In Africa, the AIDS epidemic was first reported by Clumeck et al. and Van De Perre et al. in 1984 when they described a group of Africans cared for in Belgium with profound immunosuppression and a series of patients in Kigali, Rwanda with infectious complications or generalized Kaposi sarcoma, respectively [9, 10]. Nearly equal numbers of men and women were reported in these articles, in contrast to the predominately male epidemiology described early in the United States HIV experience. The incidence of NHL in sub-Saharan Africa did not increase as markedly early in the HIV epidemic when compared to the increase seen in the US HIV population. This has been attributed to the higher rate of infectious complications that were seen in HIV-infected patients in resource-limited nations, which prevented the subsequent development of malignancies [11]. This decreased incidence may also simply reflect underreporting, attributable to more limited pathology services in combination with differences in the epidemiology of adenopathy in sub-Saharan Africa. Since the most common empiric clinical diagnosis of persistent lymphadenopathy is tuberculosis and not lymphoma, many cases may have gone unnoticed without biopsy confirmation [12]. By the mid-1990s, an increase in lymphoma risk associated with HIV was noted and at present most studies report a 5-6-fold increased risk for development of NHL in HIV-infected individuals living in Africa as seen in Figure 1 [13–16].

On average, patients in Africa present with HIV-associated complications at a higher CD4 count compared to those in resource-rich nations. With current supportive care, the burden of HIV-associated NHL is currently estimated at about 15,000 per year in the equatorial belt of Africa alone [17, 18]. The true incidence is believed to be higher, due to continued limitations in diagnosis and presentation to medical care, though the HIV-associated risk for NHL still does not seem to be as high as in resource-rich nations in the pre-cART era. Data from the International Agency for Research on Cancer shows that the incidence of NHL in most countries of Africa exceeds that of the USA by 2–5-fold. Although cancer registry data in Africa do not routinely capture the HIV status of new cancer cases, Figure 2 clearly illustrates that the incidence of NHL is highest in African countries with a high prevalence of HIV infection, though part of this trend may reflect improved diagnosis associated with HIV clinics.

The Kampala Cancer Registry is one of the oldest continually operating cancer registries in Africa and one of two which are WHO-certified. This registry was created in 1954 and enables an estimation of the impact of HIV infection on rates of malignancy in the area and serves as a tool for ongoing research [19]. The incidence of NHL in Kampala has increased 6.7% annually in men and 11% annually in women since the beginning of the HIV pandemic [20]. Linking HIV and cancer registries in Uganda showed that the incidence of NHL was 6.7-fold higher among HIV-infected persons in Uganda compared with those who are HIV negative, lending credence to the hypothesis that the increase in NHL incidence in sub-Saharan Africa is in part fueled by the HIV pandemic.

Comprehensive data on survival after a diagnosis of NHL in HIV-positive persons are lacking in both resource-rich and resource-poor regions. However, two recent studies highlight the discrepancies in survival between patients in these two settings. In the USA, 47% of HIV-positive persons diagnosed with non-CNS NHL were alive 2 years after their diagnosis [21], which was nearly identical to the survival among HIV-positive Ugandans who received cART [22]. Of note, however, is the fact that 100% of HIV-positive Ugandans with NHL who failed to receive cART died within one year of NHL diagnosis, highlighting the importance of the comanagement of HIV and cancer.

As optimal therapies become more varied for each lymphoma subtype, it is likely that the clinical significance of diagnostic accuracy will increase. An initial attempt to improve this diagnostic accuracy in a cost-effective manner was recently reported by Naresh and colleagues through a description of a diagnostic algorithm for Burkitt lymphoma. They described a three-tiered method for diagnosis that involved increasing numbers of immunohistochemical tests in each tier if the diagnosis remained unclear. They were able to confirm a diagnosis of BL versus DLBCL in 82% of the cases using the first phase of studies (CD10, bcl2, and morphology) and in 92% of cases with the addition of testing for Ki-67, CD38, and CD44 [27]. This type of data can be used to guide developing health systems in prioritizing the diagnostic procedures that are most cost-effective.

Initially, the treatment of AIDS-associated lymphoma in the United States achieved complete response (CR) rates of 53% with combination chemotherapy usually involving CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone), though these responses were tempered by a rate of relapse of 54%. Additionally, significant infectious complications were noted in 42% of the cohort, which contributed to a mortality rate of 85% during a three-year study by Ziegler and colleagues reported in 1984 [31].

The next generation of chemotherapy trials in AIDS-related lymphoma studied more aggressive chemotherapy combinations in recognition of the fact that such cases typically presented at a more advanced stage, had a higher grade at presentation, an increased risk of relapse, and an increased frequency of extranodal disease compared with non-AIDS-related lymphomas. For example, Gill and authors compared M-BACOD (high-dose methotrexate, bleomycin, doxorubicin, cyclophosphamide, vincristine, and dexamethasone) with an even more intensive regimen using high-dose cytarabine and high-dose methotrexate in combination with other agents. Rates of complete remission were 54% and 33% in the two groups, respectively, and the overall survival was 11 months in the M-BACOD group compared to 6 months in the cytarabine/methotrexate arm. This trend towards shorter survival was attributed to a higher rate of opportunistic infections in the cytarabine/methotrexate arm, with more than 75% developing this complication [32]. Similarly, Kaplan and colleagues found a high-dose regimen, COMET-A (cyclophosphamide, vincristine, methotrexate, etoposide, and high-dose cytarabine), was associated with decreased overall survival when compared to standard therapy such as CHOP in patients with HIV-associated lymphoma. Outcomes in both of these studies were best predicted by pretreatment CD4 count, performance status and presence of extranodal disease [33]. Since infectious complications overwhelmed any potential advantage to these aggressive regimens, less-intensive chemotherapy regimens were studied next.

A 1997 study by the AIDS Clinical Trials Group (ACTG 142) therefore compared lower-dose m-BACOD to standard dose m-BACOD with GM-CSF support. The lower-dose regimen was associated with decreased hematological toxicity and similar rates of response. CR rate in the low-dose arm was 39%, compared to 52% in the standard-dose arm while median overall survival was 35 weeks compared to 31 weeks, respectively. Neither of these differences was statistically significant [34]. The survival impact of regimen-related toxicity in the treatment of HIV-associated lymphomas in the pre-cART era was further clarified by a trial conducted by Mounier and colleagues where the selection of chemotherapy was adjusted based on clinical risk factors present at diagnosis. Subjects treated between 1993 and 1999 were stratified based on ECOG performance status of 2–4, a prior clinical diagnosis of AIDS, and a CD4 count less than 100cells/μL. Patients with none of these risk factors received one of two aggressive chemotherapy regimens and no differences in overall survival were noted between 4 cycles of CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone) and 3 cycles of ACVBP (doxorubicin, cyclophosphamide, vindesine, bleomycin, and prednisolone). Patients with one risk factor received either standard-dose or low-dose CHOP, again without any difference in overall survival. The third group, with either 2 or 3 risk factors, received either 4 cycles of low-dose CHOP or 12 cycles of vincristine and prednisolone. There were no differences in overall survival between regimens in any of the risk groups, though lymphoma-specific outcomes were improved in the more aggressive arms of each subset. It should be noted that after 1996 most patients were started on cART, and this subset composed about one-third of the entire study cohort. While cART did decrease the number of patients with 2-3 risk factors and improve overall survival, it did not lead to any significant differences between chemotherapy regimens in risk-stratified patients [35]. Therefore, a decrease in regimen intensity and dose adjustment by hematological parameters for high-risk patients should be considered when treating lymphoma in patients not on cART.

For a time, this report led to decreased use of rituximab therapy in patients with HIV-associated lymphoma. However, additional data was recently provided by Dunleavy and colleagues [39] and Sparano and colleagues [40] in a pair of articles published in 2010. In these studies, rituximab therapy was used to treat HIV-positive patients with DLBCL and NHL, respectively. There were no treatment-associated infectious deaths noted in the study by Dunleavy, supporting the safety of rituximab in this setting. Additionally, this study demonstrated other methods aimed at decreasing the duration of associated immunosuppression and myelotoxicity with chemotherapy regimens. The authors utilized a decreased number of cycles of infusional EPOCH in combination with “dose-dense” rituximab (given on days 1 and 5). The EPOCH was dose adjusted based on hematological parameters per protocol prior to each cycle of therapy, thereby regulating the average neutrophil nadir. Second, the study utilized interim FDG-PET scans to guide the number of chemotherapy cycles, with one cycle given after the first negative PET-CT. By using this strategy, about 80% of the patients received only three cycles of therapy with the highest number of received cycles being 5, delivered in 12% of the subjects. The CR rate was 91% with 5-year PFS and OS of 84% and 68%, respectively. Of the 10 deaths on study, half were due to lymphoma and 5 occurred while subjects remained in remission. 3 of these deaths were due to opportunistic infection and 1 developed a secondary Burkitt lymphoma. The immediate application of this treatment regimen in resource-limited settings is prevented for multiple reasons. First, a lack of hospital beds and hospital staff relative to the large numbers of patients in need of treatment prevents delivery of five days of infusional EPOCH. Second, while this regimen is dose adjusted for hematological parameters, it is delivered with G-CSF support, a therapy that is not available in most resource-poor settings. Third, PET scans, and even CT scans, are not available in routine clinical practice at the present time to rationally select patients that may safely receive fewer cycles of chemotherapy.

Even with these limitations, however, several principles can be incorporated to improve care in resource-poor settings. Dose adjustment has been investigated in trials of chemotherapy in sub-Saharan Africa with some success. One such study, and the first prospective trial investigating chemotherapy for HIV-associated lymphomas in Africa, was performed by Mwanda and colleagues, reported in 2009 [26]. This study utilized a dose-modified oral chemotherapy regimen based on a United States trial that was completed before the availability of antiretroviral therapy [41]. The dose-modified regimen included lomustine (50mg/m² day 1, cycle 1 only), etoposide (100mg/m² days 1–3), cyclophosphamide (100mg/m² days 22–26), and procarbazine (100mg/m² days 22–26). As the initial regimen required G-CSF support, doses were decreased in the protocol to prevent this requirement. Additional modifications during therapy were also outlined, in a manner similar to the infusional EPOCH regimen. These adjustments led to a 50% dose reduction in all medications for an absolute WBC count less than 3,000cells/μL or a platelet count under 100,000cells/μL. For WBC count less than 1,500cells/μL or platelets under 50,000cells/μL doses were held until counts improved. If counts did not improve within three weeks, subjects were removed from the protocol—though this did not occur in any subjects on protocol. Median CD4 count in the 49-patient cohort was 198cells/μL, 37% of patients were on antiretroviral therapy, and most had a history of previous opportunistic infection. 65% of the patients completed the two cycles determined by the protocol and a 6% treatment-related mortality was noted during the study, a significant improvement from previous studies. The reported CR rate was 58% with 78% having an objective response, numbers that are comparable to pre-cART studies in resource-rich nations. Median overall survival was 12 months, though a large variance was noted in patients on cART compared to no retroviral therapy with a median overall survival in the latter group of about 6 months.