Ординатура / Офтальмология / Английские материалы / Ocular Oncology_Albert, Polans_2003

Few prognostic risk factors are identified for melanomas confined to the iris aside from cell type and markers of proliferation.

Considerably more attention is focused on the pathology of ciliary body and choroidal melanomas because they are relatively more common than tumors confined to the iris and because of a significantly greater tendency for these tumors to follow an aggressive course. Most information about the prognostically relevant features of choroidal and ciliary body melanomas stems from intensive correlative studies that were done when most of these lesions were treated by enucleation. In the past, some experts, convinced of the inevitable progression of even the smallest pigmented lesions to life-threatening melanoma, advised enucleation to prevent any future risk to the patient’s health [3]. Subsequently, careful clinical observations documented the relative ‘‘safety’’ of observing such small lesions because many of them did not grow and did not prove to be life-threatening [4]. Thus, only a short window of time was available for ophthalmic pathologists to accumulate histological material from small uveal melanocytic lesions. Tissue from these small lesions is now seldom available except from large archival collections and in rare instances when eyes are removed and the presence of the small pigmented lesion is incidental to the event that precipitates enucleation.

Ophthalmic pathologists are likely to encounter fewer tumors for examination because of the emergence of nonsurgical, vision-sparing treatments. The innate fear of losing vision together with the instinctive fear of loss of life pose terrible dilemmas for patients with choroidal or ciliary body melanomas and their physicians. Two polls conducted by the Gallup Organization posed the following question to Americans: what disease do you fear most? The most feared disease before public awareness of AIDS and Alzheimer disease was cancer. The second most feared disease was blindness. (The Gallup Organization, Inc.: ‘‘Public Knowledge and Attitudes Concerning Blindness’’—a survey sponsored by Research to Prevent Blindness, Inc., New York, October 1965, April 1976, unpublished data.) If an ophthalmic oncologist can preserve vision while eradicating the risk to the patient’s life, then many patients will opt for a treatment that does not require removal of the eye. Indeed, a comparison between surgical enucleation and primary radiation therapy by the Collaborative Ocular Melanoma Study (COMS) showed no substantial difference in survival between enucleation and vision-sparing radiation therapy [5]. Thus, pathologists are now likely to encounter only large, advanced tumors or tumors situated within the eye in anatomical locations that preclude nonsurgical treatment without loss of vision.

Even when tissue is available, one may ask why it is necessary to classify patients into prognostic categories if there is no adjuvant treatment to prevent or delay the onset of metastases or to treat metastatic melanoma. Except for the rare case in which the tumor erodes through the anterior coats of the eye and gains access to conjunctival lymphatics—after which spread to regional lymph nodes may occur [6]—uveal melanoma spreads hematogenously (there are no lymphatics within the uvea or the uveal melanoma [7]) and preferentially to the liver [8,9]. By the time sufficient quantities of hepatic parenchyma are compromised to permit detection of metastases either biochemically (through elevated liver enzymes) or by imaging studies [10], the tumor burden is frequently quite high, adding to the challenge of the medical oncologist. Of course, one can argue that it is worthwhile to develop tissuebased prognostic indicators for metastasis in the hope that effective adjuvant

therapies will emerge from a better understanding of the molecular pathogenesis of uveal melanoma, but the pathologist must recognize that even if these prognostic indicators are validated, a substantial number of patients will still elect to be treated by modalities that do not provide any source of tissue for analysis.

Thus, the motivation for studying tissue-based prognostic markers in uveal melanoma is in part for the sake of historical interest and in part to classify into emerging treatment protocols those patients for whom tissue is available. There are, however, at least two compelling reasons why tissue based prognostic indicators should still be studied and new indicators developed: (1) tissue-based studies frequently identify tumor characteristics that can be the basis of novel investigations into the mechanisms of metastasis and (2) observations from mechanistically driven in vitro cell biological observations and animal models must be validated in human tissue for clinical relevance.

The remainder of this discussion therefore focuses on melanomas that involve the ciliary body or choroid and is divided into three parts: (1) a description of tissuebased prognostic features developed largely during the era when enucleation was the dominant form of treatment, (2) a discussion of the application of tissue-based prognostic features in the contemporary era of vision-sparing therapies, and (3) the development of new tissue-based prognostic features to identify important biological pathways for tumor progression and metastasis that can be investigated by in vitro manipulations and in animal models and, conversely, the issue of validating in vitro experiments and animal models through the study of human tissue.

II.CHOROIDAL AND CILIARY BODY MELANOMAS

A.Prognostic Histological Features in the Era of Enucleation as the Primary Treatment Modality

Larger melanomas tend to have a worse outcome than smaller tumors [11–16]. For purposes of prognostication, tumor size is measured as the largest basal dimension (LBD) in contact with the sclera (LTD for largest tumor dimension). This would seem at first to be at odds with the technique for measuring cutaneous melanomas: measuring the depth of invasion from the top of the granular cell layer of the epidermis to the deepest point of invasion (in ciliary body melanomas, the tumor height does not carry significant prognostic significance). In some studies, pathologists have measured both the major and minor axes in contact with the sclera as well as the height and have attempted to calculate tumor volume. Because of the irregular growth contours of these tumors, precise calculation of tumor volume is difficult; most pathologists, therefore, record only the LBD.

The technique by which LBD is recorded has been the subject of some controversy. Transmission of light through the eye during gross examination of the enucleation specimen typically reveals a shadow. Some pathologists choose to measure LBD from the shadow [17]. Others claim that blood or turbid fluid in a retinal detachment adjacent to the tumor might also block transmission of light and thereby lead to exaggeration of the LBD: these pathologists tend to measure LBD directly from the cut surface of the tumor [18]. If the pathologist measures LBD from the cut surface and the eye has been opened conventionally (through the pupil, optic

nerve, and tumor shadow), there is no guarantee that the cut surface of the tumor actually captures the LBD. It is altogether possible that the LBD does not lie on an axis that intersects the optic nerve. The technique by which LBD is measured would seem to be trivial were it not for the fact that the major criterion for separating patients into risk categories for treatment protocols is LBD, and it is not clear that ophthalmic oncologists restrict their measurements of LBD to an axis that intersects the optic nerve. Although differences between clinical measurements of LBD and the pathologist’s record of LBD have been attributed to the effects of fixation, it is more likely that ophthalmic oncologists and pathologists may be recording different measurements of tumor size.

To more precisely correlate clinical and pathological measurements, it has been recommended that pathologists open an eye containing a ciliary body or choroidal melanoma by first removing the cap of sclera parallel to the apex of the tumor, thereby permitting the pathologist to view the tumor from a panoramic perspective more closely resembling that viewed by the clinician; in the case of a melanoma of the posterior pole, the anterior segment is removed en bloc. The tumor can then be sectioned by direct visualization and measurement of LBD may be taken through the same axis used by the oncologist for clinical measurements [19].

For the purposes of entering prognostic data on pathology reports or for retrospective studies, either the direct measurement of LBD from the cut surface of the tumor or a measurement of LBD from the glass slides is acceptable [20].

It is widely suspected that tumor location is also associated with outcome: melanomas with a component in the ciliary body tend to have a worse outcome than tumors confined to the choroid [3,12,15,16,21]. In at least one study in which multivariate analyses were used, tumor location was not found to have an independent effect on outcome [22] (see below), but differences in chromosomal karyotyping between melanomas of the ciliary body and melanomas confined to the choroid [23–25] lend some support to the suspicion that more anteriorly situated tumors involving the ciliary body are more aggressive than more posteriorly situated tumors confined to the choroid.

Although not cited in many lists of prognostic factors, the melanoma growth pattern is associated with outcome. Diffuse melanomas [26], typically flat and encompassing large areas of the choroid and ciliary body, have an adverse outcome, as do melanomas that grow circumferentially around the major arterial circle of the iris (ring melanomas) [27]. In comparison with these two growth patterns, circumscribed melanomas have a more favorable outcome.

Melanomas that have extended through the sclera have a worse outcome than melanomas confined within the eye [28]. It was suspected for many years that invasion into vortex veins might be related to aggressive tumor behavior, and many pathologists were trained to take separate sections of vortex veins and study each for evidence of invasion. It is not clear, however, if extension of the tumor along vortex veins to reach an extraocular location is more or less ominous than direct extension of the tumor through the sclera or extension along emissary nerves. As mentioned above, uveal melanomas may invade through the sclera near the limbus, gain access to lymphatics, and thereby undergo spread to regional lymph nodes—an exceptionally rare event [6].

The morphology of the melanoma cell is undoubtedly a histologic marker of aggressive uveal melanoma behavior. The Callender classification [29] and its

In many studies, the presence of mitotic figures in choroidal and ciliary body melanomas is related to adverse outcome: the greater the number of mitoses, the worse the prognosis [16,39,40]. In many areas of tumor pathology, counts of mitotic figures are now being replaced by the calculation of proliferation indices, determined by staining histological sections for markers of proliferation (e.g., PCNA, MIB-1, Ki-67). In recent studies of the histology of uveal melanoma, proliferation indices have been related to adverse outcome [41–43].

The presence of tumor-infiltrating lymphocytes (Fig. 3) has been associated with an adverse outcome (100 tumor infiltrating lymphocytes per 20 high-power fields) in multiple studies [16,44]. By contrast, the presence of tumor-infiltrating lymphocytes in cutaneous melanomas is associated with a favorable outcome [45].

B.Prognostic Histological Features in the Era of Vision-Sparing Therapies

Recognizing that many patients may be treated without any examination of tissue, some ophthalmic oncologists advocate what is termed in the ophthalmic literature as fine-needle aspiration biopsy (FNAB) to obtain tumor tissue for the purposes of assigning patients to prognostic categories [46]. This technique has been used in some institutions to distinguish between lesions that simulate melanoma (such as metastases to the eye) and primary melanomas [47,48]; for this purpose, the

Figure 3 A cluster of tumor infiltrating lymphocytes is present between spindle melanoma cells. (H&E.)

technique has a high measure of accuracy. The indications for using FNAB to discriminate between uveal melanomas and simulating lesions may have diminished somewhat because of vastly improved accuracy in correctly diagnosing melanomas on the basis of noninvasive clinical criteria [49].

The use of FNAB to extract tumor tissue for prognostic studies is problematic because uveal melanomas, like most malignant neoplasms [50,51], tend to be heterogeneous, and the technique that ophthalmic oncologists use to extract tissue does not ensure representative sampling of the entire lesion. An ophthalmic intraocular FNAB involves only a single pass into the tumor, utilizing insertion of the needle and aspiration to secure a sample [47,48]. Although a single pass with a thin needle may provide tissue that is sufficient to distinguish melanoma from simulating lesions, it is unlikely that a restricted sample would yield information that is representative of the most malignant components of the lesion unless the feature of interest were distributed uniformly throughout the neoplasm.

It is important that ophthalmologists and pathologists understand that the technique of ophthalmic intraocular FNAB is not the same FNAB technique used in diagnostic cytology for most tumors. In conventional FNAB procedures, the pathologist, radiologist, or surgeon who performs the biopsy uses the needle to cut a thin tissue sample by inserting the needle into the tumor and withdrawing the needle. To ensure representative sampling of the tumor, the individual performing the biopsy typically reinserts the needle several times at different angles and approaches into the tumor so as to ensure representative sampling within the lesion [52].

In one histological study of an eye removed for uveal melanoma after an ophthalmic FNAB, the needle track was traced by serial sections, and it was shown that the needle track terminated in a zone of tumor populated by spindle melanoma cells, just missing a population of epithelioid melanoma cells [53]. In another study, cytomorphometry of the nucleolus was performed on both cells retrieved by ophthalmic FNAB and on histological sections of the same eye removed subsequently: significant differences existed in the cytomorphometric measurements between FNAB sample and the entire histological section [54]. Given the heterogeneity of tumors in general and uveal melanomas in specific, investigators who advocate using an ophthalmic onepass FNAB to obtain tumor tissue that is prognostically relevant must show that the prognostic marker of interest (antigenic, enzymatic, chromosomal, or otherwise) is distributed so uniformly that a random penetration of a thin needle will likely capture tumor cells that express the prognostic marker of interest.

Two other issues are relevant to a discussion of the application of ophthalmic FNAB to the prognostication of uveal melanoma. First, some investigators are concerned that tumor cells may be seeded along the aspiration track: extraocular extension by tumor, as mentioned above, is an ominous prognostic finding. Although tumor cells have been identified within the needle track [55,56], ophthalmic intraocular FNAB has been used for many years and there are no reports that would attribute dissemination of tumor to the application of this technique. Second, some uveal pigmented lesions are observed clinically for evidence of growth or change in behavior before therapeutic intervention. If an ophthalmic oncologist were to perform a one-pass FNAB and the results of the examination of tumor tissue were to indicate a relatively indolent process, would the oncologist and the patient be

prepared for repeated FNAB procedures at varying time intervals in order to follow the lesion for tissue evidence of change in behavior?

To address the two concerns about FNAB—sampling error and the need to repeat examinations periodically—ophthalmic pathologists and oncologists have collaborated to develop histological criteria that reflect the tumor’s biological behavior and can be detected by means of noninvasive imaging. The ultimate goal of this type of collaboration is the development of a noninvasive surrogate for biopsy.

Current imaging techniques do not permit the detection of cell type [57]. Even if it were possible to image individual cell sizes and shapes, it would be necessary to overcome the challenge of poor reproducibility of assignment of cell type.

Looping patterns of extracellular matrix deposition have been associated with death from metastatic melanoma [16,58–60]. Specifically, closed loops that are positive with the periodic acid–Schiff (PAS) stain and networks (at least three back- to-back loops) have been shown in repeated independent studies to be a strong prognostic marker (Fig. 4). Unlike cell type, the detection of these patterns is highly reproducible [16,60], especially if hematoxylin counterstaining is omitted and the sections viewed either with a green filter or by digital imaging and selection of the green channel. These patterns appear in metastases from uveal melanoma [61], regardless of the location of the metastatic deposit. The looping PAS-positive patterns can be reconstituted in vitro by aggressive uveal and cutaneous melanoma cells but not by nonaggressive cells, thus reinforcing a relationship between the appearance of these patterns in vitro and death from metastatic melanoma [1,62].

Figure 4 Back-to-back loops in melanoma of the choroid. (Periodic acid–Schiff without hematoxylin counterstaining.)

It is important that new tissue markers of prognostic interest be evaluated very critically. Most investigators attempt to relate the presence or absence of a marker or a quantitative assessment of a marker (how much of it is present?) to outcome— death from metastatic melanoma. In studying reports of new markers, the critical reader should look for sufficient follow-up in the study set under analysis: although most uveal melanoma metastasize within 2.5 years after enucleation, the emergence of late metastases is far from uncommon [12,77]. Therefore databases used for analysis should include cases with long follow-up intervals.

Univariate analyses may be helpful in testing for possible associations between the presence of the marker under investigation and outcome; Kaplan-Meier [78] survival curves may be helpful in illustrating differences in survival between patients whose tumor contained and did not contain the marker of interest. Many investigators select a statistical sample that is large enough to permit a multivariate analysis, allowing for an examination of the prognostic marker under investigation and other well-established markers. The Cox proportional hazards model [79] is commonly used for these purposes.

Clinicians and researchers who follow descriptions of new histological markers in choroidal and ciliary body melanoma may become confused by conflicting claims for and against the usefulness of markers. There are a number of reasons for discrepancies between claims for and against the usefulness of new histological markers. First, the data sets used by different groups to study the same marker may vary. For example, one group studying the microcirculation of uveal melanoma ‘‘enriched’’ their study set with patients who had died from metastatic melanoma [80], while other investigators, studying the same phenomenon, did not and arrived at different conclusions about the utility of the marker under consideration [81]. Second, techniques for detecting the same marker may vary subtly between groups, but the differences may account for different assessments of marker validity. For example, Foss et al. [80] reported a relationship between vascularity in choroidal and ciliary body melanomas and adverse outcome, while Lane et al. [82] and Schaling et al. [83] did not. Foss et al. [80] not only used different reagents to count foci of endothelial marker staining in these tumors from those used by Lane et al. [82] and Schaling et al. [83] but also followed a convention established previously of obtaining their counts in ‘‘hot spots’’ [84], a technique not used by groups that did not find an association between vascularity and outcome. On the other hand, Foss et al. [85] found a relationship between PAS-positive matrix patterns and outcome by univariate analysis, but these patterns did not appear in a Cox model in multivariate analysis after microvascular counts were permitted to enter the model. Makitie et al. [81] validated the observation by Foss et al. [80] concerning microvascular counts and outcome, but they also validated numerous independent observations relating PAS-positive matrix patterns to outcome (both patterns and microvascular density appeared in the Cox model published by Makitie et al. [81]). Finally, Makitie et al. [81] pointed out that the frequency distribution of PAS-positive patterns in the study published by Foss et al. [85] was markedly different from that published by groups [16,58–60,81,86] that found an independent relationship between these patterns and outcome. This therefore raised the question of whether Foss et al. [85] had identified the PAS-positive patterns using the same criteria as other investigators.

One must also be cautious in examining conflicting claims between research groups in which a marker found to be valid by one group appears to ‘‘disappear’’

when entered into a Cox model prepared by a different group. If the marker in question is found by both groups to relate to outcome in univariate analyses, it is altogether possible that the marker does not appear in Cox models because of a relationship between the marker of interest and other tumor characteristics already in the model. For example, let us assume that the expression of molecule X, when demonstrated in tissue sections, is related to adverse outcome: there is a statistically valid separation in survival in Kaplan-Meier survival curves between patients whose tumor expresses molecule X and those whose tumor lacks expression of this marker. Let us further assume that the presence of molecule X appears in Cox models from a number of groups. A new tissue marker is described—a quantification of Y—by an independent research group. Quantification of Y appears in the Cox model, but the presence of molecule X drops out of the model. This does not at all mean that the presence of molecule X is irrelevant to the pathogenesis of metastasis in choroidal and ciliary body melanomas. Rather, there may be a relationship biologically between ‘‘quantification of Y’’ and the ‘‘presence of molecule X’’ such that prognostic information contained within ‘‘presence of molecule X’’ is accounted for by the ‘‘quantification of attribute Y.’’ In fact, rather than dismiss the ‘‘presence of molecule X,’’ one should now begin to search for the new biological relationship between these two markers.

Statistical associations developed from the study of tissue sections may suggest pathogenetic mechanisms and new forms of therapy, but these associations require examination of mechanisms by in vitro or animal model studies. For example, using the example of vascularity cited above, two groups [80,81] have now established that tumors containing ‘‘hot spots’’ of high microvascular density tend to have an adverse outcome. In these two studies, microvascular density entered Cox models along with many conventional tumor characteristics, such as cell type and LBD. One might be tempted to extrapolate from these studies and conclude that the tumor characteristic measured by these investigators—microvascular density—equates with angiogenesis. Equating angiogenesis with microvascular density may indeed be valid; if so, it could be argued that antiangiogenic therapies might play an important role in the treatment of patients with high-risk uveal melanoma. If uveal melanomas at high risk for metastasis are also highly angiogenic, it might be argued that antiangiogenic therapies may play an important and effective role in the treatment of uveal melanomas. By identifying a prognostic marker in tissue studies, ophthalmic pathologists would have contributed to a new rational basis for therapy.

However, one must be exceptionally cautious in extending tissue-based prognostic associations to pathogenic mechanisms. Microarray studies and studies of gene expression in uveal melanoma now suggest that highly aggressive tumor cells may express markers typically associated with endothelial cells, including CD31 and VE cadherin [1,62,74,75,87]. It is argued that as tumor cells become genetically deregulated, they express inappropriate markers. Aggressive melanoma cells, for example, express fetal keratins 8 and 18 in vitro [88,89], a marker inappropriate for melanocytes. The masquerading of tumor cells as endothelial cells may be striking: the cell line ECV-304—which was originally reported to be an immortalized endothelial (HUVEC) cell line by virtue of expression of factor VIII, ultrastructural features such as Weibel-Palade bodies, and the formation of tubules in vitro on Matrigel [90,91]—was discovered to be a derivative of the human bladder tumor cell line T14 [92]. Thus, in equating the tumor characteristic of microvascular density