Abstract
Purpose:
Globally, disparities exist in retinoblastoma treatment outcomes between high- and low-income countries, but independent analysis of American countries is lacking. We report outcomes of American retinoblastoma patients and explore factors associated with survival and globe salvage.
Design:
Subanalysis of prospective cohort study data.
Methods:
Multicenter analysis at 57 American treatment centers in 23 countries of varying economic levels (low income=LIC, lower-middle=LMIC, upper-middle=UMIC, high=HIC) of 491 treatment-naïve retinoblastoma patients diagnosed in 2017 and followed through 2020. Survival and globe salvage rates analyzed with Kaplan-Meier analysis and Cox proportional hazard models.
Results:
Of patients, 8 (1.6%), 58 (11.8%), 235 (47.9%) and 190 (38.7%) were from LIC, LMIC, UMIC and HIC, respectively. Three-year survival rates in LICs were 60.0% (95% CI, 12.6–88.2) compared to 99.2% (94.6–99.9) in HICs. Death was less likely in patients older than four years (vs. four or younger, HR=0.45 [95% CI, 0.27 – 0.78], P=0.048). Patients with more advanced tumors (e.g., cT3 vs. cT1, HR= 4.65×109 [95% CI, 1.25×109 – 1.72×1010], P<0.001) and females (vs. males, HR=1.98 [1.27–3.10], P=0.04) were more likely to die. Three-year globe salvage rates were 13.3% (95% CI, 5.1–25.6) in LMICs and 46.2% (38.8–53.3) in HICs. At three years, 70.1% of cT1 eyes (95% CI, 54.5–81.2) versus 8.9% of cT3 eyes (5.5–13.3) were salvaged. Advanced tumor stage was associated with higher enucleation risk (e.g., cT3 vs. cT1, SHR=4.98 [95% CI, 2.36–10.5), P<0.001).
Conclusions:
Disparities exist in survival and globe salvage in American countries based on economic level and tumor stage demonstrating a need for childhood cancer programs.
INTRODUCTION
The prognosis of retinoblastoma, the most common primary pediatric eye cancer, is dependent on early diagnosis and treatment.1–3 Treatment largely aims to cure, while also prioritizing ocular salvage and vision preservation.3 Many patients in the Americas present with advanced intraocular disease that requires chemotherapy, adjunctive consolidative therapy and rarely even radiation to save the eye.3 Enucleation may be done primarily, or secondarily when efforts to save the eye have failed – for advanced unilateral Group E eyes, enucleation is the most common primary therapy.3 Success of therapy is highly related to disease burden.4 Early diagnosis to facilitate treatment is therefore integral for globe preservation and survival.
Studies have shown disparities in treatment outcomes worldwide between high- and low-income countries (HIC and LIC, respectively).2,3,5–7 Notably, data have shown higher mortality rates and globe loss among children diagnosed with retinoblastoma in LICs than in HICs.2,3 In HICs, there is nearly a 100% disease-free survival rate for retinoblastoma.8 Further, studies have shown a 9-to-10-fold higher risk of metastasis-related death in LICs than HICs.2 It should be noted, however, that systemic disease confers virtually equal mortality risk in LICs and HICs, highlighting the importance of early treatment regardless of income status.2
An initial Global Retinoblastoma Outcomes study followed 4064 children from 149 countries for three years after retinoblastoma diagnosis and explored outcomes associated with survival and globe salvage.8 Globally, patients from low-income countries experienced higher rates of death and enucleation.8 The present study is a sub-analysis that explores disparities in retinoblastoma treatment outcomes in the Americas, through analysis of 491 children from 23 countries. This is the first study to assess retinoblastoma treatment outcomes specifically in the Americas.
METHODS
This study adheres to the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) statement, as well as to the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) statement.9,10 This study is a sub-analysis of patients from the Americas included in the Global Retinoblastoma Outcome Study, a 3-year prospective analysis of retinoblastoma outcomes in treatment-naïve patients; as such, the analysis methods closely followed those in the Global Study.8 In brief, retinoblastoma treatment centers across the world were invited to participate in a cross-sectional study of all treatment-naïve patients who presented between January 1, 2017 and December 31, 2017. Next, a prospective analysis was conducted on these patients, as well as patients from additional treatment centers that were not part of the initial cross-sectional study. Data on primary and additional treatments, duration of follow-up, metastasis, globe salvage, survival outcome, and the impact of COVID-19 were collected.8 As in the Presentation Study, national income level classifications were obtained from the 2017 World Population Prospects.5,11 The study was approved by the London School of Hygiene & Tropical Medicine Observational Ethics Committee. Participating centers received local ethics approval.
Statistical Analysis
Statistical analyses were conducted using Stata/SE software (version 14.2; College Station, TX, USA). Survival analysis was used to examine both all-cause mortality and enucleation. Time to death was summarized using Kaplan-Meier estimates. Association of time to death with risk or protective factors was examined using Cox proportional hazard models. Time to enucleation was evaluated using Fine and Gray proportional sub-hazard models adjusted for the competing risk of death.12 In cases of bilateral globe loss, only the first event was included. Factors in both models included the economic group of the nation where the patient’s clinic was located; primary tumor stage (cT) and hereditary category (H) based on the AJCC Staging system,13 sex, disease laterality, family history of retinoblastoma, and age at diagnosis (fit using linear splines). Analyses were clustered by treatment center and weighted based on the inverse probability of having missing outcome data. P-values less than 0.05 were considered statistically significant after Bonferroni correction. Additional details on the global study and analysis methods are found in the Supplement.
RESULTS
The cohort included 491 treatment-naïve patients from 23 American countries, who presented to 57 treatment centers in 2017 and received or were offered treatment for retinoblastoma (Table 1A). Of these patients, 49 had missing dates of birth, and 40 had last follow-up dates missing. Of the study cohort, 1.6% (8/491) of patients were from LICs, 11.8% (58/491) were from lower-middle income countries (LMICs), 47.9% (235/491) were from upper-middle income countries (UMICs), and 38.7% (190/491) were from HICs. All countries represented in the data, identified by income level, are summarized in Supplemental Figure 1 as well as the expected number of retinoblastoma cases per country based on crude birth rates.11 The most represented countries were the USA (32.4%, 159/491), a HIC; Peru (14.9%, 73/491), an UMIC; Brazil (11.4%, 56/491), an UMIC; and Guatemala (7.5% (37/491), a LMIC.
Table 1:
Clinical diagnostic characteristics and treatment outcomes for 491 patients from 57 centers in 23 American countries
| Table 1A. Participating countries and treatment centers by national income level | |||||
|---|---|---|---|---|---|
| National Income Level | |||||
| n (%) | Low | Lower-Middle | Upper-Middle | High | Total |
| Number of countries | 1 (4%) | 5 (22%) | 12 (52%) | 5 (22%) | 23 |
| Number of centers | 1 (1.8%) | 6 (10.5%) | 21 (36.8%) | 29 (50.9%) | 57 |
| Table 1B. Clinical characteristics at diagnosis by national income level | |||||
| National Income Level | |||||
| n/N (%) | Low | Lower-Middle | Upper-Middle | High | Total |
| Age at diagnosis (months) | |||||
| Median (IQR) |
32.7 (27.6–46.0) |
21.7 (10.8–39.0) |
21.7 (9.1–32.3) |
15.3 (6.1–25.1) |
19.4 (8.3–31.9) |
| Data availablea | 6/8 (75%) | 58/58 (100%) | 226/235 (96.2%) | 152/190 (80%) | 442/491 (90%) |
| Laterality at presentation b | |||||
| Unilateral | 6/8 (75%) | 40/58 (69%) | 174/235 (74%) | 111/190 (58.4%) | 331/491 (67.4%) |
| 6/331 (1.8%) | 40/331 (12.1%) | 174/331 (52.6%) | 111/331 (33.5%) | ||
| Bilateral | 2/8 (25%) | 18/58 (31%) | 61/235 (26%) | 79/190 (41.6%) | 160/491 (32.6%) |
| 2/160 (1.3%) | 18/160 (11.3%) | 61/160 (38.1%) | 79/160 (49.4%) | ||
| Sex b | |||||
| Female | 3/8 (37.5%) | 35/58 (60.3%) | 119/235 (50.6%) | 75/190 (39.5%) | 232/491 (47.3%) |
| 3/232 (1.3%) | 35/232 (15.1%) | 119/232 (51.3%) | 75/232 (32.3%) | ||
| Male | 5/8 (62.5%) | 23/58 (39.7%) | 116/235 (49.4%) | 115/190 (60.5%) | 259/491 (52.7%) |
| 5/259 (1.9%) | 23/259 (8.9%) | 116/259 (44.8%) | 115/259 (44.4%) | ||
| Family history of retinoblastoma | |||||
| Yes | 0 | 0 | 12/235 (5.1%) | 23/189 (12.2%) | 35/490 (7.1%) |
| 0 | 0 | 12/35 (34.3%) | 23/35 (65.7%) | ||
| No | 8/8 (100%) | 58/58 (100%) | 223/235 (94.9%) | 166/189 (87.8%) | 455/490 (92.9%) |
| 8/455 (1.8%) | 58/455 (12.7%) | 223/455 (49%) | 166/455 (36.5%) | ||
| Data availablea | 8/8 (100%) | 58/58 (100%) | 235/235 (100%) | 189/190 (99.5%) | 490/491 (99.8%) |
| Clinical Tumor, Node, Metastasis, Heredity 8th Edition Staging | |||||
| Primary tumor | |||||
| cT1 | 1/6 (16.7%) | 1/58 (1.7%) | 17/231 (7.4%) | 32/189 (16.9%) | 51/484 (10.5%) |
| 1/51 (2%) | 1/51 (2%) | 17/51 (33.3%) | 32/51 (62.7%) | ||
| cT2 | 0 | 10/58 (17.2%) | 59/231 (25.5%) | 91/189 (48.1%) | 160/484 (33.1%) |
| 0 | 10/160 (6.3%) | 59/160 (36.9%) | 91/160 (56.9%) | ||
| cT3 | 1/6 (16.7%) | 33/58 (56.9%) | 134/231 (58%) | 64/189 (33.9%) | 232/484 (47.9%) |
| 1/232 (0.4%) | 33/232 (14.2%) | 134/232 (57.8%) | 64/232 (27.6%) | ||
| cT4 | 4/6 (66.7%) | 14/58 (24.1%) | 21/231 (9.1%) | 1/189 (0.5%) | 40/484 (8.3%) |
| 4/40 (10%) | 14/40 (35%) | 21/40 (52.5%) | 1/40 (2.5%) | ||
| Retinoma | 0 | 0 | 0 | 1/189 (0.5%) | 1/484 (0.2%) |
| 0 | 0 | 0 | 1/1 (100%) | ||
| Data availablea | 6/8 (75%) | 58/58 (100%) | 231/235 (98.3%) | 189/190 (99.5%) | 484/491 (98.6%) |
| Regional lymph node | |||||
| NX | 1/6 (16.7%) | 5/58 (8.6%) | 22/231 (9.5%) | 65/190 (34.2%) | 93/485 (19.2%) |
| 1/93 (1.1%) | 5/93 (5.4%) | 22/93 (23.7%) | 65/93 (69.9%) | ||
| N0 | 2/6 (33.3%) | 48/58 (82.8%) | 204/231 (88.3%) | 125/190 (65.8%) | 379/485 (78.1%) |
| 48/379 (12.7%) | 204/379 (53.8%) | 125/379 (33%) | 1/93 (1.1%) | ||
| N1 | 3/6 (50%) | 5/58 (8.6%) | 5/231 (2.2%) | 0 | 13/485 (2.7%) |
| 3/13 (23.1%) | 5/13 (38.5%) | 5/13 (38.5%) | 0 | ||
| Data availablea | 6/8 (75%) | 58/58 (100%) | 231/235 (98.3%) | 190/190 (100%) | 485/491 (98.8%) |
| Distant metastasis | |||||
| M0 | 3/6 (50%) | 50/58 (86.2%) | 218/231 (94.4%) | 190/190 (100%) | 461/485 (95.1%) |
| 3/461 (0.7%) | 50/461 (10.8%) | 218/461 (47.3%) | 190/461 (41.2%) | ||
| cM1 | 3/6 (50%) | 4/58 (6.9%) | 7/231 (3%) | 0 | 14/485 (2.9%) |
| 3/14 (21.4%) | 4/14 (28.6%) | 7/14 (50%) | 0 | ||
| pM1 | 0 | 4/58 (6.9%) | 6/231 (2.6%) | 0 | 10/485 (2.1%) |
| 0 | 4/10 (40%) | 6/10 (60%) | 0 | ||
| Data availablea | 6/8 (75%) | 58/58 (100%) | 231/235 (98.3%) | 190/190 (100%) | 485/491 (98.8%) |
| Hereditary trait | |||||
| HX | 5/7 (71.4%) | 40/58 (69%) | 163/231 (70.6%) | 38/190 (20%) | 246/486 (50.6%) |
| 5/246 (2%) | 40/246 (16.3%) | 163/246 (66.3%) | 38/246 (15.4%) | ||
| H0 | 0 | 0 | 1/231 (0.4%) | 54/190 (28.4%) | 55/486 (11.3%) |
| 0 | 1/55 (1.8%) | 54/55 (98.2%) | 5/246 (2%) | ||
| H1 | 2/7 (28.6%) | 18/58 (31%) | 67/231 (29%) | 98/190 (51.6%) | 185/486 (38.1%) |
| 2/185 (1.1%) | 18/185 (9.7%) | 67/185 (36.2%) | 98/185 (53%) | ||
| Data availablea | 7/8 (87.5%) | 58/58 (100%) | 231/235 (98.3%) | 190/190 (100%) | 486/491 (99%) |
| Table 1C. 3-year outcomes by national income level | |||||
| National Income Level | |||||
| n/N (%) | Low | Lower-Middle | Upper-Middle | High | Total |
| Enucleation * | |||||
| Yes | 4/8 (50%) | 45/58 (77.6%) | 184/235 (78.3%) | 104/190 (54.7%) | 337/491 (68.6%) |
| 4/337 (1.2%) | 45/337 (13.4%) | 184/337 (54.6%) | 104/337 (30.9%) | ||
| No | 4/8 (50%) | 13/58 (22.4%) | 50/235 (21.3%) | 82/190 (43.2%) | 149/491 (30.3%) |
| 4/149 (2.7%) | 13/149 (8.7%) | 50/149 (33.6%) | 82/149 (55.0%) | ||
| Unknown | 0 | 0 | 1/235 (0.4%) | 4/190 (2.1%) | 5/491 (1.0%) |
| 0 | 0 | 1/5 (20.0%) | 4/5 (80.0%) | ||
| Metastasis * | |||||
| Yes | 5/8 (62.5%) | 12/58 (20.7%) | 30/235 (12.8%) | 3/190 (1.6%) | 50/491 (10.2%) |
| 5/50 (10%) | 12/50 (24%) | 30/50 (60%) | 3/50 (6%) | ||
| No | 2/8 (25%) | 39/58 (67.2%) | 172/235 (73.2%) | 172/190 (90.5%) | 385/491 (78.4%) |
| 2/385 (0.5%) | 39/385 (10.1%) | 172/385 (44.7%) | 172/385 (44.7%) | ||
| Unknown | 1/8 (12.5%) | 7/58 (12.1%) | 33/235 (14%) | 15/190 (7.9%) | 56/491 (11.4%) |
| 1/56 (1.8%) | 7/56 (12.5%) | 33/56 (58.9%) | 15/56 (26.8%) | ||
| Survival Status * | |||||
| Dead | 3/8 (37.5%) | 13/58 (22.4%) | 24/235 (10.2%) | 3/190 (1.6%) | 43/491 (8.8%) |
| 3/43 (7%) | 13/43 (30.2%) | 24/43 (55.8%) | 3/43 (7%) | ||
| Alive | 2/8 (25%) | 40/58 (69%) | 183/235 (77.9%) | 178/190 (93.7%) | 403/491 (82.1%) |
| 2/403 (0.5%) | 40/403 (9.9%) | 183/403 (45.4%) | 178/403 (44.2%) | ||
| Unknown | 3/8 (37.5%) | 5/58 (8.6%) | 28/235 (11.9%) | 9/190 (4.7%) | 45/491 (9.2%) |
| 3/45 (6.7%) | 5/45 (11.1%) | 28/45 (62.2%) | 9/45 (20%) | ||
| Cause of Death | |||||
| Retinoblastoma | 3/3 (100%) | 13/13 (100%) | 18/24 (75%) | 3/3 (100%) | 37/43 (86%) |
| 3/37 (8.1%) | 13/37 (35.1%) | 18/37 (48.6%) | 3/37 (8.1%) | ||
| Tx complication | 0 | 0 | 3/24 (12.5%) | 0 | 3/43 (7%) |
| 0 | 0 | 3/3 (100%) | 0 | ||
| Other causes | 0 | 0 | 1/24 (4.2%) | 0 | 1/43 (2.3%) |
| 0 | 0 | 1/1 (100%) | 0 | ||
| Data missing | 0 | 0 | 2/24 (8.3%) | 0 | 2/43 (4.7%) |
| 0 | 0 | 2/2 (100%) | 0 | ||
| Follow-up time (months) | |||||
| Median (IQR) | 11.0 (2.6–39.8) | 30.5 (13.7–34.9) | 35.8 (24.5–40.7) | 35.2 (30.1–39.9) | 34.7 (26.6–39.8) |
| Data availablea | 6/8 (75%) | 55/58 (94.8%) | 203/235 (86.4%) | 184/190 (96.8%) | 448/491 (90.8%) |
Data are n/N (%), except where indicated otherwise. Percentages within the national income level and within the evaluated variable are shown.
Entire cohort has data available
The number of individuals for whom data were available.
Inclusion criterion: 100% reporting.
Abbreviations: IQR - interquartile range; Tx – Retinoblastoma Treatment
Clinical Characteristics at Presentation
Of the cohort, 67.4% of patients (331/491) presented with unilateral disease and 32.6% (160/491) with bilateral disease. The median age at diagnosis was 28.4 months (IQR 0.3 – 140.1 months) for patients presenting with unilateral disease (331/491, 67.4%) and 13.2 months (IQR 0.07 – 48.2 months) for patients presenting with bilateral disease (160/491, 32.6%). 47.3% of patients (232/491) were female and 7.1% (35/490) had familial retinoblastoma. By cTNMH category, 47.9% of patients were cT3 (232/484), 78.1% of patients were N0 (379/485), and 95.1% were M0 (461/485). In terms of heritable trait or the presence of an RB1 germline mutation, 50.6% (246/486) were HX (mutation unknown), 11.3% (55/486) were H0 (normal RB1 allele), and 38.1% (185/486) were H1 (bilateral/trilateral retinoblastoma, positive family history, or germline blood RB1 mutation). Presentation data were available in at least 98.6% of patient cases. The clinical characteristics at presentation, reported by national income level, and data availability, are shown in Table 1B.
Treatment
Enucleation surgery was available for all patients, and intravenous chemotherapy for 99.2% (487/491) of patients (eTable 1 in the Supplement). Detailed treatment data were available on 486 patients (eTable2 in the Supplement). Of those who received treatment, 36.0% (175/486) received intravenous chemotherapy, 13.6% (66/486) received intra-arterial chemotherapy, and primary enucleation was performed in 48.8% (235/486) of cases for 36.6% (238/651) of eyes included in analysis. Treatment refusal was reported in 4.7% (23/486) of patients and palliative treatment was reported in 1.0% (5/486) of patients.
For new tumors or tumor recurrence, additional treatments were represented as follows: 29.6% (144/486) of patients received intravenous chemotherapy, 14.8% (72/486) received intra-arterial chemotherapy, 11.3% (55/486) received intravitreal chemotherapy, 21.8% (106/486) underwent secondary enucleation/exenteration, and 32.7% (159/486) received laser or cryotherapy. Radiotherapy was given to 9.4% (46/486) of patients. Transformation to palliative therapy was reported in 0.4% (2/486) of children, and treatment abandonment was reported in 1.4% (7/486) of patients.
Outcomes
The median follow-up time was 34.7 months (IQR, 26.6–39.8), based on 90.8% (448/491) of reports (Table 1C). No patients who presented with unilateral retinoblastoma were reported to develop bilateral disease.
Survival
Death was reported in 8.8% (43/491) of the patient cohort. The mortality rate by country level is as follows: 37.5% (3/8) of patients from LICs, 22.4% (13/58) from LMICs, 10.2% (24/235) from UMICs, and 1.6% (3/190) from HICs (Table 1C). Of the 43 total deaths in the patient cohort, 86.0% (37/43) were from retinoblastoma with 5.4% (2/37) of these deaths from trilateral disease. Treatment-related complications were the cause of 7.0% (3/43) of deaths, and 2.3% (1/43) were reported as being from other causes; in 4.7% (2/43) of cases, the cause of death was not indicated. 88.4% (38/43) followed a diagnosis of metastatic spread.
Figure 1 shows the Kaplan-Meier survival estimates for the entire cohort (1A), stratified by national income level (1B), and by clinical stage at presentation (1C). For all patients, the one, two and three-year survival rates were 95.1% (95% CI, 92.5–96.8), 92.6% (89.6–94.7) and 91.4% (88.3–93.8) (Figure 1A), respectively. When considering national income level, the survival rate in LICs was 60.0% at one year (95% CI, 12.6–88.2); this rate was maintained at three years. In LMICs, the survival rate declined from 84.7% (95% CI, 71.6–92.0) at one year to 74.2% (59.7–84.2) at three years, and in UMICs survival rate dropped from 94.3% (89.9–96.8) at one year to 89.8% (84.4–93.4) at three years (Figure 1B). In comparison, for HICs the survival rate was 100% at one year, and remained 99.2% (95% CI, 94.6–99.9; Figure 1B) at three years. At three-year follow up, 22.4% of LMIC patients had died, and 10.2% of UMIC patients had died, while only 1.6% of HIC patients had died (Table 1C, Figure 1B). In examining AJCC stage, the survival rate for cT1-cT3 was >93.5% at three years, whereas for cT4 the survival rate was much lower at 48.1% (95% CI, 30.3–63.9) at one year, declining to 32.2% (15.9–49.7) at three years (Figure 1C).
Figure 1.
Survival analysis for the full study cohort, by national income level, and by clinical stage. (A) Kaplan-Meier survival plot for the entire cohort. (B) Kaplan-Meier survival plot by income group. Income Groups: LIC (Low Income Country); LMIC (Lower-Middle Income Country); UMIC (Upper-Middle Income Country); HIC (High Income Country). (C) Kaplan-Meier survival plot by AJCC tumor stage (cT1-cT4). 95% confidence intervals indicated by shaded regions. (D) Table showing one, two, and three year survival by income group and AJCC tumor stage.
Table 2 summarizes the weighted Cox proportional hazard model results for survival. National income level was not significantly associated with survival (P values>0.05), although hazard estimates reflected global results, with decreasing risk of death as a function of increasing income level.8 In adjusted analyses based on hazard ratios, children from LICs carried 3.4 times higher risk of death compared to children from HICs (Figure 1B). Similarly, age at diagnosis was not significantly associated with risk of death in patients age four years and younger (P=0.56), although the hazard model showed a trend of increasing risk with increasing year of life at diagnosis until age 4. In patients diagnosed after age four, risk of death significantly decreased with each additional year of life (HR=0.45 [95% CI, 0.27–0.78], P=0.048 for change in slope). Compared to least advanced disease by AJCC staging (cT1), more advanced stage at diagnosis (cT2, cT3, or cT4) was found to be significantly associated with all-cause mortality, with a graded increase in risk across most categories (cT2 vs. cT1, [HR= 1.1×109 (95% CI, 1.46×108 – 8.26×109), P<0.001]; cT3 vs. cT1, [HR=4.65×109 (1.25×109 – 1.72 × 1010), P<0.001]; cT4 vs. cT1, [HR= 5.98×1010], P>0.05). The mortality rate was highest for patients with extraocular cT4 disease (54.8%), while no cT1s died (P<0.0001 from Fisher’s exact test). Female sex was also found to be associated with an increased hazard of all-cause mortality (vs. male, HR=1.98 [95% CI, 1.27 – 3.10], P=0.04). Familial retinoblastoma history was not significantly associated with survival after model adjustment (HR=11.1 [95% CI, 1.66 – 74.8], P=0.16). Disease laterality and heritability (defined as bilateral or trilateral retinoblastoma, or positive blood RB1 mutation) did not have significant associations with survival. As outlined in the methods, sensitivity analyses were performed, which showed little change in risk estimates from primary analyses.
Table 2.
Summary of the clustered and weighted Cox proportional hazard model for survival*
| Coefficient | Robust standard error | Z score | P value Unadjusted (Corrected†) |
HR (95% CI) | |
|---|---|---|---|---|---|
| Income level of residence | |||||
| Low | Ref | – | – | – | 1.00 |
| Lower-middle | −0.18 | 0.22 | −0.82 | 0.41 (1.00) | 0.83 (0.54 – 1.29) |
| Upper-middle | −0.69 | 0.62 | −1.11 | 0.27 (1.00) | 0.50 (0.15 – 1.69) |
| High | −1.25 | 0.76 | −1.64 | 0.10 (1.00) | 0.29 (0.06 – 1.27) |
| All ages ‡ | |||||
| HR per month | 0.03 | 0.02 | 1.81 | 0.07 (0.56) | 1.03 (1.00 – 1.07) |
| HR per year | 0.41 | 0.23 | 1.81 | 0.07 (0.56) | 1.51 (0.96 – 2.35) |
| Age > 4 years | |||||
| HR per month | −0.07 | 0.02 | −2.89 | 0.004 (0.048) | 0.94 (0.90 – 0.98) |
| HR per year | −0.79 | 0.27 | −2.89 | 0.004 (0.048) | 0.45 (0.27 – 0.78) |
| Laterality | |||||
| Unilateral | Ref | – | – | – | 1.00 |
| Bilateral | 0.52 | 0.36 | 1.46 | 0.14 (1.00) | 1.68 (0.84 – 3.38) |
| Primary tumor | |||||
| cT1 | Ref | – | – | – | 1.00 |
| cT2 | 20.8 | 1.03 | 20.2 | <0.001 (<0.001) | 1.10×109 (1.46×108 – 8.26×109) |
| cT3 | 22.3 | 0.67 | 33.3 | <0.001 (<0.001) | 4.65×109 (1.25×109 – 1.72 × 1010) |
| cT4 | 24.8 | – | – | – | 5.98×1010 (No CI) |
| Sex | |||||
| Male | Ref | – | – | – | 1.00 |
| Female | 0.69 | 0.23 | 3.02 | 0.003 (0.04) | 1.98 (1.27 – 3.10) |
| Family history of retinoblastoma | |||||
| Negative | Ref | – | – | – | 1.00 |
| Positive | 2.41 | 0.97 | 2.48 | 0.01 (0.16) | 11.10 (1.66 – 74.8) |
| Hereditary retinoblastoma § | |||||
| H0 | Ref | – | – | – | 1.00 |
| H1 | 0.26 | 0.45 | 0.58 | 0.56 (1.00) | 1.30 (0.54 – 3.13) |
HR= hazard ratio
Overall, 43 observations were dropped from survival analysis because of missing observation time.
Corrected using Bonferroni method (multiplied by 12 for each model term).
Age included in analysis as a continuous variable.
Hereditary refers to bilateral or trilateral retinoblastoma, positive family history, or positive blood RB1 mutation. H0= non-hereditary, H1= hereditary
Metastasis
Distant metastasis at three-year follow-up was reported in 10.2% (50/491) of patients, 36% of these patients were diagnosed with cT3 disease (18/50). Of the patients with metastatic disease, 10.0% (5/491) were confirmed alive at three years. The median time of primary tumor diagnosis to metastasis was 10 months (IQR 0–57 months) based on 44.0% (22/50) of patients. Average time between diagnosed metastases and most recent follow up was 36 months (± 4.95 months) based on 40.0% (2/5) of those surviving patients with metastatic disease.
Enucleation
Of the study cohort, 68.6% (337/491) underwent enucleation (Table 1C). Both eyes were enucleated in 3.7% (18/491) of patients. For all patients with available follow-up data, the one-, two-, and three-year cumulative incidence of enucleation was 67.6% (95% CI, 63.2–71.9), 71.2% (66.9–75.3), and 72.8% (68.6–77.0), respectively. Enucleation was the primary form of treatment for 48.8% (237/486) of patients and was secondarily performed in 20.6% (100/486) of patients.
Figure 2 shows the cumulative incidence of enucleation obtained from adjusted models for the entire cohort (2A), stratified by national income level (2B), and by clinical stage at presentation (2C). When considering national income level for patients with available follow-up data, the enucleation rate at three years was 77.8% (95% CI, 38.5–99.0) for LIC patients, 86.7% (74.4–94.9) for LMIC patients, 85.7% (80.5–90.1) for UMIC patients, and 53.8% (46.7–61.2) for HIC patients. By AJCC stage, the enucleation rate at three years was 29.9% (95% CI, 18.8–45.4) for cT1 disease, 59.0% (51.3–66.9) for cT2 disease, 91.1% (86.7–94.5) for cT3 disease, and 88.1% (64.3–98.8) for cT4 disease.
Figure 2.
Cumulative incidence of enucleation and competing risk of death for the full cohort, by income level, and by clinical stage. (A) Stacked cumulative incidence plot for entire cohort. (B) Stacked cumulative incidence plots by income group. Income Groups: LIC (Low Income Country); LMIC (Lower-Middle Income Country); UMIC (Upper-Middle Income Country); HIC (High Income Country). (C) Stacked cumulative incidence plots by AJCC tumor stage (cT1-cT4). Note: Lighter color regions (e.g., LIC incidence in 2B before 1 year; cT4 incidence in 2C after 1 year) denote rates that are estimated using the last known values per group, reflecting limited follow-up data. (D) Table showing one-, two-, and three-year enucleation by income group and AJCC tumor stage.
Table 3 summarizes the clustered and weighted Fine and Gray proportional sub-hazard model for enucleation, which also accounts for the competing risk of death. More advanced primary tumor category was associated with increased hazard of enucleation, reflecting global results (e.g., cT3 vs. cT1 Subhazard ratio, SHR=4.98 [95% CI, 2.36–10.5], P<0.001). Children with bilateral retinoblastoma were less likely to have enucleation than children with unilateral disease (SHR=0.62 [95% CI, 0.46–0.84], P=0.02). Although eyes of patients from HICs were less likely to be enucleated (vs. LICs, SHR=0.37 [95% CI, 0.18–0.76], P=0.08), this effect was not significant after adjustment for multiple predictors. Other parameters including sex, familial history, hereditary status, and age at diagnosis were not significant.
Table 3:
Summary of the clustered and weighted Fine and Gray proportional subhazard model for enucleation*
| Coefficient | Robust standard error | Z score | P value Unadjusted (Corrected†) | SHR (95% CI) | |
|---|---|---|---|---|---|
| Income level of residence | |||||
| Low | Ref | – | – | – | 1.00 |
| Lower-middle | −0.27 | 0.26 | −1.04 | 0.30 (1.00) | 0.76 (0.46–1.27) |
| Upper-middle | −0.31 | 0.17 | −1.85 | 0.06 (0.77) | 0.73 (0.53–1.02) |
| High | −0.98 | 0.36 | −2.71 | 0.007 (0.08) | 0.37 (0.18–0.76) |
| All ages ‡ | |||||
| HR per month | −0.27 | 0.26 | −1.04 | 0.66 (1.00) | 1.00 (0.99–1.01) |
| HR per year | 0.03 | 0.06 | 0.44 | 0.66 (1.00) | 1.03 (0.91–1.17) |
| Age > 4 years | |||||
| HR per month | −0.01 | 0.01 | −1.34 | 0.18 (1.00) | 0.99 (0.97–1.01) |
| HR per year | −0.15 | 0.11 | −1.34 | 0.18 (1.00) | 0.86 (0.69–1.07) |
| Laterality | |||||
| Unilateral | Ref | – | – | – | 1.00 |
| Bilateral | −0.48 | 0.15 | −3.08 | 0.002 (0.02) | 0.62 (0.46–0.84) |
| Primary tumor | |||||
| cT1 | Ref | – | – | – | 1.00 |
| cT2 | 0.94 | 0.39 | 2.42 | 0.02 (0.19) | 2.57 (1.20–5.51) |
| cT3 | 1.60 | 0.38 | 4.22 | <0.001 (<0.001) | 4.98 (2.36–10.5) |
| cT4 | 0.76 | 0.39 | 1.95 | 0.05 (0.61) | 2.14 (1.00–4.58) |
| Sex | |||||
| Male | Ref | – | – | – | 1.00 |
| Female | −0.09 | 0.15 | −0.56 | 0.58 (1.00) | 0.92 (0.68–1.24) |
| Family history of retinoblastoma | |||||
| Negative | Ref | – | – | – | 1.00 |
| Positive | −0.92 | 0.36 | −2.57 | 0.01 (0.12) | 0.40 (0.20–0.80) |
| Hereditary retinoblastoma § | |||||
| H0 | Ref | – | – | – | 1.00 |
| H1 | −0.18 | 0.32 | −0.57 | 0.57 (1.00) | 0.83 (0.45–1.56) |
SHR= Subhazard ratio
Overall, 26 observations were dropped from survival analysis because of missing observation time.
Corrected using Bonferroni method (multiplied by 12 for each model term).
Age included in analysis as a continuous variable.
Hereditary refers to bilateral or trilateral retinoblastoma, positive family history, or positive blood RB1 mutation. H0= non-hereditary, H1= hereditary
Impact of the COVID-19 Pandemic on Survival and Globe Salvage
None of the deaths known to have occurred during 2020 (10%, 4/40) and none of the enucleations known to have been performed during this period (2.1%, 9/335) were associated with the pandemic or a pandemic-related delay in treatment.
DISCUSSION
Similar to the global study of retinoblastoma,8 this sub-analysis of outcomes in the Americas demonstrates a disparity in patient survival rates based on the income level of their resident country. The largest gap in survival was seen between children from LICs (60% alive at three-year follow up) and children from HICs (99.2% alive at three years); in adjusted analyses, children from LICs carried 3.4 times higher risk of death compared to children from HICs (Figure 1B, converted from HR). This disparity is smaller than what was reported globally, but this may be due to the nature of the Americas sample. Outcomes for LIC children are based on limited data from a single treatment center in Haiti, where restricted healthcare access may cause disparities in outcomes and reporting.14 Nevertheless, mortality risk was significantly reduced with increasing income level. For example, at three-year follow up, 22.4% of LMIC patients had died, and 10.2% of UMIC patients had died, compared to only 1.6% of HIC patients (Table 1C, Figure 1B).
Mortality was strongly associated with primary tumor stage at diagnosis, which also varied based on the income level of a patient’s home country. In the Americas, 67% of patients from LICs and 24% of patients from LMICs presented with extraocular cT4 disease at diagnosis, while less than 1% of HIC patients presented with advanced cT4 disease (Table 1B). The mortality rate was highest for patients with extraocular cT4 disease (54.8%), while no cT1s died (P<0.0001 from Fisher’s exact test). However, similar to the global analysis, lower income status remained a major risk factor for death independent of the stage at diagnosis. This disparity may exist due to multiple factors including limited availability of certain treatments in LICs.5,8 Limited follow-up data on patients from LICs also impacts survival estimates and interpretability of some model comparisons (e.g., very large HR estimates for all AJCC stages compared to cT1).
Age at diagnosis only predicted survival in older children, which differs from what was seen globally.8 In the Americas sample, a non-significant effect of increasing risk of death was seen for each year until age four, followed by a significant decrease in risk for each additional year older (P=0.048; Table 2). The trend of increasing risk of death in the youngest patients, who were surviving with advanced disease, was seen in both studies, although limited study power reduced significance in the Americas data.8 The trend of decreasing risk in older patients was also seen in both studies, although differences in the number of age categories assessed led to variations in how both studies report this effect.8 Globally, risk of death was stable from ages three to seven (P=0.01), and then decreased (non-significantly) after age seven, while in the Americas risk decreased significantly after age four.8 As was hypothesized in the global study,8 patients who were diagnosed at an older age may have had lesions which existed in the benign retinoma stage for longer than those lesions diagnosed in younger patients which may explain this finding. Notably, age at diagnosis was unrelated to enucleation risk in the Americas, although this trend was observed globally.
Female sex (HR=1.98, P=0.04) was associated with increased risk of all-cause mortality in the Americas, unlike the global study, which showed no effect. Mortality risk associated with female sex has been reported in other studies of retinoblastoma outcomes by our research team,15 where the increased risk to females may be associated with preferential treatment of male children in some countries as opposed to a biological mechanism. Further studies examining impact of sex on mortality in retinoblastoma patients are warranted globally.
Overall, 68.6% of patients in the Americas required enucleation; 48.8% primarily and 20.6% secondarily. Disparities in enucleation rates as a function of income were observed in the Americas, as illustrated by three-year salvage rates of 13.3% (95% CI, 5.1–25.6) in LMICs and 46.2% (38.8–53.3) in HICs (Figure 2B and 2D). Yet, the effect of income was not statistically significant in hazard models of enucleation globally or in this sub-analysis after adjustment for multiple predictors.8 Lack of access to care and treatment abandonment, especially among indigenous communities in Central American LMICs, may explain this disparity.7,16 Additional data from patients from LICs in the Americas are needed to produce stable estimates of mortality and enucleation hazard in this group.
In the larger global analysis, eyes at the lowest AJCC stage (cT1) were far less likely to be enucleated, and risk was highest for cT3 eyes, followed by cT4 and then cT2.8 Data collected from the Americas showed the same pattern, where all clinical status levels showed an increased risk for enucleation compared to cT1 (vs. cT2: HR=2.57 [95% CI, 0.46–1.27]; vs. cT3: HR=4.98 [0.53–1.02]; vs. cT4: HR=2.14) [0.18–0.76], although only the comparison between cT1 and cT3 was statistically significant after adjustment (P<0.001). AJCC stage cT3 eyes were the least likely to be salvaged (8.9% [95% CI, 5.5–13.3]), much like what was observed globally. In the Americas, eyes with stage cT4 disease (salvage rate, 11.9% [95% CI, 1.2–35.7], after one year) showed significantly reduced incidence compared to cT3 (P=0.007, unadjusted Wald test). Due to small sample size and limited follow-up data, globe salvage rates of cT4 eyes did not significantly differ from cT1 (70.1% [54.5–81.2] salvaged at three years) or cT2 cases (41.0% [33.1–48.7] salvaged at three years).
This study has many strengths. This sub-analysis is important because it is the first study of this magnitude to assess retinoblastoma outcomes specifically in the Americas. As such, these results have the unique ability to inform future clinical practices within these particular regions. Further sub-analyses of individual regions in the Americas are ongoing. This prospective study employed the same clustering and weighting methodology utilized in the analysis of global data, and many of the same sensitivity analyses were conducted, suggesting our findings are robust with respect to American retinoblastoma patients. However, limited data from LICs, which were represented by only eight patients from one country, suggest that additional data may be needed to reliably estimate risk for the most vulnerable patients. Although some hazard ratios were not statistically significant (Table 2, 3), trends in overall survival and enucleation data by national income level mirrored those of the global analysis (Figure 1, 2).8 Cohort size and geographical spread may have impacted the data, collection of treatment data was limited to treatment type or refusal, and COVID-19 impact data was limited to a caregiver survey.
In conclusion, major inequities exist in survival and globe salvage rates for retinoblastoma patients based on income status in the Americas. Overall, enucleation remains the most frequent treatment for retinoblastoma. Retinoblastoma patients from LICs are more likely to present with extraocular disease and have 3 times higher risk of death than those from HICs. Successful globe salvage is also three times more likely in HICs than LICs; cT1 eyes are five times more likely to be salvaged than cT3 eyes. Trends in this sub-analysis mirror those of the larger global study. Unique to this sub-analysis, females in the Americas with retinoblastoma are at 2 times higher risk of death compared to males. Our study reinforces the importance of international support in building high-quality childhood cancer programs for lower income American countries to ensure early diagnosis and treatment.
Supplementary Material
Supplemental Figure 1. The 23 American countries and associated number of patients and treatment centers included in analysis, categorized by income level (red= low income, orange= lower-middle income, blue= upper-middle income, green= high income). Estimated cases determined based on crude birth rate per each country’s population and retinoblastoma incidence of 1 in 17,000.11 Color map generated using mapchart.net (www.mapchart.net/americas.html).
Funding/Support:
This research was funded in part by the Queen Elizabeth Diamond Jubilee Trust. The funder had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication. The funder assisted in IRB application fees for selected retinoblastoma centers from low-income countries.
Financial Disclosures:
JLB has support from the National Cancer Institute of the National Institute of Health Award Number K08CA232344, The Wright Foundation, Children’s Oncology Group/ St. Baldrick’s Foundation, Danhakl Family Foundation, The Knights Templar Eye Foundation, A. Linn Murphree, MD, Chair in Ocular Oncology, The Berle & Lucy Adams Chair in Cancer Research, The Larry and Celia Moh Foundation, and an unrestricted departmental grant from Research to Prevent Blindness. No other authors have financial disclosures to report.
Other Acknowledgements:
This study follows the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER). The study data will become available online once all analyses are complete.
Footnotes
Declaration of Interest Statement
We declare no competing interests relevant to the present study.
REFERENCES
- 1.Kivela T The Epidemiological Challenge of the Most Frequent Eye Cancer: Retinoblastoma, an Issue of Birth and Death. Br J Ophthalmol. 2009;93(9):1129–1131. [DOI] [PubMed] [Google Scholar]
- 2.Tomar AS, Finger PT, Gallie B, et al. Global Retinoblastoma Treatment Outcomes: Association with National Income Level. Ophthalmology. 2021;128(5):740–753. doi: 10.1016/j.ophtha.2020.09.032 [DOI] [PubMed] [Google Scholar]
- 3.Truong B, Green AL, Freidrich P, Ribeiro KB, Rodriguez-Galindo C. Ethnic, Racial, and Socioeconomic Disparities in Retinoblastoma. JAMA Pediatr. 2015;169(12):1096–1104. [DOI] [PubMed] [Google Scholar]
- 4.Chung CY, Medina CA, Aziz HA, Singh AD. Retinoblastoma: evidence for stage-based chemotherapy. Int Ophthalmol Clin. 2015;55(1):63–75. doi: 10.1097/IIO.0000000000000054 [DOI] [PubMed] [Google Scholar]
- 5.Fabian ID, Abdallah E, Abdullahi SU, et al. Global Retinoblastoma Presentation and Analysis by National Income Level. JAMA Oncol. 2020;6(5):685–695. doi: 10.1001/jamaoncol.2019.6716 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Wong JR, Tucker MA, Kleinerman RA, Devesa SS. Retinoblastoma Inceidence Patterns in the US Surveillance, Epidemiology, and End Results Program. JAMA Ophthalmol. 2014;132(478–483) [DOI] [PubMed] [Google Scholar]
- 7.Luna-Fineman S, Barnoya M, Bonilla M, Fu L, Baez G, Rodriguez-Galindo C. Retinoblastoma in Central America: Report From the Central American Association of Pediatric Hematology Oncology (AHOPCA). Pediatr Blood Cancer. 2012;58:545–550. doi: 10.1002/pbc.23307 [DOI] [PubMed] [Google Scholar]
- 8.Global Retinoblastoma Study Group. The Global Retinoblastoma Outcome Study: a prospective, cluster-based analysis of 4064 patients from 149 countries. The Lancet Global Health. 2022;10(8):e1128–e1140. doi: 10.1016/S2214-109X(22)00250-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Vandenbroucke JP, von Elm E, Altman GA, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): Explanation and elaboration. Int J Surg. 2014;12:1500–1524. [DOI] [PubMed] [Google Scholar]
- 10.Stevens GA, Alkema L, Black RE, et al. Guidelines for Accurate and Transparent Health Estimates Reporting: the GATHER statement. Lancet. 2016;388:19–23. [DOI] [PubMed] [Google Scholar]
- 11.United Nations DoEaSA. World Population Prospects: the 2017 Revision. Volume I: Comprehensive Tables. https://population.un.org/wpp/Publications/Files/WPP2017_Volume-I_Comprehensive-Tables.pdf. Accessed April 15, 2023. [Google Scholar]
- 12.Fine J, Gray R. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc. 1999:496–509. [Google Scholar]
- 13.Mallipatna AC, Gallie BL, Chévez-Barrios P, et al. Retinoblastoma. In: Amin MB, Edge SB, Greene FL, et al. eds. AJCC Cancer Staging Manual. 8th ed. New York, NY: Springer; 2017. [Google Scholar]
- 14.Gage AD, Leslie HH, Bitton A, et al. Assessing the quality of primary care in Haiti. Bull World Health Organ. Mar 01 2017;95(3):182–190. doi: 10.2471/BLT.16.179846 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Fabian ID, Khetan V, Stacey AW, et al. Sex, gender, and retinoblastoma: analysis of 4351 patients from 153 countries. Eye (Lond). Aug 2022;36(8):1571–1577. doi: 10.1038/s41433-021-01675-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Dean M, Bendfeldt G, Lou H, et al. Increased incidence and disparity of diagnosis of retinoblastoma patients in Guatemala. Cancer Lett. 2014;351(1):59–63. doi: 10.1016/j.canlet.2014.04.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental Figure 1. The 23 American countries and associated number of patients and treatment centers included in analysis, categorized by income level (red= low income, orange= lower-middle income, blue= upper-middle income, green= high income). Estimated cases determined based on crude birth rate per each country’s population and retinoblastoma incidence of 1 in 17,000.11 Color map generated using mapchart.net (www.mapchart.net/americas.html).