How I treat Burkitt lymphoma in children, adolescents, and young adults in sub-Saharan Africa

It is ironic and tragic that patients with Burkitt lymphoma (BL) in sub-Saharan Africa (SSA) have not benefited from discoveries resulting from first identification of BL in Uganda in the 1950s.¹  Children in Kampala contributed to the discovery of Epstein-Barr virus (EBV),²  the MYC oncogene,³  and multiagent chemotherapy to cure aggressive hematologic malignancies.⁴  Although BL in SSA was one of the most curable cancers worldwide in the 1960s, outcomes today remain the same, whereas >90% of pediatric BL is cured in high-income countries (HICs).

This is due to weak health care systems and low cancer literacy, which contribute to late diagnosis, and inability to administer and support patients through high-intensity cytotoxic treatment. As a result, tumors are larger, disease stage higher, achievable treatment intensity lower, and treatment-related mortality higher in SSA than in HICs, all of which reduce long-term survival.

These challenges are compounded by scarce high-grade evidence to inform BL care in SSA, despite important contributions by several pediatric groups. Consequently, there is tremendous heterogeneity in BL management and reported outcomes. Here, we review the existing regional literature for BL, emphasize elements of optimal management, and highlight future directions for BL clinical research in SSA.

A 9-year-old, HIV⁻ boy in Lilongwe, Malawi presents with a 4-week history of an abdominal mass. He is malnourished with a Lansky performance score of 50. His hemoglobin is 10.0 g/dL with otherwise normal blood counts, serum albumin is 3.0 g/dL, and serum lactate dehydrogenase (LDH) is 3 times the upper limit of normal.

The first issue for this child with a typical pediatric BL presentation is to establish the diagnosis. This is challenging in SSA, given limited pediatric surgery and interventional radiology, as well as limited pathology, usually without flow cytometry, immunohistochemistry, fluorescence in situ hybridization, and cytogenetics.^5,6  However, although it is difficult to achieve a level of diagnostic certainty for BL in SSA comparable to HICs, an accurate diagnosis can often still be made.^7,8  Moreover, even for typical clinical presentations, committing children to cytotoxic treatment mandates an attempt to confirm they have BL, given the perils of clinical cancer diagnosis in SSA.^9,10 

At a minimum, fine needle aspiration (FNA), which is discouraged for lymphoma diagnosis in the United States, can confirm BL in the correct clinical context in SSA.⁸  Diagnostic accuracy of FNA is enhanced if there is consensus review by >1 pathologist and if accompanying clinical data informs the interpretation. One mechanism for achieving “enhanced” FNA interpretation is through clinicopathologic conferences, which can even be conducted internationally with relatively modest infrastructure requirements using various telepathology systems.^11-13  Improvements in diagnostic accuracy can also be achieved through flow cytometry, using ubiquitous instruments for measuring CD4 counts in HIV programs, although additional reagents, validation, and quality control are needed. Tissue biopsy and limited immunohistochemistry can also support accurate BL diagnosis in SSA.^7,11 

Given current interest in liquid biopsies for cancer screening, diagnosis, and therapeutic monitoring, it is worth noting that EBV in peripheral blood of patients with EBV-associated malignancies is largely tumor-derived,¹⁴  and could represent one of the most implementable circulating cell-free DNA technologies for cancer detection and monitoring in low-income countries (LICs).¹⁵  Although more sophisticated immunoglobulin sequencing approaches are under investigation,¹⁶  EBV DNA has existing commercial assays which can be implemented in SSA using available instruments for measuring HIV RNA.¹⁷  In Malawi, plasma EBV DNA measurement has been used to support BL diagnosis,¹⁸  not as a stand-alone tool but as an adjunct with other clinical and pathologic data to avoid confusion with other lymphoma subtypes that are also associated with EBV (Hodgkin, plasmablastic, extranodal NK/T-cell),^19,20  as well as EBV reactivation in nonmalignant diseases like malaria.

Given unavailability of molecular tools to definitively establish a BL diagnosis in most SSA settings,²¹  deriving and validating a composite diagnostic score for molecularly confirmed BL in SSA would have tremendous regional value. Such a scoring system would include data that can be easily generated locally, for example, age, clinical site, cytologic features, histologic features, immunophenotype, LDH, and plasma EBV DNA. Such a scoring system would include clear delineation of incremental costs needed to achieve incrementally higher diagnostic certainty, allowing judicious resource utilization in SSA where per capita health expenditure may be <$50 annually (vs >$9000 in the United States).²²  For instance, a child with a jaw mass and typical BL cytology in Malawi confirmed by 2 pathologists does not require additional studies to be justifiably treated for BL, even if this would be unacceptable in HICs.

A core oncology treatment principle is to tailor treatment intensity using baseline prognostic factors. This may be more important in SSA than HICs, because risks of overtreatment are higher, due to high opportunistic infection burden, frequent co-occurrence of HIV and malnutrition, and poor supportive care. However, accurate risk stratification is challenging in SSA, given crude, highly operator-dependent, nonreproducible modalities used to stratify patients. Patients with BL in SSA are frequently staged using physical examination, chest radiograph, abdominal ultrasound, and cerebrospinal fluid evaluation. Bone marrow evaluation and computed tomography (CT) may be done if available. Advanced imaging like CT with positron emission tomography (PET), which are standard staging modalities in HICs, are usually not available.

Because of these limitations, patients with BL in SSA are often understaged relative to HICs, and reproducibility of staging across SSA centers is poor. This could be addressed through uniform regional implementation of simple, quantifiable, point-of-care peripheral blood assays to enhance comparability of baseline prognostic features across cohorts. Foremost among these would be LDH and EBV DNA, which have strong, continuous relationships with prospective outcomes.^18,23  To illustrate, the International Prognostic Index (IPI) used in HICs to predict outcomes among adults with aggressive non-Hodgkin lymphoma (NHL) retains prognostic value in Malawi.^24,25  However, it is suboptimal due to dichotomized LDH as normal/abnormal in a population for whom LDH is almost always elevated, given more advanced disease than HIC cohorts, and points assigned for clinical stage and extranodal involvement which are imprecise in SSA.^24,25  Based on existing data and these considerations, an ideal simplified prognostic model for BL in SSA might include just performance status with nondichotomized LDH and/or EBV DNA, and would likely be more accurate and reproducible, and likely more reflective of true disease extent, but this requires validation in large regional cohorts. Such a simplified risk stratification scheme could obviate the need for routine CT and/or bone marrow assessment, which require time, expense, and expertise to perform and interpret, are of variable quality, cause discomfort to sick patients, and rarely change clinical management in SSA. Prognostic scores using data generated in real time to discriminate patients into clinically meaningful groups have been a cornerstone of lymphoma care in HICs for decades,^26,27  and similar derivation and validation efforts are needed in SSA rather than simply transposing HIC approaches.

After confirming the diagnosis and completing baseline evaluation, or simultaneously, the next priority is to mitigate risk of early death upon cytotoxic initiation. Fulminant tumor lysis syndrome (TLS) is fatal in many SSA settings, given limited facilities for renal replacement therapy and rasburicase unavailability. Prevention of TLS with carefully phased treatment initiation is therefore paramount. In very sick patients (Lansky performance status ≤50 or Eastern Cooperative Oncology Group [ECOG] performance status ≥3), we favor a graduated scheme of prednisone for 5 to 7 days, followed by prephase COP (cyclophosphamide, vincristine, prednisone), before initiating definitive treatment 5 to 7 days later, as is the standard approach for pediatric BL in HICs.²⁸  This approach can “rescue” patients felt to be too sick for cytotoxic treatment, and provides ample time for hydration, transfusion, and other stabilization measures before bulky, proliferative tumors are subjected to multiagent chemotherapy. For less sick patients, definitive treatment can be initiated more rapidly, 5 to 7 days after prephase COP, or even prednisone alone. Allopurinol is available in most SSA settings, and we administer this continuously throughout the prephase period.

High-intensity or continuous infusion regimens for treating BL in HICs are difficult to apply in most SSA settings, due to logistical challenges and treatment-related toxicity. Strategies in SSA fall into 3 broad categories: low-intensity approaches, higher-intensity approaches incorporating anthracyclines, and higher-intensity approaches incorporating high-dose methotrexate (Table 1). We are unaware of randomized studies comparing these strategies in SSA, and efforts to apply regimens incorporating both anthracyclines and high-dose methotrexate as in HICs have been unsuccessful. Regardless of approach, poor outcomes are mainly due to relapsed/refractory BL, although treatment-related mortality from infectious complications and/or TLS predictably increases as treatment intensity is escalated. As shown in Table 2, published data reporting efficacy and safety of these approaches with few exceptions consist of observational cohorts at single centers, with wide variation in BL diagnostic criteria, proportion of BL patients excluded, reasons for exclusion, and completeness of follow-up. These issues can have major influence on reported outcomes, making regional BL literature difficult to interpret, as illustrated in Figure 1.

First, there is inevitable referral bias among BL patients who make it to tertiary centers for care particularly from rural areas. With respect to diagnosis, including other lymphoproliferative disorders with different natural histories from BL, that cannot be distinguished from BL in many SSA settings, affects reported outcomes. Variations in time needed for diagnosis and treatment initiation, as well as subjective exclusion of patients deemed unfit for chemotherapy or other reasons, also affects frequency of early deaths and outcomes. Where pathology is limited, subjective clinical criteria are sometimes applied to support or refute BL diagnosis, including response to chemotherapy. In Malawi, children with abdominal BL diagnosed by FNA who do not respond to chemotherapy are at times empirically diagnosed as having Wilms tumors, with resection and subsequent management according to Wilms tumor protocols. Pathologic reexamination after surgical resection is often not done, although in several instances we have made efforts to rereview these specimens and found them to have immunophenotypes consistent with BL. Removing such refractory BL patients by classifying them as non-BL on clinical grounds will improve observed outcomes. Finally, treatment abandonment and/or loss to follow-up are major issues in SSA, given centralized services, long travel distances, and transportation costs. Importantly, death is a major cause of loss to follow-up in diverse populations in SSA, and censoring patients at loss to follow-up without efforts to trace all outcomes leads to overestimated survival.^29-31 

The first attempts to treat BL in SSA involved cyclophosphamide alone or with vincristine and/or low-dose methotrexate, together with intrathecal treatment to prevent leptomeningeal relapse. As shown in Table 1, these strategies have usually resulted in 5% to 10% treatment-related mortality with 50% to 60% 1-year overall survival, although uncertainty about patient selection, as well as completeness and length of follow-up raise questions about how successful such strategies are for most BL patients in SSA.^32-34  However, this may currently be the first-line strategy which best optimizes safety and efficacy at least for limited stage BL in SSA. Relapses using this approach are mainly systemic with <5% of relapses being isolated to the central nervous system.³³ 

Most patients in most SSA reports, however, have advanced BL, and due to poor outcomes for this population using low-intensity approaches, recent efforts have intensified therapy by adding anthracyclines. As shown again in Table 1, such approaches increase treatment-related mortality, typically to 15% to 20%, without clearly improving outcomes and 1-year overall survival most often reported as 40% to 50%.^23,35,36 

Of note, however, investigators in Blantyre, Malawi, have reported improved 1-year disease-free survival from 28% to 66% for stage III/IV BL by adding 2 doses of doxorubicin to a previously described regimen of cyclophosphamide, prednisone, and vincristine, with 12% treatment-related mortality and 1-year overall survival of 73%.³⁶  This represents among the best described treatment experience for this population in SSA. Importantly, of 84 children with BL during the study period, 26 (31%) were excluded, of whom 6 were treated on a different protocol, 10 died at presentation before the diagnosis was confirmed and chemotherapy started, 3 absconded during treatment, and 7 were lost to follow-up.

By comparison, in Lilongwe, Malawi, pediatric BL has recently been treated with 6 cycles of cyclophosphamide, doxorubicin, vincristine, prednisone (CHOP), guided by frequent relapse after less intensive treatment. Using this approach, 1-year overall survival in Lilongwe was 71% for stage I/II, 38% for stage III, and 25% for stage IV, in an unselected cohort using intention-to-treat analysis without excluding deaths before cytotoxic treatment, and active tracing to ascertain vital status for nearly all children.²³  Accounting for obvious environmental differences, these results are generally consistent with experience from Kenya³⁵  and from the United States in the 1980s, during which the Children’s Cancer Group found that adding daunomycin to cyclophosphamide, vincristine, low-dose methotrexate 300 mg/m², and prednisone (COMP) for nonlymphoblastic mature B-cell NHL resulted in 57% event-free survival compared with 55% for patients receiving COMP alone.³⁷  On central adjudication of cause of death in Lilongwe, 73% were from relapsed or refractory BL,²³  supported by analyses demonstrating 70% of children had detectable EBV DNA at CHOP completion suggesting persistent BL.¹⁸  These results suggest treatment failure is mainly due to inability to eradicate advanced BL using CHOP or analogous strategies, rather than excess treatment-related mortality. We have also had difficulty in Lilongwe reproducing Blantyre outcomes using the 28-day protocol with doxorubicin for stage III/IV patients.³⁸ 

These issues speak again to difficulties comparing treatment experience and outcomes in SSA even within a single country. Taken as a whole, however, these studies suggest anthracycline-based therapy can be considered for BL in SSA, and may be most appropriate for patients with advanced disease.

A more consistently encouraging strategy for advanced BL in SSA may be incorporating high-dose systemic methotrexate. High-dose methotrexate 1-3 g/m² has long been a standard core component of BL treatment in HICs, in the COPADM/CYM regimen used most typically for children (cyclophosphamide, vincristine, prednisone, doxorubicin, methotrexate induction followed by cytarabine, methotrexate consolidation), or the hyper-CVAD (hyperfractionated cyclophosphamide, vincristine, doxorubicin, dexamethasone alternating with methotrexate, cytarabine) or CODOX/M-IVAC (cyclophosphamide, vincristine, doxorubicin, methotrexate alternating with ifosfamide, etoposide, cytarabine) regimens often used in adults. However, experience with high-dose methotrexate in SSA is variable, with the French-African Pediatric Oncology Group (GFAOP) and others describing successful use of 1 to 3 g/m^2,39-41  but other studies reporting excessive treatment-related mortality of 20% to 30%.^42,43  Likely, this heterogeneity reflects differences in infrastructure, nursing, and supportive care to administer such doses of systemic methotrexate safely. Capacity for hydration, urinary alkalization, and monitoring vary markedly in SSA, and methotrexate is usually administered in settings where measuring serum drug levels is not possible, but which does not prohibit safe methotrexate administration.

GFAOP efforts in this regard are notable, as this group working across several SSA countries have described the largest experience introducing high-dose methotrexate with requisite supportive care protocols, which they found to be safe and effective at a dose of 1 g/m² in the GFA 2001 protocol, with 1-year overall survival 56%.³⁹  Encouraged by this, the GFAOP has since developed the GFA 2009 protocol increasing systemic methotrexate dose from 1 to 3 g/m² with preliminary data demonstrating 1-year overall survival of 61% and treatment-related mortality of 9% at the 3 g/m² methotrexate dose.⁴⁰ 

Notably, while implementing the GFA 2001 protocol, a parallel effort at North African sites with greater pediatric oncology capacity achieved 1-year overall survival of 75% using regimens more closely approximating those used in Europe,³⁹  with high-dose methotrexate and doxorubicin, emphasizing the important relationship between increased cytotoxic intensity and increased survival, which has after all been a guiding principle for BL treatment in HICs historically. Retrospective data from South Africa also suggest applying intensive regimens with anthracyclines and high-dose methotrexate without major modifications is feasible in middle-income countries with sufficient infrastructure and expertise.^44,45 

In HICs, the dose-adjusted etoposide, prednisone, vincristine, cyclophosphamide, doxorubicin (EPOCH) chemotherapy regimen has recently gained acceptance as a better tolerated, effective strategy without high-dose methotrexate for treating BL in adults,⁴⁶  although controversy remains about the suitability of this approach for high-risk BL.⁴⁷  Some have advocated for its application in SSA, but it is important to acknowledge the many logistical barriers which prohibit continuous infusions in most settings, including access to infusion pumps and central venous catheters, as well as limited overnight staffing to deal with extravasation, pump malfunctions, and other issues which inevitably arise. In Lilongwe, due to poor outcomes particularly among adolescents and young adults (AYAs) with BL treated with CHOP, who frequently have EBV⁻ disease,^23,44  we have begun treating such patients with a modified EPOCH strategy administering 24-hour doses over 8 hours on 4 successive clinic days, avoiding overnight administration, and with dose adjustment as per the National Cancer Institute protocol. In very early experience, this has been feasible without hematopoietic growth factors using standardized anti-infective prophylaxis (cotrimoxazole, ciprofloxacin, fluconazole), and has seemed promising for NHL subtypes like BL, plasmablastic lymphoma, and primary effusion lymphoma known to have poor outcomes after CHOP. Grade 3/4 neutropenia has occurred in all patients but is manageable in patients with and without HIV, and patients treated to date have completed a median 5 cycles at protocol dose levels typically ranging from −2 to 2 per cycle.⁴⁶  However, this experience is anecdotal with few patients and limited follow-up.

In most SSA environments, there is a cytotoxic ceiling beyond which dose escalation cannot safely occur. Therefore, targeted agents may have greater importance even than in HICs for BL treatment. Rituximab improves outcomes for aggressive CD20⁺ B-cell NHL subtypes, including for HIV⁺ patients with CD4 counts > 50 cells/µL which is the immunologic context in which HIV-associated BL typically occurs,^48-51  and including among children and adolescents with BL as demonstrated by the recent intergroup trial.⁵²  Additionally, subcutaneous administration has gained acceptance,⁵³  enhancing feasibility for settings with limited infusion capacity and potentially lowering cost,⁵⁴  and a biosimilar is commercially available worldwide with pharmacokinetic equivalence compared with the patented formulation.⁵⁵  However, cost remains an issue even for the biosimilar, with typical wholesale prices for an adult course of treatment being $3000 to $4000 US dollars, substantially exceeding annual per capital health expenditures in most SSA countries. As such, rituximab is usually sporadically applied in the private sector to the very few patients who can afford to purchase it. Apart from feasibility issues, safety and efficacy specifically in SSA have not been demonstrated, where hematopoietic growth factor support is not routine, and where opportunistic infectious complications during B-cell depletion may be higher than HICs. As a result, we and others have felt it important to prospectively assess rituximab in SSA for BL and other B-cell NHL subtypes, with our early experience from Lilongwe being encouraging to date.⁵⁶  If safe and effective, formal cost-effectiveness evaluation will be important to inform policymakers. Although very costly by SSA standards, an upfront expenditure that substantially increases long-term event-free survival might have comparable cost-effectiveness to lifelong daily antiretroviral therapy (ART) to treat HIV, which has been prioritized as an economically sound intervention in SSA. Moreover, as the movement to achieve universal HIV treatment access has shown, drug prices can be negotiated when there is sufficient demand, political will, and evidence.

A 22-year-old young man in Lilongwe, Malawi presents with a 3-week history of a bulky axillary mass. Biopsy including immunohistochemistry demonstrates BL, and he is newly diagnosed with HIV without having been on ART. His CD4 count is 550 cells per µL and HIV RNA is 4.2 log₁₀ copies per mL. He receives 4 cycles of a modified EPOCH regimen with concurrent ART and achieves a partial response. However, treatment is complicated by neutropenia which limits cytotoxic cumulative dose and intensity, and by frequently missed chemotherapy appointments, and he develops tumor progression within 2 months.

In most reports from SSA, HIV prevalence among children with BL is <5%,^{23,33,34,36,41,57}  and when HIV-associated BL occurs, it is usually among AYAs with relatively preserved CD4 counts.⁴⁵  Treatment of HIV⁺ patients with BL is generally the same as for HIV⁻ individuals, but neutropenia in settings without reliable hematopoietic growth factor availability is a major challenge.²⁴  ART should not be withheld in SSA during chemotherapy, as is sometimes advocated in HICs to avoid drug-drug interactions,⁵⁸  as this places patients at unacceptable risk for infectious complications in SSA. Interactions and overlapping toxicities with chemotherapy can be anticipated and managed, and are not so problematic with tenofovir-lamivudine-efavirenz which is currently first-line ART in most of SSA. Zidovudine should be avoided, and more caution and possible empiric chemotherapy dose reduction are needed for patients on protease inhibitor–based treatment, but integrase inhibitor-based regimens are also increasingly available in SSA⁵⁹  and preferred when possible to minimize chemotherapy interactions.⁶⁰  Moreover, there is an overwhelming literature demonstrating that earlier, continuous ART is better for individual health and population-level transmission than starting later.^61,62  ART might also exert important anti-tumor immunotherapy effects in addition to salutatory effects on HIV.^63,64 

As in HICs, AYAs are understudied and have worse outcomes in SSA than younger children with BL.^23,65  Reasons for this are multifactorial including unique disease biology which is frequently EBV⁻, uncertainty regarding best treatment approaches and whether these should follow pediatric or adult protocols, historical exclusion from clinical trials, and distinct psychosocial issues including poor treatment adherence. In SSA, given young population age structure, lower life expectancy, and general socioeconomic conditions, AYAs also often assume adult responsibilities at younger ages than HICs with greater independence and less family supervision. This may include traveling long distances alone for clinic visits and taking primary responsibility for treatment adherence. Dedicated support services and defined treatment strategies are needed for this challenging and neglected population in SSA.⁶⁶ 

Given challenges administering effective first-line treatment of BL in SSA, relapsed/refractory BL is common and hard to address in light of difficulties providing more intensive salvage chemotherapy and high-dose therapy with autologous stem cell rescue. Indeed, relapsed/refractory BL is challenging even in HICs, but effective first-line treatment makes this much rarer than in SSA. Autologous stem cell rescue is available in South Africa,^67,68  and many SSA countries have processes for externally referring patients to South Africa or India, but usually these processes are too long and/or unreliable to be a viable option for most patients with relapsed/refractory BL. For the vast majority of patients in SSA, relapsed/refractory BL is fatal. Moderate-intensity salvage chemotherapy regimens can be administered, but achievable cytotoxic intensity in the setting of relapsed/refractory BL typically produces partial responses and symptom alleviation with durations usually measured in months,^45,69  emphasizing the need for better first-line treatment to reduce the frequency with which relapsed/refractory BL occurs.

Already, molecular investigations in BL have demonstrated potential therapeutic targets beyond MYC, including B-cell receptor signaling, PI3K, and other dysregulated genes and pathways for which approved agents exist,^16,70-73  although these remain economically far out of reach of most SSA countries. Arguably, testing drugs for which there is sufficient preclinical or clinical rationale incorporated into front-line treatment in SSA, given the cytotoxic ceiling imposed by the environment, may be globally impactful and yield new, lower-intensity treatments compared with current international standards of care. Such developments would be important for SSA by outlining viable paths to cure in LICs, and important for HICs by outlining viable paths to cure that avoid short- and long-term toxicities of existing approaches. As in the 1960s, this could again represent opportunities for BL research in SSA to have truly global impact.

There are more questions than answers in this review, reflecting uncertainties in the BL literature in SSA. Our recommended approach is shown in Figure 2, but unresolved essential questions regarding best approaches to BL diagnosis and treatment in SSA, including long-term outcomes and late treatment effects, demand sustained international commitment to pursue definitive answers through regional collaboration. This is happening via initiatives led by the National Cancer Institute Center for Global Health, and care and research partnerships like GFAOP. By making necessary commitments to this neglected and vulnerable population, we can repay a 50-year-old debt to patients in SSA who contributed to the discovery of BL and fundamental insights into cancer biology. In so doing, we will ensure another 50 years do not elapse with the rest of the world moving forward while outcomes for many patients in SSA remain the same.

S.G. received support from the National Institutes of Health (National Cancer Institute grants U54CA190152, P20CA210285, P30CA016086-40S4), Lineberger Comprehensive Cancer Center (National Institutes of Health, National Cancer Institute grant P30CA016086), University of North Carolina Center for AIDS Research (National Institutes of Health, National Institute of Allergy and Infectious Diseases grant P30AI50410), and AIDS Malignancy Consortium (National Institutes of Health, National Cancer Institute grant UM1CA121947).

Contribution: S.G. and T.G.G. conceived and wrote the manuscript, and approved the final version.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Satish Gopal, UNC Project-Malawi, Private Bag A-104, Lilongwe, Malawi; e-mail: satish_gopal@med.unc.edu.