Results of a prospective study for the treatment of retinoblastoma




The objectives of this prospective study were to avoid adjuvant treatment for patients with intraocular disease and patients with postlaminar optic nerve invasion (PL-ONI) without full choroidal or scleral invasion. Adjuvant chemotherapy (Regimen 1) was given to patients with scleral invasion, PL-ONI without cut section, and full choroidal and/or scleral invasion. A more intensive regimen of higher dose intravenous chemotherapy (Regimen 2) and local radiotherapy was given to patients with PL-ONI and compromise at the cut end and to patients with overt extraocular disease.


Six-month intravenous chemotherapy included carboplatin plus etoposide alternating with cyclophosphamide plus vincristine (Regimen 1) and the same drugs at higher dosage plus idarubicin (Regimen 2). Chemoreduction with carboplatin and vincristine with or without etoposide was given to selected patients (n = 39 patients).


From 1994 to 2001, 169 patients were evaluable at the Hospital Garrahan (Buenos Aires, Argentina). One hundred eighteen patients with intraocular disease had a 5-year disease free survival (DFS) rate of 0.98, including 54 patients with choroidal invasion. None of 22 patients with isolated PL-ONI developed recurrent disease, whereas 2 of 8 patients with concomitant risk factors had tumor recurrences and died. Three of 5 patients with scleral invasion survived, and 7 of 10 patients with cut-end ONI survived. The only patient with metastatic disease that survived (n = 6) had only lymph node invasion.


Adjuvant therapy can be avoided in patients with intraocular and isolated PL-ONI. Patients with PL-ONI who also had other risk factors required intensive adjuvant therapy, such as patients with cut-end and overt extraocular disease. Metastatic disease was not found to be curable with this approach. Cancer 2004;100:834–42. © 2003 American Cancer Society.

Patients with retinoblastoma present more frequently with extraocular extension in developing countries.1 Conversely, in industrialized nations, retinoblastoma usually is diagnosed as a localized disease, and eye salvage often can be achieved by the use of chemoreduction followed by local treatment.

In recent years there has been increasing interest in determining high-risk features for extraocular recurrence after histopathologic examination of enucleated eyes.2–4 These include postlaminar and cut-end optic nerve involvement, choroidal compromise, anterior chamber invasion, and scleral disease.2–4 These features are relatively infrequent in developed countries; therefore, most reports include patients who were treated over long periods and analyses through retrospective reviews of patients who were treated with different philosophies.2–4 Many authors recommend adjuvant therapy for all patients with these putative risk factors.2, 3 However, the approach throughout our studies has been to try and identify those patients who have such a significant risk for extraocular disease that adjuvant treatment is justified, thereby avoiding the exposure of many patients to chemotherapy.1, 5 It also is not known whether adjuvant therapy actually influences the development of extraocular recurrence. The best regimen for the adjuvant treatment of these patients remains to be established. In addition, even though extraocular recurrence traditionally has been considered a catastrophic event, many patients who develop recurrences outside the central nervous system (CNS) can be rescued now with intensive therapy.6, 7 Therefore, the strategy of avoiding adjuvant chemotherapy for patients with low recurrence rates and treating aggressively those patients who develop recurrent disease may be a reasonable alternative.

The treatment of overt, extraocular retinoblastoma in developing countries also is a challenge, because results with conventional therapy are not satisfactory, and the identification of new drugs to increase the survival rate is needed.1, 8, 9 In 1987, a prospective protocol for the treatment of retinoblastoma, including 101 patients at our center, was published by our group.1 That first study from our center concluded that stage-directed therapy (based on the Grabowski and Abramson clinicopathologic classification system)10 was successful, and adjuvant chemotherapy was not needed for patients with intraocular disease, including those with choroidal invasion. Other putative risk factors that were reported in that study and in an earlier cohort of the current study, such as prelaminar optic nerve invasion, anterior chamber involvement, short optic nerve stump, or large intraocular tumors, lacked any prognostic significance.5 In the earlier study, patients with postlaminar optic nerve invasion with or without cut-end invasion had an encouraging event free survival with multimodality treatment, but patients who had distant metastatic disease fared poorly.1

The subsequent protocol was started in 1994 at our institution with the following objectives: 1) to continue avoiding adjuvant treatment for patients with intraocular disease (Stage I); 2) to withdraw chemotherapy for patients with postlaminar optic nerve invasion without involvement at the line of section (Stage IIB1) if other risk factors, such as full choroidal and/or scleral invasion, were absent; 3) to use a 6-month combination chemotherapy regimen without intrathecal chemotherapy for patients with microscopic scleral invasion (Stage IIA1), postlaminar optic nerve invasion without involvement at the line of section, and full choroidal and/or scleral invasion (Stage IIB1); 4) to use a more intensive regimen of orbital radiotherapy, higher dose intravenous chemotherapy in patients with postlaminar optic nerve invasion and compromise at the line of section, and full choroidal and/or scleral invasion (Stage IIB2); and 5) to test new drugs, such as idarubicin, in a Phase II fashion for poor prognosis patients (Stages IIA2, III, and IV) with overt extraocular disease.

During the course of this protocol, a carboplatin-based chemoreduction regimen was introduced, and eligible patients have been offered this treatment since 1995. The results of this modality in terms of eye preservation will be reported separately.


This was a prospective, nonrandomized study performed at the Hospital J. P. Garrahan in Buenos Aires, Argentina, between January 1994 and June 2001. Informed consent was obtained from parents or guardians. The Grabowski–Abramson clinicopathologic classification system was used for staging10 (Table 1). All enucleated eyes underwent standard examination of the globe, and all slides were evaluated prospectively by the same pathologist (M.T.G.D.) for stage assignment. In patients with bilateral enucleation, the eye with higher stage was considered for staging. The extent of choroidal invasion was categorized as full when tumor cells invaded the choroid > 50% of its thickness or when > 1 tumor cluster was evident. Partial invasion included all other degrees of invasion. At diagnosis, all patients underwent full ophthalmologic examinations under anesthesia, and computed tomography scans and/or magnetic resonance images were obtained of the head, including the orbit. Lumbar puncture, cell counts, and examinations of the cytocentrifugate (in 50% alcohol), bone scintigraphy (only symptomatic patients), bilateral bone marrow aspiration, and biopsies were performed in patients with disease ≥ Stage II.

Table 1. Grabowski–Abramson Clinicopathologic Classificationa
  • CFS: cerebrospinal fluid; CNS: central nervous system.

  • a

    See Grabowski and Abramson.10

I. Intraocular disease
 a. Retinal tumors
 b. Extension into to the lamina cribrosa
 c. Uveal extension
II. Orbital disease
 a. Orbital tumor
  1. Scattered episcleral cells
  2. Orbital invasion
 b. Optic nerve
  1. Tumor beyond the lamina but not up to cut section
  2. Tumor at the cut section of the optic nerve
III. Intracranial metastasis
 a. Positive CSF only
 b. Mass CNS lesion
IV. Hematogenous metastasis
 a. Positive bone marrow alone
 b. Focal bone lesions with or without positive bone marrow
 c. Other organ involvement

Treatment Outline

The ophthalmology group made the decision regarding enucleation or eye preservation. From July 1995 onward, all patients with bilateral disease and selected patients with unilateral disease in whom eye preservation with useful vision was considered were offered chemoreduction. The results of eye preservation are not reported here, but patients are analyzed separately for extraocular recurrence and survival. The need for adjuvant chemotherapy was dictated by the histopathologic staging of enucleated eyes. If overt extraocular disease was present, then neoadjuvant chemotherapy was given; and, when tumor reduction was achieved after two or three cycles of chemotherapy, all patients underwent enucleation or resection of residual mass followed by adjuvant chemotherapy. Once chemotherapy was completed, patients underwent radiotherapy to the affected sites. From 1995–1998, patients with overt extraocular extension were eligible for an up-front Phase II study testing idarubicin. One additional patient with CNS disease (Stage IIIB) received 2 cycles of topotecan at a dose of 2 mg/m2 per day on Days 1–5 as part of neoadjuvant chemotherapy.

Treatment was tailored according to clinicopathologic stage: Patients with Stage I disease did not receive any adjuvant therapy. Patients with Stage IIA1, IIA2, and IIB1 disease (with full choroidal involvement or any degree of scleral involvement) received Regimen 1. Patients with Stage IIB2, III, and IV disease received Regimen 2. Patients with Stage IIB1 with no full choroid or scleral invasion did not receive any adjuvant therapy. Descriptions of the chemotherapy regimens are shown in Table 2.

Table 2. Description of Chemotherapy Regimens
  1. Mesna: sodium mercaptoethanesulfonate; G-CSF: granulocyte–colony stimulating factor; i.v.: intravenous.

Regimen 1
 Weeks 0, 6, 12, and 18
  Cyclophosphamide, 40 mg/kg (Day 1)
  Mesna, 40 mg/kg (Day 1)
  Vincristine, 0.05 mg/kg (Day 1)
 Weeks 3, 9, 15, and 21
  Carboplatin (Day 1)
   18.7 mg/kg for patients weighing < 10 kg
   560 mg/m2 for patients weighing > 10 kg
  Etoposide (Days 1 and 2)
   3.3 mg/kg for patients weighing < 10 kg
   100 mg/m2 for patients weighing > 10 kg
 No intrathecal chemotherapy
Regimen 2
 Weeks 0, 6, 12, and 18
  Cyclophosphamide, 65 mg/kg (Day 1)
  Mesna, 60 mg/kg (Day 1)
  Vincristine, 0.05 mg/kg (Day 1)
  Idarubicin, 10 mg/m2 as a 1-hour intravenous infusion (Day 1)
 Weeks 3, 9, 15, and 21
  Carboplatin (Days 1 and 2)
   18.7 mg/kg for patients weighing < 10 kg
   560 mg/m2 for patients weighing > 10 kg
  Etoposide (Days 1–3)
   3.3 mg/kg for patients weighing < 10 kg
   100 mg/m2 for patients weighing > 10 kg
 G-CSF, 5 μg/kg, was given for approximately 7 days between cycles
 No intrathecal chemotherapy
Window therapy for patients with overt extraocular disease
 Weeks 0 and 3
  Idarubicin, 10–15 mg/m2 per day (Days 1 and 2) by 1-hour i.v. infusion
Regimens for chemoreduction (given every 3 weeks for up to 6 cycles, depending on response)
 Lower risk group (patients with bilateral disease in Groups I–III and all patients with unilateral disease)
  Carboplatin (Day 1)
   18.7 mg/kg for patients weighing < 10 kg
   560 mg/m2 for patients weighing > 10 kg
  Vincristine 0.05 mg/kg (Day 1)
 Higher risk group (patients with bilateral disease in Groups IV and V)
  Carboplatin (Day 1)
   18.7 mg/kg for patients weighing < 10 kg
   560 mg/m2 for patients weighing > 10 kg
  Vincristine 0.05 mg/kg (Day 1)
  Etoposide (Days 1 and 2)
   3.3 mg/kg for patients weighing < 10 kg
   100 mg/m2 for patients weighing > 10 kg

Radiotherapy included orbital radiotherapy at a dose of 45 grays (Gy) for patients with Stage IIA2 disease; orbital radiotherapy at a dose of 45 Gy up to the optic chiasm for patients with Stage IIB2 disease; and craniospinal radiotherapy, which was comprised of 24 Gy to the cranium and 18 Gy to the spine, for patients with Stage IIIA disease. Patients with Stage IIIB disease also received a boost of 10 Gy to the mass and no spinal radiotherapy. Patients with preauricular adenopathy received radiotherapy to the involved lymph nodes (45 Gy). Patients age < 1 year were scheduled to receive no CNS or spinal radiotherapy.

Statistical Analysis

The probability of overall survival (pOS) and the probability of disease-free survival (pDFS) were calculated according to the Kaplan–Meier method. Survival status was updated to December 2002.


Overall, 188 consecutive patients with retinoblastoma were seen at our center during the study period. Nineteen patients were not evaluable for this study: Twelve of those patients came for a second opinion after receiving treatment elsewhere and thus were not included, 4 patients were lost to follow-up soon after diagnosis, and in 1 patient with bilateral disease, the parents decided to continue treatment elsewhere, but no information regarding outcome was available. Two patients with unilateral disease (Stages IIB1 and IIB2, respectively) interrupted adjuvant treatment for > 3 months shortly after enucleation and returned to home. This was considered to be a major protocol violation, and those two patients were excluded from the analysis. Thus, 169 patients were evaluable. The median age was 24.2 months for patients with unilateral disease and 11.4 months for patients with bilateral disease. Eighty-five patients were male and 84 patients were female, and there was a family history of retinoblastoma in 9 patients (5.3%). The median follow-up was 47 months (range, 18–86 months). Seven patients underwent bilateral enucleation. Disease stage for each eye and treatment information are provided in Table 3. Overall survival (± standard error [SE]) for the whole population was 0.91 ±, 0.02). The pOS according to stage is shown in Figure 1.

Table 3. Treatment and Staging of Patients who Underwent Bilateral Enucleationa
PatientStagePreenucleation therapyAdjuvant therapyCurrent status
Right eyeLeft eyeRight eyeLeft eye
  • EBRT: external beam radiation therapy; NA: not available.

  • a

    Stage for the right eye in Patient 107 was not evaluable because of phthisis bulbi.

  • b

    This patient received both adjuvant and neoadjuvant chemotherapy.

  • c

    This patient had low-risk disease.

  • d

    For a description of Regimen 2, see Table 2.

43ICIIA1Chemoreduction (6 cycles), local therapy, EBRTChemoreduction (6 cycles), local therapy, EBRTYesAlive and disease free
92ICIIA1NoneChemoreduction (4 cycles), EBRTYesDied of extraocular recurrence
93ICICNoneChemoreduction (4 cycles), EBRT, topotecan, plaque radiotherapyNoAlive and disease free
107NAIIA2NoneNoneYesbAlive and disease free
172ICIANoneNoneNoAlive and disease-free
173IBIIB1cNoneNoneNoAlive and disease free
71ICIIB2Chemotherapy with Regimen 2, window therapy with idarubicin, oral etoposidedChemotherapy with Regimen 2, window therapy with idarubicindChemotherapy with Regimen 2, EBRTdAlive with progressive, secondary leukemia
Figure 1.

The probability of overall survival according to stage of disease in patients with retinoblastoma: Stage I, 0.99 (standard error [SE], 0.01); Stage IIA1, 0.6 (SE, 0.21); Stage IIB1, 0.93 (SE, 0.04); Stage IIB2, 0.7 (SE, 0.14); and Stage III/IV, 0.16 (SE, 0.15). Patients who did not have enucleated eyes and were without overt extraocular disease for whom stage could not be assigned were not analyzed.

Stage IA

Patients with Stage IA disease (retinal extension; n = 36 patients: 23 with unilateral disease and 13 with bilateral disease) received no adjuvant therapy after enucleation. Eleven patients with bilateral disease were treated with chemoreduction for the treatment of the fellow eye after enucleation (3 patients) or as primary therapy, ultimately requiring enucleation (8 patients; median, 4 cycles of carboplatin vincristine with or without etoposide). The 5-year pDFS and OS rate was 1.0 for patients who did and did not receive chemoreduction.

Stage IB

For patients with Stage IB disease (prelaminar optic nerve invasion; n = 13 patients: 12 with unilateral disease and 1 with bilateral disease), no adjuvant therapy after enucleation was given. Two patients received chemoreduction, both before enucleation. The 5-year pDFS was 1.0 for both groups.

Stage IC

Among 54 patients with Stage IC disease (choroidal invasion; 41 patients with unilateral disease and 13 patients with bilateral disease), 13 patients had full choroidal invasion. No adjuvant therapy after enucleation was scheduled. However, 10 patients (7 patients with bilateral disease) received chemoreduction; in 4 patients, it was prescribed after enucleation for chemoreduction of the fellow eye (median, 2 cycles). The 5-year pDFS and OS rates (± SE) were 0.98 ± 0.01 for the whole population. The 5-year DFS rate was 1.00 for patients who received chemoreduction and 0.97 ± 0.02 for patients who received no adjuvant therapy. The only event was an orbital and bone marrow recurrence in a patient with full involvement who had not received chemoreduction. After achieving a second complete remission with treatment including chemotherapy with Regimen 2, the patient developed a second recurrence and died of retinoblastoma.

Nonenucleated Patients

Of 15 patients who did not undergo enucleation (9 patients with unilateral disease and 6 patients with bilateral disease), 12 patients received chemoreduction. One patient with bilateral disease died of metastatic progression after parental refusal for enucleation. The 5-year DFS and OS rates were 0.92 ± 0.06.

Stage IIA1

Among five patients with Stage IIA1 retinoblastoma (scleral invasion and scattered episcleral cells, including three patients with bilateral disease, one patient with sporadic unilateral disease, and one patient unilateral with a family history of unilateral disease, all but one patient received adjuvant therapy with Regimen 1. The remaining patient with sporadic unilateral disease did not receive adjuvant therapy and had an orbital recurrence 5 months after enucleation. She was treated with chemotherapy on Regimen 2, limited surgery, and orbital radiotherapy and has survived event free for 62 months after recurrence. The remaining two events included one patient with CNS hematogenous metastases and orbital and preauricular recurrences and one patient with bone and femoral lymph node metastases. Both patients died of recurrent retinoblastoma. The 5-year pDFS was 0.4 ± 0.21, and the OS rate was 0.6 ± 0.21.

Stage IIB1

Thirty patients with Stage IIB1 retinoblastoma (postlaminar optic nerve invasion, cut end free of tumor) were included in the study and were subdivided further into a low-risk group and a high-risk group.

Low-risk group

The 22 patients in the Stage IIB1 low-risk group (17 patients with unilateral disease and 5 patients with bilateral disease) had no associated scleral invasion (of any degree) or full choroidal invasion. They received no adjuvant therapy after undergoing enucleation. However, four patients with bilateral disease underwent enucleation up-front and received chemoreduction for treatment of the fellow eye. They received a median of 3 cycles of carboplatin, etoposide, and vincristine (range, 2–4 cycles). The 5-year pDFS and pOS rates were 1.00 both for patients who received chemoreduction and for patients who did not.

High-risk group

The eight patients in the Stage IIB1 high-risk group (five patients with unilateral disease and three patients with bilateral disease) had associated scleral invasion (of any degree) or full choroidal invasion. Four of those eight patients had combined scleral disease (intrascleral involvement in one patient and transscleral extension in three patients) and choroidal disease. They received adjuvant chemotherapy with Regimen 1 (except for one patient who received only carboplatin plus etoposide without cyclophosphamide or vincristine for concomitant chemoreduction of the fellow eye). The 5-year pDFS and pOS rates (± SE) were 0.75 ± 0.15. Two patients with combined transscleral and choroidal involvement (unilateral disease) developed recurrences. No patients had received chemoreduction, and both developed CNS recurrences, one patient together with orbital and bone marrow recurrences and the remaining patient together with orbital disease. They both died of progressive retinoblastoma.

Stage IIB2

Ten patients with Stage IIB2 retinoblastoma (postlaminar optic nerve invasion; cut end positive for tumor cells) were included (6 patients with unilateral disease and 4 patients with bilateral disease). None had received previous chemoreduction. Four of these patients presented with overt orbital disease and received preoperative chemotherapy. All patients received Regimen 2 and orbital radiotherapy (45 Gy up to the chiasm). The 5-year pDFS rate (± SE) was 0.7 ± 0.14. Events included two deaths in patients who were in complete remission (one patient died of sepsis during a chemotherapy-induced neutropenic episode, and the other patient died of parental neglect) and one patient without overt extraocular disease at diagnosis who had a leptomeningeal recurrence. One patient (with bilateral disease) had a secondary neoplasm (acute myelogenous leukemia) and is receiving palliative treatment after a leukemic recurrence with no evidence of retinoblastoma.

Stage III and IV

There were six patients with Stage III and IV retinoblastoma, including two patients with Stage III disease (one patient with positive cerebrospinal fluid [CSF] only and one patient with both positive CSF and a mass; both patients had unilateral disease) and four patients with Stage IV disease (preauricular lymph nodes only in one patient; CNS positive, systemic metastasis in two patients; and systemic metastasis in one patient; all three patients had bilateral disease). They all received Regimen 2, and four patients received window therapy (idarubicin in three patients and topotecan in one patient). The only survivor was a patient with preauricular lymph node disease. The remaining five patients died of recurrent retinoblastoma. The pDFS and pOS rate (± SE) for patients with Stage III/IV disease was 0.16 ± 0.15.


Regimen 1 was tolerated well and caused only mild myelosuppression. There was one episode of etoposide allergy (skin rash). Only one patient experienced two episodes of fever and neutropenia (one of them with catheter infection). One patient had an abnormal audiogram, probably associated with carboplatin.

All patients who received Regimen 2 experienced severe hematologic toxicity, with a median of three episodes of fever and neutropenia. There were six documented infections, including one patient with systemic Candida infection. All patients received granulocyte-colony stimulating factor support, and seven patients received platelet or red blood cell transfusions. One patient underwent a 25% reduction in dose because unacceptable hematologic toxicity and systemic candidasis. No significant echocardiographic changes were detected.


The results of the current study provided further evidence in a large cohort that the overall survival of patients with intraocular disease is excellent without any adjuvant treatment, even when patients with some risk factors for extraocular recurrences, such as isolated choroidal invasion, are included. The introduction of chemoreduction complicated the interpretation of one of the major endpoints of this study, which was withdrawing adjuvant chemotherapy for patients with Stage I disease. The effect of chemoreduction on histopathologic features has been addressed in few patients after planned enucleation of the affected eye.11 However, in our patients, histopathologic examination of enucleated eyes was done after failure of both chemoreduction and local treatment, usually including external-beam radiation therapy. Nevertheless, its use before enucleation may downstage the affected eye, hence eliminating potential risk factors. To avoid any possible bias, patients who received chemoreduction at any time were analyzed separately for survival. After excluding those patients (n = 39 patients), only 1 of 98 patients who developed recurrent disease were withdrawn from adjuvant therapy according to the protocol guidelines. These included 44 patients with isolated choroidal invasion and 18 patients with postlaminar optic nerve involvement without concomitant full choroidal or scleral invasion.

The impact of some features, such as choroidal invasion, as risk factor for extraocular recurrence remains controversial. Some authors have suggested that adjuvant chemotherapy should be given to these patients; however, to our knowledge, most reports have been based on retrospective studies over long periods or on multivariate analyses of patients with different combinations of risk factors.2, 3 Some reports suggested that only patients with isolated, massive choroidal invasion are at greater risk for extraocular recurrence and, thus, adjuvant chemotherapy should be prescribed,2, 3 although other studies failed to find any difference in outcome between these subgroups.12 However, the definition of massive invasion is controversial and difficult to reproduce. In a multivariate analysis, Khelfaoui et al. found that massive choroidal invasion (defined as tumor cells that have destroyed the Bruch membrane and/or the presence of ≥ 3 microscopic cell clusters) was a significant risk factor for recurrence in 20 patients studied, including patients with concomitant postlaminar disease. However, only one of four patients who developed an extraocular recurrence had less than postlaminar optic nerve invasion, and that patient was retrieved successfully after treatment for the recurrence.2 Uusitalo et al., in their retrospective review from two different institutions, included 11 patients with massive choroidal involvement, which they defined as > 25% invasion of the choroid with less than postlaminar involvement.4 The patients in that study did not receive any adjuvant therapy, and none developed recurrent disease; thus, those authors recommend no adjuvant therapy for this group. In a retrospective study, Honavar et al. included eight patients who had massive choroidal infiltration as a single risk factor.3 The definition of massive choroidal invasion used was not specified, but no patient in that study developed recurrent disease. Six of those patients received adjuvant therapy, and the authors concluded that adjuvant chemotherapy is indicated in this subgroup.3 Only one patient with full choroidal invasion developed recurrent disease in the current study. Therefore, we believe that chemotherapy is not indicated clearly in this group to reduce the rate of extraocular recurrence, although chemotherapy may play a role in the treatment of patients with full invasion. Unfortunately, the realization of a randomized trial comparing chemotherapy with observation is not feasible given the low recurrence rate.

Retrolaminar invasion of the optic nerve is a well known risk factor for extraocular recurrence. However, adjuvant treatment of patients with postlaminar optic nerve invasion that does not include the line of section is controversial. Based on retrospective studies, many authors suggest that adjuvant chemotherapy should be used in all patients with postlaminar optic nerve invasion to reduce the rate of extraocular recurrence.2, 3 In our previous prospective study, only 1 of 10 patients with this feature died of recurrent disease after adjuvant chemotherapy: This patient had concomitant scleral and choroidal invasion.1 Other investigators also reported that the combination of these features is a significant risk factor for recurrence; however, when only the optic nerve is compromised, the risk of recurrence is not reported to be increased significantly.13 Our objective was to identify two different subgroups with different risks according to the presence of concomitant full choroidal and/or scleral invasion. Therefore, for this protocol, we subclassified the cohort according to the presence of these features, and adjuvant chemotherapy was given only to patients who had the combination of risk factors. Our hypothesis appears to be correct, because no patient with isolated postlaminar invasion developed recurrent disease despite receiving no adjuvant therapy. Conversely, concomitant choroidal or scleral invasion was indeed a significant risk factor for extraocular recurrence, because only patients with these features developed recurrent disease despite adjuvant chemotherapy. However, despite the strengths of the current study, its prospective design, and the relatively large patient numbers, there remain potential causes of bias, such as a relatively small number in each subgroup and the presence of unidentified risk factors. Other authors found that extension > 1 mm beyond the lamina cribrosa also may play a role.4 In the current series, all patients in this subgroup who developed recurrences had CNS involvement at the time of recurrence. Even though the chemotherapy regimen given in this protocol included drugs with proven penetration to the CNS, such as carboplatin and cyclophosphamide, it may have been insufficient for the treatment of these very high-risk patients, and the CNS may have acted as a sanctuary site. Therefore, our current policy is to increase chemotherapy intensity by prescribing higher dose chemotherapy, such as Regimen 2, with the objective of improving CNS coverage in this subset of high-risk patients.

In our previous protocol, 11 of 14 patients with invasion of the cut end of the optic nerve survived after they received adjuvant intravenous and intrathecal therapy together with orbital radiation.1 The chemotherapy used was a moderate-dose regimen that included doxorubicin, cyclophosphamide, and vincristine for 57 weeks. Although the regimen was tolerated well in the short term, the high cumulative doses of cyclophosphamide and doxorubicin prompted us to find newer, less toxic alternatives. Therefore, for the current protocol, we designed a new regimen that included the most reported active agents at the maximum conventional dose. Doz el al. pioneered the combination of carboplatin and etoposide for patients with extraocular retinoblastoma and reported an 85% response rate.14 In an attempt to intensify treatment, we choose to administer a higher those of this combination according to a high-dose carboplatin plus etoposide regimen, which is active in patients who have other malignancies with a drug-sensitivity profile similar to that of retinoblastoma.15 Cyclophosphamide and vincristine had been used in our previous protocol, but the former was administered at a higher dose in this study. Idarubicin replaced doxorubicin in this study, based in our own Phase II results, because its high response rate and its probably lower cardiotoxicity make it the anthracycline of choice for patients with this disease.16 However, although it is appealing theoretically, the introduction of this regimen for patients with cut-end invasion of the optic nerve did not improve results obviously compared with our previous protocol. However, the small cohort of patients in the subgroup included in the current study was not fully comparable with the previous cohort, because more children with overt orbital disease and most likely higher risk were included in the current series. Only one patient developed recurrent disease, and another patient died because of toxicity. One patient in this group had a secondary leukemia that may have been treatment-related, but he had also received other potential leukemogenic agents, such as oral etoposide. Because the hematologic toxicity of this regimen is high, it should be used with caution in centers that have limited resources. Nevertheless, the regimen had a substantially lower dose of anthracyclines compared with our previous regimen and may cause less treatment-related late effects.

The fact that most recurrences involved the CNS raises the question of whether intrathecal therapy is needed. When this study was designed, the only drugs that were available for intrathecal use were those used in our previous protocol, cytarabine and methotrexate, which are relatively inactive against retinoblastoma cell lines.17 Topotecan and thiotepa, both of which have possible activity against retinoblastoma, have been used successfully intrathecally and may have a role in the treatment of patients with retinoblastoma.18, 19

Thus, the decision to prescribe adjuvant therapy for patients with risk factors for recurrence is complex. Low-intensity regimens, such as our Regimen 1 or the combination of carboplatin, etoposide, and vincristine, which is proposed by many authors, may be insufficient for the prevention of recurrences in high-risk patients, especially as far as adequate CNS coverage is concerned.3, 20 Conversely, more intensive combinations, such as our Regimen 2, may be more effective, although treatment-related toxicities may outweigh the benefits of a lower recurrence rate. According to our limited data, we recommend adjuvant therapy with an intensive regimen only for patients with a significant risk of extraocular recurrence (i.e., cut-end invasion, scleral disease, and postlaminar optic nerve invasion [cut end free of tumor] with concomitant, full choroidal and/or scleral involvement). In these patients, the recurrence rate outweighs the risk of treatment-related mortality. For patients with isolated choroidal invasion and isolated postlaminar optic nerve invasion (cut end free of tumor), we estimate that the risk of recurrence is < 5% when they are given no adjuvant therapy, and adjuvant chemotherapy can be withheld. However, we acknowledge that few patients are included in each subgroup, and these results should be confirmed in a larger patient population through international collaborative studies or meta-analyses of different series, which are difficult now, because there is no widely used classification system to make data comparable.

Regimen 2 was proven to be effective for patients with orbital recurrences,8 but it was not effective for patients with distant or CNS metastatic disease, because all but one patient with only preauricular lymph node metastasis died of disease. Antonelli et al. recently reported that the addition of ifosfamide and etoposide did not result in improved survival for patients with extraocular retinoblastoma in Brazil.9 These results suggest that, even with moderately intensive regimens like that used in the current study, patients with distant metastatic disease fare poorly and may benefit from even more intensive treatments with stem cell support.6, 7