Retinoblastoma is an uncommon paediatric malignant tumour which can be fatal.1 The retinoblastoma related mortality could be due to one of three distinct causes; metastasis,2^–4 trilateral retinoblastoma,5^–9 and second malignant neoplasms (SMN).10^–14 Second malignant neoplasm14 15 and trilateral retinoblastoma develop almost exclusively in heritable cases,8 9 16 whereas metastasis occurs with equal frequency in both heritable and non-heritable retinoblastoma.17

With improvements in diagnosis and management of retinoblastoma over the past several decades,18 metastatic retinoblastoma is observed less frequently in the USA19 and other developed nations.20 21 As a result, other causes of retinoblastoma-related mortality, SMN15 and trilateral retinoblastoma,7 have become significant contributors to the retinoblastoma related mortality. In the USA, trilateral retinoblastoma contributes to more than 50% of retinoblastoma-related mortality in the first decade of life,7 and SMN contribute incrementally (1% per year) to long-term mortality.12 14 15

However, metastatic retinoblastoma continues to be a major contributor to retinoblastoma-related mortality in the developing nations.22^–24 Metastatic disease at the time of retinoblastoma diagnosis is very rare, and so staging procedures such as bone scans, lumbar puncture and bone marrow aspirations at initial presentation are not generally recommended.]25, 26] The metastasis in retinoblastoma usually occurs within 1 year of diagnosis of retinoblastoma. If there is no metastatic disease within 5 years of retinoblastoma diagnosis, the child is usually considered cured.2 3 Involvement of the central nervous system, bones and bone marrow is common.2 4 The prognosis of metastatic retinoblastoma is poor, with death usually occurring within 6 months.2 3

Since 1994, there have been changes in treatment of retinoblastoma,27 with greater emphasis on the use of chemotherapy.28 We wanted to investigate whether recent changes in the treatment of retinoblastoma had any impact on the overall survival of the patients. We analysed variations in 5-year observed actuarial survival rates of patients with retinoblastoma in the USA over a 30-year period between 1975 and 2004 using the Surveillance, Epidemiology and End Results (SEER) Program database of the National Cancer Institute.

MATERIALS AND METHODS

The National Cancer Institute SEER Public Use CD-ROM for the period 1973–2004, was used.29 The SEER Program of the National Cancer Institute is the only comprehensive source of population-based information on cancer incidence and survival rate in the USA. The SEER Program collects data from population-based cancer registries and covers approximately 26 percent of the USA. The population covered by SEER is comparable with the general USA population.30 All 17 current SEER registries were utilised. For both morphology (retinoblastoma = 9510–9513) and topography (retinoblastoma included retina and eye (C69.2, C69.9)), International Classification of Oncology (ICDO-3) codes were utilised to define the cases for the study.31 The survival rates were calculated by the Kaplan–Meier method and differences evaluated with logrank and Wilcoxon tests.32 Kaplan–Meier survival rates were reported as observed actuarial survival without any adjustment for cause of death. Cause of death was obtained and reviewed for all deceased patients. All retinoblastoma patient records were reviewed for treatments given and occurrence of SMN.

RESULTS

There were 992 cases of retinoblastoma in the SEER file for the 30-year period. Two of these patients were diagnosed as having a contralateral lesion at 4 and 5 months after the index tumour respectively. For the purposes of this study, these two were classified as bilateral cases resulting in an analytical file with 990 distinct individuals (table 1).

Surgical treatment was performed in 855 cases (86.4%), and radiotherapy was given in 173 cases (17.5%). Of the 173 cases treated with radiation, 150 patients received external beam treatment only, 18 patients received brachytherapy implants only, one patient received combination external beam/brachytherapy implant, and four patients received an unspecified type of radiation. In addition, 127 cases (12.8%) received both surgery and radiotherapy. The proportion of cases treated surgically decreased from 94.3% in 1975–9 to 84.0% in 2000–4, whereas the proportion of cases treated with radiotherapy first increased from 20.5% in 1975–9 to 34.6% in 1985–9 and then sharply decreased to 6.5% in 2000–4 (fig 1). Chemotherapy usage is not readily available in the SEER file and, therefore, could not be described in this paper.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Trends in treatment of retinoblastoma. The proportion of cases treated surgically and by radiotherapy. The proportion of cases treated with radiotherapy first increased from 20.5% in 1975–9 to 34.6% in 1985–9 and then sharply decreased to 6.5% in 2000–4.

The survival analysis could be performed on 970 patients diagnosed between 1975 and 2004 (18 patients had no follow-up, one patient was diagnosed at autopsy, and one patient was reported by death certificate only). The overall actuarial observed survival rate was 94.6% at 5 years and 92.9% at 10 years. Over the period of 30 years (1975–2004), the 5-year actuarial observed survival rate increased from 92.3% (1975–84) to 93.9% (1985–94) to 96.5% (1995–2004) (fig 2). The difference in rates was statistically significant (Wilcoxon = 6.2393, p = 0.0442).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Five-year actuarial observed survival rates from 1975 to 2004. Note the improving survival in recent years.

Causes of death according to time since diagnosis are summarised in table 2.

Of the deaths occurring more than 60 months after diagnosis, SMN were the leading cause of death (13 deaths, 65%). A total of 21 cases in which an SMN developed were reported. Osteosarcomas were the most common SMN reported, accounting for 12 out of the 21 cases (57%). Other types of sarcoma accounted for five out of the 21 cases (24%). The projected risk of SMN developing in retinoblastoma survivors was 4.3% by 18 years (fig 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Risk of second malignant neoplasms (SMN) in young survivors of retinoblastoma (osteosarcoma and overall).

DISCUSSION

With realisation of the risk of radiation therapy in causation of radiation induced SMN in children with retinoblastoma,10^–14 33 chemotherapy has become the mainstay of treatment for retinoblastoma.28 The national recent trend away from radiation therapy was evident in our analysis, wherein only 6.5% of cases received radiation therapy in 2000–4 as compared with 20.5% in 1975–9. It also appears that the change in the therapeutic approaches has translated into improved 5-year actuarial observed survival, which showed a statistically significant increase over the 30-year period, from 92.3% (1975–84) to 96.5% (1995–2004), consistent with survival rates published previously (table 3).

As expected, deaths due to metastatic retinoblastoma occurred during the first 5 years after the initial diagnosis. Due to the limited number of cases of metastatic deaths in each 5-year period, we were unable to examine the distribution of causes of death over the 30-year period. We, therefore, cannot confirm that the incidence of metastatic retinoblastoma leading to death is decreasing, only that overall survival rate is increasing. Although it has been reported that trilateral retinoblastoma contributes to 50% of retinoblastoma-related mortality in the first decade of life in the US,7 this appears unlikely, as in our population-based study, only two out of 43 deaths during the 1–60-month period postdiagnosis were attributed to trilateral retinoblastoma.

It is well known that hereditary retinoblastoma survivors, especially those treated with external beam radiotherapy, are at higher risk of developing a SMN.10^–14 33 However, because of selection bias and variability in study designs, the estimates of cumulative incidence for SMN have varied widely between the published studies.13 37 In a review of published studies, Moll et al identified only four of 11 series to be free from selection bias, reporting a cumulative incidence of SMN of 8.4% 18 years after the initial diagnosis of retinoblastoma.13 In general, registry-based studies11 13 38 have estimated a lower risk of SMN than clinic-based studies.12 33 Moreover, the risk of SMN will appear lower if the denominator includes all cases of retinoblastoma (heritable and non-heritable) as in registry-based studies, rather than only the “at risk” population (heritable cases of RB), included in clinic-based studies. The lower projected risk of developing an SMN of 4.3% by 18 years of age in our study, although similar to that reported from the UK,11 may represent an underestimation of the actual rates of SMN. Despite intense efforts by the SEER program to collect complete follow-up information (up to age 18 years in paediatric cohorts), an estimated 14% of childhood cancer cases are lost to follow-up due to persons moving outside geographic areas covered by SEER registries.39 As chemotherapy usage is not readily available in the SEER file, and we do not as yet have long-term data on SMN in those diagnosed and treated in the last decade, we are unable to conclude whether chemotherapy usage affects the risk of developing a SMN.

In summary, treatment with radiotherapy has decreased over the 30-year period (1975–2004) while overall the 5-year observed actuarial survival from retinoblastoma has increased over the same period. While all deaths due to metastatic retinoblastoma and trilateral retinoblastoma occurred within 5 years of diagnosis, all deaths due to SMN occurred more than 5 years after the initial diagnosis of retinoblastoma.