Clinical characteristics and treatment outcome of patients with visceral leishmaniasis and HIV co-infection in northwest Ethiopia

Introduction

The first case of visceral leishmaniasis (VL) and HIV co-infection was reported in 1985 (de la Loma et al. 1985), and by 2007, the number of countries reporting VL and HIV co-infections had reached 35. In East Africa, especially in Ethiopia, the proportion of VL patients with HIV co-infection has increased, despite a decline in the numbers of co-infected patients in southern Europe because of the increased use of highly active antiretroviral therapy (HAART) (Alvar et al. 2008).

HIV infection increases the risk of developing symptomatic VL in endemic areas and accelerates the development of AIDS-defining illnesses (Olivier et al. 2003). Further, it increases the risks of VL treatment failure, drug toxicity, relapse and mortality (Lopez-Velez et al. 1998; Pintado et al. 2001; Ritmeijer et al. 2001, 2006; Mengistu & Ayele 2007). Among European co-infected patients, atypical clinical and parasitological presentations occur because of multi-organ involvement and are often confused with other opportunistic infections (Rosenthal et al. 2000; Russo et al. 2003). A few cases of the first occurrence of VL (Berry et al. 2004) and post-kala-azar dermal leishmaniasis (PKDL) as an immune reconstitution syndrome after initiation of antiretroviral therapy (ART) have been reported (Gilad et al. 2001). Dermotropic Leishmania species in the New World (Leishmania braziliensis, Leishmania mexicana and Leishmania amanozensis) were identified as a cause of VL in HIV-positive patients (Alvar et al. 2008).

The national prevalence of HIV was estimated to be 2.1% (MoH 2007), and VL has been reported from at least 40 areas in Ethiopia, with the most important endemic foci being the Metema and Humera lowlands in the northwest, which account for 60% of the total burden (Hailu et al. 2006). These foci are particularly associated with high HIV co-infection rates, ranging from 18% to 31% of the cases (Ritmeijer et al. 2001; Lyons et al. 2003). Despite the high co-infection rate, inadequate information exists about the disease presentation and treatment outcome of patients with VL with or without HIV co-infection. Therefore, this study reports a retrospective review of clinical records of Gondar University and Kahsay Abera Hospitals in northwest Ethiopia to describe the clinical presentation, treatment outcome and factors associated with poor outcome in patients with and without HIV–VL co-infection.

Materials and methods

This retrospective study was conducted by reviewing the clinical records of all patients with VL ≥15 years old admitted to Gondar University Hospital (GUH) and Kahsay Abera Hospital (KAH) from January 2006 to December 2008. GUH is a tertiary-level teaching hospital with 400 beds providing services for 5 million population, while KAH has 177 beds including VL treatment shelters, providing services to about 700 000 population (400 000 residents in four surrounding districts and an estimated 300 000 annual migrants). Both hospitals are located in the northwest VL endemic regions and have dedicated VL and HIV clinics.

Ethical approval was obtained from the GUH and Liverpool School of Tropical Medicine Research Ethics Committees. Permission to review the records was obtained from GUH, KAH and Médecins Sans Frontières (MSF)-Holland. MSH-Holland has run the VL clinic of KAH since 1997 till June 2009, and all clinical records, follow-up procedures and treatments are recorded using standardised case report forms.

The case definition of a patient with VL was as follows: fever >2 weeks in the presence of wasting and splenomegaly or lymphadenopathy and malaria having been excluded (WHO 1996; MoH 2006). In addition to clinical symptoms, patients with VL had to have a positive serological test (direct agglutination test (DAT) and/or rk-39 dipstick) and/or parasitological confirmation. DAT were obtained from the Royal Tropical Institute Amsterdam (The Netherlands) and rk-39 dipsticks (DiaMedIT Leish) from DiaMed AG, Cressiers/Morat (Switzerland). Rk-39 has a sensitivity of 77% in HIV-positive and 87% in HIV-negative patients with VL, and DAT has a sensitivity of 89% in HIV-positive and 95% in HIV-negative patients with VL. The specificity of rk-39 was 99% in DAT-negative controls and 92% in DAT-negative clinical suspects in northern Ethiopia (ter Horst et al. 2009). Parasitological diagnosis was made by microscopic demonstration of Leishmania donovani bodies in Giemsa stained tissue aspirates of spleen or bone marrow, which was most commonly employed at GUH. Patients with VL were considered to be co-infected with HIV if they tested positive in 2 of 3 serial rapid HIV diagnostic tests [HIV-Determine (sensitivity 98.5–100%, specificity 97.1–100%, Abbot Diagnostics, Abbot Park, IL, USA]; Capillus (sensitivity 98–100%, specificity 99.3–100%) and Uni-Gold [sensitivity 97.6–100%, specificity 98.4–100%), Trinity Biotech, Bray, Ireland] (Plate 2007) at the time of diagnosis of the first episode of VL. Patients with previous confirmed episodes of VL or those who had received VL treatment were excluded.

Information extracted included demographic characteristics, clinical presentation, laboratory results, concomitant infections/diagnosis, treatment and treatment outcome. Anaemia was defined as haemoglobin of <13 g/dl in males and <12 g/dl in non-pregnant females (WHO 2008). A total white cell count of <4500/μl and platelet count of <150 000/μl were used to define leucopenia and thrombocytopenia, respectively. Relapse after treatment was diagnosed in patients with reappearance of signs and symptoms of VL during the first 6 months of follow-up, confirmed by detection of Leishmania parasites from tissue aspirates of spleen, bone marrow or lymph nodes.

Concomitant infections were identified using a clinical algorithm supported with laboratory examinations of clinical specimens whenever possible. Diarrhoea was defined clinically as the passage of 3 or more loose stools per day. Prescriptions of additional antimicrobial agents were recorded. Blood cultures were not available at either hospital.

First-line antileishmanial treatments used were pentavalent antimonials at 20 mg/kg body weight for 28–30 days. These included sodium stibogluconate (Albert David, Culcutta) or meglumine antimoniate (Aventis, Paris, France). Liposomal amphotericin B (AmBisome®) from Gilead Sciences Ltd. was used as second-line treatment in patients not tolerating antimonials and in critically ill patients at a dose of 3 mg/kg body weight for a total of 6–10 days.

Patients were evaluated clinically and through laboratory tests of hematologic profile (WBC, haemoglobin and platelet count) for response to treatment and toxicity to antileishmanial drugs. Parasitologic tests of cure were performed for patients whose clinical response was not satisfactory.

Treatment outcome at 6 months was classified as clinical and/or parasitological cure, treatment failure, death or lost to follow-up and was documented when follow-up records were available. Death after antileishmanial treatment and treatment failure because of lack of adequate response (i.e. persistence of fever, no reduction in size of the spleen and inadequate hematologic response) after a complete course of antileishmanial treatment were considered poor outcomes. Clinical cure (i.e. absence of fever, reduction/disappearance of splenomegaly and/or improvement in haematological parameters) and parasitic cure (i.e. clearance of the parasite from tissue aspirates at the end of complete course of treatment) were considered positive outcomes.

Double data entry was performed for information collected from the study forms, and 20% of the records were randomly selected for further data entry checking. Continuous variables were analysed using unpaired Student’s T-tests in normally distributed variables and Wilcoxon tests in variables with skewed distributions. Chi-square was used to compare categorical variables. Univariate analysis of factors associated with death or treatment failure was performed for VL patients with and without HIV co-infection and variables with P values <0.2 were selected for the backward logistic regression. The backward logistic regression model was performed to obtain independent markers of death or treatment failure. All P values were two sided, and values <0.05 were considered statistically significant for the logistic regression. Odds ratios (adjusted and unadjusted, OR and AOR) were presented with 95% confidence intervals (95% CI). spss version 16 was used for data management and statistical procedures.

Results

Discussion

Like many parasitic infections that induce the replication of latent HIV infections, the concurrent presence of leishmaniasis promotes an increase in HIV viral load (Cacopardo et al. 1996), and hence a more rapid deterioration to the AIDS stage and a shortening of survival. Concurrently, because of the immunological disturbances caused by HIV, there is an uninhibited multiplication of the parasite (Bentwich 2003), leading to further immune suppression that reduces cure rates of VL with conventional therapy.

This study confirms a very high rate of VL and HIV co-infection (38.2%) in the northwest of Ethiopia and confirms previous reports from the VL treatment centre in Humera, which had reported co-infection rates of 18.6% in 2001 (Ritmeijer et al. 2001) and 31% from 2003–2006 (ter Horst et al. 2008). About 10% of immune competent populations exposed to Leishmania infection develop clinical VL. With increasing incidence of HIV and its spread to VL endemic areas, the proportion that develops clinical VL may be much higher (Marty et al. 1994). Most patients with VL diagnosed in GUH and KAH had lived in or visited the Humera lowlands for variable periods of time and were among the high-risk groups for both VL and HIV infection.

Immunocompromised patients with VL may present atypically, e.g. without fever or splenomegaly (Rosenthal et al. 2000; Russo et al. 2003). However, most co-infected cases included in this analysis had a classical VL presentation (Lopez-Velez et al. 1998; Pintado et al. 2001; Lyons et al. 2003), probably reflecting that as this retrospective analysis targeted patients fulfilling the inclusion criteria of typical or classical VL cases, patients with atypical presentations may have been missed. In addition, in KAH, a presumptive and/or final diagnosis of VL was often made by rk-39 dipsticks – a test that has not performed well in the northern part of Ethiopia where its sensitivity and specificity were 75–84% and 70–92%, respectively (Boelaert et al. 2008; ter Horst et al. 2009). The results presented here therefore are likely to be an underestimate of the burden of HIV and VL co-infection in this population and suggest that, at least in this area, diagnostic services may need to consider revising their screening procedures to avoid missing patients without classical signs and symptoms of VL who should undergo further confirmatory using parasitologic technique or DAT tests.

The extent to which the clinical presentations of VL in HIV co-infected African patients resemble or differ from those of European patients remains an open question, even though such comparisons may not be valid considering the different species of Leishmania donovani complex (Leishmania donovani in Africa and Leishmania infantum in Europe) that cause VL. With respect to the immune status, 32 of 34 (94.1%) patients had CD₄ counts <200 cells/μl at the time of VL diagnosis and almost a third of patients had CD₄ counts <50 cells/μl, indicating that most of them were in the late stages of AIDS disease. In this respect, our data are in line with those of European reports, where CD₄ counts were <200 cells/μl in 77–90% of patients (Alvar et al. 2008). Intriguingly, we noted that the duration of illness was longer in HIV-positive than in HIV-negative patients with VL.

The study confirms that patients with HIV/VL co-infection in Ethiopia had poorer response rates to antileishmanial treatment and increased mortality. Positive outcome was achieved in 94.5% of HIV-negative patients compared to only 68.5% of co-infected patients. In a retrospective comparative study of 73 HIV-positive and 39 HIV-negative Spanish patients with VL, 90% of HIV-negative patients responded to treatment compared to only 54.8% of HIV-positive patients (Pintado et al. 2001). Similarly, other studies have demonstrated poor responses of co-infected patients, with clinical responses achieved in only 33–87% (Altes et al. 1991; Lopez-Velez et al. 1998) and parasitological responses in 38–81% of patients (Altes et al. 1991; Medrano et al. 1992).

There was a three times higher mortality rate in HIV-positive than HIV-negative patients (17.5%vs. 5.4%) in the study, and a significantly higher number of deaths occurred in pentavalent antimonial-treated VL/HIV-positive patients (P < 0.001). This could be because of two potential factors: (i) almost all patients in the co-infected group had severe concomitant infections with advanced level of immunosuppression, which might have caused death independently of VL; and (ii) an increased risk of pentavalent antimonial toxicities in co-infected patients (Alvar et al. 2008). The latter is difficult to assess because of the limited laboratory data available to confirm toxicity and the fact that concurrent infections may mimic the presentation of toxicity.

The fact that treatment failure only occurred in co-infected patients highlights the need for further identification of alternative treatments and the evaluation of the optimal dose and duration of therapy in these patients. A wide range of chemotherapeutic drugs have been used in co-infected patients, such as pentamidine (Montalban et al. 1990), combinations of pentavalent antimonial with aminosidine (Scott et al. 1992), allopurinol alone or in combination with other azole compounds (Raffi et al. 1995; Lopez-Velez et al. 1998) and others, with variable results. WHO recommends liposomal amphotericin B as the treatment of choice in co-infected patients because of its success in HIV-positive patients (Alvar et al. 2008), although some studies have demonstrated a high frequency of relapses during follow-up (Laguna et al. 1995; Russo et al. 1996), further supporting the need for combination therapies. Recent studies have shown that high pre-treatment viral loads may influence the response of antileishmanial drugs and therefore better responses could be achieved if viral replication is controlled with antiretroviral therapy (Berhe et al. 1999).

The higher tendency to relapse is one of the most prominent features of VL associated with HIV infection compared to VL in immunocompetent patients. Although very few patients were followed (19 HIV positive and 24 HIV negative), 36.8% of the co-infected patients relapsed during the first 6th months of routine follow-up. Previous reports have shown that over half of the patients relapse within the first year of initial treatment (Lopez-Velez et al. 1998; Pintado et al. 2001), and therefore a proportion of HIV-positive patients in our case series may have relapsed and failed to return or attended other health facilities in the area. The data highlight that this is a high-risk population which may require more pro-active monitoring.

Concomitant infections were routinely suspected in patients with poor clinical progress, for instance non-subsiding or recurrent fever, during VL treatment. In most cases, blood films for malaria and chest X-rays were prepared. Concomitant TB infection was seen in 27.2% of HIV-positive but only 6% of HIV-negative patients with VL. As most HIV-VL co-infected patients had severe immunosuppression under diagnosis of extra-pulmonary TB and other opportunistic infections that could occur at this stage of the disease could be additional factors contributing to death or treatment failure. In developing countries, TB is the commonest opportunistic infection, accounting for nearly 40% of clinical manifestations (Pape 2004) and 13% of all HIV-related mortality in worldwide (Corbett et al. 2003). The weakened immune system in HIV infected individuals increases the risk of acquiring TB, progression to active disease and reactivation of latent TB infection (Badri et al. 2001). However, in VL patients with or without HIV infection, the extent and pattern of TB infection has not been reported. Bacterial infections in patients with VL can be as common as 60%, with respiratory and skin infections being the commonest (Andrade et al. 1990; Berhe et al. 2001). In general, most studies demonstrate that deaths in co-infected patients with VL were because of the AIDS-related infections that complicate the course of the illness (Montalban et al. 1990; Lopez-Velez et al. 1998).

This study was based on a retrospective analysis of a case series; hence, diagnostic tests varied across patients according to their clinical presentation, and patients who did not return for follow-up were not traced. Therefore, this analysis incorporates the limitations of a retrospective case series of self-selected patients and needs to be confirmed by prospective studies examining patients and recording information using standard operating procedures and active monitoring activities. Despite these limitations, the high rate of HIV co-infection in the north of Ethiopia calls for concerted efforts of clinicians, infectologists, immunologists and public health specialists to unravel the mechanisms of pathogenesis, scale up research to identify new ways of treatment and the development of health services that address the specific needs of this high-risk population. Treatment approaches should include new ways of immune reconstitution, beyond the traditional approach of using HAART. In the short term, counselling and testing of all patients with VL for early diagnosis of HIV infection and initiation of antiretroviral therapy should be promoted. Given the shortage of diagnostic facilities for the aetiological diagnosis of concomitant infections, a high index of clinical suspicion especially for TB and empiric antimicrobial therapy is crucial. In addition, mechanisms for the active monitoring of recurrence and complications of VL should be incorporated into the routine monitoring of HAART in high HIV-VL incidence areas.