Shields CL, Honavar SG, Shields JA, Demirci H, Meadows AT, Naduvilath TJ. Factors Predictive of Recurrence of Retinal Tumors, Vitreous Seeds, and Subretinal Seeds Following Chemoreduction for Retinoblastoma. Arch Ophthalmol. 2002;120(4):460-464. doi:10.1001/archopht.120.4.460
To identify the clinical features of eyes with retinoblastomas that predict the recurrence of retinal tumors, vitreous seeds, and subretinal seeds following treatment with chemoreduction.
Prospective nonrandomized single-center clinical trial.
Ocular oncology service at Wills Eye Hospital of Thomas Jefferson University(Philadelphia, Pa) in conjunction with the division of oncology at Children's Hospital of Philadelphia.
There were 158 eyes with 364 tumors in 103 consecutive patients with retinoblastoma managed with chemoreduction between June 1994 and August 1999.
All patients received treatment for retinoblastoma with 6 cycles of chemoreduction using vincristine, etoposide, and carboplatin combined with focal treatment (cryotherapy, thermotherapy, or plaque radiotherapy) for each retinal tumor.
Main Outcome Measures
The 3 main outcome measures included recurrence of retinal tumors, recurrence of vitreous seeds, and recurrence of subretinal seeds. The clinical features at the initial examination were analyzed for their association with the main outcome measures using a series of Cox proportional hazards regressions.
All retinal tumors, vitreous seeds, and subretinal seeds showed an initial favorable response of regression during this treatment regimen. Using Kaplan-Meier estimates, at least 1 retinal tumor recurrence per eye was found in 37% of eyes at 1 year, 51% at 3 years, and no further increase at 5 years. By multivariate analysis, the only factor predictive of retinal tumor recurrence was the presence of tumor-associated subretinal seeds at the initial examination. Of the 54 eyes that had vitreous seeds at the initial examination, vitreous seed recurrence was found in 26% of eyes at 1 year, 46% at 3 years, and 50% at 5 years. By univariate analysis, the only factor predictive of vitreous seed recurrence was the presence of tumor-associated subretinal seeds at the initial examination. Of the 71 eyes that had subretinal seeds at the initial examination, subretinal seed recurrence was detected in 53% of eyes at 1 year, 62% at 3 years, and no further increase at 5 years. By multivariate analysis, factors predictive of subretinal seed recurrence included a tumor base greater than 15 mm and a patient age of 12 months or younger at diagnosis. There were no patients who developed retinoblastoma metastasis, pinealoblastoma, or second malignant neoplasms.
Chemoreduction combined with focal therapy is effective for selected eyes with retinoblastomas. Eyes with subretinal seeds at initial examination are at particular risk for recurrence of retinal tumor and vitreous seeds. Younger patients with large tumors are at risk for recurrence of subretinal seeds. Retinal tumor and subretinal seed recurrence seems to manifest within 3 years of follow-up. Close follow-up of all patients treated with chemoreduction is warranted.
CHEMOREDUCTION has become an important method in the management of retinoblastoma.1- 13 This technique has been employed in an effort to avoid enucleation and external beam radiotherapy for children with retinoblastoma, especially those with bilateral disease. Chemoreduction involves the use of intravenous chemotherapy coupled with focal treatments to induce complete regression of the tumor. Management of this disease involves attention to 3 anatomic sites of tumor, including the individual retinal tumor(s), associated vitreous tumor (termed"vitreous seeds"), and associated subretinal tumor (termed "subretinal seeds"). Retinal tumors generally respond rapidly to chemoreduction; then, subsequent consolidation with focal methods, such as cryotherapy, thermotherapy, or plaque radiotherapy, is employed. Eyes with additional vitreous seeds or subretinal seeds are managed differently using chemoreduction without focal consolidation treatments because the number of seeds is usually far beyond the capability of focal treatment methods and the multitude of tiny seeds typically respond with regression, calcification, and often complete disappearance after several months of treatment.
Prior statistical studies regarding chemoreduction for retinoblastoma have focused on the factors related to treatment failure and the need for external beam radiotherapy or enucleation.13 This information is important for the clinician in determining whether chemoreduction would be useful for a child with retinoblastoma. In this analysis, we specifically address the 3 anatomic sites of retinoblastoma, including individual retinal tumors, vitreous seeds, and subretinal seeds. We evaluated the effect of chemoreduction in controlling tumors at each of these 3 sites. This information may be useful for the retinoblastoma specialist in anticipating recurrence and providing a window of time to expect each event.
All new patients with retinoblastoma who were treated initially with chemoreduction (Table 1) at the ocular oncology service, Wills Eye Hospital, Thomas Jefferson University (Philadelphia, Pa) in conjunction with the division of oncology at Children's Hospital of Philadelphia were enrolled for this prospective clinical trial. The eligibility criteria for inclusion have been detailed in previous reports and include children with intraocular retinoblastomas in whom the involved eye of the unilaterally affected patient or either eye of the bilaterally affected patient would ordinarily require enucleation or external beam radiotherapy to cure the disease, based on published indications.1- 3,9 Any patient whose tumor(s) could be properly controlled with conservative methods alone (cryotherapy, laser photocoagulation, thermotherapy, plaque radiotherapy) were not eligible for inclusion in the chemotherapy protocol unless the opposite eye had features that required chemotherapy.1,2 Exclusion criteria for treatment with chemoreduction included biomicroscopic evidence of iris neovascularization; neovascular glaucoma; tumor invasion into the anterior chamber, iris, optic nerve, choroid, or extraocular tissues as documented by clinical, ultrasonographic, and neuroimaging modalities; or any eye in which globe prognosis was judged poor. Systemic exclusion criteria were evidence of systemic metastasis, prior chemotherapy, prior treatment for retinoblastoma, or inadequate renal, hepatic, or auditory function. The potential risks and benefits of the chemoreduction protocols were discussed with the patients' parents or guardians and informed consent was signed. The chemotherapeutic agents employed in the protocols included intravenous vincristine, etoposide, and carboplatin as presented in Table 1. The duration of treatment was 6 cycles per protocol (Children's Hospital of Philadelphia 582). The protocol was approved by the Children's Hospital of Philadelphia Committee for the Protection of Human Subjects.
All data were collected in a prospective fashion. Each retinoblastoma was measured for greatest basal dimension (in millimeters) using the indirect ophthalmoscope, and thickness (in millimeters) measured by A-scan and B-scan ultrasonography and indirect ophthalmoscopy. The proximity of the nearest tumor margin to the optic disc and foveola was recorded. The presence and extent of subretinal fluid and the presence and extent of tumor seeding in the vitreous and subretinal space were recorded.
Ocular oncologic follow-up was provided, with examination under anesthesia on a monthly basis after initiation of chemoreduction until control of the disease was achieved. Thereafter, examinations were provided every 2 to 4 months as needed. At each examination, the individual retinal tumors were measured in base and thickness. The status of the vitreous seeds, subretinal seeds, and subretinal fluid was noted. When maximum retinal tumor regression was achieved, usually at cycle 2 or 3, focal treatment using thermotherapy or cryotherapy coupled with chemoreduction9,11 was applied to all retinal tumors. Recurrent retinal tumors, vitreous seeds, and subretinal seeds were identified on follow-up and treated with methods including cryotherapy, thermotherapy, laser photocoagulation, and plaque radiotherapy to avoid external beam radiotherapy and enucleation. The focal treatment selected for each recurrent tumor or seed depended on several factors, including the tumor or seed location, status of the opposite eye, visual acuity in the involved eye and opposite eye, and, most importantly, tumor or seed size.
The clinical data were analyzed with regard to 3 main outcome measures: recurrence of retinal tumors, recurrence of vitreous seeds, and recurrence of subretinal seeds. The effect of each clinical variable recorded at the time of initial examination at the ocular oncology service on the development of these outcomes was analyzed by a series of univariate Cox proportional hazards regressions. The correlation among the variables was determined by Pearson correlations. All variables were analyzed as discrete variables. Variables that were continuous were analyzed by grouping them into discrete categories. The variables that were significant on a univariate level (P≤.05, Fisher exact and χ2 tests) were entered into the multivariate Cox regression analysis using the enter method. For variables that showed a high degree of correlation, only 1 from the set of associated variables was entered at a time into subsequent multivariate models. A final multivariate model fitted variables that were identified as significant predictors(P≤.05, Wald statistic or 95% confidence interval of the relative risk) from the initial model as well as variables deemed clinically important for the outcomes of retinal tumor recurrence, vitreous seed recurrence, and subretinal seed recurrence. In the final model, a predictor was considered to be a significant risk factor if the 95% confidence interval of its relative risk did not contain a risk value of 1.0.
There were 364 retinoblastomas in 158 eyes of 103 patients enrolled in the chemoreduction protocol. The patients received chemoreduction, as presented in Table 1, for a mean of 6 cycles(median, 6 cycles; range, 2-6 cycles). The reasons for employing fewer than 6 cycles included poor tumor response in 19 cases, patient or family compliance difficulties in 7 cases, and family preference owing to excellent control in 3 cases. The mean follow-up was 29 months (median, 28 months; range, 2-63 months). Some of the patients in this study were included in previously published studies.8,9,13 The demographic findings of this group and chemoreduction protocol have been outlined in a previous publication.13 The Reese-Ellsworth stage of each eye included group I in 9 (6%), group II in 26 (16%), group III in 16 (10%), group IV in 32 (20%), and group V in 75 (48%) eyes. The following results of the initial visit describe the 3 anatomic sites of tumors, including retinal tumor, vitreous seeds, and subretinal seeds. Overall, there was a mean of 2 retinal tumors per eye (median, 2 tumors; range, 1-10 tumors). Of the largest tumor in each eye, the mean retinal tumor base was 13 mm (median, 14 mm; range, 1-24 mm); mean tumor thickness was 7 mm (median, 7 mm; range, 1-20 mm); mean tumor distance to the optic disc was 2 mm (median, 0 mm; range, 0-20 mm); and mean tumor distance to the foveola was 2 mm (median, 0 mm; range, 0-22 mm). At the initial visit, retinoblastoma was visible in the vitreous as vitreous seeds in 54 eyes (34%) and retinoblastoma was visible in the subretinal space as subretinal seeds in 71 eyes (45%). Vitreous seeding involved a mean of 5 clock hours (median, 3 clock hours; range, 1-12 clock hours). The vitreous seeds were local in 32 eyes (60%) and diffuse in 22 eyes (40%). Subretinal seeding involved a mean of 5 clock hours (median, 5 clock hours; range, 1-12 clock hours). The subretinal seeds were mainly in the posterior pole in 12 eyes (17%), at the equator in 21 (30%), at the ora serrata in 30 (42%), and diffusely located in 8 (11%). Subretinal seeds were primarily located inferiorly in 51 eyes (72%), temporally in 1 (1%), superiorly in 2 (3%), nasally in 3(4%), in the macula in 6 (8%), and diffusely in 8 (12%). Subretinal fluid was present in 95 eyes (60%), involved a mean of 60% of the retina (median, 70%), and extended for a mean of 9 clock hours (median, 11 clock hours; range, 1-12 clock hours).
Retinal tumor recurrence of at least 1 tumor was found in 69 (44%) of 158 eyes. Using Kaplan-Meier estimates, at least 1 retinal tumor recurrence was found in 37% of eyes at 1 year, 51% at 3 years, and 51% at 5 years (Table 2). Univariate and multivariate risk factors for retinal tumor recurrence are presented in Table 3. The mean time interval from chemoreduction initiation to first retinal tumor recurrence (per eye) was 10 months (median, 9 months; range, 2-27 months) and mean time interval to last retinal tumor recurrence(per eye) was 13 months (median, 11 months; range, 4-48 months). Focal treatment of the recurrent retinal tumor included cryotherapy for 33 tumors, thermotherapy for 31 tumors, and plaque radiotherapy for 22 tumors. External beam radiotherapy was required in 14 eyes and enucleation in 19 eyes. Some tumors received more than 1 treatment method and some eyes received more than 1 treatment method.
Of the 54 eyes with vitreous seeds on initial examination, recurrence of at least 1 vitreous seed was noted in 21 (39%). Using Kaplan-Meier estimates, vitreous seed recurrence was found in 26% of eyes at 1 year, 46% at 3 years, and 50% at 5 years (Table 2). Univariate and multivariate risk factors for vitreous seed recurrence are presented in Table 4. The mean interval from chemoreduction initiation to first vitreous seed recurrence(per eye) was 14 months (median, 12 months; range, 3-37 months) and mean interval to last vitreous seed recurrence (per eye) was 21 months (median, 18 months; range, 6-50 months). The mean number of recurrent vitreous seed sites was 3 (median, 3; range, 1-8 sites). Recurrent vitreous seeds involved a mean of 6 clock hours of the fundus (median, 6 clock hours; range, 1-12 clock hours). Treatment of vitreous seed recurrence included plaque radiotherapy for 3 eyes, external beam radiotherapy for 11 eyes, and enucleation for 16 eyes. Some eyes received more than 1 treatment method.
Of the 71 eyes with subretinal seeds on initial examination, at least 1 subretinal seed recurrence was detected in 40 (56%). Using Kaplan-Meier estimates, subretinal seed recurrence was found in 53% of eyes at 1 year, 62% at 3 years, and 62% at 5 years (Table 2). Univariate and multivariate risk factors for subretinal seed recurrence are presented in Table 5. The mean interval from chemoreduction initiation to first subretinal seed recurrence (per eye) was 8 months (median, 6 months; range, 1-33 months) and time interval to last subretinal seed recurrence (per eye) was 16 months (median, 15 months; range, 5-50 months). The mean number of recurrent subretinal seed sites was 4 (median, 3; range, 1-13 sites). Recurrent subretinal seeds extended for a mean of 6 clock hours (median, 6 clock hours; range, 1-12 clock hours). The recurrent seeds were found in sites of previous serous retinal detachment in 38 eyes (95%). The recurrent subretinal seeds were located anteroposteriorly mainly at the ora serrata in 20 eyes (51%), equator in 14 (35%), macula in 3 (7%), and diffuse at all regions in 3 (7%). The main quadrant location included inferior quadrant in 28 eyes (70%), temporal in 0 (0%), superior in 1 (2%), nasal in 5 (12%), macula in 3 (7%), and diffuse in 3 (8%). Treatment of the subretinal seed recurrences included cryotherapy in 45 eyes, thermotherapy in 19, plaque radiotherapy in 22, external beam radiotherapy in 13, and enucleation in 23. Most eyes received more than 1 treatment method for multifocal recurrences.
No patient developed metastatic retinoblastoma or serious adverse effects of the chemotherapy, such as renal adverse effects, hearing loss, or second cancers. Pinealoblastoma and intracranial neuroblastic malignancy did not occur in any case.
Chemoreduction has assumed a major role in the management of retinoblastoma.13 Preliminary observations have documented dramatic regression of retinoblastoma following chemoreduction and these findings have generated enthusiasm for this treatment modality. Approximately 50% of affected eyes treated with chemoreduction are successfully preserved with avoidance of external beam radiotherapy or enucleation.13 The rate of globe preservation is best with less advanced eyes (85%), such as those in Reese-Ellsworth groups I to IV whereas with more advanced eyes, such as those in Reese Ellsworth group V, preservation is less successful at 47%.13 To reach this goal, the involved eye(s) requires tedious monitoring and possibly extensive treatment using focal measures, such as cryotherapy, thermotherapy, or plaque radiotherapy, to avoid ultimate failure.
The main concern with chemoreduction is the lack of stable tumor control following treatment. Recurrence of retinoblastoma is generally classified according to 1 of 3 anatomic sites, including tumors in the retina, vitreous, or subretinal space. This anatomic classification of tumor is important since recurrence varies in appearance, course, and management depending on its site. Additionally, eyes can have recurrence at more than 1 site. In this analysis, we found recurrence of at least 1 retinal tumor in 51% of eyes at 3 years' follow-up and no change in this percentage at 5 years' follow-up (Table 2). Those at greatest risk for retinal tumor recurrence were eyes with tumor-associated subretinal seeds surrounding the base of the tumor (Table 3). It is unclear how this factor might lead to intraretinal recurrence, but it might reflect the discohesive nature of the tumor and perhaps inadequate accessibility of the less vascular portions of the retinal mass to chemotherapy. In some instances it is difficult to differentiate between recurrent intraretinal tumor and recurrent subretinal seeds immediately adjacent to the tumor. With this in mind, perhaps some of the tumors that we judged to be recurrent were actually recurrent subretinal seeds at the retinal tumor site.
At 5 years' follow-up, vitreous seed recurrence was found in 50% of eyes and subretinal seed recurrence in 62% of eyes (Table 2). Patients at greatest risk for vitreous or subretinal seed recurrence were those who, at initial examination, were younger, had large tumor dimensions, and had tumor-associated subretinal seeds (Table 4 and Table 5). It is evident that eyes with subretinal seeds at initial examination are particularly at risk for future recurrence of retinal tumor and vitreous or subretinal seeds and should be carefully followed up. It is unclear how the presence of initial subretinal seeds is a risk factor for the eventual development of recurrent vitreous seeds. We doubt that subretinal seeds directly transgress the retina into the vitreous. On the other hand, we suspect that subretinal seeds might signify a more discohesive tumor, one that also might have macroscopic or microscopic vitreous seeds at initial examination and are at risk for recurrence. Vitreous and subretinal seed recurrence can ultimately lead to advanced tumor and loss of the eye. Thus, vitreous and subretinal seeds should be carefully documented and followed up closely. If recurrence is detected, prompt treatment is advised. Tedious fundus examination, with knowledge of the appearance of seed recurrence, is critical since seeds are often only 50 to 100 µm at the time recurrence is detected. All children receiving a chemoreduction protocol should be monitored by a retinoblastoma specialist who is able to detect minute recurrences and capable of treating the recurrences.
Recurrence is especially notable after chemoreduction is discontinued. In this study, the mean interval from discontinuation of chemoreduction to first recurrence of retinal tumor was 4 months, recurrence of vitreous seeds was 2 months, and recurrence of subretinal seeds was 2 months. Thus, monitoring of the eye is important during chemoreduction for applying focal treatments but is especially critical following chemoreduction to detect recurrence. It is reassuring to know that most children manifest their recurrent retinal tumors and subretinal seeds by 3 years after treatment with little recurrence thereafter; accordingly, follow-up can be adjusted for this time interval. Vitreous seed recurrence, however, continues to be a problem up to 5 years after treatment and potentially longer; therefore, patients with vitreous seeds at initial examination might require cautious ocular examination for many years following treatment.
Recurrence of vitreous seeds or subretinal seeds may not necessarily reflect resistance of the tumor seeds to chemotherapy. This may, on the other hand, represent inadequate penetration of the chemotherapy to relatively avascular sites in the vitreous cavity and subretinal space. Wilson et al14 demonstrated in rabbits that the application of cryotherapy to the sclera prior to intravenous delivery of carboplatin resulted in increased penetration of carboplatin into the vitreous. It is believed that any modality that disturbs the blood-retinal barrier can bring about a similar phenomenon. Thus, thermotherapy may also increase the vitreous penetration of carboplatin. Based on this, we generally apply cryotherapy at the time of institution of chemoreduction in all eyes, and at each subsequent cycle, cryotherapy or thermotherapy is applied to the tumors for consolidation, secondarily permitting exudation of the carboplatin into the vitreous cavity. Penetration of carboplatin in the subretinal space is more difficult to measure and maintain.15
There are limitations in this report. First, our group represents patients from a tertiary referral center and we may treat more advanced cases in a heroic attempt to save a remaining eye. Second, we acknowledge that some of the eyes classified as retinal tumor recurrences may have been subretinal seed recurrences adjacent to the tumors, but the clinical differentiation is challenging. Third, we employed a set protocol of chemoreduction for a set duration. There may be other chemoreduction protocols that might better control retinoblastomas. Fourth, with longer follow-up, we might detect more recurrence. However, in this comprehensive statistical analysis, important clinical information is evident to guide the ocular oncologist and pediatric oncologist in the management of affected patients and to assist in anticipating recurrence. Based on this report, we are presently treating advanced disease with more aggressive regimens for longer durations.
In summary, chemoreduction and focal conservative treatment for retinoblastoma is effective, but recurrence of at least 1 retinal tumor per eye, vitreous seeds, and subretinal seeds is found in 51%, 50%, and 62% of eyes, respectively, at 5 years' follow-up. Eyes with vitreous or subretinal seeds at initial examination can have hundreds or thousands of seeds initially. The recurrence of 1 or 2 seeds or even a group of seeds does not indicate failure of this modality since it represents only a minor degree of seeding that was manifested at the initial visit. However, the need for meticulous fundus examination of these children during and after chemoreduction is underscored. Numerous focal treatments may be necessary to adequately control the recurrences and ultimately preserve the globe and, most importantly, the patient's life.
Submitted for publication September 4, 2001; final revision received September 25, 2001; accepted December 13, 2001.
This study was supported by the Paul Kayser International Award of Merit in Retina Research, Houston, Tex (Dr J. A. Shields); the Macula Foundation, New York, NY (Dr C. L. Shields); the Eye Tumor Research Foundation, Philadelphia(Dr C. L. Shields); Orbis International, New York (Dr Honavar); and the Hyderabad Eye Research Foundation, Hyderabad, India (Dr Honavar).
Corresponding author and reprints: Carol L. Shields, MD, Ocular Oncology Service, Wills Eye Hospital, 900 Walnut St, Philadelphia, PA 19107.