The prognostic significance of perineural invasion in prostatic cancer biopsies

A systematic review


  • This work was initiated for revision of the guidance on reporting of urological cancers for the Royal College of Pathologists, United Kingdom (UK). Funding was received from the UK National Health Service Cancer Screening Programme and Cancer Research Wales for salary (H. C.) and running costs. Cancer Research UK contributed to library, computing facilities, and salary costs (P. H.).

  • The authors report no conflict of interest.


Men with clinically localized prostate cancer are faced with a wide range of treatment options, and only Gleason grading is universally used as a histopathological prognostic factor for this disease. The significance of perineural invasion in diagnostic biopsies is controversial. Opinion about whether or not it should influence treatment decisions is currently almost equally divided. To address this, the authors performed a systematic review of studies that examine the association between perineural invasion and prostate cancer recurrence. MEDLINE, Embase, and the Web of Knowledge were searched from January 1990 to December 2005. Outcomes analyzed were the development of biochemical or clinical recurrence. Twenty-one articles on the association of perineural invasion in biopsies and prostate cancer recurrence after radical prostatectomy (n = 10) or radiotherapy (n = 11) were found but none on its significance in the context of watchful waiting. Structured data extraction was performed to allow comparisons between articles and to identify sources of heterogeneity to explain discrepancies in results. The considerable variation in study design, execution, and reporting precluded meta-analysis and quantitative risk estimation, but the weight of evidence suggested that perineural invasion in biopsies was a significant prognostic indicator, particularly in specific patient groups defined by presenting serum prostate-specific antigen levels and biopsy Gleason scores. Immediate treatment rather than watchful waiting may be more appropriate for patients with localized prostatic cancer and perineural invasion. However, the data are limited, and well-designed studies that use predefined stringent protocols are required to provide robust estimates of risk to aid in treatment planning. Cancer 2007. © 2006 American Cancer Society.

With the increasing uptake of serum prostatic specific antigen (PSA) testing, growing numbers of asymptomatic men are diagnosed with clinically localized prostate cancer following transrectal ultrasound-guided prostatic biopsies.1 Because of the variable natural history of prostate cancer, treatment options range from active surveillance to different forms of radical treatment, all of which are associated with significant side effects.2 Greater understanding of the biological potential of individual tumors would allow better decisions making, but despite intensive investigations,3 the only biopsy-derived prognostic factor in routine use worldwide, because of consistent outcome study results,4–6 is Gleason grading.7 However, the range of survival probabilities for each grade is quite wide.8

Perineural invasion is an established risk factor in other malignancies9 and a recognized mechanism whereby cancer cells spread beyond the prostate10 by using the rich innervation of the posterior aspect of the prostate.11 Contact with nerves may also confer a survival advantage for these cells, as in experiments, malignant prostate cells show enhanced proliferation in culture with nerves.12 However, existing publications on the significance of perineural invasion are contradictory,13 and medical practice is almost equally divided, with 57% of experienced surgeons reporting that presence of perineural invasion did not influence their treatment decisions and 43% reporting that it did.13 Despite multiple articles involving thousands of patients, the question of why there is still uncertainty about the prognostic value of perineural invasion has not been addressed.

A systematic literature review is a useful tool for resolving uncertainties, and results of the association of perineural invasion on biopsy and the risk of prostate cancer recurrence or progression are presented here. No studies of perineural invasion in patients who opted against immediate treatment were found. Furthermore, most articles only reported indirect measures of patient outcome, such as the correlation between perineural invasion in biopsies and margin status after radical prostatectomy. To assess the prognostic value of perineural invasion rather than its predictive value relative to a specific form of treatment, this review focused on articles relating perineural invasion directly to risk of biochemical or clinical recurrence regardless of radical treatment type. Structured data extraction was performed on the articles to compare study population, spectrum of disease severity, and study design, reporting, and results. This data extraction was performed to determine whether clinical or methodological diversity among studies could be responsible for divergent results and to determine whether selecting articles that fulfilled quality criteria for study design and execution could provide a definitive answer regarding the importance of perineural invasion.


An overarching comprehensive strategy made up of 128 search terms to cover all areas of prostatic cancer pathology was developed, as it became clear that indexing of prognostic articles in histopathology was not consistent. The date range of the search was January 1990 to December 2005. The start date was selected to coincide with the gradual rise in transrectal biopsy publications, as case detection for curative treatment increased with the more widespread introduction of nerve-sparing radical prostatectomy in particular. The databases searched were MEDLINE, Embase, and the Web of Knowledge. Additionally, manual searching was conducted on recent journals and was supplemented by online searches of the Proceedings of the American Society of Clinical Oncology (ASCO), British Journal of Urology International, and Journal of Urology (the full search strategy is available from the authors).

Titles and abstracts of the 41,295 articles identified were scanned by at least 2 reviewers to identify studies relevant to histopathology and prostate cancer. The articles were maintained on a bibliographic database, which was searched for perineural invasion, yielding 128 articles for close reading. Reviews, articles relating perineural invasion to other markers rather than directly to outcomes, and articles on perineural invasion in radical prostatectomy rather than biopsy specimens were excluded. Also excluded were articles reporting on the frequency of perineural invasion in patients with local recurrence14 or on the outcomes of patients selected for small tumors and no perineural invasion,15 as neither study had a control arm. The 10 surgical16–25 and 11 radiotherapy26–36 articles on perineural invasion in diagnostic biopsies form the basis of this report.

In the absence of established methodology for the systematic review of basic pathology factors, the set of data items to extract was derived from our experience of the evaluation of clinical trials, checklists for the evaluation of diagnostic tests,37 and our knowledge of prostate cancer. Factors specific to the detection and assessment of perineural invasion in the context of prostate cancer (Table 1) were extracted in addition to standard data items applicable to the evaluation of any type of clinical study (eg, patient selection, treatment details, study design, statistical methods, Tables 2–8). These independently derived data items showed strong concordance with the checklist in the subsequently published guidance for molecular prognostic marker studies, REMARK.38 Clinical stage, presenting serum PSA levels, and biopsy Gleason scores were extracted to define study populations in terms of disease severity.4–6

Table 1. Data Items Specific to Biopsy Perineural Invasion in Prostate Cancer
 Data items extractedSignificance
Biopsy procedureBiopsy technique (needle angle, method of prostate mapping)Can affect the type and amount of prostatic tissue sampled, therefore, the number of nerves present and the identification of perineural invasion
Needle size, numbers of cores taken
Histological slide preparationNumber of levels at which the biopsies are examined
Pathological interpretationDefinition of perineural invasionConsistency of identification within and between papers
Presence of nerves stated, and analysis restricted to biopsies with nervesThe inclusion of “uninformative” tests (no nerves for evaluation) could bias results
Number of pathologists involvedInterobserver variation in histopathological diagnosis of perineural invasion
Biopsies reviewed or result taken from reportReporting of perineural invasion not mandatory so may only be identified on review
Pathologist(s) blinded to outcomeIf not, potential for bias

Data were checked by 2 reviewers, and any differences were settled through discussion.


No study reported on all data items: the proportion of missing or incomplete items ranged from 18%32 to 61%24 with a median of 39%. The 21 studies came from 14 institutions (Tables 2 and 3), with only 5 institutions contributing to the radiotherapy literature.

Table 2. Characteristics of the Study Population—Surgical Series
ReferenceOrigin & datesAge, y: Median (Mean) RangeClinical stagePSA: Median (Mean) Range. ng/mLGleason score: Median (Mean) RangeFinal patient number (lost)
  1. Positive studies are at the top of the table (Ravery to Quinn); negative studies are at the bottom (Henke to Nelson).

  2. ND indicates no data.

Ravery et al., 199424Paris, France 1989–1993ND (64.5), 42.3–75All T1/2NDND65 (35)
De la Taille et al., 199917New York, USA 1993–1998ND (61.4), 40–75.6All T1/2ND (8.02), 0.5–35ND (6.15), 2–9319 (0)
D'Amico et al., 200116Boston, USA 1989–199862 (ND), ND50% T1NDND (ND), 2–10697 (53)
50% T2
Sebo et al., 200225Rochester, USA 1995–199865 (64.5), 38–79ND6.3 (8.5), 0.6–1126 (6.5), 4–9454 (ND)
Freedland et al., 200218Los Angeles, USA 1991–2000ND (62.6), ND35% T18.7 (10.9), NDND (6.1), 2–10190 (140)
64% T2
1% T3
Quinn et al., 200323Darlinghurst, Australia 1986–199962.5 (62), 40.7–76.733.6% T19.1 (12.9), 1.0–1946.0 (5.96), 3–10696 (139)
62.2% T2
4.2% T3
Henke et al., 199819Hamburg, Germany ND63 (ND), 43–72All T1/2NDND (6.7), 6–872 (3)
O'Malley et al., 200222Baltimore, USA 1984–1995ND28.2% T1NDND (ND), 2–1078 each arm
61.5% T2
10.3% T3
Nelson et al., 200220Ann Arbor, USA 1994–1998ND (61), 39–8370.2% T1NDND (ND), 2–10588 (ND)
29.8% T2
Nelson et al., 200321Ann Arbor, USA 1994–2002ND (60.2), ND64.1% T1NDND (ND), 2–101414 (ND)
35.9% T2
Table 3. Characteristics of the Study Population—Radiotherapy Series
ReferenceOrigin & DatesAge, y: Median (Mean) RangeClinical stagePSA: Median (Mean) Range, ng/mLGleason score: Median (Mean) RangeFinal patient mumber (lost)
  1. Positive studies are at the top of the table (Bonin to Wong); negative studies are at the bottom (Grann to Merrick).

  2. ND indicates no data.

Bonin et al., 199729Philadelphia, USA 1989–199470 (ND) 51–8978.5% T1/2ab10 (21.6) 0.4–191ND (ND) 2–10484 (0)
21.5% T2c/3
Anderson et al., 199827Philadelphia, USA 1987–199469 (ND) 46–8982% T1/2abND (ND) <10ND (ND) 2–10266 (ND)
18% T2c/3
Algan et al., 199926Philadelphia, USA 1988–199470 (ND) 51–8952.7% T1/2ab35 (ND) 20–191ND (ND) 2–10129 (ND)
47.3% T2c/3
Beard et al., 200428Boston, USA 1989–1998ND41% T1NDND (ND) ND-10381 (0)
44.9% T2ab
14% T2c/3
Wong et al., 200436Scottsdale, USA 1993–199974 (ND) 46–8417% T18.1 (ND) 1–197ND239 (92)
73.7% T2
10% T3
Grann et al., 199830New York, USA 1988–199467 (ND) 48–80all T1/2NDND (ND) 4–7102 (7)
Merrick et al., 200133Washington, USA 1995–199968.0 (66.9) 48–8184% T1c/2a7.5 (9.4) ND6 (5.9) 4–8425 (ND)
16% T2b/3
Kaminski et al., 200231,Philadelphia, USA 1989–199769 (68) 45–8980.9% T1/2a8.8 (13.7) 0.4–191ND (ND) 2–10655 (ND)
19.1% T2b/3
Kestin et al., 200232Washington, USA 1991–199969 (ND) 52–85T1c-3c9.9 (ND) 0.4–58.37 (ND) 5–10160 (39)
Pollack et al., 200435Philadelphia, USA 1989–199769 (ND) 43–8641% T18.8 (ND) 0.4–146.1ND (ND) ND-10839 (0)
53% T2
6% T3
Merrick et al., 200534Washington, USA 1995–200166 (65.7) 45–8179.3% T1b-T2b6.7 (7.8) ND7 (6.6) ND512 (0)
20.7% T2c

Characteristics of the Study Population (Tables 2 and 3)

All patients had radically treated, nonmetastatic, prostate cancer diagnosed on biopsy. Two articles26, 27 selected patients on the basis of serum PSA levels, >20 ng/mL in 126 or <10 ng/mL, in the other.27 In 1 paper, patients had locally advanced prostate cancer with a serum PSA level of ≥10 ng/mL, or a Gleason sum score of ≥7, or clinical stage T2b to T3c.32 Exclusion criteria were variable and included a history of prior treatment,18, 20–23, 27–29, 35 the unavailability of biopsies for review,18, 20, 30 or the failure to obtain at least 4 biopsies.18 In 2 articles,19, 36 in which perineural invasion was not the main focus, patients were excluded if the factor of interest could not be assessed. The absence of nerves in the biopsies was not an exclusion criterion in any of the articles.

In cohort studies, sample sizes were the number of patients who fulfilled the inclusion criteria and were treated in a defined period at the relevant institution. The case-control study22 paired all patients with perineural invasion identified in their initial cohort with patients without perineural invasion, matching them for Gleason grade and PSA level.

Recruitment periods covered a period of 18 years, with no evidence of a trend toward different results over time. Radiotherapy articles tended to group stages in different ways (Table 3), precluding comparisons. One study specifically selected patients with locally advanced disease,32 otherwise the majority of patients had clinically organ-confined disease. Apart from 3 articles that selected patients on the basis of specific values,26, 27, 32 there were no obvious differences between presenting serum PSA and Gleason scores, although comparisons were limited because of missing or incomplete data on PSA levels (38% of articles) or Gleason scores (62% of articles) despite the fact that all institutions routinely used Gleason grading.

Only 1 surgical16 article and 1 radiotherapy28 article were prospective. One surgical17 and 4 radiotherapy28, 29, 34, 35 articles included consecutive patients; the proportion of patients lost to follow-up in the remainder varied from 4% at 1 extreme19 to 42.5% at the other.18

Treatment-related Factors

No article provided information on delay between diagnosis and treatment.

Surgical series

Three studies19, 21, 23 stated that all operations were by the retropubic approach, another included both retropubic and perineal prostatectomies;24 no data were available in the remainder. Only 2 studies mentioned the issue of nerve sparing,19, 22 and reasons for sacrificing nerves were not related to the presence of perineural invasion on biopsy. Details were given only in the case-control study,22 in which 89% of patients had at least 1 neurovascular bundle preserved, with similar rates reported for patients with and without perineural invasion.

A single surgeon was involved in 2 studies.17, 19 Only 1 of the 4 studies23 that explicitly involved multiple surgeons18, 20, 21, 23 compared outcomes between surgeons and found no significant differences. In 1 study,23 a proportion of patients received adjuvant radiotherapy.

Radiotherapy series (Table 4)

The effect of variations in radiation intensity was investigated in 6 articles.26, 27, 29, 31, 32, 34, 35 In unselected series of patients,29, 31, 35 radiation dose correlated significantly with biochemical control. In series of patients selected for serum PSA values27, 29 of <10 ng/mL or <20 ng/mL, treatment dose was not an independent predictor of outcome, whereas perineural invasion was significant. By contrast, when presenting PSA values were ≥20 ng/mL, patients who received a central prostate dose of ≥74 Grays (Gy) did significantly better than those who received lower doses (P = .03 in multivariate analysis).29 However, as only 10 patients in this group had perineural invasion, perineural invasion was not included as a factor in the multivariate analysis. In a further study of patients with PSA values of ≥20 ng/mL more, 41% of patients who received ≥73 Gy did not recur compared with 17% of those who received lesser doses (P < .001 in multivariate analysis),26 but perineural invasion remained independently prognostic. In the study that selected patients with locally advanced cancer,32 doses of less than 90 Gy, 90 Gy to 95 Gy, and >95 Gy were associated with biochemical control rates of 37%, 72%, and 89% respectively (P < .0001), and perineural invasion was not significant. The relation between radiation dose and outcomes was not directly investigated in 2 related reports,33, 34 but patients with perineural invasion received external beam radiation in addition to brachytherapy significantly more often than those without perineural invasion (75.9% versus 47.8% of patients, P < .001).

Table 4. Radiotherapy Treatment
ReferenceMain treatment typeTreatment variations, excluding radiation dose (% Patients affected)Median dose deliveredRange of dose delivered
  1. Positive studies are at the top of the table (Bonin to Wong); negative studies are at the bottom (Grann to Merrick).

  2. EBRT indicates external beam radiotherapy; 3D, three dimensional; ND, no data.

Bonin et al., 199729EBRT, 3D conformalPelvic lymph node irradiation if 15% risk or more of lymph node involvement6 (ND)76 Gy63–81 Gy
Anderson et al., 199827EBRT, 3D conformalPelvic lymph node irradiation if 15% risk or more of lymph node involvement6 (ND)ND63.16–78.95 Gy
Algan et al., 199926EBRT3D conformal (75.2) or conventional (24.8); prostate only (36.3) or whole pelvis (63.6)73 Gy68–79 Gy
Beard et al., 200428EBRT, 3D conformalExtended to seminal vesicles if clinically involved, pelvic lymph node irradiation if 15% risk or more of lymph node involvement (ND)NDND
Wong et al., 200436EBRT, 3D conformalAntiandrogen therapy (22)68.4 Gy64–71 Gy
Grann et al., 199830BrachytherapyNone47mCi I-12532–77 mCi
Merrick et al., 200133Brachytherapy103-Pd (50) or 125-I seed implants (50), EBRT then brachytherapy boost (36), antiandrogens (25)Monotherapy/boostND
125-I 51.7/38.8 mCi
103-Pd 207/146 mCi
Kaminski et al., 200231EBRT, 3D conformalND74.88 Gy62.60–84.09 Gy
Kestin et al., 200232EBRT + high dose brachytherapy1991–95, 3 interstitial implants (ND), from October 95, 2 implants (ND)92.0 Gy76.3–118.0 Gy
Pollack et al., 200435EBRT, 3D conformalProstate only or extended field if stage T2c/3, Gleason score 7–10 or PSA >10 ng/mL (ND)74 Gy63–84 Gy
Merrick et al., 200534Brachytherapy103-Pd (69) or 125-I (31), EBRT (55)NDND

Test Methods

Data specific to detection and assessment of perineural invasion in prostatic cancer (Table 1) were limited.

Four studies provided the median18, 23, 32 and/or the mean18, 23, 25, 32 number of biopsy cores taken from each patient, which ranged from 6 to 7 for the median and 5.5 to 7 for the mean. Six of the surgical16–19, 23, 25 and 1 of the radiotherapy32 studies gave the range of the number of cores taken, varying from 1 core,17 2 cores,19, 25, 32 3 cores23 or 4 cores16, 18 to a maximum of 6 cores,19 10 cores,23 12 cores,16, 17, 32 14 cores25 or 17 cores.18 Three articles18, 19, 25 stated needle size. No article gave details of biopsy technique, the amount of prostatic tissue obtained, or histological slide preparation methods.

Perineural invasion was defined in 7 articles,18, 27, 28, 30, 32–34 with 1 article32 restricting the definition to invasion of large nerves only. None of the 21 articles gave the proportion of patients with nerves present in their biopsies.

Twelve studies16–21, 25, 28, 32–34, 36 stated that 100% of biopsies were reviewed for the presence of perineural invasion, and 1 study30 stated that 60% of biopsies were reviewed for the presence of perineural invasion. In 5 studies,23, 26, 27, 29, 35 the slides were not reviewed but the diagnosis taken from the original report. In 3 articles,22, 24, 31 this information was not provided.

Only 2 of19, 32 the 12 articles in which slides were reviewed stated that the reviewing pathologist was blinded to patient outcome.

In 7 articles,16, 19, 28, 30, 32–34 a single pathologist reviewed slides, but in 1 article30 only 60% of cases were reviewed, so that other pathologists were presumably involved in 40%. The review was performed by 2 pathologists in a further 2 articles,17, 25 and 7 articles involved multiple pathologists.20, 21, 23, 27, 29, 35, 36 No information was given in the remaining 5 articles.18, 22, 24, 26, 31

No data were provided on intraobserver or interobserver variations in assessment of perineural invasion.

Outcome Evaluation (Tables 5 And 6)

Individual patient data that allowed reanalysis could be extracted from 5 articles,19, 20, 22, 24, 29 and data on specific subgroups of patients in a further 2 articles.16, 28

Table 5. Outcomes Evaluation—Surgical Series
ReferenceNo. of patients with PNI (%)Definition of PSA, ng/mL, recurrence (No. of measurements)No. of PSA recurrences (%)Definition of clinical recurrenceNo. of clinical recurrences (%)Follow-up: Median (Mean) Range, mo
  1. Positive studies are at the top of the table (Ravery to Quinn); negative studies are at the bottom (Henke to Nelson).

  2. DRE indicates digital rectal examination; NA, not applicable; ND, no data.

Ravery et al., 19942428 (43)>0.1 (progressive exponential rise)40 (61.5)NANAND (24), 6–48
De la Taille et al., 19991777 (24)0.2 or more, (1)46 (14.4)NANAND (25.4), 0.2–62.1
D'Amico et al., 20011653 (7)>0.1, (2 consecutive)NDNANA48 (ND), 6–104
Sebo et al., 200225103 (22.7)0.4 or more, (1)52 (11.45)PSA>0.4, local or systemic progression73 (16.1), with PSA recurrencesND (40), 0.6–68
Freedland et al., 20021833 (17)>0.2, (1)56 (30)NANA32 (40), ND
Quinn et al., 20032365 focal, 7 multifocal (10.4% overall)>0.4 (1)Included with clinical recurrencesDRE abnormal, or start of antiandrogen treatment182 (26.15) all recurrences54.9 (56.9), 1–177.9
Henke et al., 19981921 (28)0.4 or more, (3 consecutive)18 (25)NANAND (19.8), 42 maximum
O'Malley et al., 200222Case controlled>0.2, (1)35 (22.44)Biopsy/imaging positive or nodule in prostatic bed with raised PSA2 (1.3) local, 8 (5.13) systemicND (84.6 with PNI, 94.6 no PNI), ND
Nelson et al., 200220102 (17.3)>0.2, (1)71 (12)NANAND (16), ND
Nelson et al., 200321199 (16)>0.2, (1)183 (12.9)Biopsy/imaging positive or pathological fracture26 (1.8) local or systemicND (37), 2–99
Table 6. Outcomes Evaluation (Radiotherapy)
ReferenceNo. of patients with PNI (%)Definition of PSA, ng/mL, recurrence (No. of measurements)No. of PSA recurrences (%)Definition of clinical recurrenceNo. of clinical recurrences (%)Follow-up: Median (Mean) Range, mo
  1. Positive studies are at the top of the table (Bonin to Wong); negative studies are at the bottom (Grann to Merrick).

  2. ASTRO indicates American Society for Therapeutic Radiology and Oncology definition, which is 3 consecutive increases in PSA after reaching PSA nadir; DRE, digital rectal examination; NA, not applicable; ND, no data.

Bonin et al., 19972939 (8)>1.5, (2)108 (22)NANA28 (29), 2–75
Anderson et al., 19982719 (7)1.5 or more, (2)35 (13)NANA48 (ND), 2–120
Algan et al., 19992612 (9)1.5 or more, (2)NDNANA50 (49), 3–100
Beard et al., 20042886 (23)ND137 (36)NANA32 (ND), 3–204
Wong et al., 20043630 (9)ASTRONDPositive DRE or metastasis on imagingND52 (ND), 3–114
Grann et al., 19983019 (18.6)>0.1, (1)NDNANA36 (ND), 12–84
Merrick et al., 200133105 (24.7)ASTRONDAbnormal DRE/distant metastasis19 (4.5) PSA or clinical recurrences35.4 (37.1), 6–74
Kaminski et al., 200231,80 (12)ASTRONDNANA56 (ND), 24–126
Kestin et al., 20023268 (42.5)ASTRO45 (28.1)Abnormalities on imaging25 (15.6)50 (ND), 12–115
Pollack et al., 20043598 (12)ASTRONDNANA63 (ND), 2–154
Merrick et al., 200534133 (26)PSA>4 after nadir or ASTRONDNANA62 (68), minimum 48

The proportion of patients with perineural invasion varied between 7 and 12% (median, 9%) when the diagnosis was taken from the original reports and between 7 and 42.5% (median, 23%) when slides were reviewed to identify perineural invasion.

Definitions and frequencies of biochemical or clinical recurrences varied but with no clear differences between articles reporting positive or negative results.

Two of the surgical articles that reported no prognostic value for perineural invasion19, 20 had a mean follow-up period of <2 years, but there were no other clear differences between studies. When the minimum follow-up period was given, it was ≤6 months in all but 4 articles.30–32, 34

Summary of Results (Tables 7 and 8)

Six16–18, 23–25 of the 10 surgical and five26–29, 36 of the 11 radiotherapy cohort studies identified perineural invasion as a prognostic factor at least in univariate analysis (“positive” studies), whereas two orginal,19, 20 and one follow-up21 surgical and five original30–33, 35 and one follow-up34 radiotherapy articles did not (“negative” studies). The only case-control surgical study was negative.22

Table 7. Results—Surgical Series
ReferenceSelection criteriaUnivariate RR (95% CI)Multivariate analysis RR (95% CI)RFS PNI presentRFS If No PNIKaplan-Meier P (timepoint, y)
  • Positive studies are at the top of the table (Ravery to Quinn); negative studies are at the bottom (Henke to Nelson).

  • CI indicates confidence interval; HR, hazard ratio; PNI, perineural invasion; RFS, recurrence free survival; RR, relative risk; ND, no data; NS, not significant.

  • *

    Reanalysis with GraphPad Instat version 3, GraphPad Software, San Diego, Calif, from individual patient data.

  • Low risk: PSA 10ng/mL or less, biopsy Gleason score 6 or less, stage T1c, 2a and 50% or less of biopsies containing cancer; or PSA 10.1 to 20 ng/mL, biopsy Gleason score 7, stage T2b or less than 34% of biopsies containing cancer. Intermediate risk: PSA 10ng/mL or less, biopsy Gleason score 6 or less, stage T1c, 2a and greater than 50% of biopsies containing cancer; or PSA 10.1 to 20 ng/mL, biopsy Gleason score 7, stage T2b or 34 to 50% of biopsies containing cancer; or PSA greater than 20 ng/ml, biopsy Gleason score 8–10, stage T2c or less than 34% of biopsies containing cancer. High risk: PSA 10.1 to 20 ng/mL, biopsy Gleason score 7, stage T2b or greater than 50% of biopsies containing cancer; or PSA greater than 20 ng/mL, biopsy Gleason score 8–10, stage T2c or 34% or more of biopsies containing cancer.

  • Estimated from graph.

Ravery et al., 199424Unselected1.507 (1.01–2.23), P = .048*NDNDNDND
De la Taille et al., 199917UnselectedND2.25 (1.19–4.23), P = .01224%64%.0003 (5)
D'Amico et al., 200116Intermediate and high riskNDND (ND), NSNDNDNS (5)
Low riskNDND (ND), significant82%95%.04 (5)
Sebo et al., 200225Unselected3.01 (1.90–4.78), P < .0011.96 (1.19–3.22), P = .008NDNDND (ND)
Freedland et al., 200218UnselectedNo RR but HR = 1.91 (1.06–3.46), P = .03ND (ND), NSNDNDND (ND)
Quinn et al., 200323Unselected1.87 (1.23–2.84), P = .0031.40 (0.90–2.16), P = .1350% (focal PNI), 0% (multifocal PNI)70%.009 (5)
PSA 10 ng/mL or moreND2.120 (1.258–3.573), P = .005NDNDND
Henke et al., 199819Unselected2.232 (1.03–4.81), P = .064*ND (ND), NSNDNDND
O'Malley et al., 200222Case controlledNDND60%375%.13 (ND)
Nelson et al., 200220Unselected1.500 (0.908–2.47), P = .13*NDNDNDND (ND)
Nelson et al., 200321UnselectedND (ND), NSNDNDNDND (ND)
Table 8. Results—Radiotherapy Series
ReferenceSelection CriteriaUnivariate RR (95% CI)Multivariate RR (95% CI)RFS PNI presentRFS if No PNIKaplan-Meier P (timepoint, y)
  • Positive studies are at the top of the table (Bonin to Wong); negative studies are at the bottom (Grann to Merrick).

  • CI indicates confidence interval; HR, hazard ratio; PNI, perineural invasion; RFS, recurrence-free survival; RR, relative risk, ND, no data; NS, not significant.

  • *

    Re-analysis with GraphPad Instat version 3, GraphPad Software, San Diego,USA from individual patient data.

  • Assumes that the consistent statement that perineural invasion was a poor prognostic factor is correct.

  • On the basis of 1992 AJCC T-stage, pretreatment PSA level and biopsy Gleason score, no further details.

  • §

    Confirmed by Dr Wong.

  • Estimated from graph.

Bonin et al., 199729Unselected1.84 (1.19–2.84), P = .01*ND (ND), P = .6862%79%.01 (3)
PSA < 20 ng/mL2.87 (1.61–5.10), P = .002*ND (ND), P = .00265%88%.0009 (3)
Anderson et al., 199827PSA < 10 ng/mLNDND (ND), P = .000859%80%.0005 (5)
Algan et al., 199926PSA ≥ 20 ng/mLNDND27%36%Univariate 0.43, multivariate 0.01 (4)
Beard et al., 200428Unselected1.73 (1.33–2.25), P = .0002*ND47%69%.001 (5)
Low riskNDND50%80%.04 (5)
Intermediate riskNDND70%75%.72 (5)
High riskNDND (ND), NS29%53%.03 (5)
Wong et al., 200436UnselectedNDND (ND), P = .2248%§64%§.01 (5)
Grann et al., 199830Localized cancer, Gleason scores 4–7NDND50%70%.31 (5)
Merrick et al., 200133UnselectedNDND93.1%94.8%.50 (5)
Kaminski et al., 200231UnselectedNDNDNDNDND
Kestin et al., 200232Locally advanced cancerNDND61%75%.07 (5)
Pollack et al., 200435UnselectedND0.92 (0.61–1.41), P = .710NDNDND
Merrick et al., 200534UnselectedNDND94%94.9%.67 (8)

The analysis of results varied, the majority calculating the relative risk (RR)17, 23, 25, 35 and/or recurrence-free survival percentages by using the Kaplan-Meier method16, 17, 23, 26–30, 32–34, 36 but with no confidence interval (CI) for differences in survival. Three negative studies21, 22, 31 did not provide detailed statistics.

Surgical series

In multivariate analysis with serum PSA and biopsy Gleason score, 2 studies18, 24 found that perineural invasion was not an independent prognostic factor. However, these may have been biased by the high percentage of patients excluded from final analysis (35% in 124 to 42% in the other18).19, 24 Studies that included a larger number of patients (Table 2) and that excluded fewer patients (0%–17%)16, 17, 23 reported that perineural invasion was independently prognostic. In 1 study,23 perineural invasion was significant on univariate analysis including all patients, but this finding was maintained in multivariate analysis only in patients with PSA values >10 ng/mL. These authors found that outcomes in the overall group were particularly poor for those patients with multifocal rather than focal perineural invasion, but only 7 patients were in the multifocal group, precluding further analysis. Perineural invasion was an independent variable in a regression model including only Gleason score and a proliferative marker, MIB-1, the focus of 1 article.25

Radiotherapy series

More than two-thirds of the studies that used external beam radiotherapy26–29, 36 but none of those that used brachytherapy30, 32–34 showed prognostic significance for perineural invasion. Two articles26, 36 appear to contain errors in their data for recurrence-free survival. In 1 article,26 it was given as 27% in the absence of perineural invasion versus 36% if perineural invasion was present. However, the presence of perineural invasion was consistently discussed as a positive predictor of recurrence, although in multivariate but not univariate analysis, which suggests that the figures were transposed. In the other article,36 the abstract section gives the recurrence-free survival as 64% for patients without perineural invasion and 48% for patients with perineural invasion, but the results section states the opposite. The abstract results section data have been confirmed as correct by 1 of the authors. Two articles showed a greater significance for perineural invasion in patients with PSA values below28, 29 or above28 certain limits. Negative articles did not provide subgroup analyses.

In multivariate analysis, in a group of patients with PSA values <10 ng/mL,27 perineural invasion was independently associated with recurrence (P = .0008) together with Gleason score (P = .0496) and stage (P = .0153). Bonin et al.29 also demonstrated independent significance for perineural invasion in patients with a serum PSA level of <20 ng/mL. Treatment dose was not an independent predictor of outcome in these 2 selected series.27, 29 Bonin et al.29 were not able to analyze separately patients with PSA values >20 ng/mL because of their small number of patients. However, Algan et al.26 found that perineural invasion was an independent prognostic factor in addition to treatment dose, clinical stage, and Gleason score, but not serum PSA. Beard et al.28 performed a multivariate analysis only in the high-risk group and reported no prognostic value for perineural invasion, only for PSA levels of ≥20 ng/mL and Gleason scores of ≥8. Wong et al.36 found that only the risk group, defined by Gleason grade and PSA level, and the percentage of biopsies positive for cancer were independently associated with recurrence.


This is the first time, to our knowledge, that a systematic approach has been applied to the investigation of the prognostic value of a standard histopathological factor. Even the widely used UICC TNM clinical and pathology staging system,39 which has been valuable in achieving international consistency, was historically based on consensus from expert opinion,40 an approach that may not detect sources of variability, bias, and error. It is only by examining the literature for all essential data items in the context of a systematic review that the lack of key information and the variability in virtually all parameters across all studies are clearly shown.

The most striking finding concerning study design was that none of the investigators considered that the presence of nerves in the biopsies was a prerequisite for patient inclusion. Nerves are not systematically sampled; 1 study documented their presence in only 45% of cases, the number of nerves present in each case varying from 0 to 16 (median, 0).41 Variations may be due to biopsy technique, as the greatest density of nerves is posterior.11 The number of cores obtained and the number of sections taken at different levels through the tissue influence the surface area of prostatic tissue and, therefore, the potential number of nerves available for examination, as up to 23% of the core length may not be sampled if only one histological section is taken.42 No data were provided on biopsy technique or slide preparation for histological study. Only one-third of articles provided information on the number of cores taken, which varied markedly within and between studies. Therefore, a wide variation in the proportion of biopsies with nerves is likely, and this could be a large factor in obscuring the true significance of perineural invasion. Cases with no nerves in the biopsy sample are essentially uninformative, and their inclusion in the “no perineural invasion” group for analysis could erroneously distort data and underestimate the value of the “perineural invasion test.”

There were marked variations in the frequency of perineural invasion, which appeared to be higher when biopsies were reviewed to detect it.16–21, 25, 28, 32–34, 36 As pathologists are not required to routinely report perineural invasion, its frequency and/or significance may be underestimated in studies that collate this information from original reports.23, 26, 27, 29, 35 No article provided data on the reproducibility of perineural invasion diagnosis, which may be a further factor in confounding results, particularly because several studies17, 18, 22–27, 29, 31, 35, 36 involved multiple pathologists.

The highest frequency of perineural invasion was seen in patients with locally advanced disease.32 This is in keeping with the recognized role of perineural invasion in prostate cancer spread,10 and it may be that once spread beyond the gland has occurred, perineural invasion has no further prognostic role.32 However, whether other differences in disease severity between the populations investigated could be responsible for reported differences in the significance of perineural invasion is difficult to assess because of the lack of consistent data on population characteristics (Tables 2 and 3).

An important issue for patients opting for radical prostatectomy is whether sacrificing the neurovascular bundle when perineural invasion is present reduces the risk of subsequent biochemical recurrence, but no study considered nerve sparing as a confounding factor in the analysis of results. Treatment modality, radiation field, and intensity varied within and between radiotherapy studies. Six articles26, 27, 29, 31, 32, 35 found significant associations between radiation dose and outcomes, but both articles that selected patients with low PSA values27, 29 found that perineural invasion, but not treatment dose, was an independent predictor of outcome. A further study of patients with PSA values of ≥20 ng/mL also found that it was an independent predictor.26 Negative articles did not perform subgroup analyses. Variations in treatment in the positive articles appeared to be protocol-driven on the basis of disease severity (Table 4). However, in some negative studies, differences in seed type33, 34 and number of interstitial implants32 were dependent on the time period within the study. Also, although the presence of perineural invasion was not a factor determining selection for supplemental radiation in 2 related articles,33, 34 significantly more patients with perineural invasion received such boosts, possibly reducing the significance of perineural invasion.

In general, results of statistical analyses were inconsistently presented, and confidence intervals for differences in recurrence-free survival rates were not provided. There was reliance on P values to dichotomize results into statistically significant or not, which can be misleading if the sample size is too small to detect true differences in outcomes. Sample size was determined by numbers of patients and data available rather than by statistical power considerations, as would be required for the design of clinical trials. The relatively short and variable minimum and average follow-up periods could also affect analyses. It has been established that patients who have not suffered a biochemical relapse 2 years after prostatectomy have a 90% recurrence-free survival rate,43 but the mean follow-up period was <2 years in 2 negative studies.19, 20 Finally, none of the negative studies assessed the importance of perineural invasion in specific patient groups. Instead, these studies combined all patients regardless of clinical stage, PSA levels, or biopsy Gleason scores and performed Cox regression analysis. Yet a limitation of the Cox model is that it can be difficult to represent complex interactions in which a variable has a different effect depending on the level of another variable.44 This may explain why studies that analyzed groups of patients defined by specific PSA levels23, 26, 27, 29 or other factors, such as clinical stage, Gleason grade, and biopsy tumor extent,16 found that perineural invasion was an independent prognostic factor.

Should perineural invasion influence treatment decisions, as currently thought by 47% of surgeons surveyed?13 Considering that the importance of perineural invasion is likely to have been underestimated by the inclusion of uninformative test results (biopsies with no nerves) and that despite this, the majority of studies, including those performing multivariate analyses, ascribed prognostic significance to perineural invasion, the weight of evidence suggests that it should. Furthermore, studies controlling for other factors, such as presenting PSA levels,16, 27–29 found prognostic significance in specific patient groups, which was not considered in the negative studies. Nevertheless, determining the magnitude of risk associated with perineural invasion is difficult, as all of the articles have methodological shortcomings. Given that study design and variability are the fundamental problems, pooling of results in a meta-analysis would not provide a greater degree of certainty. Furthermore, only 24% of studies provided individual patient data, and a meta-analysis of such a selected group could produce misleading results. From a patient's perspective (Sandy Tyndale-Biscoe), examination of data led to the conclusion that men with perineural invasion in their biopsies may be more inclined to choose immediate treatment rather than surveillance. However, the important question of whether perineural invasion is an absolute contraindication for a nerve-sparing radical prostatectomy was not addressed in this series of articles.

Concerns have been expressed about the quality of prognostic studies, but these have mainly focused on research into molecular markers.45 The failure to translate much of the research on these markers into routine clinical use was attributed to poor study design, sparking the initiative to develop the REMARK guidance.38 However, this review suggests that similar problems also affect studies of commonly reported pathology factors, currently, although to a variable extent, used in patient management.13 Publications on perineural invasion span 2 decades, yet the significance of perineural invasion remained uncertain. Through a systematic appraisal, this review has, for the first time, identified sources of variations in studies, any or all of which could explain the discrepant results for perineural invasion as a prognostic factor. Further studies should be designed to fulfill the quality criteria highlighted by this review. Patients and clinicians who treat them require clear information not only about whether there is a risk associated with a given factor, but also about what the magnitude of that risk may be in different treatment contexts and whether strategies exist to reduce it. This will only be achieved if the same stringent rules are applied to the investigation of histopathological prognostic factors, as to the design of clinical trials.