Purpose The purpose of this investigation is to describe the clinical, imaging, histologic and flow dynamic characteristics of orbital cavernous haemangioma.
Methods In this clinicopathologic series, clinical features were obtained from patient records. All imaging studies were reviewed. All specimens were reviewed with haematoxylin and eosin, and 10 were subject to a staining protocol including: Movat Pantachrome, periodic acid Schiff, D2-40, CD31, GLUT-1, Ki-67, vascular endothelial growth factor receptor 1 (VEGF-r1) (flt-1), VEGF-r2 (Flk-1), VEGF, anti-smooth muscle actin (SMA), CD20, CD4, CD8 and CD68. Imaging and pathology were reviewed in a systematic fashion.
Results Clinically, lesions were more common in middle-aged females presenting with axial proptosis and pain. One-third demonstrated signs of optic nerve dysfunction. Dynamic imaging revealed focal early and diffuse late enhancement. Lesions demonstrated slow growth at 0.2 cm3/year. Histologically, all lesions demonstrated large vascular channels with mature-appearing endothelium and abundant stroma. Three salient features were noted and characterised: thrombosis, nests of perivascular hypercellularity and expanded stromal elements. Acute thrombosis was a feature of each specimen (<10% of channels). Fibrin clots were lined by a layer of CD31+ endothelium. Perivascular hypercellular areas stained uniformly with CD31 and less so with VEGFr2. Additionally, focal areas of Ki67+ and CD68+ cells were found in these regions. Expanded stroma contained CD31+ microcapillary networks and stained diffusely with anti-SMA.
Conclusions Cavernous haemangioma demonstrate clinical features and growth characteristics of a benign mass. Dynamic imaging highlights their slow flow vascular nature. Histologically, the hypercellularity and stromal changes identified can be understood within the pathogenic context of thrombosis and recanalisation in an organised lesion.
Statistics from Altmetric.com
Cavernous haemangiomas are common, representing 5% of orbital masses.1 They are slow growing, causing axial proptosis in a middle-aged individual.1–3 Diagnosis is usually made by CT or MRI1 ,3 and treatment is surgical resection, usually with very good results.1 ,2 ,4 ,5 Histologically, they are finely encapsulated masses containing a complex network of dilated vascular channels lined by mature, flattened endothelial cells surrounded by abundant fibrous stroma.6
Despite extensive experience with the management of these lesions, relatively little literature is dedicated to describing their pathophysiology. Without such a discussion, the designation ‘haemangioma’ has persisted, however, this is not based on pathophysiologic classification; rather, it is historical.
The purpose of this investigation is to elucidate the clinical, imaging and histologic features of cavernous haemangioma, and to contextualise these features within the wider vascular malformations literature. A discussion of pathophysiology, growth mechanisms and reclassification schema will be presented. For the remainder of this report, ‘cavernous haemangiomas’ will be referred to as cavernous malformations.
This study was approved by the University of British Columbia, Office of Research Services, Clinical Research Ethics Board, and procedures were performed in accordance with the 1964 Declaration of Helsinki.
All cases of cavernous malformation diagnosed and managed in a single clinician's quaternary academic orbit practice (JR) between December 1999 and January 2009 were included in this retrospective cross-sectional cohort study. This time period was selected to include only digitally stored imaging series, which became the practice at our institution in 1999. Cases were excluded if the diagnosis was equivocal on imaging, histologically incorrect, or if imaging was incomplete.
Each patient had a full ophthalmic and standardised orbital examination. Goldmann visual fields and imaging were performed at the discretion of the treating surgeon. Information including demographics, type and duration of symptoms, Snellen visual acuity, pupillary reactions, eyelid position, extraocular motility, strabismus and globe position were extracted from patient records and entered into a standardised electronic database.
Records for initial presentation and most recent follow-up were included for all patients, as was type, timing and outcome of any intervention. In patients followed clinically for greater than 1 year prior to any treatment, or those that had no intervention, pertinent information from each clinical visit was entered into a longitudinal natural history database.
All radiologic series within the University of British Columbia and Vancouver Coastal Health systems or from outside sources, of cavernous malformations identified in the clinical database were included. Each study was reviewed in a standard fashion by a neuroradiologist (MKSH) with special interest in orbital disease and the orbital surgeon (JR). Patients presenting to the orbital service with imaging from outside sources were not reimaged in our institution and these studies were directly reviewed. Some cases presenting more recently without imaging were assessed using dual phase early (30 s—arterial) and late (70 s—venous) contrast enhanced CT scans with or without Valsalva.7
Non-contrast CT and/or MRI series were evaluated for static structural characteristics, such as location (intra/extraconal, orbital quadrant, extension, focality), morphology (borders, homogeneity, shape, contour), size (triplanar extent) and effect on tissues (infiltration, modelling, optic nerve compression, bony changes).
Contrast enhanced CT and MRI series were evaluated for the intensity and location of contrast enhancement within the lesions as well as the characteristics of contrast filling over time.
For those who underwent dynamic dual phase CT imaging, series were evaluated for features related to the extent and intensity of early filling as well as change during late phases. Additionally, any change in maximum triplanar extent of the lesion was calculated and noted for series in which Valsalva was performed.
Any patients with multiple imaging series greater than 6 months apart were included in growth analysis. Lesion volume was calculated using the manual region of interest volume calculator in the OsiriX (Pixmeo, Geneva, Switzerland) image-processing suite. Change in volume from presentation to most recent follow-up imaging was calculated and presented as percentage of lesion volume and raw cubic volume increase in cm3.
Specimens were obtained from the Department of Pathology and Laboratory Medicine, Anatomic Pathology archives, Vancouver General Hospital. All were drawn from the submissions of a single surgeon (JR). Standard slides stained with haematoxylin and eosin were examined on all cases. Ten specimens with sufficient and representative tissue were selected for extended conventional staining, including Movat pentachrome and Periodic Acid Schiff and immunostaining protocols. Immunohistochemical staining was performed by a routine avidin-biotin immunoperoxidase method on a BenchMark XT autostainer (Ventana Medical Systems, Tucson, Arizona, USA). Stains were performed for smooth muscle actin (SMA), D2-40 (lymphatics),8 ,9 CD31 (vascular endothelium),8 GLUT-1 (infantile haemangioma),10 Ki-67 (cell proliferation),11 vascular endothelial growth factor receptor 1 (VEGFr1), VEGF-r2) (Flk-1), VEGF,12 L26 (CD20) (B- cells), CD3 (T cells), CD4 (Th cells), CD8 (Tc Cells), CD68 (monocytes).
Control slides for SMA, CD3, CD4, CD8, CD20, CD31 CD68, Ki67 and D2-40 contained tonsil, appendix, liver and pancreas. The control slide for Movat Pentachrome was skeletal muscle, for periodic acid–Schiff was small intestine, for GLUT-1 was endothelioma, for VEGFr1 and VEGFr2 was placenta and for VEGFA was kidney. All controls were appropriately stained.
Samples were described in a standard format by two ocular pathologists (JR and VAW). Features related to the overall morphology and tissue components were noted. Endothelium was described qualitatively, and any areas of papillary endothelial hyperplasia were noted. Sclerosis was defined through the identification of collagen deposition in the stroma on movat pentachrome staining.
The character, completeness and cellular components of the capsule were documented as were the percent of channels with thrombosis, and areas of active fibrin clot were counted on Movat pentachrome low-power slides. Elastic lamina were also identified on Movat pentachrome staining.
SMA, CD31, D2-40, VEGFr1, VEGFr2 and VEGFA positive cells were documented in terms of their tissue distribution, while GLUT-1 and Ki67 were additionally described in terms of intensity of staining on a subjective scale from 0 (no staining) to (extensive). These reviews were performed on medium power fields (×4).
Tissue distribution, focality and the intensity of staining (on the same 0–4-point scale) of inflammatory cells were recorded, and any germinal centres were noted.
Data was presented primarily in descriptive form. Comparisons were made using Student t test or χ2 analysis where appropriate. All analyses were performed with IBM SPSS Statistics for Mac V.19.0 (SPSS, an IBM Company, Somers, New York, USA).
Forty-one patients were screened for study. In one case, the imaging and clinical behaviour was suggestive of schwannoma, and this case was excluded. Another had imaging that could not be located due to radiologic archiving, and this also was excluded. The remaining 39 patients were included in the study.
Mean (SD) age at presentation was 48.7 years (12.4) and 66.7% (26) were female. The mean (SD) length of follow-up for the entire sample was 18.39 months (20.9). Clinical features are noted in table 1.
The most common presenting complaint was pain in 38.5% (15) of patients. Seventy-three per cent (11) of these patients described it as retrobulbar or periorbital, while the remainder had a more diffuse headache. The only symptom, sign or radiologic feature to be associated with pain was male sex. Males were 4.34 times (95% CI 1.06 to 17.86) more likely to complain of pain on presentation than women (p<0.05).
The next most common presenting complaint was vision loss (33.3% (13)). In 76.9% (10) of these cases, vision loss was related to optic nerve dysfunction secondary to a compressive optic neuropathy. Hyperopic shift was responsible for visual changes in a single case and other aetiologies not related to the primary complaint including cataract and keratoconus each accounted for a single case (table 1). Diplopia and gaze-evoked amaurosis were reported in 10.3% (4) and 6.3% (2) of patients, respectively. Not surprisingly, proptosis was found in 79.5% (31) of the sample. Just over 10% of patients complained of diplopia and duction deficits were evident in 20.0% (8); however, tropias were rare and found in only 5.4% (2) of cases.
Optic nerve dysfunction in the form of a demonstrable relative afferent pupillary defect (RAPD) was noted in 33.3% (13) of patients. These cases were more likely to have a posteriorly located lesion (OR (95% CI) 10.67 (2.05 to 55.52)). In particular, lesions involving the superior orbital fissure (OR (95% CI) 6.67 (1.18 to 37.78)) and/or the optic canal (OR (95% CI) 15.0 (1.52, 148.31)) were much more likely to be associated with an RAPD. Individuals with an RAPD also tended to demonstrate optic nerve compression on imaging (OR (95% CI) 13.5 (1.47 to 123.74) (table 2).
More than 75% of patients underwent surgery to remove the malformation (table 1). Three patients were treated successfully with stereotactic fractionated radiotherapy, and these patients are the subject of a separate report.13 None were found to have additional systemic vascular malformations, and no local recurrence was found.
No clinical variable was found to be statistically different for individuals whose lesion was subject to pathologic analysis relative to those who were not (p > 0.05).
Static imaging features
A total of 61 imaging studies were reviewed. MRI with contrast was the most commonly performed study at 24.6% (15). Less commonly performed scanning sequences included CT with contrast, CT without contrast and CT angiogram at 18% (11), 13.1% (8) and 14.8% (9), respectively. Biphasic dynamic CT imaging was performed in 18% (11) of cases and in an additional 4.9% (3) dynamic CT studies were performed with Valsava.
Over 75% of lesions were located at least partially in the intraconal space. Cavernous malformations were found in all quadrants and were predominantly located in the middle (48.7%) or posterior orbit (20.5%). Individuals presenting with very large lesions involving most of the orbit made up 10.3% (4) of the sample (table 3).
Cavernous malformations were found to involve the superior orbital fissure, the inferior orbital fissure and any part of the optic canal in 18.4% (7), 10.5% (4) and 15.8% (6) of cases, respectively. Compression of the optic nerve was found in more than half the sample (54.3% (19)). When lesions abutted the globe, they tended to indent rather than mold to the eye (table 3).
Cavernous malformations demonstrated smooth edges in almost all cases and did not infiltrate local tissues. They were typically elliptical or spherical in shape (77.0% (30)) and had variable internal density (table 3).
Dynamic imaging features
The majority (81.5%) of lesions showed focal or multifocal enhancement on arterial phase imaging, while they tended (69%) to enhance more diffusely on venous phase assessment (table 4). In 95.0% of cases, the filling pattern increased in area from arterial to venous phase imaging.
The intensity of the enhancement was complementary to this pattern (following target vascular blood pool intensity), with 65.4% (17) of lesions demonstrating high intensity on arterial phase and 89.3% showing moderate intensity on venous phase (table 4). Overall, 57.9% of lesions demonstrated a reduction in intensity from early to late in the sequence (see figure 1 for examples). For the three cases in which Valsalva manoeuvre was performed, there was no increase in maximum triplanar extent.
Six patients were followed without treatment for a mean of 2.9 years (range 0.7–5.5). Three went on to have surgery; the first due to a two-line loss of vision after 3 years of careful follow-up, and the other two had surgery due to intractable chronic pain at 12 and 32 months after presentation, respectively. No patient developed a new or worsening RAPD or visual field defect over this time period. Four of these patients had multiple sequential imaging studies. The mean (range) of follow-up for sequential scans was 1.5 years (0.5–2.9). The mean (SD) rate of lesion expansion over this time was 13.8% (15.2%) or 0.2 cm3 (0.3) per year.
Ten specimens were subject to extended histologic analysis. Morphologically, well-described features of large ectatic vascular channels with abundant stroma were evident in each case. This ‘classic’ pattern was uniform in 70% of cases (figure 2A), while the remainder demonstrated a lobular organising structure for the vascular channels (figure 2B). Diffuse sclerosis was found in half the cases (figure 2C), and a minority (20%) had some small foci of intralesional fat.
Each specimen had an external capsule, which was complete in 90% of cases. The capsule contained small vessels (figure 2D) in 60% of cases and small nerves in 20%. Dilated, ectatic vascular loops were found contained within the capsule of one case.
Channels almost exclusively had normal-appearing, mature, CD31 positive endothelium. The majority (>50%) of vessel walls were diffusely thickened and stained positively for α-SMA. Only one lesion had a few isolated areas of D2-40 positive lymphatic vasculature within the lesion. This sample was from the subcutaneous eyelid where lymphatics are common.
Some thrombosis, acute or chronic, was found in each specimen, although representing less than 10% of channels in all cases (figure 3A, B). Acute fibrin clots (as defined by purple fibrin staining on Movat pentachrome) were noted in 90% of samples. Fifty percent of the time, there were between one and five channels with acute thrombosis per specimen. These areas of thrombosis were lined in all cases by a thin layer of normal-appearing CD31 positive (figure 3B) endothelium.
Hypercellularity was noted in areas of thrombosis and intrastromally. These were CD31 positive, indicating that they contained vascular endothelial cells (figure 3C). Thrombotic regions also demonstrated small foci of Ki67 positivity (figure 3D) and CD 68 positive cells (figure 3E) in 60% and 40% of cases, respectively.
VEGFr2 staining (figure 3F) in areas of CD 31 positive hypercellularity was found in 60% of samples. VEGFr1 was nonspecific, and positive staining was found diffusely throughout the lesion in each case. VEGFA staining was found in endothelial cells, and was associated with endothelial hypercellularity in areas of thrombosis for half of the specimens (figure 3G).
Each lesion demonstrated areas of abundant stroma. These regions were mostly paucicellular and diffusely α-SMA positive (figure 3H). Smooth muscle bundles within the stroma were found in 60% of cases. Sparse areas of CD31 positive (figure 3D) hypercellularity were evident intrastromally in >50% of samples and some sparse CD68 positive cells were distributed in these areas 40% of the time. The stroma did not demonstrate significant Ki67 positivity and GLUT-1 staining was uniformly negative.
There was very little inflammation identified. As noted, some focal areas of CD68 positive cells were found around areas of thrombosis and endothelial hypercellularity. Lymphocytes were sparsely and unevenly distributed throughout most lesions. There was a preponderance of CD3+, CD4+ and CD8+ cells relative to CD20+, and these cells tended to be concentrated near the periphery of the lesion close to the capsule. In approximately 1/3 of cases, some CD4+ and/or CD8+ cells were found in the areas of hypercellularity associated with thrombosis as noted above. Germinal centres were rare.
Clinically, cavernous malformations are slow growing, non-infiltrative lesions. In our series, radiologic growth rate was approximately 10–15% per year. They typically cause symptoms by mass effect mostly presenting as asymptomatic axial proptosis, and rarely lead to duction deficits or diplopia. In our series, we did note, however, that they were associated with reports of orbital and/or headache pain in more than 1/3 of patients.
Surgical management is typical if symptomatic, and is usually definitive.1 ,2 ,4 ,5 Although rare case reports of local recurrence can be found in the literature,14 this is the exception rather than the rule and likely due to incomplete resection. We did not find a single case of regrowth.
In our series, all lesions were isolated, although these lesions can be multiple and systemic.15 Such cases are not thought to be metastatic but, rather, coincidence16 or related to syndromic multisystem disorders such as Maffucci's syndrome17 or Blue Rubber Bleb Nevus Syndrome.18
On static imaging, cavernous malformations appear well circumscribed, ovoid or elliptical, with smooth borders and have variable internal consistency. When abutting the globe, they tend to indent it, rather than mold to or infiltrate it. They may remodel bone but do not erode it or produce reactive hyperostosis.
As vascular lesions, cavernous malformations demonstrate characteristic dynamic imaging patterns. Biphasic CT angiography protocols, such as described in this investigation, reveal focal, high-intensity, early filling with more moderate intensity, diffuse filling late.7 They do not expand with Valsalva. This pattern is characteristic of a slow flow venous malformation without arterio-venous shunting or distensible components.19 Studies of time-resolved MRI sequences20 and older direct venography1 confirm this claim. These features can be useful in discriminating them from true neoplasms, such as schwannoma, haemangiopericytoma or solitary fibrous tumours.20 ,21
Histologically, cavernous malformations demonstrate features typical of slow flow venous lesions. The cellular components are mature, do not tend towards dysplasia or hypercellularity, stain mostly negative for markers of cellular proliferation (such as Ki67), and are clearly differentiated from proliferating infantile haemangiomas in that they do not express GLUT-1.10 ,22 ,23 They are well encapsulated and do not infiltrate local tissues, although as we and others have demonstrated,24 they can infrequently incorporate local structures into the fibrous capsule.
Many of the key histologic features found in these lesions were described in the 1970s by Harris and Jakobiec, who noted ‘proliferation of endothelial cells around small clots or fibrin papillae’22 and the presence of myofibroblasts. These myofibroblastic elements were later described by Garner as an ‘apparently haphazard distribution of blocks of smooth muscle within the fibrous trabeculae’.25
The use of modern staining techniques as in this study and others,23 allow us to contextualise these findings in terms of natural processes occurring in slow flow venous systems, mainly thrombosis and recanalisation.
There are multiple pathways to intravascular clot formation in general, however, it is likely that in cavernous malformations the mechanism is related to vascular stasis and eddy formation. Such a microenvironment is susceptible to spontaneous thrombosis,22 ,26 ,27 possibly explains the occurrence in cavernous lesions.
After fibrin clot formation, in the first 24 h, a layer of endothelial cells forms over the matrix. This is presumably to sequester the prothrombotic elements from humoral factors in order to stop the thrombosis and begin resorption. During this period, multiple proangiogenic factors are produced by a number of cells including activated platelets, endothelial and polymorphonuclear cells. These factors include matrix mettaloproteases (MMP-2and MMP-9) that serve the dual purpose of degrading extracellular matrix and promoting endothelial growth and other proangiogenic factors, such as VEGF and basic fibroblast growth factor (bFGF).
Natural clefts in the thrombus appear and new vascular channels are formed along these fault lines as locally and humoral derived endothelial cells proliferate to line the spaces.28–30 Monocytes are also recruited and appear to play a critical role in the process of clot resolution by degrading fibrin, phagocytosing debris and secreting chemotactic and angiogenic modulators.31
As the active neovascularisation fades over the next 4–8 weeks, fibroblastic components and collagen deposition predominate.32 A localised thickening of the intimal tissue follows the resolution of the thrombus. This thickening is mainly composed of myofibroblastic tissue and the residual neovascular network.
Our study demonstrates that each of these processes is evident in different channels of cavernous malformations simultaneously. Thrombosis with extensive fibrin can be seen infrequently, although lesser amounts of residual fibrin can be seen in multiple areas of earlier thrombosis. This had been noted previously22 ,23 and can be seen with the Movat stain (figure 3A,B).
The endothelial lining of acute clots and the neovascular networks are demonstrated with CD31 staining (figure 3C). These networks appear to be highly active in the periluminal space (during clot resolution) and less active in the interstitial spaces, presumably as these older areas have undergone regression (Figure 3C). Ki67 concentration in the areas of periluminal hypercellularity (figure 3D) likely represents the proliferative phase of this process.28 ,33 Monocytes are also noted in these regions (figure 3E), and are known to have a role in the clot resolution and recanalisation.33 ,34 Proangiogenic regulators, such as VEGFA and its receptor VEGFr2 (figure 3F and 3G),12 are also upregulated in these areas. These are likely participating in the neovascular remodelling, rather than acting as an engine of growth as others have asserted.35
Finally, stromal remodelling and proliferation of α-SMA positive cells follows, leading to the stiffening and expansion of the interstitial spaces.36 This myofibroblastic infiltration dominates the late stages, and is widely evident in areas of expanded interluminal stroma (figure 3H). As a result of the back-to-back venous channels present in this lesion, instead of focal vein wall remodelling, the interchannel spaces expand.
As vascular lesions, cavernous malformations should be understood in the context of current vascular malformation categorisation schema. The modern International Society for the Study of Vascular Anomalies (ISSVA) schema is based on the work of Mulliken and Glowacki,37 which was later modified to include descriptions of flow characteristics38 and some rare lesions.39
The ISSVA classification has three levels of division (table 5). In the first, vascular tumours are differentiated from malformations. Malformations are then subdivided by flow characteristics into slow flow and fast flow. Finally, within these categories, lesions are defined by their components as arterial, venous, lymphatic or mixed.40 According to ISSVA classification, we contend that cavernous malformations be considered slow flow venous lesions.
Of note, we accept that it is not completely clear if cavernous malformations contain exclusively venous elements. In fact, our finding of focal/multifocal inflow in the late arterial phases is suggestive of some arterial component. Additionally, both earlier arteriography studies1 and more modern Time-Resolved Imaging of Contrast KineticS (TRICKS)20 MRI imaging series have demonstrated small puddles of contrast in the arterial phase. However, despite these suggestive flow dynamics, arterial elements are not evident histologically. No vessels have an internal elastic lamina, which one would expect in arterioles of this size, and the thicker walled vessels in the lesions are likely not arterioles but rather the result of thrombosis and recanalisation as described. Should there be some arterial component, it is clearly minor in the overall lesion behaviour and composition, and for simplicity it appears reasonable to consider cavernous malformations as primarily (if not exclusively) venous.
Despite some confusing historical nomenclature,41 our results and those present in the literature demonstrate that so-called cavernous haemangioma are slow flow venous lesions. The moniker ‘cavernous haemangioma’ should be abandoned in favour of the term ‘cavernous venous malformation’ when referring to these lesions.
Contributors DBR: design, acquisition, analysis and interpretation, manuscript drafting, final approval. MKSH: acquisition, analysis and interpretation, manuscript revising, final approval. JR: design, analysis and interpretation, manuscript revising, final approval. VAW: acquisition, analysis and interpretation, manuscript revising, final approval. PL: acquisition, manuscript revising, final approval. YHY: analysis and interpretation, manuscript revising, final approval. All authors agree to be accountable for all aspects of the work.
Funding This project was funded by a grant from the Physicians’ Services Incorporated (PSI) Foundation.
Competing interests None
Ethics approval The University of British Columbia, Office of Research Services, Clinical Research Ethics Board.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.