Is high grade prostatic intraepithelial neoplasia still a risk factor for adenocarcinoma in the era of extended biopsy sampling?

Is high grade prostatic intraepithelial neoplasia still a risk factor for adenocarcinoma in the era of extended biopsy sampling?

Pathology (June 2010) 42(4), pp. 325–329 PROSTATE Is high grade prostatic intraepithelial neoplasia still a risk factor for adenocarcinoma in the er...

438KB Sizes 0 Downloads 6 Views

Pathology (June 2010) 42(4), pp. 325–329

PROSTATE

Is high grade prostatic intraepithelial neoplasia still a risk factor for adenocarcinoma in the era of extended biopsy sampling? JENNIFER L. MERRIMEN*, GLENN JONES{

AND

JOHN R. SRIGLEY{{

Pathology Downloaded from informahealthcare.com by University of Virginia on 10/27/13 For personal use only.

*Dalhousie University, Halifax, {McMaster University, Hamilton, {Mount Sinai Hospital, Toronto, Canada

Summary Aims: There is controversy regarding the role of high grade prostatic intraepithelial neoplasia (HGPIN) on prostatic needle biopsy (PNB) as a risk factor for prostatic adenocarcinoma. We utilise a large Canadian database to determine whether HGPIN detected on extended PNB is a significant risk factor for prostatic adenocarcinoma. Methods: Pathological findings from PNBs from 12 304 men who underwent initial PNB during an 8 year period were analysed. Patients were included in the study if their initial diagnosis was HGPIN alone or a benign diagnosis, if at least one follow-up PNB was performed, and if both the initial and follow-up PNB contained at least 10 prostate cores. Results: In the benign group of 105 patients and the HGPIN group of 120 patients, 14.1% and 20.8% were diagnosed with prostatic adenocarcinoma, respectively. When the HGPIN group was further subdivided into unifocal (1 core) and multifocal (2 cores) groups, 9.4% and 29.9% developed prostatic adenocarcinoma, respectively (p 5 0.0001). Cox regression analysis adjusting for age and prostate specific antigen (PSA) confirms the significance of HGPIN as a risk factor for prostatic adenocarcinoma (p ¼ 0.0045). Conclusions: Patients with an initial diagnosis of multifocal HGPIN on extended PNB are at a greater risk for subsequent prostatic adenocarcinoma than those with unifocal HGPIN or benign diagnoses. Key words: Prostate cancer, adenocarcinoma, prostatic intraepithelial neoplasia, risk, extended biopsy.

neoplasia (HGPIN) and/or atypical small acinar proliferation (ASAP).2 HGPIN is a neoplastic transformation of the lining epithelium of pre-existing prostatic ducts and acini without stromal invasion which shares a clinical, morphological and genetic relationship with prostatic adenocarcinoma (Fig. 1).3,4 Using a large study population, a recent study from this group shows that HGPIN, especially when multifocal, is a significant risk factor for subsequent prostatic adenocarcinoma.5 The PNBs in the mentioned study were from a mixture of sextant and extended protocols, reflective of a change in PNB practice during the study period. Studies have shown that thorough sampling of the prostate gland (i.e., extended PNB protocols) correlates with an increased rate of cancer detection6–8 and the contemporary approach to PNB typically yields 10 or more prostate cores. Here, we utilise a large Canadian PNB database to address whether HGPIN detected on extended PNB protocols is a significant risk factor for prostate cancer.

MATERIALS AND METHODS Patients and data collection The biopsies were acquired from the practices of 28 community-based urologists working in a large Canadian city. Clinical and pathological information on 15 268 PNBs performed between May 1999 and June 2007 was collected. The biopsies were performed typically for PSA or DRE abnormalities. PNBs were interpreted by three experienced pathologists with expertise in urological pathology. The clinical variables included patient age and PSA. The pathological variables included sign-out pathologist, diagnosis, and extent of sampling (number of sites biopsied and total number of cores). The data were entered into a Microsoft Access database.

INTRODUCTION

Study design

In Canadian men, prostate cancer represents approximately 28% of all new cancer diagnoses and is responsible for approximately 11% of cancer deaths in men.1 Many men over the age of 50, or younger if there is a family history of prostate cancer, elect to undergo annual screening by digital rectal examination (DRE) and serum prostate specific antigen (PSA) level. Prostatic needle biopsies (PNBs) are routinely conducted to evaluate patients with abnormalities on these screening tests. PNB protocol is continually evolving; currently, extended protocols generating 10–14 cores are routinely performed. In 5–10% of cases, the biopsies do not reveal clearly benign or clearly malignant findings but show high grade prostatic intraepithelial

PNBs were initially excluded if there was a history of previous treatment including chemotherapy, radiation, and previous radical prostatectomy (i.e., prostatic bed biopsies), or if there was a history of previous PNB which was not available in the database. After these exclusions, 12 304 patients underwent initial PNB during the study period with the following distribution of diagnoses: benign (4938; 40.1%), HGPIN (1283; 10.4%), ASAP (329; 2.7%), and prostatic adenocarcinoma (5754, 46.8%). Patients were assigned to a diagnostic category based on the pathology results of the initial PNB, regardless of whether HGPIN, ASAP or prostatic adenocarcinoma was diagnosed on subsequent biopsies. HGPIN diagnostic criteria were those described by Bostwick et al.4 Patients with ASAP alone, or ASAP with HGPIN, were included in the ASAP diagnostic category. Patients with HGPIN or ASAP and a diagnosis of adenocarcinoma were included in the prostatic adenocarcinoma diagnostic category.

Print ISSN 0031-3025/Online ISSN 1465-3931 # 2010 Royal College of Pathologists of Australasia DOI: 10.3109/00313021003767306

326

MERRIMEN et al.

Pathology (2010), 42(4), June

Patients were eligible for this study if the initial diagnosis was HGPIN alone (unifocal or multifocal) or was benign. Patients were included in the analysis if at least one follow-up PNB was performed and if both the original and follow-up biopsies contained at least 10 cores.

Kaplan-Meier method from the date of the first biopsy until the date of an event or censoring, with contrasts assigned by log-rank tests. Cox regression analyses were conducted, complementing the logistic regression analyses. Statistical significance means p 5 0.05 with all alpha set as twotailed and with a value of 0.05.

Statistical analysis

Pathology Downloaded from informahealthcare.com by University of Virginia on 10/27/13 For personal use only.

Data were analysed using STATA 11 (StataCorp, USA) statistical software. Descriptive statistics included mean and ranges (see tables), and PSA values were transformed into their logarithm for analysis. Contrasts for distributions were assessed by t-test, analysis of variance and KruskallWallis rank tests. Logistic regression was used to test for predictor of the occurrence or absence of an event (e.g., subsequent cancer) and to provide odds ratios relative to a baseline patient group. Plots were generated using

Fig. 1 An example of HGPIN, micropapillary pattern, on prostate core. (A) Low power and (B) medium power (H&E stain).

Table 1

RESULTS HGPIN and benign study population characteristics During the study period 4938 patients (40.1%) had a benign diagnosis and 1283 patients had a diagnosis of HGPIN. Of these patients, 1409 underwent at least one subsequent PNB. When limiting the study population to only patients with at least 10 cores taken on the initial and follow-up PNB, 225 patients remained. The mean, median, and range of core numbers in the initial and follow-up PNBs was 10.9, 10, and 10–16, respectively. The benign group consisted of 105 patients and the HGPIN group consisted of 120 patients. Subdividing the HGPIN further to reflect extent of HGPIN involvement revealed 53 patients with one core involved by HGPIN, 53 patients with two cores involved by HGPIN, nine patients with three cores involved by HGPIN and five patients with four cores involved by HGPIN (Table 1). Patient characteristics including age, PSA, and logPSA for the overall study population and study groups are listed in Table 1. The majority of patients (218, 96.9%) underwent a single follow-up biopsy, while five (2.2%), one (0.4%), and one (0.4%) underwent two, three and five follow-up PNBs, respectively, in the study period. An analysis of variance (ANOVA) showed that patients with any HGPIN underwent first repeat biopsy 0.5 years earlier than patients in the benign group (p 5 0.00005), but there is no difference in mean time to subsequent biopsy within the three subgroups of HGPIN (range 0.72 to 0.79 years; ANOVA p ¼ 0.92). Initial ANOVA analyses showed an association between greater patient age and larger extent of HGPIN (data not shown). This association was expected due to the biological association between HGPIN and advancing age. ANOVA analyses showed no association between HGPIN and the study variables of PSA, logPSA or pathologist examining the specimen (data not shown). Fisher’s exact test also showed no association between reviewing pathologist and diagnosis of or extent of HGPIN (data not shown).

HGPIN as a prostatic adenocarcinoma risk factor in the setting of extended PNB sampling There were 39 prostatic adenocarcinoma occurrences on follow-up in the 225 patients who underwent extended PNB

Patient characteristics of benign and HGPIN groups showing means and ranges of study variables

No. patients Mean age, years Serum PSA, ng/mL* LogPSA No. follow-up PNBs Years to first follow-up PNB

60.5 6.46 1.75 1 1.25

Benign

HGPIN on 1 core

105 (45.5–78.4) (0.68–19.65) (–0.38–2.98) (1–1) (0.31–5.71)

63.2 7.58 1.79 1.08 0.72

53 (47.0–80.0) (0.63–43.0) (–0.46–3.76) (1–5) (0.08–2.09)

HGPIN on 2 cores

64.3 5.99 1.73 1.09 0.76

53 (43.7–79.5) (1.20–10.46) (0.18–2.35) (1–3) (0.08–3.84)

HGPIN on 42 cores

66.4 8.86 1.95 1.14 0.79

14 (54.3–74.9) (3.97–37.6) (1.37–3.62) (1–2) (0.08–2.75)

*PSA was available for 99, 45, 47, and 13 patients of the above groups, respectively. HGPIN, high grade prostatic intraepithelial neoplasia; PNB, prostatic needle biopsy; PSA, prostate specific antigen.

Total study population

62.4 6.75 1.77 1.05 0.98

225 (43.7–80.0) (0.63–43.0) (–0.46–3.76) (1–5) (0.08–5.71)

Pathology Downloaded from informahealthcare.com by University of Virginia on 10/27/13 For personal use only.

SIGNIFICANCE OF HGPIN ON EXTENDED PNB

sampling (Table 2). When assessing the risk for prostatic adenocarcinoma on follow-up biopsy, Pearson chi-squared test and log-rank test for trend revealed a significant difference between the four groups (p ¼ 0.0006 and p ¼ 0.0001, respectively) (Table 2), with increasing prostatic adenocarcinoma risk associated with increasing numbers of cores involved by HGPIN. Kaplan-Meier plot showing time to cancer detection for the benign, unifocal HGPIN (one core), HGPIN on two cores and HGPIN on three or more cores (Fig. 2), shows similar hazards for the benign and unifocal HGPIN groups and similar hazards for the HGPIN on two cores and HGPIN on greater than two cores groups, with the latter two groups showing a higher percentage of patients developing prostatic adenocarcinoma. Logistic regression analysis to assess the impact of patient age at initial biopsy, PSA and pathologist examining the specimen, showed that none of these factors, either independently or combined, was predictive of prostatic adenocarcinoma on follow-up biopsy (Table 3). Logistic regression analysis to determine the impact of each of the HGPIN subgroups using the benign group as the control or baseline group is presented in Table 4. There was a significantly increased risk for prostatic adenocarcinoma when two or more cores were involved by HGPIN. HGPIN on two cores and HGPIN on greater than two cores carried odds ratios

Table 2 Risk of prostatic adenocarcinoma on extended PNB after benign and HGPIN diagnoses on initial PNB

Group Benign Unifocal HGPIN HGPIN on 2 cores HGPIN on 42 cores

No. cores with HGPIN

No. patients

No. patients with PCa on follow-up

Percentage with PCa

0 1 2 42

105 53 53 14

14 5 15 5

13.33% 9.43% 28.30% 35.71%

p value ¼ 0.0001 by log rank test for trend. HGPIN, high grade prostatic intraepithelial neoplasia; PCa, prostatic adenocarcinoma; PNB, prostatic needle biopsy.

Fig. 2 Kaplan-Meier curve showing time to prostatic adenocarcinoma diagnosis for benign, unifocal HGPIN (1 core), HGPIN on 2 cores and HGPIN on 42 cores subgroups (log rank test p ¼ 0.0006).

327

(OR) of 2.57 and 3.61, respectively. Unifocal HGPIN did not carry a significantly higher risk for prostatic adenocarcinoma than the benign group in this analysis. A multivariable logistic regression analysis that included age at initial biopsy reduced the statistical significance of HGPIN as predictive of subsequent cancer adjusted by age, because of the biological association of age and HGPIN. Proportional hazards assumptions were confirmed for all possible predictive variables for Cox regression analyses. Univariable Cox regression analyses demonstrated that age, logPSA, and initial pathologist, were not statistically associated with the subsequent diagnosis of cancer, accounting for the temporal distributions of those events (Table 3). However, baseline pathology (benign, HGPIN in 1 core, HGPIN in 2 cores, and HGPIN in 3 or more cores) was statistically associated with cancer both on univariable (p ¼ 0.0023; Table 4) and multivariable analyses (p ¼ 0.0045, being adjusted for age, logPSA, and reporting pathologist; Table 5). Again, significance was reduced slightly for HGPIN as a predictor when accounting for age, because the presence of HGPIN and advancing age were mutually associated.

DISCUSSION HGPIN is pathogenically related to prostatic adenocarcinoma but there is debate in the literature regarding its current role as a risk factor for subsequent detection of prostatic adenocarcinoma. Two large literature reviews have summarised published studies and shown a wide range of percentage of patients developing prostatic adenocarcinoma after a HGPIN diagnosis (2–100%).2,5 Most previous studies addressing the impact of HGPIN on the development of prostatic adenocarcinoma have been limited by flaws in study design, especially related to creating a valid control group of patients, and small study populations.2 In addition, the majority of studies looking at this issue were performed when prostate sampling typically yielded six prostate cores. Recent studies addressing extended PNB protocols yielding 21 cores have also questioned the significance of HGPIN as a risk factor for prostatic adenocarcinoma.9–11 These studies were also limited by small study size and showed little analysis regarding the extent of HGPIN. Our previous results, using a large study group, showed that patients with a diagnosis of HGPIN on initial PNB are at a greater risk of prostatic adenocarcinoma than patients with a benign diagnosis on initial PNB.5 This risk is independent of age, PSA, extent of sampling and sign-out pathologist.5 Notably it was also independent of the extent of initial sampling of the prostate; that range of sampling included patients with greater than nine cores sampled, but the overall mean number was approximately eight for all patients.5 We have also shown that the risk of subsequent prostatic adenocarcinoma varies with number of sites involved with HGPIN, such that the risk of developing prostatic adenocarcinoma generally increases with greater numbers of sites showing HGPIN and becomes significant when two or more sites are affected.5 We concluded that follow-up should be more rigorous in patients diagnosed with multifocal HGPIN on initial PNB. Many studies have shown that the diagnosis of prostatic adenocarcinoma on PNB correlates with extent of

328

MERRIMEN et al.

Pathology (2010), 42(4), June

Table 3 Multivariate logistic and Cox regression analyses assessing the impact of patient age, PSA and reporting pathologist on the prediction of prostatic adenocarcinoma Logistic regression Variable

Odds ratio (95% CI)

Patient age LogPSA Pathologist A Pathologist B p value for model

Table 4

1.02 1.58 0.99 1.66

Cox regression p value

(0.96–1.07) (0.73–3.46) (0.38–2.59) (0.69–4.02)

0.536 0.248 0.996 0.259 0.341

1.01 1.20 1.03 1.63

(0.96–1.06) (0.59–2.44) (0.42–2.50) (0.73–3.65)

p value 0.751 0.616 0.952 0.230 0.688

Multivariate logistic and Cox regression analyses assessing the impact of extent of HGPIN compared to the benign control group Logistic regression

Pathology Downloaded from informahealthcare.com by University of Virginia on 10/27/13 For personal use only.

Hazard rate (95% CI)

Extent of HGPIN Unifocal HGPIN HGPIN on 2 cores HGPIN on 42 cores p value for model

Cox regression

Odds ratio (95% CI)

p value

Hazard rate (95% CI)

p value

0.68 (0.23–1.99) 2.57 (1.13–5.83) 3.61 (1.06–12.35)

0.479 0.024 0.041 0.0146

1.26 (0.45–3.53) 3.50 (1.68–7.28) 4.23 (1.52–11.81)

0.656 0.001 0.006 0.0023

HGPIN, high grade prostatic intraepithelial neoplasia.

Table 5 Multivariate Cox regression analysis, adjusting for patient age, PSA and reviewing pathologist, showing the impact of extent of HGPIN on the risk of developing of prostatic adenocarcinoma Variable Patient age Log PSA Pathologist A Pathologist B Unifocal HGPIN HGPIN on 2 cores HGPIN on 42 cores

Hazard rate (95% CI) 0.98 1.43 1.14 1.85 1.76 4.15 5.01

(0.93–1.03) (0.68–3.00) (0.46–2.83) (0.82–4.19) (0.58–5.32) (0.58–5.32) (1.48–16.93)

p value 0.378 0.350 0.773 0.138 0.317 0.001 0.009

*p value for model ¼ 0.0045. HGPIN, high grade prostatic intraepithelial neoplasia; PSA, prostate specific antigen.

prostate gland sampling,6–8 indicating a high degree of sampling bias in prostatic adenocarcinoma diagnosis. This issue, specifically related to sampling and HGPIN, has also been addressed.12 Our previous study had many patients with less extensive prostate gland sampling on initial PNB, with an average of 7.5 and 8.1 cores in our benign and HGPIN groups, respectively.5 Although subsequent discovery of cancer was not statistically associated with the number of cores obtained at baseline biopsy, we did not conduct specific, detailed subgroup analyses for the degree of sampling on biopsies done subsequent to the baseline biopsy. Therefore, herein we restricted our analysis to those patients who had 10 or more cores on both their initial and follow-up PNBs, to assess whether HGPIN is significant in the era of extended PNB protocols. This could confirm prior findings, but strengthen the conclusion in relation to adhering to present protocols for biopsies that call for a routine of at least 10 cores per procedure.

As expected, when limiting our analysis to patients with extended PNBs only, we again found similar results to our initial study, despite a smaller sample size. Patients with HGPIN on one core only had no significant increase in risk of prostatic adenocarcinoma on follow-up PNB when compared to patients with a benign diagnosis. However, when patients had two or more cores involved by HGPIN, there was a statistically significant increased risk of prostatic adenocarcinoma. Logistic regression analysis also suggested a similar trend to that seen in our larger study,5 where there was a higher risk of prostatic adenocarcinoma development in patients with three or more cores affected (OR ¼ 3.6) as compared to patients with HGPIN on two cores (OR ¼ 2.6). Respective hazard rates from Cox regression analyses, which more explicitly take into account the temporal patterns of such diagnoses, were 5.0 and 4.1, and were with similar p values for statistical significance; again, HGPIN in only one core was not statistically different from a baseline diagnosis of benign findings. Other investigators have also noted the relationship between extent of HGPIN on initial PNB and subsequent diagnosis of prostatic adenocarcinoma.13–16 Overall, 97% of our patients underwent only two biopsies an average of 0.98 years apart. However, there was a longer delay to first follow-up biopsy in the benign group, which potentially introduced a bias towards the development of cancer, since prostatic adenocarcinoma had more time to develop; despite this, the benign group had a low rate of cancer development. In contrast, the HGPIN groups were older and underwent first follow-up biopsy earlier than the benign group, which raised the issue of a temporal bias towards detection of cancer related to greater sampling of a gland already harbouring prostate cancer. To assess bias, we conducted both logistic (time-independent) and Cox regression (time-dependent) analyses. Any bias

SIGNIFICANCE OF HGPIN ON EXTENDED PNB

Pathology Downloaded from informahealthcare.com by University of Virginia on 10/27/13 For personal use only.

was not a large contributing factor in our findings, since these two analyses gave similar results. The overall pattern of findings indicated that HGPIN was a predictor of cancer when compared to a benign diagnosis, and when there was greater involvement with HGPIN the odds ratio or hazard rate proportionately increased. In summary, when analysing only patients with extended PNB protocols on both initial and follow-up biopsy, we confirm a significant risk of detecting prostatic adenocarcinoma on follow-up biopsy when HGPIN is present on two or more cores in the initial sample. As with our larger mixed sample study population,5 when limiting the analysis to extended PNB protocols only, the results suggest that follow-up should be more rigorous in patients with multifocal HGPIN. We recommend that patients with HGPIN in two or more prostate cores on extended biopsy sampling undergo repeat biopsy within one year. Address for correspondence: Dr J. L. Merrimen, Division of Anatomical Pathology, Queen Elizabeth II Health Sciences Centre, Rm 724, Mackenzie Building, 5788 University Avenue, Halifax, NS, Canada B3H 1V8. E-mail: [email protected]

References 1. Canadian Cancer Society/National Cancer Institute of Canada. Canadian Cancer Statistics 2008. Toronto: Canadian Cancer Society/ National Cancer Institute of Canada, 2008. 2. Epstein JI, Herawi M. Prostate needle biopsies containing prostatic intraepithelial neoplasia or atypical foci suspicious for carcinoma: implications for patient care. J Urol 2006; 175: 820–34. 3. Bostwick DG, Qian J. High-grade prostatic intraepithelial neoplasia. Mod Pathol 2004; 17: 360–79. 4. Bostwick DG, Norle´n BJ, Denis L. Prostatic intraepithelial neoplasia: the preinvasive stage of prostate cancer. Overview of the prostate committee report. Scan J Urol Nephrol Suppl 2000; 205: 1–2.

329

5. Merrimen JL, Jones GJ, Walker D, et al. Multifocal high grade prostatic intraepithelial neoplasia is a significant risk factor for prostatic adenocarcinoma. J Urol 2009; 182: 485–90. 6. Presti JC Jr, Chang JJ, Bhargava V, Shinohara K. The optimal systemic prostate biopsy scheme should include 8 rather than 6 biopsies. Results of a prospective clinical trial. J Urol 2000; 163: 163–7. 7. Chen ME, Troncoso P, Johnston DA, et al. Optimization of prostate biopsy strategy using computer based analysis. J Urol 1997; 158: 2168– 75. 8. Ravery V, Goldblatt L, Royer B, et al. Extensive biopsy protocol improves the detection rate of prostate cancer. J Urol 2000; 164: 393–6. 9. Amin MM, Jeyaganth S, Fahmy N, et al. Subsequent prostate cancer detection in patients with prostatic intraepithelial neoplasia or atypical small acinar proliferation. Can Urol Assoc J 2007; 1: 245–9. 10. Ploussard G, Plennevaux G, Allory Y, et al. High-grade prostatic intraepithelial neoplasia and atypical small acinar proliferation on initial 21-core extended biopsy scheme: incidence and implications for patient care and surveillance. World J Urol 2009; 27: 587–92. 11. Campos-Fernandes J-L, Bastien L, Nicolaiew N, et al. Prostate cancer detection in patients with repeated extended 21-sample needle biopsy. Eur Urol 2009; 55: 600–9. 12. Herawi M, Kahane H, Cavallo C, Epstein JI. Risk of prostate cancer on first re-biopsy within 1 year following a diagnosis of high-grade prostatic intraepithelial neoplasia is related to the number of cores sampled. J Urol 2006; 175: 121–4. 13. De Nunzio C, Trucchi A, Miano R, et al. The number of cores positive for high grade prostatic intraepithelial neoplasia on initial biopsy is associated with prostate cancer on second biopsy. J Urol 2009; 181: 1069–75. 14. Roscigno M, Scattoni V, Freschi M, et al. Monofocal and plurifocal high-grade prostatic intraepithelial neoplasia on extended prostate biopsies: Factors predicting cancer detection on extended repeat biopsy. Urology 2004; 63: 1105–10. 15. Akhavan A, Keith JD, Bastacky SI, et al. The proportion of cores with high-grade prostatic intraepithelial neoplasia on extended-pattern needle biopsy is significantly associated with prostate cancer on sitedirected repeat biopsy. BJU Int 2007; 99: 765–9. 16. Netto GJ, Epstein JI. Widespread high-grade prostatic intraepithelial neoplasia on prostatic needle biopsy: A significant likelihood of subsequent diagnosed adenocarcinoma. Am J Surg Pathol 2006; 30: 1184–8.