1 – 1

01 ]

No

HR

1.67

1 [ 0

.93,

2.

99 ] ,

P =

0.08

3

Stam

ey e

t al

. [ 24

] 32

665

[ 35 –

79 ]

7.3

[ 0.3

– 146

] N

/ApN

1: 1

0% p

T3A:

53%

, pT

3B: 2

1%43

(13.

2%)

4463

[ 36 –

133 ]

Ye

s P

= 0.

005

Svat

ek e

t al

. [ 25

] 42

561

5.7

53%

pN1:

3%

pT3

A: 3

3%,

pT3B

: 16%

28 (1

1.2%

)18

37N

oN

/A

Whi

ttem

ore

et a

l . [ 2

6 ]

214

N/A

N/A

100%

pT2:

76%

, pT3

: 21%

pT

4: 1

%, p

N1:

2%

12 (5

.6%

)35

N/A

Yes

RR 2

.49

[ 1.0

9, 5

.65 ]

P

= 0.

03

Yosh

imot

o et

al .

[ 27 ]

12

5N

/A < 1

0.0:

56.

3% 1

0 – 20

: 31

.1%

> 20

: 12.

6%

39%

pT2:

56%

pT3

: 39%

pT

4: 4

.9%

26 (2

2.6%

)49

87N

oN

/A

N/A

, dat

a no

t rep

orte

d in

pub

licat

ion;

* onl

y m

ean

data

repo

rted

in p

ublic

atio

n; †

whe

re m

edia

n, ra

nge

not r

epor

ted

in p

ublic

atio

n.

L V I A S A P R O G N O S T I C F A C T O R I N P R O S T A T E C A N C E R

© 2 0 1 2 T H E A U T H O R S

B J U I N T E R N A T I O N A L © 2 0 1 2 B J U I N T E R N A T I O N A L 5

QUALITY OF THE REPORTING

Figure 3 provides a summary of the adherence of the 19 publications to the REMARK recommendations for tumour marker prognostic studies. (adapted from McShane et al . [ 7 ] ). While overall the quality of reporting indicated compliance with the majority of reporting standards, the absence of some critical elements, (such as the defi nition of LVI and the relationship of LVI to at least one standard prognostic value or the results of univariate and multivariate analyses), was noted.

DISCUSSION

Our systematic review of recent literature reveals confl icting evidence regarding LVI as an independent prognostic factor for disease progression post-radical prostatectomy. Of the studies that were of high quality and of largest size, the prognostic value of LVI was often statistically signifi cant, indicating the potential independent value of LVI when other common prognostic factors are controlled for in the predictive model. The strength of the association, however, was relatively modest, and indeed the largest study did not reach statistical signifi cance. These weak associations indicate that the prognostic value of LVI is of borderline clinical signifi cance. An exception to this general trend is the report of May etal [ 20 ] .

who found a stronger association between LVI and recurrence (HR 4.39) with a sample of 412 patients. Additionally, this publication met a high number of the REMARK quality indicators for reporting.

One way of addressing this variation in study fi ndings is systematically to pool results using standard meta-analytic techniques. Unfortunately, however, this approach was not possible in this setting owing to substantial variation between studies in various elements of their design. In addition, concern is raised regarding bias of the published literature resulting from the selective publication of studies based on their results with negative studies being published less for a number of reasons. Pooling the results of published studies can clearly lead to an overestimation of prognostic signifi cance [ 28 ] . In this analysis, we compared the outcome of those papers that mentioned LVI in the title with those that did not, and illustrated this potential bias ( Fig. 2 ). Moreover, since multiple studies can be published on the same patient cohort, systematic reviews of retrospective studies are also susceptible to study population overlap between studies. Overlap of patient cohorts occurred more often than we had anticipated, with some studies having partial patient overlap and some with near-complete overlap.

If a broad review of the literature produces the highest level of evidence for readers, how can readers ensure that the review itself meets scientifi c standards? Oxman & Guyatt [ 29 ] have proposed guidelines to assist readers to assess the scientifi c quality of review papers that focus on: the defi nition of the question, the comprehensiveness of the search strategy, methods of choosing the papers included in the analysis and methods of combining the results and coming to a fi nal conclusion. In the case of the current analysis we complied with the Oxman & Guyatt guidelines. For example, the study question and methods were clearly stated, our search strategy was comprehensive, our methods were explicit and we examined the quality of the primary studies.

Our analysis, however, does have some limitations. Our decision to include only publications from 2000 forward (in an attempt to maximize the likelihood of LVI being explicitly addressed in the pathologic

examination of the resected prostate), excluded previously published articles addressing the prognostic signifi cance of LVI. We note, however, that inclusion of such studies would only likely add further variation in study quality, and would not improve our ability to pool study results. Our search strategy did not attempt to explore unpublished analyses. As noted, our conclusions were also compromised by possible reporting bias in studies evaluating potential prognostic factors and by studies with partial patient overlap.

Finally, we note that in the period since this systematic literature review was completed additional relevant papers have been published that add to the uncertainty of LVI as an independent prognostic indicator. Mizuno et al [ 30 ] found LVI to be positively associated with biochemical failure on univariate analysis. However, in a Cox proportional hazard model, LVI did not have an independent association with biochemical failure. In 2008 Yamamoto et al . [ 31 ] found that LVI was an independent predictor of PSA failure in a subset of prostate cancer patients with pT3aN0 disease. Finally, Hwang Gyun Jeon et al . [ 32 ] concluded that LVI was positively associated with biochemical failure-free survival on univariate analysis ( P 0.001), but was not shown to be an independent factor associated with biochemical failure free survival on multivariate analysis.

There is insuffi cient evidence to conclude that LVI is an important independent prognostic factor for disease progression post-radical prostatectomy. A systematic review of the literature provides confl icting results and illustrates substantial variation in study methodology. Even in the large studies in which a statistically signifi cant association between LVI and biochemical recurrence was noted, the strength of the association was modest. Thus, clinicians should not overweigh the importance of LVI status on clinical prognostication. The use of LVI status as a strong predictor of clinical outcomes is not recommended.

CONFLICT OF INTEREST

None declared. Source of funding: Dr Brundage is supported by Cancer Care Ontario.

FIG. 2. Bar graph showing number of studies concluding that LVI is an independent prognostic factor, stratifi ed by study type (LVI-focused report or not).

0

10

20

30

40

50

60

70

80

90

100

LVI-focused study

Focus of Study

%

LVI is not found to be anindependent prognostic factorLVI is found to be an independentprognostic factor

Not an LVI-focused study

N G E T A L .

© 2 0 1 2 T H E A U T H O R S

6 B J U I N T E R N A T I O N A L © 2 0 1 2 B J U I N T E R N A T I O N A L

FIG. 3. Quality assessment of publications.

May

, 200

6

Loeb

, 200

6

Her

man

, 200

0

Ito,

200

3

Shar

iat,

2004

Baba

ian,

200

1

Ferr

ari,

2004

Chen

g, 2

005

Qui

nn, 2

001

Epst

ein,

200

0

Stam

ey, 2

000

Mia

n, 2

002

Hua

ng, 2

007

Lee,

200

5

Bayd

ar, 2

008

Enge

rs, 2

006

Whi

ttem

ore,

200

8

Svat

ek, 2

008

Yosh

imot

o, 2

008

Tota

l (%

'yes

')

METHODS

Patient characteristics:

Inclusion criteria 19 (100)

Treatment received described 18 (95)

Histopathological Review:

Definition of LVI provided 13 (68)

Protocol for specimen review 17 (89)

Quality assurance (e.g., 2nd review) 13 (68)

Clinical Follow-up:

Definition of recurrence 18 (95)

Follow-up protocol described 14 (74)

Study Design:

Method of case selection described 18 (95)

Lists all candidate variables for model

17 (89)

Statistical Analysis:

Specifies statistical methods 19 (100)

Specifies handling of missing data 11 (54)

RESULTS

Data:

Range and duration of patient follow- up described.

15 (79)

Analysis and Presentation:

Shows the relation of the marker to at least 1 standard prognostic variable

9 (47)

Known prognostic factors significant on multivariate analysis

Stage 7 (37)

Pre-op PSA 6 (32)

Gleason Score 15 (79)

Uni- or multi-variate analysis of association between LVI and recurrence with estimated effect size

12 (63)

DISCUSSION

Interprets the results in the context of the pre-specified hypotheses and other relevant studies.

16 (84)

Studies had LVI-focus 8 (42)

Legend: = "yes"'

= "no"'

L V I A S A P R O G N O S T I C F A C T O R I N P R O S T A T E C A N C E R

© 2 0 1 2 T H E A U T H O R S

B J U I N T E R N A T I O N A L © 2 0 1 2 B J U I N T E R N A T I O N A L 7

REFERENCES

1 Sutcliffe P , Hummel S , Simpson E et al . Use of Classical and novel biomarkers as prognostic risk factors for localised prostae cancer: a systematic review . Health Technol Assess 2009 ; 13 : 1 – 242

2 D ’ Amico AV , Whittington R , Malkowicz SB et al . Biochemical outcome after radical prostatectomy, external beam radiation therapy, or interstitial radiation therapy for clinically localized prostate cancer . JAMA 1998 ; 280 : 969 – 74

3 Soerjomataram I , Louwman MW , Ribot JG , Roukema JA , Coebergh JW . An overview of prognostic factors for long-term survivors of breast cancer . Breast Cancer Res Treat 2008 ; 107 : 309 – 30

4 Magi-Galluzzi C , Evans AJ , Delahunt B et al . International Society of Urological Pathology (ISUP) Consensus Conference on Handling and Staging of Radical Prostatectomy Specimens. Working group 3: extraprostatic extension, lymphovascular invasion and locally advanced disease . Modern Pathol 2011 ; 24 : 26 – 38

5 Harnden P , Shelley MD , Clements H et al . The prognostic signifi cance of perineural invasion in prostatic cancer biopsies: a systematic review . Cancer 2007 ; 109 : 13 – 24

6 Cookson MS , Aus G , Burnett AL et al . Variation in the defi nition of biochemical recurrence in patients treated for localized prostate cancer: the American Urological Association Prostate Guidelines for Localized Prostate Cancer Update Panel report and recommendations for a standard in the reporting of surgical outcomes . J Urol 2007 ; 177 : 540 – 5

7 McShane LM , Altman DG , Sauerbrei W et al . Reporting recommendations for tumor marker prognostic studies . J Clin Oncol 2005 ; 23 : 9067 – 72

8 Henson DE , Hutter RV , Farrow G . Practice protocol for the examination of specimens removed from patients with carcinoma of the prostate gland. A publication of the Cancer Committee, College of American Pathologists. Task Force on the Examination of Specimens Removed From Patients With Prostate Cancer . Arch PathLab Med 1994 ; 118 : 779 – 83

9 Babaian RJ , Troncoso P , Bhadkamkar VA , Johnston DA . Analysis of clinicopathologic factors predicting outcome after radical prostatectomy . Cancer 2001 ; 91 : 1414 – 22

10 Baydar DE , Baseskioglu B , Ozen H , Geyik PO . Prognostic signifi cance of lymphovascular invasion in clinically localized prostate cancer after radical prostatectomy . TSWJ 2008 ; 8 : 303 – 12

11 Cheng L , Jones TD , Lin H et al . Lymphovascular invasion is an independent prognostic factor in prostatic adenocarcinoma . J Urol 2005 ; 174 : 2181 – 5

12 Engers R , Ziegler S , Mueller M , Walter A , Willers R , Gabbert HE . Prognostic relevance of increased Rac GTPase expression in prostate carcinomas . Endocr Relat Cancer 2007 ; 14 : 245 – 56

13 Epstein JI , Partin AW , Potter SR , Walsh PC . Adenocarcinoma of the prostate invading the seminal vesicle: prognostic stratifi cation based on pathologic parameters . Urology 2004 ; 56 : 283 – 8

14 Ferrari MK , McNeal JE , Malhotra SM , Brooks JD . Vascular invasion predicts recurrence after radical prostatectomy: stratifi cation of risk based on pathologic variables . Urology 2004 ; 64 : 749 – 53

15 Herman CM , Wilcox GE , Kattan MW , Scardino PT , Wheeler TM . Lymphovascular invasion as a predictor of disease progression in prostate cancer . Am J Surg Pathol 2000 ; 24 : 859 – 63

16 Huang SP , Huang CY , Wang JS et al . Prognostic signifi cance of p53 and X-ray repair cross-complementing group 1 polymorphisms on prostate-specifi c antigen recurrence in prostate cancer post radical prostatectomy . Clin Cancer Res 2007 ; 13 : 6632 – 8

17 Ito K , Nakashima J , Mukai M et al . Prognostic implication of microvascular invasion in biochemical failure in patients treated with radical prostatectomy . Urol Int 2003 ; 70 : 297 – 302

18 Lee KL , Marotte JB , Ferrari MK , McNeal JE , Brooks JD , Presti JC Jr . Positive family history of prostate cancer not associated with worse outcomes after radical prostatectomy . Urology 2005 ; 65 : 311 – 5

19 Loeb S , Roehl KA , Yu X et al .

Lymphovascular invasion in radical prostatectomy specimens: prediction of adverse pathologic features and biochemical progression . Urology 2006 ; 68 : 99 – 103

20 May M , Kaufmann O , Hammermann F , Loy V , Siegsmund M . Prognostic impact of lymphovascular invasion in radical prostatectomy specimens . BJU Int 2007 ; 99 : 539 – 44

21 Mian BM , Troncoso P , Okihara K et al . Outcome of patients with Gleason score 8 or higher prostate cancer following radical prostatectomy alone . J Urol 2002 ; 167 : 1675 – 80

22 Quinn DI , Henshall SM , Haynes AM et al . Prognostic signifi cance of pathologic features in localized prostate cancer with radical prostectomy: implications for staging systems and predictive models . J Clin Oncol 2001 ; 19 : 3692 – 705

23 Shariat SF , Khoddami SM , Saboorian H et al . Lymphovascular invasion is a pathological feature of biologically aggressive disease in patients treated with radical prostatectomy . J Urol 2004 ; 171 : 1122 – 7

24 Stamey TA , Yemoto CM , McNeal JE , Sigal BM , Johnstone IM . Prostate cancer is highly predictable: a prognostic equation based on all morphological variables in radical prostatectomy specimens . J Urol 2000 ; 163 : 1155 – 60

25 Svatek RS , Karam JA , Roehrborn CG , Karakiewicz PI , Slawin KM , Shariat SF . Preoperative plasma endoglin levels predict biochemical progression after radical prostatectomy . Clin Cancer Res 2008 ; 14 : 3362 – 6

26 Whittemore DE , Hick EJ , Carter MR , Moul JW , Miranda-Sousa AJ , Sexton WJ . Signifi cance of tertiary Gleason pattern 5 in Gleason score 7 radical prostatectomy specimens . J Urol 2008 ; 179 : 516 – 22 ; discussion 522

27 Yoshimoto M , Joshua AM , Cunha IW et al . Absence of TMPRSS2:ERG fusions and PTEN losses in prostate cancer is associated with a favorable outcome . Modern Pathol 2008 ; 21 : 1451 – 60

28 Montori VM , Smieja M , Guyatt GH . Publication bias: a brief review for clinicians . Mayo Clinic Proc 2000 ; 75 : 1284 – 8

29 Oxman AD , Guyatt GH . Guidelines for reading literature reviews. CMAJ . Can Med Assoc J 1988 ; 138 : 697 – 703

30 Mizuno R , Nakashima J , Mukai M

N G E T A L .

© 2 0 1 2 T H E A U T H O R S

8 B J U I N T E R N A T I O N A L © 2 0 1 2 B J U I N T E R N A T I O N A L

et al . Tumour length of the largest focus predicts prostate-specifi c antigen-based recurrence after radical prostatectomy in clinically localized prostate cancer . BJU Int 2009 ; 104 : 1215 – 8

31 Yamamoto S , Kawakami S , Yonese J et al . Lymphovascular invasion is an independent predictor of

prostate-specifi c antigen failure after radical prostatectomy in patients with pT3aN0 prostate cancer . Int J Urol 2008 ; 15 : 895 – 9

32 Jeon HG , Bae J , Yi JS , Hwang IS , Lee SE , Lee E . Perineural invasion is a prognostic factor for biochemical failure after radical prostatectomy . Int J Urol 2009 ; 16 : 682 – 6

Correspondence: Michael Brundage, Queen ’ s Cancer Research Unit, 10 Stuart Street, Level 2, Kingston, ON, Canada K7L 3N6. e-mail: michael.brundage@krcc.on.ca

Abbreviations : LVI , lymphovascular invasion ; PNI , perineural invasion.