GROSS PATHOLOGY AND GROWTH PATTERNS
Macroscopically, retinoblastoma has a white encephaloid or brain-like appearance, which is not surprising since the tumor arises from the retina, a peripheral colony of brain cells
(Fig. 12-3). Lighter flecks of calcification or necrotic tumor usually are evident grossly in the tumor. Necrotic retinoblastomas found in infants with aseptic orbital cellulitis typically have a blood-tinged, orange, or soupy, grayish necrotic appearance.
Tumors with endophytic, exophytic, mixed, or indeterminate and diffuse infiltrating growth patterns occur. Endophytic retinoblastomas arise from the inner layers
of the retina, which remains attached. The tumor invades the vitreous cavity and can seed the anterior chamber
(Figs. 12-3A,B and 12-5B). A pseudohypopyon of tumor cells can develop if there is extensive seeding of the anterior chamber. Hence, endophytic tumors can be confused with primary inflammatory disorders such as toxocariasis, mycotic endophthalmitis, or granulomatous uveitis.
Exophytic retinoblastomas arise from the outer layers of the retina and cause retinal detachment
(Fig. 12-3B,C). The detached retina is often highly elevated, and its vessels are visible behind the lens on clinical examination. Exophytic retinoblastomas are usually confused clinically with simulating lesions such as Coats disease that cause an exudative retinal detachment. Strictly endo- or exophytic retinoblastomas are actually relatively uncommon; most tumors have a mixed endophytic-exophytic or indeterminate growth pattern. About 1.4% of retinoblastomas diffusely thicken the retina without forming a distinct mass
(Fig. 12-4). This rare diffuse infiltrating growth pattern usually is found in older children (mean age 6 years) who typically present with pseudoinflammatory signs and invariably have unilateral sporadic tumors. During gross or histopathologic examination, it is nearly impossible to distinguish multifocal primary lesions spawned by germline mutations from secondary tumors that result from tumor seeding.
HISTOPATHOLOGY
Under low magnification, retinoblastoma appears as a basophilic mass with pink and purple foci that arises from and destroys the retina and fills part or all of the vitreous cavity
(Fig. 12-5). The basophilic areas are composed of viable retinoblastoma cells. The poorly differentiated neuroblastic cells appear blue because they have intensely basophilic nuclei and scanty cytoplasm
(Fig. 12-5A). Numerous mitoses and fragments of apoptotic nuclear debris usually are present. Retinoblastoma grows rapidly and has a marked propensity to outgrow its blood supply and undergo spontaneous coagulative necrosis. This usually occurs when the proliferating cells have extended about 90 to 110 µm away from a blood vessel
(Fig. 12-5D). The necrotic tumor cells lose their basophilic nuclear DNA and become pink or eosinophilic. The residual viable cells typically form cuffs or sleeves around vessels, imparting a
multilobulated or papillary appearance to some tumors
(Fig. 12-5C,D). These perivascular cuffs were called pseudorosettes by some in the past. Foci of dystrophic calcification develop in the necrotic parts of the tumor in many cases
(Fig. 12-6). Histopathologically, the calcific foci appear reddish-purple in hematoxylin and eosin sections, and the presence of calcium can be confirmed by the von Kossa or alizarin red stains. Electron microscopy suggests that calcification probably begins in the mitochondria of necrotic cells. Clinically, the demonstration of calcification by ultrasonography or computed tomography can help to differentiate retinoblastoma from other simulating lesions.
Intensely basophilic deposits of DNA released from necrotic tumor cells are another characteristic histopathologic feature of retinoblastoma
(Fig. 12-7). DNA deposition generally is found in eyes with extensively necrotic tumors. Typically, the DNA deposits surround retinal or iris vessels or are found in the trabecular meshwork or in basement membranes such as the lens capsule or the internal limiting membrane (ILM) of the retina.
Retinoblastoma cells often collect on the inner surface of Bruch membrane causing focal retinal pigment epithelial (RPE) detachments. Such sub-RPE deposits of tumor cells do not constitute choroidal invasion
(Fig. 12-8F). During histopathologic examination, care must be taken to distinguish artifactual contamination of the choroid (and epibulbar tissues) from true invasion
(Fig. 12-8C,D). The presence of a mixture of viable and necrotic cells is one feature that serves to identify artifactual seeding. Foci of extramedullary hematopoiesis also can be confused with uveal invasion in rare instances
(Fig. 12-8E).
More than 40% of eyes enucleated for retinoblastoma have iris neovascularization, which may produce NVG, iris heterochromia, and even secondary buphthalmos. NVG is almost three times more common in eyes with high-risk features such as massive choroidal invasion or retrolaminar optic nerve invasion.
TUMOR DIFFERENTIATION: ROSETTES AND FLEURETTES
Varying degrees of retinal differentiation occur in retinoblastoma. This is evident as Homer Wright and Flexner-Wintersteiner rosettes and photoreceptor differentiation (fleurettes) as well
(Figs. 12-9 and 12-10). Flexner-Wintersteiner rosettes represent an early attempt at retinal differentiation. Histologically, these rosettes are composed of a ring of cuboidal cells surrounding a central lumen
(Fig. 12-9B). The lumen, which corresponds to the subretinal space, contains hyaluronidase-resistant acid
mucopolysaccharide (AMP) similar to photoreceptor matrix AMP. The cells surrounding the lumen are joined near their apices by intercellular connections (zonulae adherentes), analogous to the external limiting membrane of the retina. Cilia, which exhibit the 9 + 0 pattern of microtubular doublets found in the central nervous system, project into the lumen. Cilia are hypothesized to be the precursor of photoreceptor outer segments. Although highly characteristic of retinoblastoma, Flexner-Wintersteiner rosettes are not pathognomonic because they do occur in malignant medulloepitheliomas and some pineal tumors.
Homer Wright rosettes (named after James Homer Wright) indicate neuroblastic differentiation
(Fig. 12-9A). They lack a central lumen, and their constituent cells encompass a central tangle of neural filaments. Wright rosettes are relatively nonspecific. They also occur in neuroblastoma and are a characteristic feature of cerebellar medulloblastoma.
Retinoblastoma tends to become less well differentiated with age. Numerous Flexner-Wintersteiner rosettes typically are found in eyes enucleated from younger infants, while tumors in older children tend to be poorly differentiated. A recent study found that the mean age at enucleation for tumors with many rosettes was 10.4 months; for moderate rosettes, 18.3 months; for sparse rosettes, 20.4 months; and for poorly differentiated tumors, 33.9 months. These differences were statistically significant. The presence of many Flexner-Wintersteiner rosettes in a tumor may have prognostic significance. In one series, patients who had moderately well-differentiated tumors that contained abundant Flexner-Wintersteiner rosettes had about a sixfold better prognosis than did those whose tumors lacked rosettes.
About 15% to 20% of retinoblastomas harbor very well-differentiated foci of actual photoreceptor differentiation
(Figs. 12-10 and 12-13A). Such areas contain aggregates of neoplastic photoreceptors called fleurettes by Ts’o, Zimmerman, and Fine, who described them in 1970. Photoreceptor differentiation typically is found in areas of viable tumor that appear relatively eosinophilic and paucicellular compared to adjacent undifferentiated retinoblastoma on low-magnification microscopy of H&E-stained sections
(Fig. 12-10A). The term “fleurette” denotes a bouquet-like arrangement of cytologically benign cells joined by a series of zonulae adherentes comprising a short segment of neoplastic external limiting membrane. Neoplastic photoreceptor inner segments evident as bulbous eosinophilic processes form the “flowers” of the bouquet
(Figs. 12-10B and 12-13A). Electron microscopy has disclosed stacks of cellular membranes representing early outer segment differentiation in some cases. The demonstration and characterization of photoreceptor differentiation firmly established that retinoblastoma was not a retinal glioma and affirmed that the adoption of the name retinoblastoma by the American Ophthalmological Society in 1926 at Frederick Verhoeff’s suggestion was indeed appropriate.
A retinal tumor composed entirely of photoreceptor differentiation is now thought to be a benign variant of retinoblastoma called a retinoma or retinocytoma. The cells constituting retinoma/retinocytoma are quite bland compared to those of retinoblastoma with a low nuclear to cytoplasmic ratio, finely dispersed chromatin, and no apoptosis, mitoses, or necrosis. Calcification is found, but occurs in viable parts of the tumor.
Retinomas or retinocytomas initially were thought to represent spontaneously regressed retinoblastomas on clinical grounds because they resemble retinoblastomas that have regressed after radiation therapy
(Fig. 12-11A). They have a translucent “fish flesh” appearance, contain abundant calcification that has been likened to cottage cheese, and are surrounded by a ring of RPE depigmentation. Retinomas/retinocytomas generally are small tumors that are found in eyes that retain useful vision. They may be found incidentally or are discovered in a parent or sibling when the detection of retinoblastoma in a child prompts examination of other family members. Retinocytomas are relatively resistant to radiation, as are other benign tumors. Therefore, it is not surprising that foci of photoreceptor differentiation are found more often in eyes that are enucleated after external beam radiotherapy or chemotherapy.
There is evidence that retinoma/retinocytoma is a precursor of retinoblastoma. Rare cases of clinically documented malignant transformation have been reported, and photoreceptor differentiation has been observed repeatedly at the base of endophytic retinoblastomas in enucleated eyes. Based on the results of molecular genetic studies, Dimaras recently has redefined retinoma as a precancerous lesion characterized by the loss of function of both copies of the RB1 gene, but lacking the additional genomic changes characteristic of retinoblastoma.
NATURAL HISTORY AND PROGNOSTIC FACTORS
Retinoblastoma is a highly malignant neoplasm that grows relentlessly and is invariably fatal if untreated. The tumor arises from the retina and invades the vitreous cavity and/or the subretinal space. Tumor cells eventually breach the Bruch membrane and invade the choroidal stroma
(Figs. 12-8C and 12-12B,C). Choroidal vessels serve as a major route for distant hematogenous dissemination, and massive choroidal invasion is an indication for adjuvant
chemotherapy in many centers. Involvement of the anterior chamber, iris stromal, and trabecular meshwork also is thought to affect prognosis adversely, but the actual magnitude of this effect is uncertain. Of 297 previously untreated eyes enucleated for retinoblastoma at the Wills Eye Hospital, 16.5% were found to have uveal invasion, which was classified as massive (>3 mm in diameter) in about half. Histopathologic findings that are considered high-risk for metastasis are summarized in
(Fig. 12-12).
Retinoblastoma has a marked proclivity to invade the optic nerve, and optic nerve invasion is an extremely important prognostic factor
(Fig. 12-12A). The tumor can travel along the optic nerve to the brain, or malignant cells may be dispersed along the neuraxis if they gain access to the cerebrospinal fluid. Of the 297 untreated eyes in the Wills Eye Hospital series, 38.7% had some degree of optic nerve invasion. Retrolaminar optic nerve invasion, an important indication for adjuvant chemotherapy, was present in about 10.4%. Mortality rates correlate directly with the depth of optic nerve invasion. In one series, 10% died if there was superficial invasion of the nerve head only (grade I) and 29% if the tumor reached and invaded the lamina cribrosa (stage II). Mortality rose to 42% when there was retrolaminar invasion (grade III) and 78% when the tumor extended to the surgical margin (grade IV). Hence, surgeons should try to obtain as a long a segment of optic nerve as possible when enucleating an eye that is known or suspected to harbor retinoblastoma. In addition, enucleation should not be performed by an inexperienced surgeon.
The Children’s Oncology Group recently conducted a large prospective study that investigated the role of adjuvant chemotherapy in patients with unilateral retinoblastoma. Histopathologic features in that study that were considered
high risk and served as an indication for adjuvant chemotherapy included massive posterior uveal invasion, retrolaminar optic nerve invasion, and any degree of concurrent choroidal and optic nerve invasion. Massive choroidal invasion was defined as >3 mm in diameter. Only one of the 93 patients who had high-risk features and received chemotherapy developed metastases. Central review of the ocular histopathology disclosed that a significant number of cases (15.7%) had been misclassified by pathologists at the contributing hospital
(Fig. 12-8B).
Mendoza et al. have reported that severe anaplasia is an additional risk for metastasis and death in patients with retinoblastoma who do not have standard high-risk features. In their grading scheme, mild anaplasia was characterized by moderate nuclear pleomorphism, occasional mitotic figures, and early retinal differentiation evident as Flexner-Wintersteiner and/or Homer Wright rosettes
(Fig. 12-13). Tumors with moderate anaplasia were poorly differentiated and had enlarged nuclei, definite pleomorphism, and frequent mitotic figures. Severely anaplastic tumors were poorly differentiated and had cells with large nuclei that were extremely pleomorphic and angular, rhomboid, or fusiform in shape. Cell wrapping and numerous mitotic figures were additional criteria.
After retinoblastoma has filled the globe and destroyed its contents, it extends extraocularly. Anteriorly, the tumor extends through the aqueous outflow pathways, preexisting emissarial canals, or perforations in the cornea. Posterior segment tumors that have invaded the choroid can extend extraocularly through emissarial canals or can invade the orbit by directly infiltrating and destroying the sclera. Secondary buphthalmos and staphylomas caused by secondary glaucoma can facilitate extrascleral extension. Secondary closed-angle glaucoma caused by pupillary block or iris neovascularization is relatively common in eyes with retinoblastoma. Forty-three percent of eyes enucleated for retinoblastoma have iris neovascularization, and 26% have NVG. In addition, NVG is much more common in eyes with high-risk histopathologic features.
Retinoblastoma typically metastasizes hematogenously to lungs, bones, brain, and other organs. Cervical and preauricular adenopathy can develop when tumors with extensive anterior segment involvement gain access to lymphatics in the conjunctival stroma. Lymphatics are not present in the orbit. Metastatic disease usually becomes evident within 1 or 2 years after therapy. Late metastasis is so rare in retinoblastoma that it should raise the possibility of a second independent primary tumor such as pineoblastoma.
Retinoblastoma occasionally undergoes spontaneous regression. A typical bona fide example of spontaneous regression is a phthisical eye that contains foci of calcified retinoblastoma cells
(Fig. 12-11B). Both the tumor regression and phthisis bulbi probably are caused by extensive ischemic necrosis in an eye with severe NVG. Many lesions that previously were thought to be spontaneously regressed tumors actually are retinomas or retinocytomas.
MOLECULAR GENETICS AND THE RETINOBLASTOMA ONCOGENE
Retinoblastoma is a hereditary cancer: 5% to 10% of the tumors are inherited in what appears to be a classic mendelian autosomal dominant trait; carriers transmit the tumor to one half of their offspring
(Fig. 12-14B). Bilaterality is a characteristic feature of heritable retinoblastoma. Bilateral tumors occur in about 60% of patients who have germline mutations in the
RB1 gene. The average age of patients with familial retinoblastoma is about 1 year. A germline mutation should be suspected when retinoblastoma is found in a very young infant.
The great majority of retinoblastomas (85%) are sporadic tumors that arise in patients with a negative family history
(Fig. 12-14A). About 75% of these sporadic tumors are caused by somatic mutations in retinal cells. Sporadic retinoblastomas caused by somatic mutations are invariably unilateral, unifocal tumors that cannot be passed on to progeny, and they tend to occur in older infants (mean age 2 years). The remaining 25% of sporadic cases are caused by germinal mutations and represent new familial cases that are transmissible.
A small number (<5%) of retinoblastomas occur in infants who have a variety of congenital anomalies and are found to have deletions in the long arm of chromosome 13 that are evident in karyotypic analysis. In addition to retinoblastoma, the 13Q deletion syndrome comprises severe mental retardation and other anomalies including microcephaly, hypertelorism, ptosis, micrognathia, deformed low-set ears, a wide nasal bridge, cardiac anomalies, anal atresia, microphthalmia, colobomas, and cataracts. The association of retinoblastoma with the 13Q deletion syndrome initially suggested that the retinoblastoma gene was located on chromosome 13.
The paradigmatic human recessive oncogene or tumor suppressor gene, the retinoblastoma or RB1 gene, is located in the 14 band of the Q or long arm of chromosome 13 (13q14). The RB1 gene is 180,388 base pairs in length, and its protein product pRB comprises 928 amino acids. pRB is abundant in the nucleus, where it is involved in control of the cell cycle. During the G1 or resting phase of the cell cycle, pRB is bound to transcription factors such as E2F. Phosphorylation of pRB causes release of E2F. Uncomplexed E2F, in turn, activates a variety of other genes and transcription factors that are important in the initiation of DNA synthesis (S phase). Absence of pRB causes continual cell division and lack of terminal differentiation. The RB protein is phosphorylated by cyclin D2 and its cyclin-dependent kinase (cdk) partner. Certain oncoviruses cause tumors by synthesizing proteins (e.g., adenoviral protein E1A and SV40 large T protein) that bind to pRB and inactivate it.
Classically, the
RB1 gene is thought to cause cancer when its protein product is absent or dysfunctional
(Fig. 12-15). Healthy persons have two normal or wild-type RB1 genes. Both alleles of the
RB1 gene are absent or inactivated in the retinoblastoma cells. Carriers of familial retinoblastoma
are heterozygous for the
RB1 gene. Although the pRB produced by a heterozygote’s single functional gene is sufficient to inhibit tumorigenesis, heterozygotes are at substantial risk to develop retinoblastoma. Although retinoblastoma appears to be inherited clinically as an autosomal dominant trait, the gene is recessive at the molecular level
(Fig. 12-16).
The genotype of a child with familial retinoblastoma includes one functional copy of the
RB1 gene
(Fig. 12-15B). The second copy of the gene is mutated and encodes dysfunctional RB protein or it may be absent. Retinoblastoma develops when the solitary remaining wild-type gene is lost or inactivated, by chance, in a cell within the developing retina
(Fig. 12-15C). Cytogenetic mechanisms responsible for gene inactivation (and resultant homozygosity for the recessive allele) include chromosomal loss or deletion, somatic recombination, and point mutation. The spontaneous mutation rate of the normal wild-type
RB1 gene is 1 in 10 million or greater. It is estimated that 100 million mitoses occur during the growth and development of each retina. Hence, it is highly probable that the second functional copy of the
RB1 gene will be lost or inactivated in at least one retinal cell in a carrier with a heterozygous genotype. Furthermore, it is equally probable that gene inactivation and tumorigenesis will occur in both eyes
(Fig. 12-13D).
Immune surveillance appears to arrest some tumors, despite the appropriate additional mutations. This probably is responsible, in part, for the incomplete penetrance of the RB1 gene, which is estimated to be ˜80% (in other words, there is an 80% chance that one tumor will develop in one eye). A few families with low-penetrance retinoblastoma have been reported. They have reduced levels of wild-type RB protein or mutant RB protein that retains partial activity.
About one third of patients with retinoblastoma have bilateral tumors
(Figs. 12-1B and 12-17). The sporadic occurrence of bilateral tumors indicates that the affected patient has a germline mutation and is capable of transmitting the disease to one half of his or her offspring. Unfortunately, the opposite is not true. As a result of incomplete penetrance and expressivity, 10% to 15% of unilateral, sporadic retinoblastomas are caused by potentially transmissible, germline mutations. Statistics used for genetic counseling
(Table 12-1) reflect both the effect of gene penetrance and the proportion of familial, chromosomal deletion and sporadic somatic and germinal retinoblastomas in the population.
Molecular genetics readily explains several hitherto puzzling clinical features of retinoblastoma. For example, retinoblastoma is predominantly a tumor of early childhood. Although rare adult cases have been reported, the mean age at diagnosis is 18 months and the tumor is extremely rare after age 4 years. Cytogenetic misadventures that cause gene inactivation generally occur during cellular division. Most mitotic activity in the retina actually ceases before birth, making neoplastic transformation in older persons highly unlikely. Retinoblastomas caused by germline mutations typically develop at an earlier age than do retinoblastomas caused by sporadic somatic mutations (mean age 12 months vs. 23 months), presumably because only a single RB1 gene allele, rather than two, must be inactivated. Knudson graphically compared the ages of patients who had unilateral and bilateral tumors with the logarithm of the proportion in each group as yet undiagnosed. His results led him to postulate that two separate events or “hits” are necessary for the development of sporadic retinoblastomas (“two-hit hypothesis”). The curve for bilateral, hereditary tumors is a simple exponential relation, consistent with a single gene inactivation or “hit” superimposed on the inherited genotypic defect.
Patients who are heterozygous carriers of familial retinoblastoma are predisposed to develop other malignant tumors. A survivor of bilateral retinoblastoma has a 20% to 50% chance of developing a second tumor within 20 years (Armed Forces Institute of Pathology series: 26% within 30 years). These secondary nonocular tumors include osteogenic and other soft tissue sarcomas, carcinomas of the upper respiratory passages, malignant melanomas, and carcinomas of the skin. In the past, many second tumors occurred within the field of irradiation many years after external beam radiotherapy for intraocular retinoblastoma. However, a 500-fold increase in the incidence of osteogenic sarcoma in the nonirradiated femur has been reported.
Some of the most interesting secondary nonocular neoplasms that develop in patients with germline mutations are tumors of the pineal gland or parasellar region that resemble ectopic retinoblastomas. This association between pinealoma (also termed pineoblastoma) and bilateral hereditary retinoblastoma has been termed “trilateral retinoblastoma.” The pineal gland, which serves as a “third eye” in some primitive reptiles, shares antigenic determinants with the retina and exhibits transient photoreceptor differentiation in the neonatal rat. Photoreceptor differentiation has been identified in human pineal tumors.
The RB1 gene has been implicated as a contributing factor in a variety of other systemic malignancies, including breast, lung, and bladder cancer. This is not surprising, considering the RB1 gene’s fundamental role in control of the cell cycle.
Recent molecular genetic studies suggest that the initial concepts outlined above concerning the
RB1 tumor suppressor gene and its role in the pathogenesis of retinoblastoma are an oversimplification. Dimaras showed that both copies of the
RB1 gene are inactivated and Rb protein is absent in retinoma/retinocytoma, which is now considered to be a benign precursor lesion of retinoblastoma.
Additional mutations are necessary for malignant transformation of retinoma into retinoblastoma. The retinoblastoma gene plays an important role in the regulation of cellular division in all cells in the body, yet only retinoblastoma, a relatively rare neoplasm that affects a small population of highly differentiated cells, results from the inactivation of both copies of the
RB1 gene. The highly retinal-specific predisposition imposed by mutations in the
RB1 gene is hypothesized to result from the specific pattern of expression of other genes in the unidentified cell of origin in the developing human retina rather than
RB1.
Although it is generally accepted dogma that retinoblastoma is initiated by mutations in both alleles of the
RB1 gene, extensive genetic testing fails to reveal mutations in the
RB1 gene in ˜2% of tumors from probands who have unilateral tumors and a negative family history for retinoblastoma. Rushlow and coworkers have proposed that
RB1+/+ tumors could to be initiated by amplification of the NMYC oncogene. They state that these tumors tend to be large and invasive, are diagnosed at a very young age, and have characteristic histopathologic features. They are composed of undifferentiated cells that have large, prominent nuclei, multiple nucleoli, extensive apoptosis, and little calcification. Although Homer Wright rosettes may be present, they lack Flexner-Wintersteiner rosettes, which usually are abundant in retinoblastomas removed from young children. The nuclei of these tumors are immunoreactive for RB1 protein. Affected children are not at risk for developing tumors in their fellow eye, and siblings are not at risk for retinoblastoma as well. This topic is somewhat controversial as significant amplification of the NMYC gene is known to occur in other neoplasms and has been found in a number of retinoblastomas with mutations in the
RB1 gene
(Fig. 12-18).
THE CLASSIFICATION AND TREATMENT OF RETINOBLASTOMA
There are several staging systems for retinoblastoma. The older Reese-Ellsworth classification has prognostic significance for retention of an eye, maintenance of sight, and the control of local disease but is complicated and recently has become less useful because it is based on the response to external beam radiotherapy, a treatment modality that has fallen out of favor and is now used infrequently. The newer International Classification of Retinoblastoma (ICRB) is a relatively simple, practical classification based on clinical findings that was designed primarily to evaluate the potential response to modern therapy including chemoreduction. The International Classification includes five stages A through E. Tumors with localized subretinal seeds are placed in group C and more diffuse seeding in group D. Group E tumors generally require enucleation. They fill more than 50% of the globe or have opaque media, NVG, or high-risk features such as postlaminar optic nerve invasion, significant choroidal invasion, and invasion of the anterior chamber, sclera, or orbit.
A detailed discussion of the treatment of retinoblastoma is beyond the scope of this chapter. A number of reports dealing with modern treatment modalities are referenced below. In short, there have been major changes in the treatment of retinoblastoma in developed countries in recent years. A concerted effort has been made to avoid the use of external beam radiotherapy, which leads to disfiguring facial deformities and predisposes retinoblastoma gene carriers to secondary malignant neoplasms such as soft tissue sarcomas in the field of radiation. In addition, fewer eyes with retinoblastoma currently are being enucleated because there has been a shift at major centers to eye and even sight-sparing therapy using chemotherapy.
Chemotherapeutic treatment regimens include the intravenous administration of chemotherapeutic agents, direct targeting of the tumor by administration via the ophthalmic artery, and most recently intravitreal injection of drugs such as melphalan to target vitreous seeds. In some cases, intravenous chemotherapy is used to shrink tumors so they become amenable to additional treatment or “consolidation” with local therapeutic agents such as cryotherapy, infrared laser transpupillary thermotherapy, or plaque brachytherapy. This type of chemotherapy is called chemoreduction and typically employs multiple cycles of vincristine, etoposide, and carboplatin. It has been used frequently to treat patients with bilateral retinoblastoma. Chemoreduction can help eradicate tumors and save eyes that once required enucleation.
Intravenous chemotherapy is also used in an adjuvant fashion when histopathologic examination of eyes treated by enucleation discloses certain high-risk features that are believed to place a patient at high risk for metastasis (see above). Intravenous chemotherapy kills retinoblastoma cells that have metastasized from the primary intraocular tumor. In addition, it may decrease the frequency of pineoblastoma, the potentially fatal retinoblastoma-like tumor of the pineal gland that can develop in patients with germline mutations in the RB1 gene. Adjuvant chemotherapy is one of the factors that is responsible for improved survival in recent years.
During the past several years, both of these forms of intravenous chemotherapy have been supplanted, to some extent, by intra-arterial chemotherapy (IAC), which is used extensively at some major centers. Also termed ophthalmic artery chemosurgery, IAC of retinoblastoma initially was developed by Kaneko et al. in Japan where enucleation is abhorrent for cultural reasons. In this procedure, chemotherapeutic drugs, typically melphalan, are directly administered to the tumor-containing eye via a catheter inserted within or in close proximity to the ophthalmic artery. The catheter is introduced into the femoral artery and advanced under fluoroscopic guidance. Several treatment sessions usually are required.
The topic of IAC remains controversial. IAC can totally eradicate retinoblastomas in many instances, and spectacular treatment results have been reported. Theoretically, the technique should decrease potential complications of systemic chemotherapy since most of the drug is administered solely to the eye. However, IAC has its limitations and does not work in all cases. Furthermore, the technique theoretically carries a higher risk for metastatic disease compared to intravenous chemotherapy because the drug does not reach extraocular tissue where metastatic cells can reside. About 20% of a series of eyes that were enucleated for retinoblastoma at the Wills Eye Hospital were found to harbor high-risk features such as massive choroidal or postlaminar optic nerve invasion that are indications for adjuvant chemotherapy. These high-risk features will not be detected if IAC is performed, and tumor cells that have metastasized to extraocular sites will not be killed. Similarly, IAC should not effectively control pineoblastoma.
Other complications of IAC such as ischemic chorioretinal and RPE atrophy have been observed clinically and confirmed histopathologically. Intravascular foreign bodies and thrombosed vessels have been found in some cases, stressing the need for skilled operators and meticulous technique. Pulsatile optic nerve and choroidal blanching, retinal artery narrowing, and retinal artery precipitates have been observed during drug infusion in infants and experimental animals, raising additional concerns about IAC and the long-term vascular toxicity of melphalan.
The technique of intravitreal chemotherapy was reported after the last edition of this text was published. This technique uses direct intravitreal injections of chemotherapeutic drugs, usually the alkylating agent melphalan, to treat vitreous seeds of retinoblastoma, which are one of the most important causes of treatment failure and lost eyes. In the past, clinicians were cautioned not to stick needles in eyes containing retinoblastoma because the risk of extraocular extension and orbital dissemination
was considered high. Such complications have not been encountered, however, because the new intravitreal injection technique incorporates safeguards including multiple cycles of postinjection freeze-thaw cryotherapy to the injection site.
In developed countries, retinoblastoma is one of the great success stories of pediatric ocular oncology. Survival rates from the primary tumor approach 100% at major centers, and secondary tumors are now the primary cause of death. Unfortunately, this is not the case in developing countries, where patients frequently present with advanced stages of the disease and therapy generally is palliative
(Fig. 12-2).