Outcome of children treated for relapsed acute lymphoblastic leukemia in Central America




Outcomes for relapsed childhood acute lymphoblastic leukemia (ALL) have not been documented in resource-limited settings. This study examined survival after relapse for children with ALL in Central America.


A retrospective cohort study was performed and included children with first relapse of ALL in Guatemala, Honduras, or El Salvador between 1990 and 2011. Predictors of subsequent event-free survival (EFS) and overall survival (OS) were examined.


There were 755 children identified with relapsed disease. The median time from diagnosis to relapse was 1.7 years (interquartile range, 0.8-3.1 years). Most relapses occurred during (53.9%) or following (24.9%) maintenance chemotherapy, and the majority occurred in the bone marrow (63.1%). Following the initial relapse, subsequent 3-year EFS (± standard error) and OS were 22.0% ± 1.7%, and 28.2% ± 1.9%, respectively. In multivariable analysis, worse postrelapse survival was associated with age ≥ 10 years, white blood cell count ≥ 50 × 109/L, and positive central nervous system status at the original ALL diagnosis, relapse that was not isolated central nervous system or testicular, and relapse < 36 months following diagnosis. Site and time to relapse were used to identify a favorable risk group whose 3-year EFS and OS were 50.0% ± 8.9% and 68.0% ± 8.1%, respectively.


Prognosis after relapsed ALL in Central America is poor, but a substantial number of those with favorable risk features have prolonged survival, despite lack of access to stem cell transplantation. Stratification by risk factors can guide therapeutic decision-making. Cancer 2013. © 2012 American Cancer Society.


Outcomes for children with acute lymphoblastic leukemia (ALL) have greatly improved in recent decades in high-income countries (HIC), where 76% to 86%1 survive without recurrence. When relapse occurs among children with ALL, overall survival (OS) in HIC is generally 15% to 50%, depending on relapse and initial disease features.2-8 In this setting, the most important factors associated with poor OS after relapse include duration of remission prior to relapse (worse for relapses less than 12, 18, or 36 months after diagnosis), bone marrow site of relapse, and T cell disease.2-8

In low- and middle-income countries (LMIC) such as those in Central America, outcomes have been worse than in HIC. A recent study of pediatric ALL in El Salvador demonstrated a 5-year event-free survival (EFS) of 56% among those with standard-risk ALL and 49% among those with high-risk ALL.9 However, in LMIC, the outcomes following relapse and factors that predict subsequent EFS after relapse have not been documented. Knowledge of outcomes following relapse and predictors of relapse are important, because this information may suggest children in whom aggressive retreatment may be considered and others where a palliative approach would be more appropriate. Furthermore, some variables may be unique to the LMIC setting such as relapse following abandonment of therapy.

Our objectives were to describe EFS and OS among children with first relapse of ALL in Guatemala, Honduras, and El Salvador, and to identify predictors of subsequent EFS and OS.


Study Population and Setting

Research ethics approval was obtained from The Hospital for Sick Children, Toronto, Canada, and at each participating center. The patient sample was derived from 4 primary centers in 3 Central American countries, namely Hospital Nacional de Niños Benjamin Bloom in San Salvador, El Salvador; Unidad Nacional de Oncología Pediátrica in Guatemala City, Guatemala; Hospital Escuela in Tegucigalpa, Honduras; and Hospital Rivas in San Pedro Sula, Honduras. These centers and their satellites provide the majority of pediatric oncology care in their respective countries, and typically treat 315 children with ALL per year combined. The patient sample included children with relapsed ALL who were younger than 20 years at initial diagnosis, and who experienced a first relapse of their ALL between September 15, 1990, and April 30, 2011. Exclusion criteria included L3 ALL and induction failure as a first event.

Initial therapeutic regimens varied by site and time period. Prior to 2000, each site used different protocols. Subsequently, the vast majority of patients at each of the 3 countries were treated according to common clinical management protocols: the GHS LLA 2000 from 2000 to 2007 (all sites), and the LLAG-2007 (Guatemala) or the AHOPCA LLA-2008 (El Salvador and Honduras) from 2007 and/or 2008 forward. The GHS LLA 2000 protocol was based on the St. Jude Total XIIIB10 and Total XV11 backbones. Important modifications included the omission of etoposide, changing the high-dose methotrexate to 2 g/m2 for standard-risk and 3 g/m2 for high-risk patients, each infused over 3 hours, and the use of only 2 risk groups (standard and high). Standard-risk patients were defined as those aged 1 to < 10 years at diagnosis who presented with an initial white blood cell count (WBC) of <50 × 109/L, DNA index ≥ 1.16, and the absence of any high-risk features (central nervous system [CNS] or testicular involvement, T cell immunophenotype, M3 bone marrow on day 15, or M2/M3 bone marrow on day 36). The 2007/2008 protocols were both based on the ALL IC-BFM (Berlin-Frankfurt-Münster) 2002 Trial,12 with the major modifications including the omission of etoposide in both the LLAG-2007 and the AHOPCA LLA-2008 trials, the omission of daunorubicin in induction, and omission of a phase 1b block during remission induction (cyclophosphamide, cytarabine, and mercaptopurine) for standard-risk patients in the AHOPCA LLA-2008 study. Both studies used 3 risk groups (standard, intermediate, and high) to stratify patients. Standard-risk was defined as B lineage ALL, patient aged 1 to 6 years, initial WBC < 20 × 109/L, and absence of high-risk features (T cell immunophenotype, age < 1 year at diagnosis, CNS 3 disease, testicular infiltration, day 8 prednisone response with blasts ≥ 1 × 109/L, day 15 M3 marrow, day 33 M2/M3 marrow, high-risk translocation [t(4;11), t(9;11), t(1;19)], or hypodiploidy [< 45 chromosomes]). Intermediate risk was defined as B lineage ALL, with age > 6 years or WBC count ≥ 20 × 109/L, and absence of high-risk features.

When relapse occurred, therapeutic regimens varied considerably. In some cases, patients were managed with palliative intent, whereas in others, curative therapy was pursued using a variety of different relapse regimens. Notably, hematopoietic stem cell transplantation was not available, nor were clofarabine or forms of asparaginase other than native E. coli asparaginase. Management of relapse with palliative intent was defined as initiation of palliative therapy within 1 week of the date of relapse diagnosis.

The care at each of the centers was provided at no financial cost to the families. Accommodation, subsidized transportation, and food were provided free of charge for families that lived far from the hospital in order to reduce abandonment of therapy. To further reduce abandonment, clinic social workers contacted, by telephone and home visits, those families who did not adhere to clinic visits. Funding for these programs was provided by local nongovernmental organizations and international partnerships.

Data Source

The data source was the Pediatric Oncology Networked Database (POND) (www.pond4kids.org/ddd). POND is an online, multilingual clinical database created by the International Outreach Program at St. Jude Children's Research Hospital, Memphis, Tennessee, to meet the needs of oncology centers in LMIC for cancer registration, outcome monitoring, and quality improvement.13 In addition, data can be anonymously shared for research purposes, as was done for this study. The data contained within POND were abstracted from patient charts in real time by trained data managers at each of the sites, and confirmed by the treating oncologists. An audit of POND data from Honduras showed that the accuracy for basic data fields was 99%.14

Outcome and Predictor Variables

The primary outcomes were EFS and OS after relapse. EFS was calculated from the time of first relapse until the next event or date of last contact. Events included subsequent relapse, second malignant neoplasm, treatment refusal, abandonment (at least 4 consecutive weeks of missed appointments during active therapy), or death. OS was calculated based on the time from first relapse until death or last contact.

Potential predictors of survival were derived from variables shown to be associated with outcome in studies of relapsed ALL in the HIC setting, as well as variables considered to be possibly important in the LMIC setting. The following variables at diagnosis were examined: age ≥ 10 years, race, initial WBC ≥ 50 × 109/L, DNA index, B versus T lineage, CNS status, and disease risk category. CNS positivity was defined as CNS 3 status, whereas CNS 1 and CNS 2 were considered CNS-negative. Relapse characteristics included time to relapse, site of relapse, and phase of therapy at relapse. Particular to this setting, abandonment of therapy has affected outcome in Central America,9 and thus abandonment of therapy prior to relapse was also considered. Finally, diagnosis time period was examined, as defined by changes in the common treatment protocols (1990-1999, 2000-2007, and 2008-2011).

Statistical Analysis

Survival was described with the Kaplan-Meier method. The association between factors at initial diagnosis and relapse and subsequent EFS and OS was analyzed using the Cox proportional hazards model. Variables significant in univariate analysis were included in multivariable models. From the variables independently associated with survival in multivariable analyses, a favorable risk group was created that used a minimum number of variables in order to simplify adoption and to be relevant to a larger number of children. All statistical analyses were performed using the SAS statistical program (SAS-PC, version 9.3; SAS Institute, Cary, NC). All tests of significance were 2-sided, and statistical significance was defined as P < .05.


During the study period, there were 759 children with relapsed ALL who were identified. One patient with mature B cell leukemia and 3 who experienced induction failure were excluded, and the remaining 755 patients were analyzed. Demographic characteristics of the patient population can be seen in Table 1. The majority of relapses (63.1%) occurred in the bone marrow. The most common phases of therapy in which relapse occurred was during maintenance (50.5%) and following completion of therapy (24.5%).

Table 1. Demographics of Children With First Relapse of Acute Lymphoblastic Leukemia in Central America (N = 755)
Characteristic at Diagnosis or RelapseValue
  • Abbreviations: CNS, central nervous system; IQR, interquartile range; n, number in each category; N, number at risk and only displayed when missing data; WBC, white blood cell count.

  • a

    Risk category assignment as per protocol, see text for full description.

  • b

    Other includes chloroma, spleen, pleura, ovary, and thorax. Combined means any combination of sites.

Characteristics at Initial Diagnosis 
Median age, y (IQR)7.11 (4.0-10.6)
Male, n (%)472 (62.5)
Race, n (%), N = 746 
 Hispanic223 (29.9)
 Indigenous69 (9.3)
 Caucasian6 (0.8)
 Mixed race448 (60.1)
Initial WBC ≥50 × 109/L, n (%), N = 661168 (25.4)
High DNA index (≥1.16), n (%), N = 47674 (15.5)
Histology, n (%), N = 714 
 B lineage641 (89.8)
 T lineage73 (10.2)
Central nervous system status, n (%), N = 678 
 CNS 1585 (86.3)
 CNS 252 (7.7)
 CNS 341 (6.1)
Risk categorya, n (%), N = 702 
 Standard233 (33.2)
 Intermediate55 (7.8)
 High or very high414 (60.0)
Diagnosis time period, n (%) 
 1990-1999118 (15.6)
 2000-2007501 (66.4)
 2008-2011136 (18.0)
Characteristics at Relapse 
Median years to first relapse (IQR)1.66 (0.8-3.0)
Site of first relapse, n (%), N = 721 
 Bone marrow455 (63.1)
 Central nervous system159 (22.0)
 Testicular31 (4.3)
 Other (extramedullary)b11 (1.5)
 Combined sitesb65 (9.0)
Phase of therapy at first relapse, n (%) 
 Consolidation/intensification76 (10.1)
 Maintenance381 (50.5)
 Follow-up185 (24.5)
 Abandoned113 (15.0)

There were 103 patients managed with palliative intent following first relapse. Among patients who did not experience another event, the median follow-up time was 1.9 years (interquartile range [IQR], 0.6-4.9 years). At 3 years, EFS ± standard error after relapse for the entire cohort was 22.0% ± 1.7% and OS after relapse was 28.3% ± 1.9% (Fig. 1). Death, treatment refusal, and abandonment typically occurred within the first 6 months after relapse (Table 2), but the median time to second relapse, when it occurred as the second event, was 10.5 months (IQR, 5-24 months). A total of 139 patients experienced a second relapse: 118 subsequently died and 21 were alive at the last patient encounter (median follow-up of 7 months; range, 1-46 months). All patients who abandoned therapy after relapse subsequently died. The most frequent cause of death was progressive disease (70.3%). At 3 years, postrelapse EFS and OS for patients treated with curative intent (defined as those who did not start palliative treatment within 1 week of relapse diagnosis) and patients who did not abandon therapy after relapse were 26.2% ± 2.0% and 31.5% ± 2.1%.

Figure 1.

Event-free survival (EFS) and overall survival (OS) are shown for children with relapsed acute lymphoblastic leukemia in Central America.

Table 2. Outcomes Following First Relapse in Children With Acute Lymphoblastic Leukemia in Central America (N = 755)
  1. Abbreviations: IQR, interquartile range n, number experiencing the event; N, total number at risk and only displayed when missing data; SE, standard error.

Event-free survival at 3 y ± SE22.0 ± 1.7
Overall survival at 3 y ± SE28.3 ± 1.9
Subsequent event following first relapse, n (%) 
 Abandonment46 (6.1)
 Refusal of treatment4 (0.5)
 Relapse139 (18.4)
 Death370 (49.0)
 None196 (26.0)
Median years to next event (IQR) 
 Abandonment0.22 (0.09-0.52)
 Refusal of treatment0.45 (0.29-0.57)
 Relapse0.87 (0.45-1.96)
 Death0.19 (0.08-0.44)
 None1.90 (0.58-4.87)
Cause of death, n (%), N = 492 
 Infection96 (19.5)
 Hemorrhage34 (6.9)
 Progressive disease346 (70.3)
 Other16 (3.3)

Factors associated with EFS and OS following a first relapse, in univariate analysis, are illustrated in Table 3. Worse EFS and OS was observed in children who had presented initially with T lineage ALL; age ≥ 10 years; WBC ≥ 50 × 109/L; DNA index < 1.16; positive CNS status; intermediate-, high-, or very high-risk disease; and those diagnosed in the latest time period of therapy. Relapse characteristics associated with worse survival were relapse earlier than 36 months after diagnosis, and relapse site that was not isolated CNS or testicular. Children who experienced a relapse without previous abandonment of therapy did not have worse EFS or OS, compared with those who relapsed after abandonment.

Table 3. Univariate Analysis of Event-Free and Overall Survival for Children With Relapsed Acute Lymphoblastic Leukemia in Central America
 Event-Free SurvivalOverall Survival
CharacteristicHR95% CIPHR95% CIP
  1. Abbreviations: CI, confidence interval; CNS, central nervous system; HR, hazard ratio; REF, reference value; WBC, white blood cell count.

Characteristics at Diagnosis
Age >10 y1.571.32-1.88<.00011.611.33-1.95<.0001
Sex, male0.960.81-1.130.5980.940.79-1.130.532
 Caucasian or mixed raceREF  REF  
Initial WBC      
 <50 × 109/LREF  REF  
 ≥50 × 109/L1.591.31-1.93<.00011.491.20-1.840.0003
DNA index      
 ≥1.16REF  REF  
 B lineageREF  REF  
 T lineage1.461.12-1.900.0051.531.16-2.020.003
Central nervous system status      
 Negative (CNS 1 or CNS 2)REF  REF  
 Positive (CNS 3)1.61.14-2.260.0071.591.08-2.330.018
Risk category      
 Standard riskREF  REF  
 Intermediate risk1.731.19-2.510.0042.061.40-3.020.0002
 High risk/very high risk1.691.39-2.05<.00011.741.41-2.14<.0001
Diagnosis time period      
 1990-1999REF  REF  
Characteristics at Relapse      
Time to relapse <36 mo2.221.79-2.74<.00012.391.90-3.01<.0001
Site of first relapse      
 Isolated CNS or testesREF  REF  
 Not isolated CNS or testes1.611.32-1.96<.00011.891.51-2.37<.0001
Phase of therapy at relapse      
 MaintenanceREF  REF  
Not abandoned prior to relapse1.130.90-1.430.3021.150.90-1.470.263

Variables included in multivariable analysis were age ≥ 10 years at diagnosis, WBC ≥ 50 × 109/L at diagnosis, T lineage histology, CNS positivity at diagnosis, time period of diagnosis, not isolated CNS or testicular relapse, and relapse before 36 months. Two factors were not considered in the multivariable model: DNA index was not included because data were available on too few patients, and risk group was collinear with initial age and WBC.

Table 4 illustrates that in multivariable analysis, independent poor prognostic variables were age ≥ 10 years at diagnosis, initial WBC ≥ 50 × 10/L, initial CNS positivity, relapse before 36 months, and not isolated CNS or testicular relapse. Time period of therapy and T cell immunophenotype were no longer significant in the model.

Table 4. Multivariable Analysis of Event-Free and Overall Survival for Children With Relapsed Acute Lymphoblastic Leukemia in Central America
 Event-Free SurvivalOverall Survival
CharacteristicHR95% CIPHR95% CIP
  1. Abbreviations: CI, confidence interval; CNS, central nervous system; HR, hazard ratio; REF, reference value; WBC, white blood cell count.

Age at diagnosis >10 y1.421.18-1.710.00021.451.20-1.770.0002
WBC at diagnosis      
 <50 × 109/LREF  REF  
 ≥50 × 109/L1.471.20-1.810.00021.351.08-1.690.008
Histology at diagnosis      
 B lineageREF  REF  
 T lineage1.050.79-1.390.751.160.86-1.560.327
CNS status at diagnosis      
 Negative (CNS 1 or CNS 2)REF  REF  
 Positive (CNS 3)1.521.07-2.160.0211.531.03-2.250.034
Diagnosis time period      
 1990-1999REF  REF  
Time to relapse < 36 mo2.191.76-2.74<.00012.351.85-3.00<.0001
Site of first relapse      
 Isolated CNS or testesREF  REF  
 Not isolated CNS or testes1.791.46-2.19<.00012.11.67-2.63<.0001

In order to define a good risk group, we used site of relapse and time to relapse because they had the most extreme hazard ratios with regard to EFS and OS. We did not incorporate additional factors such as initial diagnostic features because they added only marginally to the ability to distinguish between groups, and would have limited the favorable-risk group to a very small number of patients. Using the 2 features at the time of relapse, the favorable risk group consisted of 41 (5.4% of the initial cohort) children who experienced relapse in either a testicular or CNS site, at least 36 months from diagnosis. For this group, the 3-year postrelapse EFS was 50.0% ± 8.9% and OS was 68.0% ± 8.1%.


We found that children with relapsed ALL in Central America had poor outcomes overall, although some children appear to have prolonged survival. For those who experience a second relapse, some may live relapse-free for several months to several years. We determined predictors of survival and identified a good risk group with EFS of 50% and OS of 68%. These findings can be used to guide therapeutic decision-making following relapse for children with ALL in Central America.

For children with relapsed ALL in Central America, EFS and OS are inferior to those described in recent reports from HIC.2-8 There are multiple reasons for this inferior survival. First, ALL outcomes are generally inferior in LMIC, and this survival gap is typically attributed to higher treatment-related mortality and abandonment of therapy.15 In addition, salvage treatments are limited in LMIC. In particular, access to newer expensive drugs may be difficult or impossible, and hematopoietic stem cell transplantation is not available in the region. Finally, given the costs and burden associated with treatment of relapsed ALL, parents and health care professionals may choose a palliative approach earlier in a resource-limited context.

We found that some children had prolonged survival, and even for those who did relapse a second time, most of those relapses occurred between 6 and 24 months; this relapse-free time may be important to families. Thus, our findings suggest that aggressive treatment may be beneficial for at least some Central American children with ALL relapse, particularly for those who have favorable prognostic features. However, there are important factors to consider when deciding to treat a child with relapsed ALL, including preferences and the socioeconomic context of the family.

The strengths of this study include the large sample size, which permitted robust modeling of predictors of postrelapse survival, and that all the centers included (with their satellites) provide the majority of the pediatric oncology care in their respective countries. This approach allowed highly generalizable, population-level analysis of all children experiencing relapse.

However, our study must be interpreted in light of its limitations. First, we observed that all children who abandoned treatment after relapse died. If some children with ALL abandoned primary therapy, relapsed, and never returned to the health care system, these children would have been excluded from our study. Thus, this exclusion may have biased our results. Although the magnitude of this problem is unknown, prior data may be helpful. In a previous study from El Salvador, we found that 19 of 443 (4.3%) of children with ALL abandoned therapy and did not return for care.9 These data provide an estimation of the number of children with relapsed ALL who may have been excluded from the analysis. Second, our study only included relapsed patients; therefore, we do not have an estimation of relapse rates and how these may have changed over the study time period. An additional limitation is that relapse therapy varied without a coordinated relapse protocol and thus, we could not evaluate the role of therapy on relapse outcomes. Therapy is one of the most important prognostic factors in relapsed ALL, and introduction of a common relapse protocol for the region is an important future goal. Finally, we do not have an accurate measure of the proportion of patients treated with a curative intent who achieved a second remission; such data would provide important prognostic information for specific patient subgroups.

In conclusion, we found that outcomes for relapsed ALL in Central America are poor. However, some children have prolonged survival, and even among those who relapse again, they may live relapse-free for many months. Consequently, aggressive treatment may be warranted for at least some children with relapsed ALL, particularly in those with better prognostic features. Future studies should evaluate treatments received and quality of life in order to gain more insight into optimal treatments in the relapsed ALL setting in the LMIC context. Use of common relapse protocols would facilitate better understanding of relapsed ALL outcomes.


The Pediatric Oncology Group of Ontario funded the data management program through which the data were initially collected.


Dr. Sung is supported by a New Investigator Award from the Canadian Institute of Health Research.