Article Text
Abstract
Aims: To analyse the distribution of posterior uveal melanoma origin sites and to visualise the patterns of tumour extent in the ocular fundus.
Methods: Data on the size and location of 110 posterior uveal melanomas were converted into a database of two-dimensional retinal charts by means of computer-drawing software. The initiation sites (geometric tumour centres) were entered into corresponding sectors of the retinal chart. The extent of the tumours was visualised by merging the charts and displaying the number of overlapping tumours on colour-coded contour maps.
Results: Seventy-one initiation sites (65%) were located in the temporal and 39 (35%) in the nasal hemisphere (p = 0.002). Seventy-six initiation sites (69%) were located posterior and 34 (31%) anterior to the equator. More initiation sites anterior to the equator were observed in large versus small/medium tumours (p = 0.0003), in tumours with a largest basal diameter/height ratio <2 versus ⩾2 (p = 0.002), in tumours with a ruptured versus intact Bruch membrane (p = 0.03), in tumours with a mixed/epithelioid versus spindle cell type (p = 0.02) and in tumours leading to metastatic disease (p = 0.003).
Conclusion: The temporal hemisphere posterior to the equator is the preferential area of melanoma occurrence and growth. For larger and more aggressive tumours, there is a shift towards more peripherally located initiation sites.
Statistics from Altmetric.com
The most common primary intraocular malignancy is melanoma, which may arise in any part of the uveal tract with increasing frequency from the iris to the ciliary body and the choroid. The posterior segment of the eye (choroid and ciliary body) is the most common site for uveal melanomas accounting for 95–98% of the cases, whereas only 2–5% are located to the iris.1–3 Although many prior reports on posterior uveal melanoma include information about the topographical distribution of the tumours, only a few authors have focused on this topic. Li et al showed that the sites of tumour origin concentrated in the macular area and decreased progressively with increasing distance from the macula.4 On the contrary, Schwartz et al found no preferential location of posterior uveal melanoma, which occurred with equal frequency in all quadrants of the eye.5 The aims of the present study were to analyse the distribution of posterior uveal melanoma origin sites and to visualise the patterns of tumour extent in the ocular fundus.
METHODS
Patients
The clinical records of 111 consecutive patients with posterior uveal melanoma referred to Haukeland University Hospital between 1 January 1993 and 31 December 2005 were reviewed. The incidence, clinical findings, tumour characteristics and results of treatment have been described.6 Briefly, 60 women and 51 men were included in the study, with a median age of 66 years (mean, 64; range, 24–96 years) at the time of diagnosis. Twenty-seven patients (24%) developed systemic metastases. The median follow-up time was 36 months (mean, 52 months; range, 2 months to 13 years). None of the patients were lost to follow-up.
Tumour data
Tumour dimensions were determined by ophthalmoscopy, fundus photography, ultrasonography, fluorescein angiography, and computed tomography or magnetic resonance imaging. Data including the largest basal diameter and distance of the nearest tumour margin to the foveola and optic disc margin were also estimated clinically by using the horizontal diameter of the optic disc as a reference of 1.75 mm.7 8 Tumour size was scored according to both the TNM classification and the criteria of the Collaborative Ocular Melanoma Study (COMS).9 10 In order to discriminate between flat and more sphere-shaped uveal melanomas, the tumours were categorised according to their largest basal diameter/height ratio. The growth pattern was defined as flat for tumours with a diameter/height ratio ⩾2 and spherical for those with a ratio <2. The Bruch membrane was considered ruptured if ultrasonography revealed a mushroom-shaped tumour. The presence of exudative retinal detachment was defined as subretinal fluid involving at least one quadrant of the fundus. In patients who underwent enucleation, histopathological examination determined the classification of tumour cells grouped into spindle, mixed, or epithelioid cell types,11 scleral invasion of melanoma cells, and the integrity of the Bruch membrane. For the purpose of this study, a re-evaluation of all the archival specimens was performed.
Tumour location
Tumour location was determined by a combination of the abovementioned methods. In addition, 62 melanomas eligible for brachytherapy were analysed in a three-dimensional treatment-planning programme (Plaque Simulator, BEBIG Isotopen- und Medizintechnik GmbH, Berlin, Germany) and examined by trans-scleral illumination for accurate tumour location.12 13 The presumed sites of tumour origin (in the following also referred to as “initiation sites”) were defined as the geometric centre of each tumour, which for the round and oval tumours corresponded to the midpoint of the largest tumour diameter. For tumours with an asymmetrical shape, the initiation site was assessed by judgement of the centremost point. The initiation sites were then categorised according to their meridional location (quadrants and hemispheres defined by a horizontal and vertical line passing through the fovea) and to their concentric zone location (central to the equator, between the equator and ora serrata, and between the ora serrata and limbus). Melanomas that were too large to obtain a detailed assessment of tumour location were excluded.
Tumour mapping
Based on a thorough evaluation of the original fundus drawings, images (fundus photographs, fluorescein angiograms, ultrasound scans and computed tomography or magnetic resonance images) and the 62 diagrams created by the treatment-planning programme, each tumour was redrawn with azimuth equidistant projections on a standardised retinal drawing chart with a macular centre surrounded by circles representing the equator, ora serrata and limbus.4 8 All drawings were done by one of the authors (JK), and special care was taken to correct for circumferential distortion in the periphery when calculating retinal diagrams from tumour parameters.14 Thereafter, the drawing tools of the computer software PowerPoint (Microsoft, Redmond, WA) were used to manually convert all the tumour drawings into a database of identical two-dimensional retinal charts of right eyes. By means of the software packages AVS (Advanced Visual Systems, Waltham, MA) and Matlab (The MathWorks, Natick, MA), the collection of digital tumour drawings was merged, filtered and finally converted into a contour map of the fundus, displaying the number of overlapping tumours with different colour codes. Separate contour maps were made for various subgroups of patients and tumours. Although the maximum number of overlapping tumours differed between these groups, each contour map was labelled with the same colour scale ranging from blue to red. The dark blue colour indicated areas without any tumours, and the dark red colour displayed the area with the maximum number of overlapping tumours.
Statistical methods
The chi-square goodness-of-fit test was used to analyse the distribution of the initiation sites under the null hypothesis that the melanomas were uniformly distributed in the ocular fundus. Due to possible sources of error when calculating the area of each concentric zone, a statistical analysis of the distribution of all tumours was only performed for their meridional location. This was done under the assumption that each quadrant or hemisphere includes an equal area of the choroidal and ciliary sphere. However, for the comparison between two groups of patients or tumours with different characteristics (binary variables), the Fisher exact test was used to analyse both the meridional distribution (between corresponding hemispheres) and the concentric zone distribution (anterior and posterior to the equator) of the tumour initiation sites. The Bonferroni correction for multiple comparisons was applied if necessary. For all tests, two-tailed p values less than 0.05 were considered to be statistically significant.
RESULTS
Distribution of tumour initiation sites
Accurate data on tumour location were obtained in 110 patients. In one patient, the tumour was too large to determine its extent and geometric centre. There was a significant nasotemporal asymmetry in the distribution of the initiation sites, as 71 (65%) were located in the temporal and 39 (35%) in the nasal hemisphere (p = 0.002). The distribution of the initiation sites in the superior and inferior hemisphere was 56 (51%) and 54 (49%), respectively (p = 0.85). Seventy-six initiation sites (69%) were located posterior and 34 (31%) anterior to the equator.
The number of tumours and the distribution of their initiation sites according to patient and tumour characteristics are presented in table 1. In females, the number of initiation sites in the upper and lower hemisphere was 23 and 36, respectively, while the corresponding figures in males were 33 and 18 (p = 0.008). Otherwise, neither the hemispherical nor the concentric distribution differed significantly between genders, between patients in the age group ⩽60 years and >60 years, or between right and left eyes. The hemispherical distribution of the uveal melanoma initiation sites was not significantly different in patients with and without metastatic disease, but there was a significant shift towards more initiation sites anterior to the equator in the patients who later developed melanoma metastases (p = 0.003).
The hemispherical distribution of the initiation sites did not differ significantly between T1-2 and T3-4 tumours, or between small/medium and large tumours. A shift towards more initiation sites anterior to the equator was observed in T3-4 versus T1-2 tumours (p = 0.004) and in large versus small/medium tumours (p = 0.0003). More tumours with a spherical growth pattern (largest basal diameter/height ratio <2) were located anterior to the equator compared with those with a flat growth pattern (largest basal diameter/height ratio ⩾2) (p = 0.002). Similarly, there was a shift towards more initiation sites anterior to the equator in tumours with a ruptured versus an intact Bruch membrane (p = 0.03). The distribution of initiation sites was not significantly influenced by the presence or absence of exudative retinal detachment. Histopathological classification of the tumour cell type was determined in 54 eyes. The concentric distribution showed more initiation sites anterior to the equator in tumours with a mixed/epithelioid cell type compared with those with a spindle cell type (p = 0.02). The distribution did not differ significantly between eyes with and without scleral invasion of tumour cells. After adjustment with Bonferroni corrections for multiple comparisons, the differences in tumour location related to gender, the condition of Bruch membrane and the presence of various tumour cell types were no longer significant at the p<0.05 level.
Patterns of tumour extent
The median largest basal tumour diameter was 13.3 mm (range, 4.4–21.0 mm), and the median tumour height was 7.0 mm (range, 1.5–15.0 mm). The anterior tumour margin was located at the iris in seven patients (6%), the ciliary body in 27 (24%), between the ora serrata and equator in 35 (32%) and posterior to the equator in 42 (38%). The median distances from the nearest tumour margin to the foveola and to the rim of the optic disc were 4.0 mm (range, 0–16.2 mm) and 4.0 mm (range, 0–16.0 mm), respectively. The patterns of tumour extent, visualised by the computationally merged retinal charts, corresponded clearly with the regional distribution of the initiation sites. The maximum number of overlapping tumours was 45, and this area was located temporal to the macular region at the boundary between the upper and lower hemisphere (fig 1). The upper nasal quadrant was the region with the lowest number of overlapping tumours.
The patterns of tumour extent according to various patient characteristics are presented in fig 2. When comparing the merged retinal charts from females and males, the superoinferior asymmetry of the tumour initiation sites became evident (fig 2A,B). In patients ⩽60 years of age, the tumours seemed to be more diffusely spread within the posterior pole compared with those in the older age group (fig 2C,D). The patients who subsequently developed metastatic disease had a more peripheral location of the uveal melanomas, and most of the overlapping tumours were located in the lower temporal region of the fundus (fig 2E,F).
The patterns of tumour extent according to various tumour characteristics are presented in fig 3. The merged retinal charts showed that more tumours classified as small/medium were located posterior to the equator compared with the large tumours (fig 3A,B). Similarly, a more posterior location was found in tumours with a flat growth pattern (largest basal diameter/height ratio ⩾2) (fig 3C,D), and in tumours with an intact Bruch membrane (fig 3E,F). When analysing the enucleated eyes with regard to different histopathological features, more tumours with a spindle cell type were located posterior to the equator than those with a mixed/epithelioid cell type (fig 3G,H). Although there was no significant difference in the distribution of the initiation sites between tumours with and without scleral invasion of melanoma cells, the merged retinal charts indicated that more tumours without scleral ingrowth were located anterior to the equator (fig 3I,J).
DISCUSSION
Most clinical studies of posterior uveal melanoma provide detailed information about the concentric tumour location, such as ciliary body involvement and proximity to the optic nerve and macula. Some authors have also included data on the meridional location, showing that uveal melanomas are usually located temporally,15–20 while others have found a more symmetrical distribution between the temporal and nasal hemisphere.5 21 According to Li et al,4 the density of uveal melanoma occurrence is concentrated in the macular area and gradually decreases with distance from the macula to the ciliary body. The authors propose that this pattern is consistent with the dose distribution of solar light on the retinal sphere and a lower melanin concentration in the macular and perimacular retinal pigment epithelium (RPE). Their findings provide support for the highly controversial hypothesis that ultraviolet (UV) light exposure plays a role in the induction of uveal melanoma. Several previous studies have examined the relationship between sunlight exposure and the risk of uveal melanoma,22 and a recent review concludes that the existing data are incomplete and conflicting, even though some case-control studies may suggest an association.23 But why do uveal melanomas have a predilection to arise in the temporal fundus hemisphere? In some way, it must be related to morphological features of the eye wall and/or an asymmetrical exposure to potential oncogenic agents. One possible explanation can be found in the study by Panda-Jonas et al,24 showing that the RPE cell density is significantly higher in the nasal region than in any other region of the fundus. Supposing that the melanin containing RPE cells play a protective role against UV light-induced malignant transformation of choroidal melanocytes, the relatively low RPE cell density in the temporal hemisphere may account for the nasotemporal asymmetry found in the present and other studies.
Patients who subsequently developed metastases had significantly more uveal melanoma initiation sites anterior to the equator compared with those without metastatic disease, and a similar central-to-peripheral trend was also found for large tumours and tumours with a mixed/epithelioid cell type. Anterior tumour location, large tumour size and malignant cytology are well-known risk factors for metastatic disease in posterior uveal melanoma.25 26 However, there are essential interactions between all these parameters, and the present study gives no answer to the question of whether peripheral tumour origin per se leads to more aggressive or malignant melanomas. Interestingly, Shields et al found that the inferotemporal tumour location was a significant risk factor for metastases after brachytherapy for large posterior uveal melanomas.18 The merged fundus drawings in our study also revealed that most of the melanomas in patients who later developed metastases were located in the lower temporal part of the fundus.
According to Shammas and Blodi,27 scleral invasion by tumour cells is highly influenced by the largest tumour diameter, and we therefore expected a more peripheral location of the tumours with scleral ingrowth. However, we found no significant difference in the concentric distribution of the initiation sites between tumours with and without scleral invasion of melanoma cells. Conversely, the merged fundus drawings indicated that more tumours with scleral invasion were located posterior to the equator. A possible explanation for this finding could be that cell invasion usually takes place along the scleral emissaries. As the vortex veins and a number of ciliary vessels and nerves pierce the posterior pole of the eye, they could promote the scleral ingrowth of melanoma cells in this region.
Given the limited number of patients and the multiplicity of subgroup analyses in this series, some of the differences between the groups may possibly be explained by chance alone. Other potential limitations of the study include spherical distortions when the tumour parameters were converted to the retinal charts, the assumption that each quadrant or hemisphere comprises an equal area of the fundus, and the definition of the initiation sites implying that the growth of uveal melanomas is isotropic.
In summary, we found a highly significant nasotemporal asymmetry in the distribution of uveal melanomas. Both the location of the initiation sites and the merged fundus drawings demonstrated that the temporal hemisphere posterior to the equator is the preferential area of melanoma occurrence and growth. For larger and more aggressive tumours, there is a shift towards more peripherally located initiation sites. Although the merged fundus drawings had limitations regarding detailed analyses of tumour location, they provided additional information and illustrated clearly the differences between various groups of patients and tumours.
REFERENCES
Footnotes
Funding: The study was supported by grants from The Norwegian Cancer Society.
Competing interests: None.
Ethics approval: The study was approved by The Norwegian Social Science Data Services, and followed the ethical standards for clinical research in accordance with the Helsinki Declaration.