Children with cancer in middle-income countries have inferior outcomes compared with similar children in high-income countries. The magnitude and drivers of this survival gap are not well understood. In the current report, the authors sought to describe patterns of clinical presentation, magnitude of treatment abandonment, and survival in children with sarcoma in Central America.
A retrospective review was conducted of hospital-based registries from national pediatric oncology referral centers. Patients with newly diagnosed osteosarcoma, Ewing sarcoma, rhabdomyosarcoma (RMS), and soft tissue sarcoma (STS) between January 1, 2000 and December 31, 2009 were included. Survival analyses were performed first using standard definitions of overall survival (OS) and event-free survival (EFS) and then with abandonment included as an event (abandonment-sensitive OS and abandonment-sensitive EFS).
In total, 785 new cases of pediatric sarcoma were reported (264 diagnoses of osteosarcoma, 175 diagnoses of Ewing sarcoma, 240 diagnoses of RMS, and 106 diagnoses of STS). The rate of metastatic disease at presentation was high (osteosarcoma, 38%; Ewing sarcoma, 39%; RMS, 29%; and STS, 21%). The treatment abandonment rate also was high, particularly among patients with extremity bone sarcomas (osteosarcoma, 30%; Ewing sarcoma, 15%; RMS, 25%; and STS, 15%). Of 559 patients who experienced a first event, 59% had either recurrent or progressive disease. The 4-year OS rate (±standard error) was 40% ± 3%, and the EFS rate was 30% ± 2%; however, these rates decreased further to 31% ± 2% and 24% ± 2%, respectively, when abandonment was taken into account.
Eighty percent of children with cancer live in low-income and middle-income countries (LMCs).1 In developed countries, childhood cancer survival has steadily improved, reaching 81% for all children with cancer and 70% for children with sarcoma.2, 3 The improved survival of patients with pediatric sarcoma in high-income countries (HICs) reflects improved access to chemotherapy, surgery, and radiation as well as better diagnostic techniques, risk-adapted approaches, and optimization of supportive care and palliation.4 In developing countries, advances have been hampered by low socioeconomic conditions, which result in delayed diagnosis and fractionated care. Information on clinical presentation, epidemiology, delivery of cancer care, and outcomes among children with sarcoma in LMCs is scarce5-8; a detailed analysis of all these elements is needed.
International organizations have called for action to address cancer in the developing world.9-11 Great variability in cancer outcomes has been documented not only between developed and developing countries but also between countries with similar government investment in health care,12, 13 highlighting the need to improve our understanding of regional disparities and to develop new models of care for children with cancer in LMCs. Successful models of care for the treatment of childhood leukemia in resource-limited settings have been developed.14 However, an expanded model that accounts for the complexity and multidisciplinary nature of solid tumor care is critically needed. Pediatric sarcomas collectively represent only between 6% and 16% of pediatric cancers2, 3; however, because of the complexity of diagnostic and treatment, they can serve as a benchmark to study health care systems and the influence of quality-improvement or multidisciplinary care paradigms on cancer outcomes in LMCs.
The Central American Association of Pediatric Hematologists and Oncologists (AHOPCA) is a regional collaboration that has successfully implemented protocol-based treatment, studied causes of treatment abandonment, and improved survival in childhood leukemia in the region.15, 16 However, despite parallel development of pediatric leukemia and solid tumor programs, improvements in outcomes for children with solid tumors have not always matched those for children with pediatric leukemia.
In Guatemala, the implementation of a soft tissue sarcoma protocol based on results from Intergroup Rhabdomyosarcoma Study IV (IRS-IV) proved deliverable; yet, despite access to chemotherapy, surgery, and radiation, the outcomes reported in developed countries were not reproducible in this lower income setting.8, 17 Similar challenges in matching published outcomes have been documented for the treatment of rhabdomyosarcoma in Turkey18; for the treatment of osteosarcoma in Brazil19; and for the treatment of other solid tumors in Nigeria, Morocco, and South Africa.20-22 The objective of our current study was to describe patterns of clinical presentation, magnitude of treatment abandonment, and survival in children with sarcoma in Central America to document and guide the development of focused initiatives aimed at addressing the expected survival gap.
MATERIALS AND METHODS
A retrospective review was conducted of previously collected, deidentified data obtained from hospital-based registries at 7 pediatric oncology centers in Central America. Based on the nature of the data, this study was reviewed and was considered exempt from further review by the Office of Human Subject Research Studies at the Dana Farber Cancer Institute.
Six of the 7 pediatric oncology centers involved were pediatric cancer national referral centers and members of AHOPCA: National Children's Hospital, Costa Rica; Benjamin Bloom National Children's Hospital, El Salvador, National Pediatric Oncology Unit, Guatemala; Maternity and Children's Hospital, Honduras; “La Mascota” Children's Hospital, Nicaragua; and Children's Hospital of Panama, Panama. These centers serve the general population, including children from indigenous backgrounds, economically constrained households, and distant locations. Data from the Pediatric Specialties Hospital, the second largest center in Panama, also were included.
Data from hospital-based cancer registries are stored in the Pediatric Oncology Networked Database (POND4KIDS).23 Data collection and entry were approved previously by local ethics committees. Data have been prospectively collected since 2005 and were collected retrospectively for cases diagnosed before 2005. Data from the Pediatric Specialties Hospital in Panama were provided in an Excel format without linking identifiers.
Patients with newly diagnosed osteosarcoma, Ewing sarcoma, rhabdomyosarcoma (RMS), and soft tissue sarcomas (STS) between January 1, 2000 and December 31, 2009 were included. Patients who had sarcoma as a secondary cancer were excluded. An age cutoff was not pre-established, and all patients who presented to the pediatric oncology center and were entered into the database were analyzed.
Patients had received multimodal treatment on a variety of published or regionally adapted protocols. Standard chemotherapy agents, pediatric surgeons, radiation therapy, and subsidies for pediatric cancer care were available for all countries. For example, RMS treatment was based on IRS-IV; and osteosarcoma treatment was based on amputation and chemotherapy with cisplatin, doxorubicin, ifosfamide, and etoposide. Treatment with limb-sparing procedures and high-dose methotrexate were only reliably available in Costa Rica.
Costa Rica is the only AHOPCA-member country with a national cancer registry. Therefore, the pediatric sarcoma incidence reported in Costa Rica (9.7 cases per million)24 for the period from 1984 to 1992 is the best regional estimate available. In the United States, the reported incidence is higher (16 cases per million).25 On the basis of the population aged <15 years of age (14.8 million in 2009), between 143 and 237 new cases of pediatric sarcoma would be expected per year in the region during the analytic period. Because the incidence of sarcoma increases during adolescence, the number expected to present to a pediatric center may be higher and depends on the maximum age allowed at the center.
Outcomes and Important Definitions
AHOPCA centers have a shared agreement on cancer outcomes reported into POND4KIDS: treatment refusal, treatment abandonment, relapse, progressive disease, secondary malignancy, lost to follow-up and death.26Treatment refusal and treatment abandonment are not usually included in the analysis of cancer outcomes in HICs, but these are noteworthy outcomes in LMCs.27Treatment refusal involves the upfront decline of treatment and does not allow initiation of therapy. Treatment abandonment occurs after the initiation of treatment and has been defined in AHOPCA as missing 4 or more consecutive weeks of treatment. Refusal and abandonment were analyzed as 1 event (“abandonment”), as recommended by the International Society of Pediatric Oncology Working Group on Treatment Abandonment.28Date of abandonment was defined as the last date that the patient was known to participate in cancer care. Palliative care was considered continued care until death and was not counted as abandonment. Patient loss to follow-up after completion of therapy not counted as abandonment.
Subdivision of Data
To evaluate changes over time, patients were divided into 2 treatment eras (2000-2004 and 2005-2009) based on the existence of more mature twinning, more established psychosocial strategies and socioeconomic supports, and improved access to multimodal therapy in the second era. To evaluate the role of disease extent on outcomes, patients were divided based on the presence or absence of metastases at diagnosis into those with localized or metastatic disease. To evaluate the role of disease location on outcomes, patients were divided according to 2 disease locations: bone sarcoma (osteosarcoma and Ewing sarcoma of bone located in extremity or other [skull, chest wall, spine, or pelvis]) and soft tissue sarcoma (extraosseous Ewing sarcoma, RMS, and STS in the extremity or other [located elsewhere]).
Descriptive statistics were used to describe patient characteristics. The chi-square test was used to compare proportions. Cumulative incidence of abandonment curves were generated, and the Pepe-More test was used to determine statistically significant differences between the curves (P < .05). Four survival analyses were performed: 1) event-free survival (EFS), in which the events were defined as relapse, progressive disease, secondary malignancy, or death; 2) abandonment-sensitive EFS (AEFS), in which the events were defined as abandonment of therapy, relapse, progressive disease, secondary malignancy, or death; 3) overall survival (OS), in which the event was defined as death, and 4) abandonment-sensitive OS (AOS), in which the events were defined as abandonment and death. The time to an event was calculated from the time of diagnosis until the first event or the last contact if no event occurred. Survival results are presented as 4-year point estimates ± standard errors based on the stability of the curves at that time point. Kaplan-Meier curves were generated, and a log-rank test was used to test for statistically significant differences between the curves. P values < .05 were considered significant.
Seven hundred ninety-one children were diagnosed with sarcoma in the 7 pediatric oncology centers between January 1, 2000 and December 31, 2009. Six patients (0.7%) were excluded because their sarcoma represented a second malignancy. Among the children who were included in the analysis, 323 (43%) were diagnosed between 2000 and 2004, and 458 (61%) were diagnosed between 2005 and 2009 (4 patients had data missing). The patients were from Costa Rica (115 patients), El Salvador (102 patients), Guatemala (259 patients), Honduras (94 patients), Nicaragua (151), and Panama (53 from the Children's Hospital of Panama and 11 from the Pediatric Specialties Hospital). Of the 785 patients who were included in the analysis, 182 (23%) abandoned treatment. A summary of patient characteristics is presented in Table 1 according to abandonment status. The median age of the cohort was 10.2 years (range, 0.1-23.6 years), and the mean age and age range were similar between the countries analyzed.
Table 1. Patient Characteristics by Treatment Abandonment Status
Abandonment Status, N = 785
Patients Who Did Not Abandon Treatment, n = 603
Patients Who Abandoned Treatment, n = 182
Abbreviations: SD, standard deviation.
Total no. of patients
Age: Mean±SD, y
Metastasis at diagnosis, %
Location in extremity, %
Total no. of patients
Age: Mean±SD, y
Metastasis at diagnosis, %
Primary bone, extremity, %
Extraosseous, extremity, %
Total no. of patients
Age: Mean±SD, y
Metastasis at diagnosis, %
Surgical group, %
Tumor histology, %
Location in extremity
Soft tissue sarcoma
Total no. of patients
Age: Mean±SD, y
Metastasis at diagnosis, %
Location in extremity, %
Burden of Disease
A high proportion of patients in these LMCs presented with metastatic disease (osteosarcoma, 38%; Ewing sarcoma, 39%; RMS, 29%; and STS, 21%) compared with patients in HICs (20%, 25%, 18%, and 15%, respectively).17, 29-31 A decrease in metastatic disease was noted by era for all diagnoses, but the trend was not statistically significant (osteosarcoma, from 39% to 38%; Ewing sarcoma, from 42% to 37%; RMS, from 33% to 26%; and STS, from 25% to 19%). Only 17% of RMS tumors were classified as surgical group I and II (compared with 33% in HICs).17 According to histology, the proportion of alveolar subtype RMS in LMCs was similar to that in HICs (30% vs 32%, respectively).17 However, an analysis by country revealed a higher rate of ARMS in Guatemala compared with the other countries (48% in Guatemala vs 17% in El Salvador, 22% in Nicaragua, 27% in Costa Rica, 28% in Honduras, and 29% in Panama).
Twenty-three percent of patients abandoned treatment. The abandonment rate differed significantly by diagnosis (P = .0094) and was higher for osteosarcoma and RMS (30% and 25%, respectively) than for Ewing sarcoma and STS (15% each). The abandonment rate decreased between the 2 eras (P = .0193), but the decrease was significant only for RMS and STS (P = .011 and P = .0411, respectively) (see Table 2). The cumulative incidence of abandonment was not significantly different by era (P = .2312) (Fig. 1A) or disease extent (P = .4760) (Fig. 1B). Pairwise comparisons between diagnoses revealed a higher cumulative incidence of abandonment for osteosarcoma compared with Ewing sarcoma (P = .0008) or STS (P = .0028), but it was not significantly different from RMS (Fig. 1C). The cumulative incidence of abandonment in the first 6 months was steeper for osteosarcoma compared with all other diagnoses, including RMS. Furthermore, 54% of patients with osteosarcoma and 56% of patients with Ewing sarcoma who abandoned therapy did so in the first 3 months after diagnosis compared with 30% of patients with RMS and 41% of patients with STS.
Table 2. Proportion of Patients Who Abandoned Treatment by Disease and Treatment Era
Disease Type by Treatment Era
No. of Patients
Patients Who Abandoned Treatment, %
Soft tissue sarcoma
Pairwise comparisons by disease location revealed a higher cumulative incidence of abandonment for patients with bone sarcoma located in the extremity compared with other locations (P = .0006) (Fig. 1D). For sarcomas that arose in soft tissues, the cumulative incidence of abandonment was not significantly different by disease location in the extremity versus elsewhere (P = .6880). The cumulative incidence of abandonment in the first 6 months was steeper for patients with bone sarcoma in the extremity than for tumors located elsewhere. Fifty-three percent of patients with bone sarcoma located in the extremity who abandoned therapy did so in the first 3 months after diagnosis compared with 45% of those with soft-tissue sarcoma located in the extremity, 34% of those with soft-tissue sarcoma located elsewhere, and 22% of those with bone sarcoma located elsewhere.
Of 559 patients who experienced a first event, 329 (59%) experienced progressive disease or relapse, 161 (29%) abandoned therapy, 68 (12%) died, and 1 developed a secondary malignancy. The median follow-up for patients who remained alive at last contact was 2.2 years (range, from 0 days to 10.7 years). Of all 785 patients, 8 were excluded from the EFS analysis, and 7 were excluded from the OS analysis because of missing data.
Figure 2 provides an overview of survival analyses for pediatric sarcoma in Central America. The estimated 4-year OS and EFS rates were 40% ± 3% and 30% ± 3%, respectively. With abandonment included as an event, the estimated AOS and AEFS rates decreased to 31% ± 2% and 24% ± 2%, respectively. The 4-year OS estimates by disease were 31% ± 4% for osteosarcoma, 36% ± 6% for Ewing sarcoma, 44% ± 5% for RMS, and 53% ± 8% for STS. By disease location, the 4-year OS estimates were 32% ± 4% for bone sarcomas and 46% ± 4% for soft-tissue sarcoma. In contrast, the 5-year survival rate in HICs is 68% for malignant bone tumors and 73% for soft tissue sarcomas.3
Significant differences in survival were noted by disease extent, diagnosis, and disease location, but not by treatment era, even for patients with localized disease (see Table 3 and Fig. 3). Assuming a negative outcome after abandonment, failing to include it as an event would have overestimated survival by 3% to 14% (ie, OS vs AOS and EFS vs AEFS). A difference ≥10% between OS and AOS was noted for localized disease, RMS, and STS. No differences ≥10% were observed between EFS and AEFS.
Survival differed significantly by diagnosis. Patients with osteosarcoma and Ewing sarcoma experienced the poorest outcomes. In patients with osteosarcoma, this was particularly strong when abandonment was included as an event (AOS for osteosarcoma vs RMS, P = .0007; vs STS, P = .0011; vs Ewing sarcoma, P = .0669). Disease location in bone versus elsewhere was associated with decreased survival, particularly when abandonment was taken into account (AOS for extremity vs other, 28% vs 35%; P = .0239; AEFS for extremity vs other, 22% vs 27%; P = .0376).
Analysis by Country
Treatment abandonment and survival differed by country. Details of this analysis and correlations with cancer-infrastructure and health and economic indicators will be reported separately. In brief, the treatment abandonment rate ranged from 1% in Costa Rica to 38% in Honduras, and the AOS rate ranged from 23.5% in Panama to 47.1% in Costa Rica.
This retrospective review of pediatric sarcoma outcomes in Central America documents a significant survival gap despite improved access to multimodal therapy and indicates that a high burden of metastatic disease at diagnosis, a high rate of treatment abandonment, and difficulty with upfront treatment effectiveness are likely drivers for the poor outcomes observed. To our knowledge, this study reviews the largest cohort of pediatric sarcoma cases from the developing world and provides relevant information for comprehensive cancer survival analysis in developing countries.
Because of a lack of regional cancer registries, the generalizability of our cohort cannot be fully ascertained. However, based on reported cancer incidence from Costa Rica, the cohort may represent up to 66% of the expected cases, particularly during the second analytic period. Therefore, our results are likely the closest approximation to the true overall clinical presentation and outcomes of children with pediatric sarcoma in Central America.
One interesting variation in disease distribution was noted; although RMS histology for our LMC cohort was similar compared with that in HICs, the proportion of alveolar RMS was notably higher (48%) in Guatemala. This supports prior observations of increased incidence of retinoblastoma, decreased incidence of neuroblastoma, and a high proportion of alveolar RMS in Guatemala.8, 32 Comprehensive etiologic studies merging cancer epidemiology and tumor biology in LMCs are lacking, and potential genetic and environmental modifiers of tumor behavior have not been fully explored. Studies that confirm pathologic diagnostic accuracy followed by studies that integrate genetics, environment, and cancer biology would be of great value to the region.
A high proportion of patients presented with metastatic disease. In patients with sarcoma, advanced disease translates into higher upfront risk-group assignment, more chemotherapy, difficulty achieving complete surgical resection, and increased reliance on radiation therapy to achieve adequate local control. Therefore, advanced disease has an impact on outcomes and imposes further pressure on already overburdened health care systems. Advanced disease at presentation is most often attributed to delayed diagnosis, and this is the leading argument for national media and educational campaigns. The short-term impact of such programs has been documented in Honduras, where a retinoblastoma education program linked to a national vaccination campaign increased referrals and decreased presentation with extraocular spread.33 Bundled strategies that educate child advocates, the general public, and health care providers on the referral process, treatment availability, and improved curability of low-stage disease have the potential to impact this driver of poor outcomes and merit further support.
Central America has pioneered research on the delivery of pediatric cancer care in resource-limited settings and has greatly improved survival and treatment abandonment for children with leukemia.15, 16 However, treatment abandonment in pediatric sarcoma remains a concern, with rates consistently higher for pediatric sarcoma (average, 20%) than for leukemia (average, 6%) in recent years (AHOPCA; unpublished data provided by Dr. Baez).
The current study demonstrates that, although the proportion of patients with pediatric sarcoma who abandon therapy has decreased over time, these patients remain at high risk for treatment abandonment. Also, the failure to include treatment abandonment as an event leads to the overestimation of OS and EFS.
Relevant socioeconomic and demographic drivers of abandonment have been described previously.15, 27, 34 Our current study focuses on the clinical characteristics that influence abandonment. The occurrence and timing of abandonment appear to be mostly influenced by the primary diagnosis and disease location, and not by age, sex, or disease extent. Patients with bone tumors of the extremities (particularly osteosarcoma) abandon treatment most frequently and do so earlier than patients with other types of sarcoma, presumably reflecting a rejection of amputation. Unfortunately, because of the high cost of endoprosthesis, a lack of bone allografts, and limited skilled personnel, amputation is the primary modality available for local control in all countries except Costa Rica. Strategies for strengthening the rehabilitation and reintegration programs for children who have undergone amputation and possibly increasing access to limb-sparing procedures may have a great impact.
Patients who have RMS share a similar long-term incidence of abandonment with patients who have osteosarcoma; however, abandonment accumulates over a longer period, suggesting that the drivers of early and late abandonment for bone and soft-tissue tumors are different. All in all, these findings suggest that abandonment is driven by factors beyond sociodemographics, that they may be possibly disease/treatment specific, and that the development of retention models designed for these high-risk patient populations is needed.
However, metastatic disease and abandonment do not fully explain the poor long-term outcomes observed. Survival did not improve significantly between eras, even for patients with localized disease and despite improved access to conventional therapy. Furthermore, 42% of all patients experienced progressive disease or relapse as their first event, suggesting fundamental difficulty in the effective delivery of upfront therapy. Focus on capacity building, quality improvement, communication, and coordination of multidisciplinary care is warranted to maximize upfront cancer care effectiveness and outcomes. This is particularly relevant for the management of solid malignancies, in which sophisticated local control measures are at the core of therapy. Successful restructuring of care to effectively deliver complex, frontline therapy has been reported for the treatment of nonmetastatic Ewing sarcoma35 and osteosarcoma36 in Chile.
In conclusion, we have identified a high burden of metastatic disease, a persistently high rate of treatment abandonment, and difficulty with upfront treatment effectiveness as major barriers to the optimal care of children with sarcoma in Central America. Although advances in care delivery for children with hematologic malignancies have resulted in significant improvements in outcomes, children with sarcomas have not benefited from those gains, likely because of deficiencies in the delivery of multidisciplinary care that is key in the management of solid malignancies. Several initiatives are ongoing to address these issues. The Pan-American Health Organization is working regionally on general and health care provider education using the Integrated Management of Childhood Illness method. Collaboration with major academic centers in the United States, Italy, and Brazil is allowing for further education and capacity-building opportunities in limb-salvage surgery as well as early referral of selected patients for limb-sparing procedures. Finally, web-based case discussions, detailed infrastructure assessment, quality-control initiatives, assessment of local control effectiveness, and further analysis of inherent drivers of abandonment in pediatric sarcoma are the focus of the current AHOPCA Sarcoma Initiative and the core of our future studies.
We gratefully acknowledge support from the Dana-Farber/Children's Hospital Cancer Center Global Health Initiative and the St. Jude Children's Research Hospital International Outreach Program, and we thank the Central American Association of Pediatric Hematologists and Oncologists data managers and clinical staff for their inspiring work.
Dr. Friedrich was supported by Pathophysiology of Human Blood Cells (T32 HL 7574) and Program in Cancer Outcomes Research Training (R25 CA092203), both National Institutes of Health training grants.