The significance of perineural invasion in early-stage cervical cancer

The significance of perineural invasion in early-stage cervical cancer

Gynecologic Oncology 123 (2011) 561–564 Contents lists available at SciVerse ScienceDirect Gynecologic Oncology journal homepage: www.elsevier.com/l...

290KB Sizes 0 Downloads 85 Views

Gynecologic Oncology 123 (2011) 561–564

Contents lists available at SciVerse ScienceDirect

Gynecologic Oncology journal homepage: www.elsevier.com/locate/ygyno

The significance of perineural invasion in early-stage cervical cancer Karim S. ElSahwi a,⁎, 1, Emma Barber a, 1, Jessica Illuzzi a, Natalia Buza a, b, Elena Ratner a, Dan-Arin Silasi a, Alessandro D. Santin a, Masoud Azodi a, Peter E. Schwartz a, Thomas J. Rutherford a a b

Department of Obstetrics, Gynecology and Reproductive Sciences, Yale University School of Medicine, CT, USA Department of Pathology, Yale University School of Medicine, CT, USA

a r t i c l e

i n f o

Article history: Received 27 July 2011 Accepted 27 August 2011 Available online 2 October 2011 Keywords: Perineural invasion Cervical cancer Risk factors Adjuvant therapy

a b s t r a c t Introduction. Cervical cancer spreads directly and through lymphatic and vascular channels. Perineural invasion is an alternative method of spread. Several risk factors portend poor prognosis and inform management decisions regarding adjuvant therapy. Objective. To evaluate the incidence and significance of PNI in early cervical cancer. Methods. Retrospective chart review of early-stage cervical cancer patients (IA–IIA) from 1994 to 2009. Results. One hundred ninety two patients were included, 24 with perineural invasion in the cervical stroma (cases) and 168 without (controls). The mean age of the cases was 53 years, versus 45.9 in the controls (P = 0.01). PNI was associated with more adjuvant therapy (P = 0.0001), a higher stage (P = 0.005), a larger tumor size (≥ 4 cm) (P b 0.0001), lymphovascular space invasion (P = 0.002), parametrial invasion (P b 0.0001) and more tumor extension to the uterus (P = 0.015). On multivariate analysis using an adjusted hazard ratio, risk factors for recurrence included grade (HR, 95% CI; 3.61, 1.38–9.41) and histopathology (HR, 95% CI; 2.85, 100–8.09). Similarly, risk factors for death included grade (HR, 95% CI; 3.43, 1.24–9.49) and histopathology (HR, 95% CI; 3.71, 1.03–13.33). Perineural invasion was not identified as an independent risk factor for either recurrence or death. The mean follow up time was 56 months. There was no significant difference in recurrence (P = 0.601) or over-all survival (P = 0.529) between cases and controls. Conclusion. While perineural invasion was found to be associated with multiple high-risk factors, it was not found to be associated with a worse prognosis in early cervical cancer. © 2011 Elsevier Inc. All rights reserved.

Introduction In 2010, there were 11,070 new cases of cervical cancer and 3870 cancer-related deaths in the United States [1]. Most deaths from cervical cancer are due to a failure to diagnose the disease prior to an advanced stage. Based on clinical findings, early stage refers to stage IA, IB1 and non-bulky IIA [2]. In this setting radiation therapy yields equivalent results to radical hysterectomy and lymph node dissection. However the surgical approach allows ovarian preservation and offers better sexual function. It also makes resection of enlarged lymph nodes possible, and facilitates histopathological evaluation of the parametria and lymph nodes [3]. Lymph node involvement, parametrial involvement, and/or the presence of positive margins on surgical resection specimens are indications for adjuvant therapy to prevent locoregional recurrence. Large tumor size, deep stromal

⁎ Corresponding author at: Dept. Ob/Gyn & Reproductive Sciences, Yale University School of Medicine, FMB Building, Room 328, 333 Cedar St. P.O. Box 208063, New Haven, CT 06520-8063, USA. Fax: + 1 203 785 6728. E-mail addresses: [email protected], [email protected] (K.S. ElSahwi). 1 K. S. ElSahwi and E. Barber contributed equally to this manuscript. 0090-8258/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.ygyno.2011.08.028

invasion, and lymphovascular space invasion are also emerging as prominent risk factors for recurrence of disease in published trials [4,5]. While it is generally accepted that cervical cancer spreads directly as well as through vascular and lymphatic channels, a less well studied method of cancer spread is perineural invasion (PNI) [6]. Described in the literature since the mid-1800s, perineural invasion has been identified in approximately 90% of surgical specimens in cancers of the head and neck, prostate, and pancreas and in up to 33% of colorectal cancer specimens. PNI has been found to be independently associated with high recurrence rates and poor survival in these cancers. Indeed, assessment of PNI status is a required component of the pathologic analysis of head and neck cancers and significantly affects surgical strategies and adjuvant treatment, although the rationale for changing therapy based on PNI status is thus far anecdotal [7]. Memarzadeh et al. investigated the significance of PNI and lymphovascular space invasion in early-stage cervical cancer in an effort to better describe the histologic characteristics of disease status in the “parametria”. In patients with tumors more than 4 cm in size, the presence of PNI more than doubled the risk of recurrence or death (RR 2.5; P b 0.05). In node-negative patients, PNI was found to more than triple the risk of recurrence (RR 3.7, P b 0.05) [6]. Aware of the emerging evidence regarding the role of PNI in cervical cancer, we set out to study the incidence and significance of PNI in our

562

K.S. ElSahwi et al. / Gynecologic Oncology 123 (2011) 561–564

patient population with early cervical cancer, and whether it has influenced our treatment decisions over the years. Patients and methods Approval was obtained from the Human Investigative Committee at The Yale School of Medicine. In this retrospective cohort study, electronic medical records were searched for the terms “cervical cancer” and “hysterectomy” from 1994 to 2007. We included all patients with stage IA to IIA with squamous cell carcinoma, adenosquamous, or adenocarcinoma of the cervix. Rare histologic cancer types, such as small cell and neuroendocrine tumors, were excluded. Patient demographics, comorbidities, cancer status, and survival data were abstracted. Hospital inpatient records, outpatient clinic records, operative records, radiographic imaging reports, and pathology reports were reviewed. For the purpose of this study, a secondary review of histopathology slides of cases with PNI or parametrial invasion was conducted by a gynecologic pathologist, who was blinded regarding patient outcome. Tumor characteristics including histopathology, grade, stage, size, depth of invasion, and parametrial stromal or lymph node invasion were recorded. The presence of pelvic and paraaortic lymph node metastases, as well as vaginal and uterine extension, were noted. Also abstracted was the presence of lymphovascular invasion in the cervix and parametrium, as well as the presence of perineural invasion in the cervix and parametrium. Hematoxylin and eosin staining was used for identification of PNI, which was defined as the presence of cancer cells in or immediately surrounding nerve bundles. Parametrial extension was defined as any invasion of cancer cells into stroma, lymph nodes, lymphovascular spaces, or perineural invasion in the parametrium. In the final analysis, metastases in either the pelvic or paraaortic lymph nodes or both were collectively defined as lymph node metastases. Of the 192 patients included in this study, most underwent a class III radical hysterectomy and retroperitoneal lymph node dissection (n = 159). Thirty two patients underwent a class II radical hysterectomy, and one patient underwent a class I extrafascial hysterectomy. External beam pelvic radiation therapy with or without brachytherapy was given in an adjuvant fashion when warranted. When chemoradiation was recommended, cisplatin was the agent used as a radiosensitizer. Seven patients received neoadjuvant chemotherapy for unresectable clinical stage IB disease. Parametrial extension and/or the presence of lymph nodes fixed to major pelvic vessels determined the need for neoadjuvant chemotherapy, and were identified either at an initial surgical exploration or by radiographic imaging. All patients underwent definitive surgery after 3–7 cycles of chemotherapy. We used the Yale University platinum bleomycin methotrexate–fluorouracil (PBM–PFU) alternating protocol [8]. The dates and sites of recurrence were recorded including date of the last follow up visit and the present status of all patients. Survival curves, generated by the Kaplan–Meier method, were compared with the log-rank test. Multivariate analysis using a stepwise regression model was also performed. Chi-square was used as a test of association, with the application of the two-tailed Fisher exact test where appropriate. P b 0.05 was considered statistically significant. Statistical analysis used SAS 9.2 software (SAS Institute, Cary, NC). Results Two hundred thirty five patients with early stage cervical cancer (IA–IIA) were treated at Yale-New Haven Hospital from 1994 to 2009. Forty-one subjects had stage IA disease, and were excluded as they had no deep stromal or parametrial invasion. Two more patients were excluded due to lack of sufficient information, leaving 192 subjects in our study. Cases with PNI or parametrial invasion in the original pathology report were reviewed. PNI was confirmed in all cases originally reported. PNI was also found in six additional cases with

parametrial involvement. Twenty four (12.5%) patients were found to have PNI either in the cervix (23) or in both the cervix and the parametrium [1], twenty-three patients with stage IB, and one with stage IIA. One hundred sixty eight patients had no evidence of PNI and were classified from stage IB1 to IIA (Table 1). The mean age in years of subjects with PNI was 53 years, compared to 45.9 years in subjects without PNI (P = 0.01). The distribution of tumor histology was comparable between the two groups (P = 1); adenocarcinoma and adenosquamous carcinoma were grouped together during data analysis and compared to squamous cell carcinoma. There was significantly more stage IB2/IIA disease in subjects with PNI compared to subjects without PNI (37.5% vs. 11.9%, P = 0.005) (Table 1). Twenty subjects (83.3%) in the PNI group and 139 subjects (82.7%) in the non-PNI group underwent a class III radical hysterectomy (P = 0.77). The remaining patients underwent either a class II radical hysterectomy (12.5% and 17.3%, respectively) or an extrafascial hysterectomy (n = 1 in the PNI group). Neoadjuvant chemotherapy was used in 2 subjects with PNI (8.3%) and 5 subjects without PNI (3%) (P = 0.21). Patients in the PNI group received significantly more adjuvant therapy than those without PNI. Seventeen subjects with PNI (70.8%) received either radiation or chemoradiation compared to 52 patients without PNI (31%), (P = 0.0001) (Table 2). Chi square test was used to measure the association between PNI and known high-risk factors. It was found that 10 subjects with PNI (41.7%) vs. 18 subjects without PNI (10.7%) had tumors ≥4 cm (P b 0.0001). Lymphovascular space invasion was more common in the PNI group (n = 18/24, 75%) than in non-PNI group (n = 67/168, 39.9%), (P = 0.002). Eleven patients (45.8%) in the PNI group had parametrial invasion as compared to 5 (3%) in the non-PNI group (P b 0.0001). Ten patients in the PNI group (41.7%) had tumor extension to the uterus as compared to 31 patients (18.5%) in the nonPNI group (P = 0.015). There was no difference between the two groups regarding lymph node metastasis, depth of invasion, tumor grade, or extension to the vaginal margin (Table 2). Cox proportional hazard regression modeling was used to identify factors associated with a high risk of recurrence (Table 3) and death (Table 4) in the study cohort. Risk factors for recurrence in unadjusted models were found to be advancing stage (HR, 95% CI; 3.22, 1.71–6.03), advancing grade (HR, 95% CI; 4.004, 1.73–9.29), uterine Table 1 Patient characteristics by perineural invasion (PNI) status.

Mean age (years) Stage (%) IB1 IB2 IIA Histology (%) SCCb Adeno/adenosquamous carcinoma Neoadjuvant (%) Surgery (%) Radical hysterectomyc Otherd Missing Any adjuvant therapy (%) Type of adjuvant therapy Chemoradiation Radiation Chemotherapy Missing

Subjects with PNI (n = 24)

Subjects without PNI (n = 168)

P valuea

53 (± SD)

45.9 (± SD)

0.01 0.005

15 (62.5) 8 (33.3) 1 (4.2)

148 (88.1) 18 (10.7) 2 (1.2)

14 (58.3) 10 (41.7) 2 (8.3)

98 (58.3) 70(41.7) 5 (3)

20 (83.3) 3 (12.5) 1(4.2) 17 (70.8)

139 (82.7) 29 (17.3) 0 52 (31)

12 (70.6) 5 (29.4) 0 0

15 (28.8) 26 (50) 9 (17.3) 2 (1.2)

1.00

0.21 0.77

0.0002 0.0001

Data are mean (± standard deviation) for age and n (%) for all other categories. a Chi-square test. b SCC, squamous cell carcinoma. c Class III radical hysterectomy. d Class II radical hysterectomy, extrafascial hysterectomy.

K.S. ElSahwi et al. / Gynecologic Oncology 123 (2011) 561–564 Table 2 Pathological findings by perineural invasion (PNI) status.

Size ≥ 4 cm (%) LVSIa (%) Parametrial invasion (%) Uterine Extension (%) LNb metastases (%) Depth of Invasion N 2/3 (%) Positive vaginal margin (%) Grade (%) 1 2 3

563

Table 4 Risk of death on multivariate analysis.a

Subjects with PNI

Subjects without PNI

n = 24

n = 168

10 18 11 10 5 18 4

(41.7) (75) (45.8) (41.7) (20.8) (75) (16.7)

18 (10.7) 67 (39.9) 5 (20.8) 31 (18.5) 20 (11.9) 105 (62.5) 13 (7.7)

1 (4.2) 13 (54.2) 10 (41.7)

31 (18.5) 88 (52.4) 49 (29.2)

P value

0.0001 0.002 0.0001 0.015 0.21 0.26 0.24 0.14

Histopathology Stage Grade Size PNIc LVI Uterine extension Parametrial extension LN metsd Adjuvant therapy Age a

Data are n (%). a LVSI, lymphovascular invasion. b LN, lymph node.

b c d

extension (HR, 95% CI; 3.92, 1.59–9.66), lymphovascular space invasion (HR, 95% CI; 3.52, 1.27–9.79), size ≥4 cm (HR, 95% CI; 4.92, 1.97–12.27), and adjuvant therapy (HR, 95% CI; 3.2, 1.26–8.13). Only advancing grade (HR, 95% CI; 3.61, 1.38–9.41) and histopathology (HR, 95% CI; 2.85, 1.00–8.09) remained statistically significant independent risk factors, and adeno/adenosquamous carcinoma portending a higher risk of recurrence compared to squamous carcinoma. Similarly, tumor size ≥4 cm (HR, 95% CI; 4.55, 1.75–11.83), advancing stage (HR, 95% CI; 3.04, 1.59–5.83), advancing grade (HR, 95% CI; 3.82, 1.55–9.43), lymph node metastasis (HR, 95% CI; 4.27, 1.57– 11.58), uterine extension (HR, 95% CI; 3.15, 1.22–8.18), and lymphovascular space invasion (HR, 95% CI; 3.57, 1.16–10.99) were found to be risk factors for death in this population in unadjusted analyses. Again, only advancing grade (HR, 95% CI; 3.43, 1.24–9.49) and adeno/ adenosquamous histology (HR, 95% CI; 3.71, 1.03–13.33) remained independent risk factors for death in adjusted proportional hazard models. Perineural invasion was not identified as an independent risk factor for either recurrence or death in bivariate or multivariate analytic models. In this retrospective analysis, the mean follow up time was 66 months. Kaplan–Meier curves for recurrence and survival were henceforth constructed and compared using the log-rank test. There was no significant difference in recurrence (P = 0.601) or over-all survival (P = 0.529) between subjects with PNI and those without PNI in our study population (Figs. 1 and 2).

Unadjusted hazard ratio

Adjusted hazard ratiob

(95% Confidence interval)

(95% Confidence interval)

1.19 3.04 3.82 4.55 1.44 3.57 3.15 1.41 4.27 1.76 1.01

0.96 2.35 3.43 1.22 0.99 1.93 2.71 0.49 3.71 0.26 1.02

(0.46–3.09) (1.59–5.83) (1.55–9.43) (1.75–11.83) (0.44–4.46) (1.16–10.99) (1.22–8.18) (0.4–4.96) (1.57–11.58) (0.68–4.58) (0.97–1.05)

(0.32–2.59) (0.68–8.15) (1.24–9.49) (0.32–4.72) (0.25–3.87) (0.48–7.8) (0.71–10.41) (0.09–2.61) (1.03–13.33) (0.06–1.16) (0.98–1.07)

Cox proportional hazard. Adjusted for all variables in the table. Perineural invasion. Lymph node metastasis.

on our management strategies. The incidence of PNI in early-stage cervical cancer (IB1–IIA) was 12.5%. PNI was significantly associated with larger tumors (≥4 cm) (P = 0.0001) as well as with higher stage disease (P = 0.005). There was also an association with LVSI (P = 0.002) and uterine extension (P = 0.015). Moreover, patients with PNI received significantly more adjuvant therapy than patients without PNI (P = 0.0001). This, at first glance, seemed to imply that PNI was associated with more aggressive disease; however, on multivariate analysis, PNI was not found to be an independent high-risk factor that predicted either recurrence (HR, 95% CI; 0.74, 0.16–3.4) or death (HR, 95% CI; 0.99, 0.25–3.87). It is worth pointing out that this remained to be true even after controlling for adjuvant therapy in a Cox proportional regression model. Whether our patients with PNI received more adjuvant therapy because of this particular histological characteristic or because of its association with more established markers of poor prognosis is not clear. Upon constructing Kaplan–Meier time-to-event curves, PNI also failed, in our study, to portend recurrence or death after a mean follow up of 56 months. We compared our results to a prior report on PNI in cervical cancer by Memarzedah et al. The latter authors reported a 7% incidence of PNI in 93 patients with stage IA2–IIA, followed over a median of 33 months. On univariate analysis, the presence of parametrial PNI doubled the risk of recurrence or death,

Discussion We investigated the incidence of PNI in 192 cases with early-stage cervical cancer, and sought to evaluate its significance and its effect

Histopathology Stage Grade Size PNIc LVI Uterine extension Parametrial extension LN metsd Adjuvant therapy Age a b c d

Unadjusted hazard ratio

Adjusted hazard ratiob

(95% Confidence interval)

(95% Confidence interval)

2.53 3.22 4.004 4.92 1.387 3.52 3.92 1.097 2.604 3.2 1.01

2.85 1.563 3.61 1.73 0.74 1.31 1.73 0.71 1.05 1.007 1.00

(0.99–6.42) (1.71–6.03) (1.73–9.29) (1.97–12.27) (0.40–4.77) (1.27–9.79) (1.59–9.66) (0.25–4.76) (0.94–7.25) (1.26–8.13) (0.97–1.04)

Cox proportional hazard. Adjusted for all variables in the table. Perineural invasion. Lymph node metastasis.

(1.00–8.09) (0.51–4.77) (1.38–9.41) (0.5–5.94) (0.16–3.4) (0.33–5.17) (0.504–5.95) (0.11–4.49) (0.32–3.45) (0.28–3.66) (0.96–1.05)

Survival Distribution Function

Table 3 Risk of recurrence on multivariate analysis.a

1.00

0.75

0.50

0.25

0.00 0

25

50

75

100

125

150

175

200

camos STRATA:

CPNI=0

Censored CPNI=0

CPNI=1

Censored CPNI=1

Fig. 1. Recurrence in patients with PNI vs. patients without PNI. Kaplan–Meyer curves showing no significant difference in recurrence between patients with perineural invasion (red).

564

K.S. ElSahwi et al. / Gynecologic Oncology 123 (2011) 561–564

Survival Distribution Function

1.00

0.75

0.50

0.25

0.00 0

25

50

75

100

125

150

175

200

alivemos STRATA:

CPNI=0

Censored CPNI=0

CPNI=1

Censored CPNI=1

Fig. 2. Survival in patients with PNI vs. patients without PNI. Kaplan–Meyer curves showing no significant difference in survival between patients with perineural invasion (red) and those without (black).

with a risk ratio of 2.64 (95% CI, 1.0, 6.3) (P = 0.39). On multivariate analysis, parametrial PNI was found to be a predictor of recurrence (P b 0.05) [6]. The study, however, evaluated the presence of PNI in the parametrium, and the authors did not comment on the presence of PNI in the cervical specimen. In our study, PNI was found in the parametrium in only 1 case, and in the cervix in all 24 cases. Furthermore, parametrial extension was defined in our study as extension of cancer cells into the stroma or lymphovascular spaces in the parametrium, or as parametrial PNI. The definition of neither parametrial extension nor PNI, in the Memarzedah study, was clearly stated. It is possible that parametrial extension confounded the risk attributed to PNI at least in their univariate analysis. In our pathology review, PNI was defined as the presence of cancer cells in or immediately surrounding nerve bundles. The Memarzedah study did not offer a more precise definition. The above discrepancies may not be responsible for the different results encountered, but these findings do highlight the importance of reaching a consensus regarding a standardized definition and reporting of PNI in cervical cancer histopathology reports. A precise definition of PNI has been lacking in the literature, and this has been identified as a barrier to proper reporting. Recently Liebig et al. proposed that finding tumor cells within any of the 3 layers of the nerve sheath or tumor foci outside of the nerve with involvement of ≥33% of the nerve's circumference are sufficient features for defining PNI [7]. PNI is now recognized as a distinct pathologic entity that can be observed in the absence of lymphatic or vascular invasion. New evidence suggests the presence of a complex signaling interaction involving neurotrophic growth factors between nerves and invading tumor cells in its pathogenesis. The reason certain carcinomas exhibit a predilection for PNI and others do not remains unknown. PNI is better described in head and neck cancers where the incidence is as high as 80% in squamous cell

carcinomas. Indeed this is one of the only malignancies in which assessment for PNI status is required by the College of American Pathologists as part of the histopathology report. Though PNI status in head and neck cancers has been reported to impact surgical strategy, this remains anecdotal. PNI has been found to have a high incidence and to adversely impact recurrence and survival in other cancers as well. These include adenocarcinomas of the prostate, colon, and pancreas [7]. The diagnosis of PNI, however, continues to be challenging, as in hematoxylin and eosin stained slides, small PNI foci may be difficult to see or may be hidden behind inflammatory cells or large mucinous pools. It has been shown that on review of slides from patients with head and neck squamous cell carcinomas, the detection of PNI almost tripled from 30% to 82% when slides were stained with the nervespecific antigen Protein S 100 [7,9]. We did not use this staining technique in our study. In conclusion, PNI seems to have a low incidence in early-stage cervical cancer (12.5%). Our data show that it seems to be associated with larger tumors and higher stage, but that it does not adversely impact recurrence or survival on multivariate analysis or by log rank test after a mean follow up of 66 months. A standardized definition and reporting method as well as a histopathology slide review using immunohistochemical nerve-specific antigens should strongly be considered in future research.

Conflict of interest statement The authors have nothing to declare.

References [1] Jemal A, Siegel R, Xu J, Ward E. Cancer statistics, 2010. CA Cancer J Clin Sep–Oct 2010;60(5):77–300. [2] Benedet JL, Bender H, Jones III H, Ngan HY, Pecorelli S. FIGO staging classifications and clinical practice guidelines in the management of gynecologic cancers. FIGO committee on gynecologic oncology. Int J Gynaecol Obstet 2000 Aug;70(2): 209–62. [3] Piver MS, Rutledge F, Smith JP. Five classes of extended hysterectomy for women with cervical cancer. Obstet Gynecol 1974 Aug;44(2):265–72. [4] Peters III WA, Liu PY, Barrett II RJ, Stock RJ, Monk BJ, Berek JS, et al. Concurrent chemotherapy and pelvic radiation therapy compared with pelvic radiation therapy alone as adjuvant therapy after radical surgery in high-risk early-stage cancer of the cervix. J Clin Oncol 2000 Apr;18(8):1606–13. [5] Sedlis A, Bundy BN, Rotman MZ, Lentz SS, Muderspach LI, Zaino RJ. A randomized trial of pelvic radiation therapy versus no further therapy in selected patients with stage IB carcinoma of the cervix after radical hysterectomy and pelvic lymphadenectomy: a gynecologic oncology group study. Gynecol Oncol 1999 May;73 (2):177–83. [6] Memarzadeh S, Natarajan S, Dandade DP, Ostrzega N, Saber PA, Busuttil A, et al. Lymphovascular and perineural invasion in the parametria: a prognostic factor for early-stage cervical cancer. Obstet Gynecol 2003 Sep;102(3):612–9. [7] Liebig C, Ayala G, Wilks JA, Berger DH, Albo D. Perineural invasion in cancer: a review of the literature. Cancer 2009 Aug 1;115(15):3379–91. [8] Chambers SK, Lamb L, Kohorn EI, Schwartz PE, Chambers JT. Chemotherapy of recurrent/advanced cervical cancer: results of the Yale University PBM–PFU protocol. Gynecol Oncol 1994 May;53(2):161–9. [9] Kurtz KA, Hoffman HT, Zimmerman MB, Robinson RA. Perineural and vascular invasion in oral cavity squamous carcinoma: increased incidence on re-review of slides and by using immunohistochemical enhancement. Arch Pathol Lab Med 2005 Mar;129(3):354–9.