Evolution in Regression Patterns Following Chemoreduction for Retinoblastoma

Objective  To evaluate the change in regression pattern following chemoreduction and tumor consolidation therapy (thermotherapy or cryotherapy) for retinoblastoma

Methods  Retrospective medical record analysis was completed for 557 retinoblastomas (239 eyes of 157 patients) that were treated with chemoreduction and showed regression to 1 of 5 patterns (type 0, no visible remnant; type 1, completely calcified remnant; type 2, completely noncalcified remnant; type 3, partially calcified remnant; and type 4, atrophic chorioretinal flat scar). Evolution of these regression patterns was observed over time.

Results  Immediately following 6 cycles of chemoreduction, types 0 (2%), 1 (30%), 2 (3%), 3 (33%), and 4 (32%) regression patterns were found. During a mean follow-up period of 56 months (median, 48 months; range, 18-145 months), there was no change in regression patterns classified as type 0, 1, or 4. However, there was evolution of regression pattern types 2 and 3. Over time, type 2 tumor scars either remained stable (41%) or evolved to type 4 (41%), 3 (9%), or 1 (9%) scars. Type 3 tumor scars remained stable (74%) or evolved to type 1 (26%) scars.

Conclusion  Following chemoreduction and tumor consolidation therapy, retinoblastoma regression patterns types 2 and 3 can slowly evolve over time into a slightly different appearance, even without additional treatment. Ophthalmologists should be familiar with these regression patterns and their evolution.

Following chemoreduction and tumor consolidation therapy, retinoblastoma shows regression into 1 of 5 clinical patterns. Regression patterns types 2 and 3 can further slowly and spontaneously evolve into other patterns, with disappearance of the noncalcified portion, without additional treatment.

Chemoreduction (CRD) is an established treatment of intraocular retinoblastoma that leaves the globe intact, with regressed, often calcified retinal scars.^1-3 The regressed scars are classified into 1 of 5 regression patterns. These patterns are based on the clinical features and reflect the ophthalmoscopic appearance of the mass and presence or absence of calcium within the mass. The patterns include type 0, disappearance of the tumor; type 1, purely calcified scar; type 2, purely noncalcified scar; type 3, partially calcified scar; and type 4, atrophic scar of chorioretinal atrophy.^4-9 These patterns, initially described by Dunphy in 1964,⁴ have been used for many decades to judge tumor response to various therapies. Most retinoblastomas managed with CRD and adjuvant methods show type 3 or 4 regression.⁵

One might presume that regression patterns remain stable throughout the patient's life, but we have previously recognized slow involution of spontaneously regressed retinoblastoma (retinocytoma) in which a type 3 pattern evolved into a type 1 pattern over 20 years.⁶ Subsequently, we observed a similar slight evolution in regression patterns following CRD and tumor consolidation therapy (thermotherapy or cryotherapy). In this study, we analyze the regression patterns of retinoblastomas controlled with CRD and document the evolution in patterns over time.

A retrospective medical record review was performed on all patients with retinoblastoma treated with CRD in the Ocular Oncology Service, Wills Eye Institute, Thomas Jefferson University, and the Division of Oncology at The Children's Hospital of Philadelphia. Patients who had received treatment for retinoblastoma or initial CRD at other clinics were excluded. The 157 patients who met the inclusion criteria (239 eyes, 557 retinoblastomas) were included. The CRD protocol included intravenous vincristine (0.05 mg/kg), etoposide (5 mg/kg), and carboplatin (18.6 mg/kg) given every 28 days for 6 cycles. This protocol has been previously published.¹⁰

Each patient's age at diagnosis, hereditary pattern of retinoblastoma, tumor laterality, and affected eye were noted. All eyes were classified according to the Reese-Ellsworth classification⁷ and the International Classification of Retinoblastoma.^10-12 Nasal dimension, thickness, proximity to the optic disc, and foveola were analyzed. Tumor regression patterns after completion of CRD were graded as type 0, no visible remnant; type 1, completely calcified remnant; type 2, completely noncalcified remnant; type 3, partially calcified remnant; and type 4, atrophic chorioretinal flat scar.

Patients were examined every month before the cessation of CRD and every 2 to 6 months after using general anesthesia with pupillary dilation for any recurrences or changes. Tumor consolidation was provided for all tumors using thermotherapy for posterior or cryotherapy for peripherally located tumors at each examination while the patient was participating in the CRD protocol. The response of the tumor to the therapy was determined according to the shrinkage of the tumor (in terms of base diameter and thickness). Manual drawings and the RetCam II Digital Retinal Camera (Clarity, Pleasanton, California) were used to archive the conditions of each eye and tumor. Earlier examination pictures were compared with the actual examination images for any change. This study was approved by the institutional review board at Wills Eye Institute.

The medical records of 157 patients (93 male, 64 female) with retinoblastoma who were treated with CRD and adjuvant methods between July 1994 and March 2007 were evaluated (Table 1). The mean age of the patients at presentation was 9 months (range, <1 to 48 months). Sporadic retinoblastoma was seen in 129 of 157 cases (82%) and familial retinoblastoma in 38 cases (18%).

The mean basal diameter of all retinoblastomas was 7 mm (median, 4 mm; range, 1-25 mm), and the thickness was 4 mm (median, 3 mm; range, 1-24 mm) (Table 1). The mean basal diameter for retinoblastomas in the macula was 11 mm; macula to equator, 5 mm; equator to ora serrata, 3 mm; and diffuse tumors, 18 mm. The mean tumor thickness for retinoblastomas in the macula was 6 mm; macula to equator, 3 mm; equator to ora serrata, 2 mm; and diffuse tumors, 9 mm.

Chemoreduction was given in each case for 6 cycles with 1 cycle per month. At termination of CRD, the retinoblastoma showed regression types 0 (10 tumors; 2%), 1 (167 tumors; 30%), 2 (22 tumors; 3%), 3 (183 tumors; 33%), and 4 (175 tumors; 32%) (Table 2). At 6 months after termination of CRD, 59% of type 2 regressed tumors were stable, while 36% evolved to regression pattern type 4 and 5% evolved to type 1. At this time, 82% of type 3 tumors remained stable and 18% evolved to type 1. At 12 months after termination of CRD, there were no visible further changes. During a mean follow-up of 56 months (median, 48 months; range, 18-145 months), none of the type 0, type 1, or type 4 regressed tumors changed. Type 2 tumors remained stable (41%) or evolved to type 4 (41%), 3 (9%), or 1 (9%). Type 3 tumors remained stable (74%) or evolved to type 1 (26%) (Figure 1 and Figure 2).

Retinoblastoma regression patterns following external beam radiotherapy (EBRT) were first mentioned by Dunphy in 1964.⁴ He classified tumor regression as type 1 when the tumor assumed a glistening white appearance resembling cottage cheese.⁴ Type 2 had a homogeneous, translucent fish-flesh appearance and was difficult to distinguish from viable tumor.⁴ Type 3 was a combination of type 1 and 2 characteristics. Occasionally, EBRT leads to complete disappearance of a tumor, leaving only minor changes in the retinal pigment epithelium as the only indication of the original tumor site. This was termed type 0.^6-8 Similar patterns have been described following plaque radiotherapy.⁹ In some instances, plaque radiotherapy leads to total tumor destruction with completely flat chorioretinal atrophy, no tumor remnant, and exposure of the underlying sclera. This pattern, which had not been seen after EBRT, was termed type 4.⁸ Type 4 is now most commonly found after cryotherapy sessions, a treatment introduced after Dunphy's original description of regression patterns.^5-9

A detailed analysis of regression patterns following EBRT was provided by the Moorfields Eye Hospital in 1993.¹³ In the article, 180 retinoblastomas were evaluated, and regression patterns of type 0 (18%), 1 (50%), 2 (17%), and 3 (14%) were identified.¹³ More recently, the Wills Eye Institute described regression patterns following CRD of 557 retinoblastomas.⁵ They identified type 0 in 2%, type 1 in 13%, type 2 in 5%, type 3 in 23%, and type 4 in 57%. Correlation of regression pattern with tumor size revealed that tumors with an initial thickness of 3 mm or less showed regression most often to type 4 (92%), 3 to 8 mm to type 3 (34%) or 4 (40%), and greater than 8 mm to type 1 (40%) or 3 (49%). Regarding location, retinoblastoma in the macular area most often regressed to types 1 (25%), 3 (40%), or 4 (27%), whereas extramacular retinoblastoma most often regressed to type 4 (76%).⁵

Since the original description and classification of retinoblastoma regression patterns, there has been little information regarding the stability of various regression patterns. We found that regression patterns following CRD can evolve over time. The most stable regression patterns were types 0, 1, and 4. However, types 2 and 3 showed change over time without additional treatment. Both of these patterns involve noncalcified portions, and the evolution involves reduction in the noncalcified portion with complete disappearance of noncalcification or transition from noncalcified to increased calcification slowly over time. Type 2 regression patterns remained stable in 41% of cases, changed to type 4 with complete disappearance of all “fish flesh” in 41%, and changed to type 3 (9%) or type 1 (9%) with transformation of the fish flesh into calcified component. Similarly, type 3 regression patterns remained stable in 74% of patients and evolved to develop more calcification into type 1 in 26%. In all cases, the transformation was gradual and only documented on repeated follow-up examinations with careful comparison using ophthalmoscopy and fundus photography.

The long-term stability of regression patterns in 89 retinoblastomas treated with EBRT and followed up for a minimum of 7 years revealed slow evolution over time.¹⁴ Considering those that showed change between completion of EBRT and final evaluation (7 or more years after treatment), the number of tumors with type 1 regressions increased by 10%, type 2 decreased by 19%, type 3 decreased by 8%, type 4 increased by 10%, and the number of tumors that disappeared completely increased by 7%. It was observed that type 2 regression pattern had the greatest potential for change following EBRT, similar to the conclusions in our series following CRD. We found no comparative change in types 0, 1, and 4 regression patterns at termination compared with 6 and 12 months after termination of CRD (Table 2). At 6 months following termination of CRD, type 2 showed stability in 59% and evolution to type 4 in 36% and to type 1 in 4%; type 3 remained stable in 82% and showed evolution to type 1 in 18%. Comparatively, at 12 months following termination of CRD there was no further visible change from 6 months earlier. At a mean follow-up of nearly 5 years after termination of CRD, type 2 remained stable in 41% and evolved to type 4 in 41%, type 1 in 9%, and type 3 in 9%. At the date last seen at a mean of 5 years' follow-up, type 3 regression patterns remained stable in 74% of cases and evolved to type 1 in 26%. Although most of the evolution in regression pattern occurred in the first 6 months following termination of CRD, minor changes continued to take place over the long-term. Most likely, regression patterns types 2 and 3 have “retinocytomalike” noncalcified areas that have been previously described histopathologically following CRD.¹⁵ These benign cellular remnants could be the source of gradual change in appearance of the scar. Types 0, 1, and 4 do not manifest these remnants.

In conclusion, retinoblastoma regression patterns can evolve over time following EBRT and CRD with consolidation therapy. The most evident changes are the slow disappearance of the noncalcified fish-flesh portion or the transformation of noncalcified portion to calcified. The physician treating a child with retinoblastoma should be familiar with regression patterns, their frequency, and their evolution, and be able to differentiate the various regression patterns from incomplete tumor response or tumor recurrence.

Correspondence: Carol L. Shields, MD, Ocular Oncology Service, Wills Eye Institute, 840 Walnut St, Ste 1440, Philadelphia, PA 19107 (carol.shields@shieldsoncology.com).

Submitted for Publication: May 23, 2010; final revision received September 28, 2010; accepted September 29, 2010.

Financial Disclosure: None reported.

Funding/Support: This study was supported by the Eye Tumor Research Foundation, Philadelphia, Pennsylvania (Dr Shields).

Previous Presentation: This study was presented as a poster at the World Ophthalmology Congress; June 28 through July 2, 2008; Hong Kong, China.