Clinicopathologic Features of Retinoblastoma After Primary Chemoreduction

Background  Primary chemotherapy is a new treatment approach in retinoblastoma, aiming to avoid radiogenic adverse effects, such as second tumor–associated mortality, as observed following external beam irradiation.

Objective  To describe the clinical and histopathologic regression pattern after primary chemotherapy in retinoblastoma.

Methods  Five patients with sporadic bilateral retinoblastoma underwent planned enucleation of their functionally blind eye after 2, 3 (in 2 patients), 4, and 6 courses of primary chemotherapy with carboplatin, etoposide, cyclophosphamide, and vincristine. The eyes were examined histopathologically, using light microscopy and immunohistochemical analysis with proliferation markers.

Results  One patient had a type 1 (cottage cheese) regression and 4 patients had either a type 2 (fish flesh) or a type 3 (combined) regression pattern. Histopathologic examination revealed a complete tumor necrosis in 1 patient with type 1 regression after 3 courses of chemotherapy and in 1 patient with type 3 regression after 4 courses of chemotherapy. The remaining 3 patients with type 2 or type 3 regression had histologically still active proliferative tumor cells after 2, 3, and 6 courses of chemotherapy.

Conclusion  This article correlates histopathologically the clinically described efficacy of primary chemotherapy in the treatment of retinoblastoma, underlining, however, the necessity of careful observation and the use of ancillary treatment whenever there is no complete tumor regression.

RETINOBLASTOMA is the most common intraocular tumor in childhood, occurring in 1 of 17000 to 24000 live births,^1-3 independent of race and sex.⁴ With modern methods of treatment, the survival rate is more than 90%, while untreated, retinoblastoma is always fatal.⁵ Retinoblastoma may occur as a hereditary or nonhereditary tumor. Tumors in nonhereditary retinoblastoma are typically solitary and unilateral with no family history and no detectable chromosomal abnormalities. About 40% of retinoblastoma are caused by a germ line mutation and include the cases with a positive family history.⁶ Most children with hereditary mutations develop multifocal retinoblastoma in both eyes. As many as 15% of sporadic unilateral retinoblastomas may be hereditary.⁷

Until recently, the standard therapy for bilateral retinoblastoma in most ophthalmic oncology centers included external beam radiotherapy (EBR) in the majority of patients; chemotherapy was reserved for extraocular disease.⁸ Several studies^5,9-11 have shown that EBR is a powerful tool in the management of advanced retinoblastoma, including stages Va and Vb of the Reese-Ellsworth classification system. Adverse effects of EBR, such as midface growth inhibition, which occurs in 90% of the patients after EBR,¹² seemed to be tolerable in view of preserving an only functional eye.^12,13 The effect of EBR on the prevalence of secondary malignant tumors, however, remained unclear until the retinoblastoma group at Cornell University, New York, NY, published their results¹⁴ on long-term survival of children with hereditary retinoblastoma after EBR. In this study, the risk of secondary nonocular malignant tumors increased 6-fold after EBR, imposing a 35% risk of secondary cancers at the age of 30 years.¹⁴ Consequently, the need for alternative treatment modalities in eye-salvaging therapy of retinoblastoma without EBR is evident. This new treatment course ideally should have the same local tumor control rate, but a lower morbidity. Alternatives avoiding the long-term complications of EBR include local treatment with photocoagulation and cryocoagulation and plaque radiotherapy making use of low energy γ-ray plaques^15,16 or β-ray plaques.¹⁷ These treatment modalities, however, are limited by tumor size and location. Large tumors, tumors with vitreous seeding, or tumors at the posterior pole cannot be treated with these modalities alone and still maintain the tumor control rates comparable to EBR. Combining local treatment modalities with systemic chemotherapy offers an attractive alternative, as large tumors and even bilateral disease may become accessible to laser treatment, cryocoagulation, or brachytherapy, avoiding the potential complications of EBR.^18-22 The aim of this article is to compare the clinical and histopathologic findings of tumor regression in retinoblastoma after primary chemotherapy to investigate the efficacy of this treatment.

Five patients with bilateral nonfamilial retinoblastoma are described who received primary chemotherapy for tumor reduction (chemoreduction) and enucleation of their more severely affected eye after 2, 3 (in 2 patients), 4, and 6 courses of chemotherapy.

All patients were white; 3 were boys and 2 were girls. At the initial examination, the patients underwent complete ophthalmologic evaluation under general anesthesia, as well as computed tomographic examination of the orbits and magnetic resonance imaging of the orbits and brain. Laboratory analysis of cerebrospinal fluid and bone marrow aspirate was performed to rule out metastatic disease. Each eye was categorized according to the Reese-Ellsworth classification system. Table 1 shows the patient data at initial examination.

Chemotherapy was performed according to the treatment regimen as shown in Table 2 and repeated every 3 weeks. After course 3, the whole regimen was repeated, starting with course 1 at day 64. Table 3 shows the number of chemotherapy courses these children received before enucleation.

Chemotherapy was performed for tumor size reduction (chemoreduction) to enable ancillary local treatment for the residual tumors. After each chemotherapy course, the children had a complete ophthalmologic examination to evaluate the response to treatment. If they had further tumor regression, treatment was continued. If it was judged that an eye was beyond salvage, enucleation was advised.

Systemic chemotherapy was tolerated generally well by all patients. Typical adverse effects, such as mild reversible myelosuppression and alopecia, occurred in all patients.

The enucleated eyes were immersed in a 4% buffered formaldehyde solution. After macroscopic examination and photography, the eyes were transilluminated and cut in such a plane to include the bulk of the tumor in the main pupil-optic nerve block. Serial sections were obtained in several levels through the pupil-optic nerve block and the calottes. Sections (4 µm thick) were made and stained with hematoxylin-eosin and periodic acid–Schiff.

Immunohistochemical analysis was performed to evaluate the proliferative potential of tumor cells using the monoclonal antibodies (MoAb) PC-10, which binds to the proliferating cell nuclear antigen, and MIB-1, which binds to the Ki-67 antigen. The macrophage response was assessed by using the MoAb PG-M1. The alkaline phosphatase antialkaline phosphatase technique was performed with an antibody dilution of 1:10 for MIB-1, 1:50 for PC-10, and 1:100 for PG-M1. The MoAbs PC-10 and PG-M1, as well as the alkaline phosphatase antialkaline phosphatase kit, were obtained from DAKO Diagnostika GmbH, Hamburg, Germany; the MoAb MIB-1 was obtained from Dianova GmbH, Hamburg. The amount of proliferating cells was determined semiquantitatively by estimating the percentage of positive stained cells divided by the entire cell number as counted in 10 high-powered fields (×400 magnification) by 1 observer (N.E.B.). In all cases, positive controls were obtained by staining lymphomas with known proliferative activity. Negative controls were obtained by assessing nonaffected areas within the same section (eg, cornea, sclera, lens, choroid), as well as by omission of the primary MoAb.

All patients described had nonfamilial bilateral retinoblastoma and Reese-Ellsworth stage V disease in the enucleated eyes. Three patients had vitreous seeding (Reese-Ellsworth stage Vb) at the initial evaluation, and all patients had subtotal or complete retinal detachment (Table 1). One patient had secondary glaucoma and heterochromia at the initial evaluation, and 1 patient had esotropia with retrohyaloid hemorrhage. Tumor height before treatment varied in the enucleated eyes from 6.0 to 12.0 mm (mean, 9.4 mm). In all patients, there was tumor reduction observed after various courses of chemotherapy, resulting in tumor height from 3.0 to 6.1 mm (mean, 5.1 mm) (Table 3). Serous retinal detachments were either reduced (patients 1 and 2), markedly reduced (patient 4), or resolved after chemotherapy (patients 3 and 5), which was not anticipated before commencing chemotherapy because of detachment severity in most patients. After having received 3 courses of chemotherapy, patient 3 developed a vitreous hemorrhage that prevented further evaluation of tumor regression. In the remaining 4 patients where ophthalmoscopic evaluation was possible before enucleation, the tumor regression pattern was similar to that described after EBR, but occurred more quickly after initiation of therapy. The main effect of chemotherapy was present after only 2 courses.

Enucleation of the more severely affected eye in patients 1 and 2 was performed after 2 and 3 courses of chemotherapy, respectively, because these patients had persisting retinal detachment, indicating that they were functionally blind. Nevertheless, in both eyes, a massive tumor reduction was observed. Especially in patient 2, who was initially evaluated as having vitreous seeding, the tumor appeared to be in complete type 1 regression on clinical evaluation. Patient 1 had a small fish-flesh–appearing tumor nodule near the optic nerve, resulting in a type 3 regression pattern (Table 3).

In patients 3 and 4, enucleation was postponed after 3 and 4 courses of chemotherapy, respectively, because choroidal and optic nerve involvement could not be ruled out on ultrasound, computed tomographic, and magnetic resonance imaging scans. It was judged that enucleation after chemoreduction would be the safer option, limiting the metastatic potential in the case of choroidal involvement and the cerebral spread in the case of optic nerve involvement, if either were present. In patient 5, enucleation was advised after 3 courses of chemotherapy, when it became evident that the more severely affected eye could not be salvaged. The patient's parents refused enucleation at this point and consented only after the sixth course. In patient 3, it was obvious on clinical evaluation that there was still active tumor present after 3 courses of chemotherapy, since there was no sign of a cottage cheese–appearing reaction observed. However, the tumor was markedly reduced from 9.3 to 6.1 mm with a type 2 regression. Patients 4 and 5 both showed some degree of cottage cheese–appearing reaction, resulting in type 3 tumor regression. Especially in patient 4, there was a massive tumor reduction from 12.0 to 5.0 mm. Only a few small nodules of residual fish-flesh tumor were present after 4 courses of chemotherapy (Table 3).

In all patients, the same regression pattern was observed in the fellow eye. These retained eyes received ancillary local treatment by ruthenium brachytherapy, cryocoagulation, laser treatment, or thermotherapy. To date, none of these fellow eyes have had to be enucleated, and all patients are free of metastatic disease.

Macroscopic examination of all enucleated eyes showed various patterns of tumor regression that were similar to the clinical evaluation. Macroscopically, the most pronounced tumor regression was observed in patients 1, 2, and 4, who had massive tumor shrinkage resulting in calcification, which had been clinically visible having a cottage cheese appearance.

Only patient 2 had a pure cottage cheese–appearing type 1 tumor regression with no residual fish-flesh tumor appearance. Microscopically, the tumor in this patient was calcified and necrotic and completely replaced by reactive gliosis surrounding cystic spaces (Figure 1). Migrated retinal pigment epithelial cells and macrophages were scattered in the gliotic tumor, along with lipid exudations forming cholesterol clefts. There were no vital tumor cells present. Clusters of necrotic retinoblastoma cells were present in the vitreous in areas of the clinically described vitreous seeding before chemotherapy. These histopathologic features corresponded to a type 1 regression.

In contrast to patient 2, patients 1, 4, and 5 had some degree of fish-flesh–appearing reaction along with cottage-cheese type regression (combined type 3 regression pattern) (Figure 2, A). Patient 3 presented exclusively a fish-flesh–appearing regression (type 2 regression pattern) (Figure 2, B).

In patients 1 and 4, the bulk of the tumors had identical histological features as observed in patient 2 with the exception that there were small foci of vital-appearing retinoblastoma cells at the tumor base near the optic nerve in conventional hematoxylin-eosin and periodic acid–Schiff stains. These foci had a moderate tumor differentiation with the presence of numerous Flexner-Wintersteiner rosettes and contained cells with mainly round nuclei that had finely granular nuclear chromatin. These areas had, on macroscopic and clinical observation, a fish-flesh appearance. Whether these cells were truly vital and proliferating was evaluated by using immunohistochemical analysis with the proliferation markers PC-10 and MIB-1.

A common feature in all the patients who had some degree of cottage cheese regression and retinal detachment was the presence of small clusters of subretinal necrotic calcified retinoblastoma cells overlying the Bruch membrane.

In the fifth patient who also had a combined type 3 regression at initial evaluation, macroscopic examination revealed a residual retrohyaloid hemorrhage that partially obscured the posterior pole. Most of the residual tumor had a fish-flesh appearance, except at its apex that was cystic with small foci of cottage cheese–appearing regression. On microscopic examination, the tumor appeared mainly viable except for the cystic apex that contained foci of basophilic granular necrotic tumor corresponding to the small cottage-cheese–appearing areas. The tumor was composed mainly of pleomorphic undifferentiated retinoblastoma cells with large and hyperchromatic nuclei. Flexner-Wintersteiner rosettes were only rarely present, and mitotic figures were abundant. In multiple retinal areas, there was also a diffuse flat exophytic infiltration, but choroidal or optic nerve invasion could not be observed.

Patient 3 was the only one who had a pure fish-flesh–appearing type 2 regression pattern. On macroscopic examination, there was a diffuse multinodular thickening of the entire retina and a prominent endophytic retinoblastoma at the inferior nasal retinal quadrant that protruded into the vitreous with overlying retinal seeding (Figure 2, B). On microscopic examination, there was necrosis in the center of the main tumor, but calcification was not present (Figure 3, A). The tumor was well differentiated with the presence of Flexner-Wintersteiner rosettes and mainly composed of viable retinoblastoma cells with large and hyperchromatic nuclei. Mitotic figures were abundant. Over the tumor apex, there were clusters of vital-appearing pleomorphic retinoblastoma cells producing vitreous seeding. Infiltration of the tumor into the choroid could not be observed. The entire inner retinal circumference was covered by a multilayered sheath of retinoblastoma cells, accounting for retinal seeding. A marked invasion of the prelaminar part of the optic nerve was present at the optic disc with vital retinoblastoma cell clusters infiltrating around the central retinal artery and vein (Figure 3, B).

General pathologic features included the following: only patients 1 and 4 had circular anterior synechiae in the anterior segment. Peripheral superficial iris neovascularization and posterior synechiae producing a pupillary membrane were present only in patient 1, along with pigment clumping and dispersion between the plicae of the ciliary body. The retina in all cases had some degree of diffuse reactive gliosis with cystoid edema. In patients 1 and 2, who had the most pronounced retinal detachment, there was marked photoreceptor outer segment atrophy. Segmental cystic optic nerve atrophy was also a common feature. None of the patients had a laminar or postlaminar optic nerve invasion.

No proliferating tumor cells could be demonstrated using the MoAbs PC-10 and MIB-1 in any tumor sections in patients 2 and 4 after 3 and 4 courses of chemotherapy, respectively (Table 3). It was also notable that the apparent vital tumor areas in conventionally hematoxylin-eosin–stained sections in patient 2 did not show any reactivity for proliferation markers even after repeated immunohistochemical staining. In addition, negative controls did not show any false-positive staining (Figure 4, A). Immunoreactivity with the macrophage marker PGM-1 showed an accumulation of macrophages around vessels and around clusters of calcification.

In patient 1, immunohistochemical analysis using the MoAbs PC-10 and MIB-1 showed a proliferation rate of 10% within the vital area at the tumor base after 2 courses of chemotherapy (Figure 4, B). In patients 3 and 5 after 3 and 6 courses of chemotherapy, respectively, immunohistochemical analysis with MoAbs PC-10 and MIB-1 still showed a massive proliferation rate of 90% within the vital areas of the tumor (Figure 4, C).

The clinical regression patterns of retinoblastoma after radiotherapy are well documented and classified into 4 types. Type 1 regression is characterized by rapid tumor reduction that takes on the appearance of cottage cheese. In type 2 regression, the tumor changes from pink to gray and loses its vascularity. This regression pattern has been compared with the appearance of fish flesh. Type 3 regression is a combination of the first 2 patterns and is probably the most commonly seen. In type 4, there is a complete tumor regression within a flat chorioretinal scar. The latter regression pattern is most commonly encountered after radioactive plaque brachytherapy, whereas the first 3 patterns are usually observed after EBR.²³ Histopathologic reports^23,24 on these regression patterns have been scarce and are mainly confined to unsuccessful brachytherapy and EBR or the rare event of spontaneous regression of retinoblastoma. The absence of clinicopathologic reports on tumor regression of retinoblastoma is mainly caused by the highly successful cure rates of these treatment modalities with an overall survival of 95% for all retinoblastoma patients and ocular cure rates of more than 90% in Reese-Ellsworth stages I through III.¹¹ Any new approach to the management of retinoblastoma has to be compared with these results, which are mainly based on the success rate of high-precision EBR.¹¹ There is, however, a definite need for a new approach as there is no doubt that EBR enhances the risk of secondary malignant tumors dramatically. The recently published results^18-25 of phase 1 and 2 studies on chemotherapy combined with ancillary therapy are promising and demonstrate that these treatment modalities have the potential to produce the same or better cure rates when compared with EBR. Primary chemotherapy may result in complete tumor regression indistinguishable from tumor regressions detectable after EBR. In nearly all patients, intraocular retinoblastomas will show at least a response to chemotherapy, resulting in retraction of juxtamacular or juxtapapillar tumors from these vital structures.²²

The use of chemotherapy to reduce tumor volume before surgery is a common practice in pediatric oncology.²⁶ In applying this strategy in the management of bilateral retinoblastoma, it seems appropriate to conserve both eyes at first. The decision to enucleate 1 eye can be postponed at least until early response to primary chemotherapy has been assessed. It is then easier to judge which eye is salvageable and which not. This rationale is based on the observation of dramatic tumor regression in eyes with advanced retinoblastoma after use of full chemotherapy and focal treatment, resulting in useful vision.

The chemotherapy regimen used in our patients was in concordance to the recently published studies^19-22 that used carboplatin as the mainstay of treatment, with the only difference being the inclusion of cyclophosphamide in the protocol. These studies have shown the efficacy of carboplatin, vincristine sulfate, and etoposide in chemoreduction for retinoblastoma. After intravenous administration, carboplatin enters the human eye and is able to form positive-charged platinum complexes, which are highly reactive with DNA, forming platinum-DNA adducts. The events that are most lethal to cells appear to be interstrand cross-links and intrastrand guanine-guanine cross-links.^27,28 Etoposide, derived from the North American May apple, exerts its cytotoxicity through interrupting the breakage reunion reaction of DNA topoisomerase type II and is most effective against late S phase and early G₂ phase cells.²⁹ The vinca alkaloid vincristine sulfate is lethal to mitotic cells by binding to tubulin and preventing the polymerization of tubulin subunits into microtubules, resulting in mitotic cells being arrested in metaphase from which cell death rapidly ensues.³⁰ Cyclophosphamide is one of the main alkylating agents being used that exerts its toxicity through covalent binding to DNA, thus impeding DNA replication. An increased incidence of secondary primary tumors has been attributed to the use of cyclophosphamide in children with RB1 mutations; therefore, its use should probably be avoided when possible.^31,32 Nevertheless, it is an effective chemotherapeutic drug in patients with advanced retinoblastoma³³ and that is why it was used in these 5 children.

As has recently been published,^20-22 it is possible to reduce chemotherapeutic treatment to a 3-drug regimen (carboplatin, etoposide, and vincristine) or even a 2-drug regimen (carboplatin and etoposide)¹⁹ thus minimizing toxic effects and possible second malignant neoplasms.

Enucleation was performed in our patients after 2, 3 (in 2 patients), 4, and 6 courses after commencement of chemotherapy in anticipation of a reduction in tumor size. All patients had advanced Reese-Ellsworth stage Va or Vb tumors in which optic nerve or choroidal invasion could not have been ruled out preoperatively with certainty.

Since it is difficult, if not impossible, to determine the viability of treated tumor cells based on conventional histological analysis alone, we used immunohistochemical analysis to determine the proliferative activity of retinoblastoma after various courses of primary chemotherapy. The MIB-1 antibody is a recently developed MoAb against a recombinant fragment of the Ki-67 antigen, which enables the assessment of cell-proliferating activity.³⁴ Similarly, the PC-10 MoAb detects an epitope on the proliferating cell nuclear antigen.³⁵ The concentration of proliferating cell nuclear antigen increases during the S phase of the cell cycle, thus providing an index for proliferative activity, which was determined semiquantitatively by estimating the percentage of positive stained cells divided by the entire cell number (Table 3).

In patients 2 and 4, after 3 and 4 courses of chemotherapy, respectively, we found a complete tumor regression without any proliferative activity. In patient 2, there was clinically a regression type 1 present that corresponded to a histologically confirmed complete tumor necrosis. In serial sectioning, no viable tumor cells could be observed. There was total tumor necrosis with residual calcification, retinal gliosis, and fibrosis present. In patient 4 after 4 courses of chemotherapy, there was a more complex, but highly interesting, finding. On conventional histological analysis, there were still areas of apparently viable tumor cells present at the tumor base, showing a Flexner-Wintersteiner rosette differentiation. Immunohistochemical analysis, however, demonstrated that the tumor cells in these areas had lost their proliferative activity. A false-negative result was ruled out by repeating the staining procedure 3 times with positive controls. Clinically, this tumor had a type 3 regression with a cottage cheese appearance in the areas of tumor necrosis and calcification. The area of apparently viable tumor on conventional histological analysis corresponded to a clinical fish-flesh appearance.

In patients 1, 3, and 5 after 2, 3, and 6 courses of chemotherapy, respectively, we found incomplete tumor regression with the presence of a still-viable tumor having maintained its proliferative activity. In patient 1, the result of incomplete tumor regression was anticipated, since it is unlikely that only 2 courses of chemotherapy can induce complete tumor necrosis. Enucleation was, however, encouraged in this early stage of therapy since there was massive tumor reduction. Only about 10% of the apparently viable tumor cells were actually in a proliferative state, allowing the conclusion that in this patient the chemotherapeutic regimen was highly effective. A completely different observation could be made for patients 3 and 5 after 3 and 6 courses of chemotherapy, respectively. Although there was a significant tumor reduction present in both patients, approximately 90% of the apparently viable tumor cells were in a proliferative state, even after 6 courses of chemotherapy in patient 5.

These results indicate the heterogeneity of retinoblastoma cells that results in different sensitivity to the same treatment. One of the major obstacles to the ultimate success of cancer chemotherapy is the ability of malignant cells to develop resistance to cytotoxic drugs. Unfortunately, the phenomenon of multidrug resistance is most commonly observed; this confers on the cells the ability to withstand exposure to lethal doses of many structurally unrelated agents. It has been shown that multidrug resistance is caused by the overexpression of a membrane-associated energy-dependent drug efflux pump, the P-glycoprotein. The P-glycoprotein is encoded by the mdr1 gene, and cells with multidrug resistance often show amplification of this gene. It has already been shown that the effect of P-glycoprotein is reversible with high concentrations of cyclosporine and might even be diminished in the future by using MoAbs against P-glycoprotein.¹⁸

There are no data showing which tumors can be classified as cured after chemotherapy alone as stated by some authors^20,22 or whether there is a definite need for ancillary therapy despite the regression patterns of the intraocular tumors as stated by others.²¹

Similar to the effect of EBR, it has also been observed in primary chemotherapy that it is not equally effective in all patients. Therefore, chemotherapy alone is not a reliable treatment for all retinoblastomas. It has the potential to become a new first-line treatment approach if combined with adjuvant brachytherapy, photocoagulation, or cryocoagulation, hence avoiding the well-known potential of developing secondary cancers after EBR in children with the germinal retinoblastoma gene mutation. The great advantage of chemoreduction in retinoblastoma seems to be the ability to move the tumor margins away from visually vital structures, such as the optic disc and the foveola. The residual tumors can thereafter be destroyed without vision loss by using adjuvant local brachytherapy, photocoagulation, or cryocoagulation.

Accepted for publication March 25, 1998.

Reprints: Nikolaos E. Bechrakis, MD, Department of Ophthalmology, Klinikum Benjamin Franklin, Free University of Berlin, Hindenburgdamm 30, 12200 Berlin, Germany.