Practical Approach to Management of Retinoblastoma

Objective  To present a simplified approach to management of retinoblastoma usingbasic clinical features.

Design  In a prospective, nonrandomized, single-center clinical trial, 158 eyesof 103 patients with retinoblastoma were managed with 6 cycles of chemoreduction(vincristine sulfate, etoposide, and carboplatin). The eyes were classifiedaccording to the Reese-Ellsworth classification and were also grouped on thebasis of clinical features as follows: group 1, tumor only; group 2, tumorplus subretinal fluid; group 3, tumor plus focal seeds (3a, focal subretinalseeds; 3b, focal vitreous seeds); group 4, tumor plus diffuse seeds (4a, diffusesubretinal seeds; 4b, diffuse vitreous seeds); and group 5, neovascular glaucomaor invasive retinoblastoma.

Main Outcome Measure  Treatment success (avoidance of enucleation and external beam radiotherapy).

Results  According to the Reese-Ellsworth classification, chemoreduction wassuccessful in 100% of group Ia, 100% of group Ib, 86% of group IIa, 100% ofgroup IIb, 91% of group IIIa, 100% of group IIIb, 50% of group IVa, 77% ofgroup IVb, 50% of group Va, and 27% of group Vb. There was erratic correlationof the Reese-Ellsworth classification with treatment success. In contrast,the simplified grouping system displayed a smooth, nonerratic correlationfor treatment success, with 100% success for group 1, 91% for group 2, 59%for group 3, and 12% for group 4 (group 5 always managed by primary enucleation).When all 6 subcategory groups were analyzed, there was consistent correlationfor treatment success of 100% for group 1, 91% for group 2, 68% for group3a, 54% for group 3b, 17% for group 4a, and 11% for group 4b.

Conclusion  This practical approach to retinoblastoma using basic clinical featuresis predictive of treatment success for eyes in which modern conservative therapyfor retinoblastoma is used.

The management of retinoblastoma is complex and includes enucleationand nonenucleation (conservative) techniques. Currently, the most popularconservative technique is chemoreduction.^1-12 Chemoreductioninvolves intravenous chemotherapy to reduce the tumor size, followed by focalconsolidation with cryotherapy or thermotherapy to permanently devitalizeeach retinoblastoma.^6,13 Chemoreductionis most successful for tumors without associated subretinal fluid or tumor-relatedseeding.⁸ Success, defined as globe salvage,is found in 85% of treated patients by 5 years when the tumor is less advanced(Reese-Ellsworth groups I to IV) and 47% when the retinoblastoma is more advanced(Reese-Ellsworth group V).⁷

Despite these informative results regarding therapy, most clinicianshave difficulty applying Reese-Ellsworth classification to their practice,as it is complex and few ophthalmologists are able to recall its 10 categories.¹⁴ Reese and Ellsworth formulated this classificationapproximately 40 years ago as a method of predicting globe salvage after externalbeam radiotherapy. In a subsequent publication, Ellsworth stated that theReese-Ellsworth classification was a "purely arbitrary scheme" with the intentionto quantify tumor during an era when eyes with advanced retinoblastoma werebeing saved.¹⁵ This classification may no longerbe so useful in predicting globe salvage.

New and improved methods of therapy for retinoblastoma, such as cryotherapy,laser photocoagulation, thermotherapy, chemothermotherapy, chemoreduction,and custom-designed plaque radiotherapy, have subsequently been developed.External beam radiotherapy currently is used far less often because of long-termradiation complications and the success of the newer methods. Current therapieshave allowed eyes, previously judged by the Reese-Ellsworth classificationas probable failures, to be uncomplicated successes. For example, an eye previouslyclassified as group IIIa (any lesion anterior to the equator) would be "doubtful"for globe salvage by the Reese-Ellsworth classification, but current therapiessuch as cryotherapy, chemoreduction, or plaque radiotherapy would likely salvagethe eye, resulting in a more favorable outcome. The doubtful prognosis forReese-Ellsworth group IIIa eyes stemmed from failure of external beam radiotherapyto effectively treat tumors anterior to the equator and was not an inherentquality of retinoblastoma located anterior to the equator. The doubtful prognosiswas related to the specific method of treatment, external beam radiotherapy,but since this therapy is used much less often today, there is a need to useother criteria for judging success of treatment.

On the basis of extensive clinical experience with retinoblastoma andits complexities and current management, we have developed a practical approachfor grouping eyes with retinoblastoma, and this approach may be applicableto future classifications of retinoblastoma. Chemoreduction is currently theleading conservative treatment modality for retinoblastoma; in this study,we applied our grouping to a large cohort of patients with retinoblastomatreated with chemoreduction and evaluated its predictive value for treatmentsuccess.

All new patients with retinoblastoma who were treated with initial chemoreduction(institutional review board approval [The Children's Hospital of Philadelphia]582) on the Ocular Oncology Service, Wills Eye Hospital, Thomas JeffersonUniversity, Philadelphia, Pa, in conjunction with the Division of Oncologyat The Children's Hospital of Philadelphia were identified. Patients eligiblefor treatment with chemoreduction⁶ were childrenwith retinoblastoma in whom either eye would ordinarily require enucleationor external beam radiotherapy for cure of the disease on the basis of publishedindications.^10,11 Any patientwhose tumor(s) could be properly controlled with focal methods alone (cryotherapy,laser photocoagulation, thermotherapy, chemothermotherapy, and/or plaque radiotherapy)was not eligible for inclusion in the chemoreduction protocol. Exclusion criteriafor treatment with chemoreduction included evidence of iris neovascularization,neovascular glaucoma, extensive hyphema or vitreous hemorrhage, or tumor invasioninto the anterior chamber, iris, optic nerve, choroid, or extraocular tissuesas documented by clinical, ultrasonographic, and neuroimaging modalities.Exclusion criteria from a systemic standpoint were evidence of systemic metastasis,previous chemotherapy, or inadequate organ function of the kidney, liver,or ear. Patients who received previous treatment for retinoblastoma were notincluded in this database. The chemotherapeutic agents included intravenousvincristine sulfate, etoposide, and carboplatin in the following dosages:vincristine sulfate, 1.5 mg/m² (0.05 mg/kg for children ≤36months of age and maximum dose ≤2 mg); etoposide, 150 mg/m² (5mg/kg for children ≤36 months of age); and carboplatin, 560 mg/m² (18.6 mg/kg for children ≤36 months of age). All 3 drugs were givenon day 0 and only etoposide on day 1 in each cycle. The planned duration oftreatment was 6 months, with a cycle every 28 days. The potential risks andbenefits of the chemoreduction protocol were discussed with the patient'sfamily and informed consent was signed.

Each affected eye was classified according to the Reese-Ellsworth classification(Table 1). In addition, we groupedeyes on the basis of clinical features as in Table 2 (Figure 1). Ourgrouping was created to practically quantify retinoblastoma and its associatedfeatures without the need to refer to a list of complex qualification criteria.Proceeding from lowest to highest grouping was meant to imply worse ocularprognosis. Each eye was placed in the group with most advanced features; forexample, an eye with focal vitreous seeds and diffuse subretinal seeds wouldbe catergorized as group 4a, not 3b.

Ocular oncologic follow-up was provided at examination with the patientunder anesthesia every 1 to 2 months after initiation of chemoreduction untilcomplete control of the disease was achieved. Thereafter, examinations wereprovided every 2 to 4 months as needed. Adjuvant treatment to the regressedretinoblastomas with thermotherapy or cryotherapy was provided while the patientwas receiving the chemoreduction protocol. The final ocular outcome, needfor enucleation or external beam radiotherapy (failure) or no need for enucleationor external beam radiotherapy (success), was recorded at the date of mostrecent examination. The predictive value for success in treatment with chemoreductionby means of the Reese-Ellsworth classification and our simplified groupingwas compared.

During the 62-month period of this study, there were 364 retinoblastomasin 158 eyes of 103 patients enrolled in the chemoreduction protocol. The Reese-Ellsworthclassification and our grouping of each eye are listed in Table 3. The mean follow-up was 26 months (median, 20 months; range,1-83 months). The success rate was then calculated on the basis of the Reese-Ellsworthclassification and our grouping (Figure 2 and Figure 3). When theeyes were assessed by means of the 5 major Reese-Ellsworth groups, there wasa general decreasing correlation of success with each higher group, with theexception that group III fared slightly better than group II (Figure 2A). When the 10 subcategory groups of the Reese-Ellsworthclassification were assessed, the success for each group was as follows: 100%success for group Ia, 100% for group Ib, 86% for group IIa, 100% for groupIIb, 91% for group IIIa, 100% for IIIb, 50% for IVa, 77% for IVb, 50% forVa, and 27% for Vb. As is evident, there was erratic correlation, as groupsIIb and IIIb fared as well as Ia and Ib; group IVb fared better than IVa;and group Va fared the same as IVa (Figure2B). Reese-Ellsworth groups Ia, Ib, IIb, and IIIb showed 100% success.

In contrast, the 4 major groups in our system showed a consistent progressivedecrease in success rate with increasing group number as follows: 100% successfor group 1, 91% for group 2, 59% for group 3, and 12% for group 4 (Figure 3A). Group 5 eyes were not eligiblefor chemoreduction and were managed by primary enucleation in all cases. Withthe use of the 6 subcategory groups of our system, there was a consistentgeneral decreasing trend for success with higher group number without exceptionas follows: 100% success for group 1, 91% for group 2, 68% for group 3a, 54%for group 3b, 17% for group 4a, and 11% for group 4b (Figure 3B).

More than 40 years ago, Reese and Ellsworth presented their classificationof retinoblastoma at the 67th annual session of the American Academy of Ophthalmology.¹⁴ For the first time, they proposed a scheme for groupingeyes with retinoblastoma according to quantity and location of tumor(s) aswell as associated features such as vitreous seeding (Table 2). Their goal was to create guidelines to predict successof external beam radiotherapy for retinoblastoma. They tested their classificationscheme on 164 patients with retinoblastoma treated with "x-ray and TEM [chemotherapy]"between 1953 and 1960 and found success (globe salvage) in 85% of group Ipatients, 60% of group II, 52% of group III, 68% of group IV, and 23% of groupV.¹⁴ In the discussion of their paper, Bettmanstated that "it should be crystal clear that the basic therapy for retinoblastomais irradiation with x-ray" and further queried why the Reese-Ellsworth classificationfailed with regard to group IV, which fared better than group III, contraryto its presumed value. Ellsworth replied in the discussion that he acknowledgedthis deficiency and would like to improve the classification but would waitfor a future time to change the scheme. In a later communication, Ellsworth¹⁶ admitted that the classification had deficiencies;for example, an eye with relatively small tumors with localized vitreous seedingwould be categorized as "very unfavorable" in group Vb, but in reality, itsprognosis was much better than the classification implied. He stated thatmodification of the Reese-Ellsworth classification would be ultimately necessary.^16,17 However, this classification hasremained unchanged for 4 decades and still remains the most commonly usedgrouping scheme for retinoblastoma.

In 1983, Hopping¹⁷ presented the Essenclassification of retinoblastoma (Table4). This accounted for newer methods of clinical examination suchas binocular ophthalmoscopy and newer treatments such as light coagulation,cryocoagulation, and cobalt and ruthenium applicators.¹⁸ Inan evaluation of 273 eyes staged by the Essen classification, de Sutter etal¹⁸ found globe salvage for 97% of group 1,99% of group 2, 100% of group 3, 69% of group 4, and 27% of group 5. Thus,this classification showed favorable but identical results for groups 1, 2,and 3, reducing its predictive power. This thoughtful classification was elegantin its detail, but was also difficult to recall or apply to clinical practice.Vague words without precision were used, such as "tumors near the macula,even if small," "moderately sized tumors," and "limited seeding." The definitionof macula, details of proximity of tumor to the macula,exact tumor size, and quantification of seeding were not provided. In addition,it appeared that the Essen classification, primarily designed to predict globesalvage, attempted to consider visual outcome to some extent, as a tumor nearthe macula, even if small, was categorized as group 2 rather than group 1.The blending of 2 outcomes may have decreased the predictive power for a singleoutcome. One advantage of our current grouping is that it is based on thesingle outcome of treatment success (avoidance of enucleation and externalbeam radiotherapy), without factoring in visual or systemic outcome.

Our practical grouping system was created to provide a simple approachthat could be easily remembered and could reliably predict treatment success(avoidance of enucleation and external beam radiotherapy). This approach aroseduring the past decade and evolved during patient treatment as it became apparentthat eyes with tumor alone were successfully treated with chemoreduction,no matter the size, number, or location of the tumor(s). In fact, an eye withtumor alone was found to have 100% success, so that the tumor size, number,or location became irrelevant (Figure 3).

Before layout of the grouping system, we first performed a MEDLINE searchfor all published reports on retinoblastoma and selected for review thosethat provided statistical analysis for success of chemoreduction in largegroups of patients with retinoblastoma.^6-9,13 Wefound that the most relevant clinical features that were predictive of chemoreductionfailure (need for enucleation or external beam radiotherapy) included thepresence of subretinal fluid, subretinal tumor seeds, and vitreous tumor seedsat the initial examination.^6-8 Thus,we based our grouping on these predictive factors. It should be noted thatthe Reese-Ellsworth classification did not account for subretinal fluid orsubretinal seeds. When analyzing our grouping system with regard to treatmentsuccess, we found impressive predictive power, with treatment success in 100%of group 1 eyes, 91% of group 2 eyes, 59% of group 3 eyes, and 12% of group4 eyes (Figure 3A). Further subdivisioninto 6 groups allowed for more refinement in predictive ability (Figure 3B). In contrast, the Reese-Ellsworthclassification showed erratic correlation with treatment success, as 100%of eyes were successes in groups Ia, Ib, IIb, and IIIb, decreasing its predictivepower; in addition, group IVb fared better than group IVa and groups IVa andVa fared the same (Figure 2B).

We suspect that eyes with subretinal seeds and vitreous seeds showedgreater risk of recurrence and treatment failure owing to inadequate exposureto chemotherapy because of deficient blood supply to the seeds. We suspectthat patients with tumor plus subretinal fluid (group 2) showed greater failurethan those with tumor alone (group 1) because of possible subclinical subretinaltumor seeds within the fluid, accounting for a higher rate of tumor recurrence.

There are limitations to this grouping system. First, it is designedto predict success (avoidance of enucleation and external beam radiotherapy)with current conservative (nonenucleation) therapies, especially chemoreduction.It was created to predict globe prognosis, not life or visual prognosis. Concerningsuccess of treatment, group 1 eyes showed complete success, with continueddecreasing success with each group to approximately 10% success in group 4eyes (Figure 3). Group 5 eyes werealways managed with primary enucleation. Second, our evaluation of the applicabilityof the grouping system in this report was directed to eyes treated with initialchemoreduction and not any of the other less commonly used conservative methods(plaque radiotherapy, cryotherapy, laser photocoagulation, and thermotherapy),^10,19 but we believe that it will applyto all methods. Third, because of the simplicity of this grouping, there maybe details that are not included but may be useful in predicting globe outcome.Our grouping system is based on clinical factors identified by previous studiesto be predictive of treatment failure (subretinal fluid, subretinal seeds,and vitreous seeds).^6-8 Althoughlarger tumors are generally more likely to be associated with poor prognosticfactors like subretinal fluid or seeds, we did not find that tumor featuresalone, such as size, location, or number, were specifically predictive offailure. In fact, all patients with tumor alone (group 1) showed success,despite variations in size, location, or number per eye (Figure 3). Fourth, we tested this grouping in 158 eyes, but a largercohort of patients might be useful to better assess its predictive power.

In summary, we have used an innovative, practical method to group eyeswith retinoblastoma on the basis of expected success of conservative treatmentssuch as chemoreduction and focal therapies. We hope that this informationwill be beneficial to the design of future retinoblastoma classifications.²⁰ This grouping system is concise and easy to recall,provides prediction for treatment success, and, we hope, will be useful forclinicians to estimate outcomes of management for retinoblastoma.

Corresponding author and reprints: Carol L. Shields, MD, Ocular OncologyService, Wills Eye Hospital, 840 Walnut St, Philadelphia, PA 19107 (e-mail: carol.shields@shieldsoncology.com).

Submitted for publication March 18, 2003; final revision received August27, 2003; accepted September 15, 2003.

This study was supported by the Eye Tumor Research Foundation, Philadelphia(Dr C. L. Shields); the Macula Foundation, New York, NY (Dr C. L. Shields);the Rosenthal Award of the Macula Society, Barcelona, Spain (Dr C. L. Shields);and the Paul Kayser International Award of Merit in Retina Research, Houston,Tex (Dr J. A. Shields).

This study was presented at the 11th International Society for Retinoblastomaand Genetic Eye Diseases meeting; May 22, 2003; Paris, France; presented aspart of the William Smiddy Memorial Lecture; October 11, 2002; Baltimore,Md (Dr C. L. Shields); and presented as part of the Donders Lecture; April10, 2003; Rotterdam, the Netherlands (Dr C. L. Shields).