Morphologic Pattern, Frequency, and Spatial Distribution of Lymphovascular Invasion Foci in Radical Prostatectomy Specimens

Abstract

Introduction. Lymphovascular invasion (LVI) is an adverse pathological finding in radical prostatectomy (RP) specimens associated with increased risk of metastatic disease. Its variable incidence may be attributed to underreporting. We characterized the location, quantity, and morphology of LVI foci in RP specimens and assessed the relationship between LVI and cribriform and intraductal carcinoma and metastatic risk. Methods. Two pathologists reviewed retrospectively 54 RP specimens reported as positive for LVI. Ambiguous cases were confirmed by immunostaining for ERG, CD31 and D2-40. Results. In 4/54 (7.4%), LVI could not be confirmed. Main mimickers of LVI were retraction artifact and dislodged tumor cells. Based on our review, the most important criteria to distinguish LVI from its mimickers were a corrugated lining of vascular spaces, endothelial nuclei bulging into the lumen, and presence of proteinaceous material. The LVI frequency per case ranged from 1 to 109 (median 7.5). In 47/50 (94%) cases with LVI, the associated carcinoma comprised cribriform pattern and/or intraductal carcinoma. The most common morphology of LVI foci was cribriform, occurring in 43/50 specimens, representing 469/843 (56%) of LVI foci. Most LVI foci were intraprostatic and located at the carcinoma-stroma interface. Particularly the risk of bone metastases during follow-up was independently associated with higher frequency of LVI foci (P = .009). Conclusions. The detailed description of prostatic LVI, and awareness of their predominant location and morphology may help its identification and improve the diagnostic accuracy of LVI in pathology reporting. The clinical impact of LVI quantification in prostate cancer needs validation by further studies.

Introduction

Lymphovascular invasion (LVI) in radical prostatectomy (RP) and biopsy specimens is associated with increased grade, advanced stage and unfavorable clinical outcome.^1–6 After adjustment for grade and pathological stage, presence of LVI remained prognostic for metastatic disease. Therefore, the International Society of Urological Pathology (ISUP) recommends integration of LVI in routine reporting of RP specimens. ⁷ The frequency of LVI reported in RP series varies widely from 5% to 53%. ⁸ This variation may be attributed to variations in prostate cancer population characteristics in each series, but may also be due to differences in pathologists’ reporting, since there are several pitfalls in diagnosing LVI ⁹ and pathologists often consider its diagnosis challenging and subjective.

Several publications have established that among the four Gleason grade 4 sub-patterns (ie poorly-formed, fused, glomeruloid and cribriform), cribriform pattern prostate cancer is particularly associated with regional and distant metastatic disease. Both cribriform pattern 4 prostate cancer and intraductal carcinoma are independent prognosticators for metastatic disease after taking into account grade and pathological stage.^10,11 The relationship between Gleason sub-pattern and LVI has, to our knowledge, yet to be investigated. Similarly, we are not aware of any study focused on the architectural features of LVI foci. Here we describe the association of LVI with cribriform pattern prostate cancer and intraductal carcinoma as well as the morphologic features of LVI and its distribution within and outside the prostate gland. Our findings may help improve the diagnostic accuracy of LVI in pathology reporting. Since we noticed an unexpectedly large variation in number of LVI in our RP series, we also explored the relationship between number of LVI at RP and the risk of regional or distant metastatic disease at follow-up.

Material and Methods

RP Specimens

A total of 54 RP specimens reported positive for LVI (2004-2020) were retrieved from the archives of our institution. Their global Gleason score/Grade group, pathological stage and lymph node status were recorded. Twenty three cases were submitted in toto and the remaining cases were partially submitted per protocol, that is the entire posterior and two transverse anterior sections if weighing over 40 grams. Patient files were reviewed to obtain information regarding treatment and development of metastatic disease and its location during follow-up. Approval from the Institutional Review Board was obtained for this retrospective study. A urogenital pathology fellow first reviewed all slides for presence of LVI, cribriform and/or intraductal carcinoma. Slides with LVI were annotated, number of LVI foci were counted and subsequently the selected slides were reviewed together with a urogenital pathology subspecialty pathologist at a multi-header microscope to confirm LVI. If there was uncertainty or disagreement (30 slides), immunostaining was performed using anti-podoplanin antibody (clone D2-40) as marker for lymph vessels and anti-ERG antibody (clone EPR-3864) as marker for endothelial cells. CD31 was performed in selected cases to further highlight the endothelial cells. In selected cases (two slides), additional markers, including PSA (clone ER-PR8) and CD68 (clone PG-M1) were employed to confirm the prostatic nature of the focus. Consecutive lymphovascular channels separated by scant intervening stroma counted as one LVI focus. For each LVI focus its anatomical location (intraprostatic, extraprostatic, seminal vesicle (SV); including muscle sleeve and surrounding perivesicular adipose tissue, and spatial relationship to the carcinoma was recorded: a) intratumoral, b) at the advancing edge, that is attached to tumor, c) adjacent to but separated from tumor by normal tissue (<5 mm), d) > 5 mm away from tumor. The morphology of the LVI foci was recorded as follows: 1) cribriform, 2) simple glands, 3) cell aggregates, 4) complex/fused glands 5) solid sheet and 6) individual cells, (See Figure 1 for description of each pattern). Features of LVI-positive vascular structures, including vein, lymph vessel, presence of a muscle layer and presence of erythrocytes, fibrin, necrosis or proteinaceous material were also recorded. We did not distinguish lymphatic vessels from small venules or veins because we found histological criteria too inaccurate. Therefore, we decided to use the generic term “lymphovascular (space) invasion” to encompass both.

Figure 1.

Morphologic sub-patterns of LVI foci. (A) Cribriform pattern: Confluent sheet of contiguous tumor cells with easily visible multiple lumina (H&E, x100). (B) Simple glandular: Presence of non-fused glands (H&E, x400). (C) Cell aggregate: Small collection of malignant cells without intervening spaces (H&E, x400). (D) Complex glandular: Fused glands (arrow) (H&E, x100). (E,F) Solid sheet: Confluent sheet of contiguous tumor cells without lumina, and individual tumor cells present singly and discohesive. Note the corrugated shape of the lymphovascular space (E: H&E, x200. F: ERG, x100).

Statistics

Outcome parameters were lymph node metastatic disease and bone metastatic disease at follow-up after RP. Spearman rank correlation was performed for correlation analysis and Fisher exact test is used to calculate the LVI cut-off number. Using the calculated cut-off, multivariable analysis was performed taking into account Gleason score/Grade group and pathological stage. The sample size is small so overfitting is an issue, however multivariable analysis using Firth logistic regression, ie a penalized regression, was performed as an exploratory step to determine if LVI retains an association with metastatic disease endpoints in presence of confounders. A P-value of < .05 was considered statistically significant.

Results

Frequency of LVI

Slide review and immunostaining confirmed LVI in 50/54 (92.6%) cases. The clinicopathological features of the 50 LVI-positive RP specimens are listed in Supplemental Table 1. Among them, 45 revealed cribriform pattern 4 prostate cancer (90%), 32 intraductal carcinoma (64%) and 47 cribriform pattern and/or intraductal carcinoma (94%). The average percentage of the prostate gland involved by carcinoma among LVI positive cases was 22% (2%-90%). The majority of LVI-positive cases (27/50) showed SV invasion (pT3b) and 14 had lymph node metastasis (pN1) at RP. Gleason score was not applicable in three cases due to hormone therapy effects. Foci indeterminate for LVI at initial review were confirmed in 29/32 slides with immunostaining, demonstrating 231 of 246 tumor emboli in D2-40 positive (lymphatic) vessels. LVI foci were further verified by ERG and/or CD31 immunostaining. D2-40 and CD31 showed continuous linear staining in the true lymphatic spaces, while ERG showed a discontinuous nuclear staining of the endothelial cells. The number of foci was consequently changed in 5 slides (from 1 to 68, 2 to 9, 3 to 29, 13 to 32, and 15 to 27). A total of 843 LVI foci were seen in 293 slides with a range of 1-18 LVI positive slides per case (average of 5.86 slides/case). The mean and median number of LVI foci per case was 16.86 and 7.5, respectively, with a maximum of 109. Supplemental Figure 1 shows the case distribution of number of LVI foci. Spearman rank correlation suggests a significant positive association of 0.84 (P = .01) between pathological stage and the number of LVI foci (Supplemental Table 1).

Morphological Descriptors of LVI

During review, the following criteria were deemed most essential (Table 1): In the majority of LVI, we observed that the lymphovascular spaces are dilated, display a corrugated contour (Figure 1E and F, Figure 2A and B) as compared to the smooth outline of the clefting/retraction artifacts surrounding both benign and malignant acini (Figure 2C and D). Nuclei of myofibroblastic stromal cells are within the smooth contour of the retraction artifact while endothelial cell nuclei are often protruding into the vascular spaces. In case of retraction artifact, the malignant glands are often in close vicinity, whereas LVI foci are farther apart, separated by intervening prostatic stroma. In five cases, we observed that tumor emboli fully occupied the lymphovascular spaces, with a minimal slit between the endothelial cells and the emboli, rendering their recognition as LVI very challenging (Figure 2E and F). Additional features supportive of true LVI were the presence of fibrin, necrosis, the disruption of lymphovascular valves by the tumor embolus, the presence of crushing artifact especially at the edge of tumor emboli, attachment to the lymphovascular wall or valve, thrombosis, and inflammatory cells (Figure 3A-D). The lymphovascular spaces may contain proteinaceous material with an eosinophilic hue that can be homogenous or non-homogenous (Figure 3D and E). In one case, LVI extensively involved larger prostatic arteries. The associated foamy and pigmented histiocytes gave a false impression of organized thrombi, however, the tumor emboli were recognized by their basophilia, glandular formation and wall destruction (Figure 3F).

Table 1.

Criteria Used to Define LVI and the Frequently Encountered Imitators of LVI.

Major criteria	Supportive features	Mimickers related to tumor	Mimickers related to normal structures
- Dilated channels - Corrugated surface - Endothelial cell nuclei protruding into the lumen - Proteinaceous material	- Fibrin - Necrosis - Valve disruption - Crushing artifact of tumor emboli - Attachment to the wall or valve - Thrombosis - Inflammatory cells	- Retraction artifact of tumor - Intraductal carcinoma - Mechanical displacement of tumorous tissue into vascular spaces or prostatic duct - Tumor encroachment with vascular spaces - Tumor hemorrhage - Tumor angiogenesis	- Retraction artifact of normal acini and perineural spaces - Mechanical displacement of benign tissue into vascular spaces or prostatic duct - Mechanical displacement of RBCs into clefting artifact - Organized thrombi - Autolysis of seminal vesicle - Dilated prostatic ducts with sloughed epithelium, proteinaceous material and foamy cells

Figure 2.

The major histopathologic criteria of LVI. (A,B) Extensive LVI with different morphologies (cribriform, simple glandular and complex glandular). The true lymphovascular spaces show irregular outlines. ERG immunostaining confirmed presence of endothelial cells, excluding retraction artifact (A: H&E x40. B: ERG, x40, insert x200). (C) Retraction artifact around benign acini (arrow) (H&E, x100). (D) Carcinoma glands with retraction artifact (H&E, x200). (E,F) LVI fully occupying the lymphatic spaces, with only a slit space separating the endothelial cells from the tumor embolus (arrow) (A: H&E, x200. B: ERG, x200).

Figure 3.

Features of true LVI. (A) Fibrin (arrow) associated with cribriform LVI (H&E, x80). (B) Necrosis and wall destruction (arrows) with attachment to the wall (H&E, x100). (C) Crush artifact (long arrow) and permeation through the vascular valve (short arrow). The vascular valve is the luminal folding lined by a single layer of endothelial cells (H&E, x200). (D) Attachment to the vessel wall (short arrow) with the presence of non-homogeneous proteinaceous material and RBCs (long arrow) (H&E, x100). (E) Homogenous proteinaceous material (arrow) (H&E, x100). (F) Tumor embolus involving an artery. The tumor embolus is distinguished by its basophilia (H&E, x100).

Basal cells mimicking endothelial cells in intraductal carcinoma or (dilated) benign ducts or SV with detached epithelial cells or foamy macrophages represented imitators of LVI (Table 1, Figure 4A). The basal cells are more basophilic, and can be readily recognized at low power examination (objective x40 or x100), while the endothelial cell nuclei are amphophilic and not easily identified at low power as they are scant and line the vascular spaces in a discontinuous fashion. Mechanical displacement of tumor fragments into lymphovascular spaces or dilated acini/ducts was another commonly encountered pitfall in our series (Figure 4B and C). We found that a number of cases showed organized thrombi in large vessels (Figure 4D); in such cases, the associated macrophages mimicked tumor emboli (Figure 3F); particularly when the prostate carcinoma displayed a clear cell morphology. We observed that the carcinomatous foci frequently encroach the lymphovascular spaces without real penetration, the endothelial lining wraps around the tumor foci and is present on one side only as confirmed by immunostaining (Figure 4E and F). Hemorrhage into carcinoma and neovascularization of larger tumor areas may also superficially resemble LVI.

Figure 4.

LVI mimickers in RP specimens. (A) Intraductal carcinoma with retained basal cells seen on low power examination (H&E, x40). (B) Mechanical displacement of malignant cells into true lymphovascular space (long arrow). Presence of detached fragments in the surrounding area (short arrow) and absence of any supporting features of LVI (such as fibrin, attachment) argue against a diagnosis of LVI (H&E, x100). (C) Mechanical displacement of malignant glands into a dilated acinus (H&E, x200). (D) Organized thrombus showing foamy cells. The macrophages show prominent nucleoli (see insert) (H&E, x40. Insert: H&E, x400). (E, F) Tumor encroachment of a vascular space; the tumor focus is protruding into a vascular space giving the impression of LVI; ERG highlighting the endothelial cells at the outer surface of the tumor with no permeation into the vessel (arrow). The ERG positivity within the tumor is evidence of a tumor vasculature (E: H&E, x100, F: ERG, x100).

Anatomical and Spatial Distribution of LVI

At case level, the most common anatomical location of LVI was intraprostatic (n = 46), followed by the SV “the surrounding loose connective tissue and muscle sleeve” (n = 27) and extraprostatic (n = 11), with neurovascular bundle location in only one case. Location adjacent to the carcinoma (n = 41) or within the outer advancing edge of the carcinoma (n = 24) was most common. In 20 cases, LVI was identified at more than 5 mm distance from the carcinoma. Intratumoral LVI foci were identified in 20 cases. Relative to the total number of LVI foci (843), 47% of the tumor emboli were adjacent to the tumor, 29% in the SV, 8% intratumoral, 7% > 5 mm away from the tumor, 6% within the advancing edge of tumor and 3% in an extraprostatic location.

Morphological Patterns of LVI Foci

The frequency of morphological patterns of LVI foci is listed in Supplemental Table 2. Cribriform pattern of LVI foci was noted in 43 cases, followed by simple glandular and cell aggregates (n = 29 each), complex glandular (n = 21), solid sheet (n = 19) and individual cells (n = 5). Cribriform pattern represented 56% (469/843) and simple glandular LVI 16% (137/843) of intravascular foci, followed by cell aggregates and solid sheet, complex glandular pattern and individual cells. Necrosis was observed in 11 foci. Among the five RP specimens without cribriform pattern of carcinoma, two nevertheless displayed cribriform morphology within the lymphovascular emboli.

Prognostic Impact of Quantitative LVI Assessment

Among patients with LVI, follow-up data (median 66 months, ranging from two months to 10 years) were available in 44/50 (88%) patients. Twelve showed biochemical recurrence and 15 had metastasis to bone and/or lymph nodes at follow-up after RP. No significant correlation was found between number of LVI and nodal status at time of RP (P = .09). In univariate analysis of the hormone treatment naïve patients with LVI (n = 47), we found that the number of LVI is strongly and positively correlated with both lymph node and bone metastatic disease at follow-up with a cut-off of 20 and 9 LVI foci, respectively. Multivariate analysis further showed that LVI quantification using these cut-off values could independently predict bone metastasis (Table 2).

Table 2.

Prognosticators of Metastatic Disease at Follow-up After RP.

Parameters	Lymph node metastasis at follow-up			Bone metastasis at follow-up
	OR	95% CI	p-value	OR	95% CI	p-value
# of LVI	13.85	2.54-115	0.002	17.25	1.81-2308	0.009
pT	0.23	0.02-3.08	0.24	0.38	0.00-74.06	0.62
GG	4.34	0.85-22.82	0.08	0.49	0.04-3.33	0.28

GG: Grade group.

Discussion

Several previous studies established that presence of LVI at RP is associated with adverse pathological parameters and unfavorable clinical outcome.^1–5 In this small RP series we confirmed this relationship of LVI with pathological stage and poor outcome. We further noted that the unfavorable sub-pathologies, cribriform and/or intraductal carcinoma, were nearly always present in our LVI positive prostate cancers. Similarly, Kryvenko and Epstein ⁹ reported a frequent association of LVI with intraductal carcinoma, documented in 42% of their cases. Likewise, we confirmed that LVI was associated with high tumor volume, that is 22% in our study as compared to 25% in the study by Kryvenko and Epstein. ⁹ Since prostate cancers with cribriform and/or intraductal carcinoma are more likely to harbor LVI, they could be selectively searched for presence of LVI. More striking was our finding that the vast majority of LVI foci had a cribriform morphology. Although this may suggest that carcinoma with cribriform morphology would have a stronger propensity for LVI, another explanation may be that the lymphovascular environment exerts a morphogenetic influence resulting in the development of cribriform architecture. Since in two of our cases the carcinoma lacked a cribriform component with cribriform pattern exclusively in LVI foci, the latter seems a better explanation.

The identification of LVI in RP specimens is not a simple process owing to the unique prostate tissue composition; fibromuscular stroma largely composed of myofibroblasts and smooth muscle cells admixed with prostatic glands with a wide variation in size and density and mainly thin-walled lymphovascular channels lacking a muscle layer. Nuclei of stromal myofibroblasts resemble endothelial cell nuclei, and along retraction artifacts, the myofibroblast nuclei may easily be confused with endothelial cell nuclei. Similarly, prostatic acini and ducts are enveloped by an outer layer of very flat basal cells that can falsely interpreted as the endothelial lining of a vascular channel when an artificial cleft appears between the basal and luminal cell layer, especially when the glands are atrophic and have their inner glandular lining sloughed off in the lumen. To avoid overcalling of LVI in our series, we used several strict criteria; an unambiguous presence of carcinoma component within endothelial lined spaces, based on presence of a corrugated instead of a smooth lining with endothelial cells protruding into the lumen. Awareness of retraction artifact, dislodged inner glandular cell lining due to poor fixation, perineural invasion and tissue displacement as main mimickers of LVI may help improve the accuracy of LVI diagnosis (Table 1). The most challenging LVI mimicker is the periacinar clefting artifact that was seen frequently in prostatic carcinoma and occasionally in benign acini. Clefting artifact may occur due to altered basement membrane component, absence of basal cells, and stromal changes.^12–15 Some authors believe that tumors with extensive retraction artifact are associated with more aggressive behaviour. ¹⁶

Mechanical displacement of tumor fragments into lymphovascular spaces was commonly encountered in our series as well. This might be attributed to transfer of tissue during gross cutting of the prostatic gland or the organ manipulation during the surgery as suggested by studies on uterine carcinomas.^17,18 In the case of absence of any supporting features of LVI, the presence of detached tissue (malignant or benign) elsewhere in the same slide would favor the interpretation of contamination and argue against LVI.

Difficulty in recognizing LVI occurred particularly in cases of cribriform morphology LVI when there was hardly any visible space between the endothelial lining and the carcinoma embolus, filling up and distending the vascular space. This became apparent when reviewing D2-40 stained slides resulting in some cases in a substantial increase in LVI foci after comparison with the corresponding H&E slide. Since D2-40 staining may give spurious results as it also stains basal cells ¹⁹ we further confirmed LVI using ERG and/or CD31 staining in selected cases.

Tumor encroachment of the lymphovascular spaces can hardly be distinguished from true LVI, where the tumor foci superficially appear to be invading the adjacent lymphovascular spaces. As illustrated in Figure 4F, the ERG positivity within the tumor is evidence of a tumor vasculature and therefore it is debatable whether this actually represents LVI. In some cancers, it is thought that endothelial cells may be carried with the thrombus. ²⁰

A few previous prostatectomy studies reported on the anatomical and spatial distribution of LVI. Similar to our findings the most common location of LVI was intraprostatic, mainly in the stroma immediately surrounding the carcinoma with rare LVI foci within the carcinoma.^2,9,21–23 In our cohort, LVI foci were mostly adjacent to the tumor and only 8% of LVI foci were seen in the center of carcinoma. Possibly, this preferential peritumoral location of LVI can be attributed to the reported increase in lymph vessel density of the peritumoral tissue as compared to the intratumoral tissue.^22,24 The next most frequent location was the surrounding loose stroma and muscle sleeve around the SV. Neurovascular bundle and extraprostatic fat were uncommonly involved. This may guide pathologists by focusing on the tumor-stroma interface when searching for LVI foci.

Few prostate cancer studies have tried to quantify LVI.^1,4,9,25 We observed an unexpectedly large variation in the density of LVI foci among our RP specimens (Supplemental Figure 1). Quantification of LVI foci appeared rather subjective, particularly when multiple LVI foci are in close vicinity raising the possibility that they may be in the same tangentially cut lymphovascular vessel. Therefore, we decided to count multiple and consecutive lymphovascular channels with scant intervening stroma as one focus. Other methods of LVI quantification have also been proposed: In studies on cervix carcinoma, Roman et al ²⁶ defined the quantity of LVI as the proportion of slides showing LVI, whereas Chernofsky et al ²⁷ quantified LVI based on the combination of percentage of slides with LVI and the total number of LVI foci. Dekker et al ²⁸ calculated the tumor burden of lymphovascular space invasion in breast carcinoma by multiplying the number of tumor cells in the largest tumor embolus with the number of peritumoral LVI (LVI tumor burden) and found that this measure had a higher sensitivity and specificity to predict disease-free survival than the counting of LVI foci only. In our small RP series, the number of LVI foci was independently associated particularly with the risk of bone metastatic disease during follow-up, but also with lymph node metastatic disease at follow-up. One previous RP study also demonstrated an association between extensive LVI (defined as more than 6 LVI foci) and disease recurrence/death ¹ and another study reported more numerous LVI foci in RP specimens with positive lymph nodes (range of 1-21 LVI) as compared to lymph node negative cases (range 1-4 LVI foci). ⁹ Our finding that extensive LVI rather than presence of LVI is independently associated with bone metastatic disease at follow-up (Table 2) requires validation in a larger series. Its clinical importance could be that the subset of patients with extensive LVI may benefit from treatment intensification.

In conclusion, the strong association of LVI with presence of cribriform and intraductal carcinoma, its predominant cribriform morphology and its common localization at the carcinoma–stroma interface may help focus the search for LVI in prostatectomy specimens. Rigorous histopathological criteria, awareness of potential mimickers and application of immunostaining may further improve the identification and quantification of LVI. Our preliminary data further suggest that extensive LVI is an important prognosticator for bone metastatic disease at follow-up.