Visceral leishmaniasis and HIV in Tigray, Ethiopia


Suzi Lyons, Department of Public Health and Epidemiology, University College Dublin, Earlsfort Terrace, National University of Ireland, Dublin 2, Ireland. Tel.: +353 1 716 5550; Fax: +353 1 716 7407; E-mail: (corresponding author).
Dr Hans Veeken, Medical Department, Médecins Sans Frontières, Max Euweplein 40, 1001 EA, Amsterdam, The Netherlands. E-mail:
Dr Jean Long, Department of Community Health and General Practice, Trinity Centre for Health Sciences, The Adelaide and Meath Hospital Incorporating The National Children's Hospital, Tallaght, Dublin 24, Ireland. Tel.: +353 1 608 1087; E-mail:


Objectives To identify characteristics that increased the risk of mortality in Ethiopian visceral leishmaniasis patients in a treatment programme managed by Médecins sans Frontières, in Tigray, Northern Ethiopia.

Methods Retrospective review of a cohort of 791 patients treated for visceral leishmaniasis.

Results The cohort displayed all the classical signs and symptoms of the disease. The case fatality rate was 18.5% (146) (95% CI: 15.8–21.3%). Logistic regression showed that individuals who experienced at least one episode of vomiting or haemorrhage were more likely to die than those who did not. A subcohort of individuals who tested human immunodeficiency virus (HIV)-positive were more than four times more likely to die than those who tested HIV-negative (OR 4.5, 95% CI: 1.8–11.4).

Conclusion This study identifies characteristics associated with death in this population and highlights the devastating effect of co-infection with visceral leishmaniasis and HIV in the African context.


Visceral leishmaniasis is a systemic parasitic illness, transmitted primarily by the sandfly from animal or human reservoirs (WHO 1990; Bryceson 1996). Untreated, the disease is usually fatal, but with appropriate treatment the case fatality rate drops to 10% or less (Bryceson 1996; Veeken et al. 2000). Visceral leishmaniasis is endemic in Ethiopia, with a patchy distribution in the southern and north-western lowlands (Maru 1979; WHO 1996). The causative parasite is Leishmania donovani, but the definitive reservoir remains unknown, although anthroponetic transmission has been implicated (WHO 1996). Tigray Province, the endemic region in the north–west of Ethiopia, has historically reported only sporadic cases. Since 1970 the number of cases in the region has increased, which appears to correspond to an extensive programme of agricultural development with its annual influx of migrant workers (Mengesha & Abuhoy 1978; Maru 1979; Berhe et al. 2001). In response to this increase in cases, Médecins sans Frontières (MSF) opened a visceral leishmaniasis treatment centre in the region in 1997. Treatment is provided free of charge for all. Proximity to the frontline of the conflict between Ethiopia and Eritrea meant the site had to be moved twice for security reasons.

Human immunodeficiency virus (HIV) is a major public health issue in Ethiopia and in 2002 it was estimated that 6.4% of adults were HIV-positive (UNAIDS et al. 2002). The influx of a high risk group, migrant workers, into Tigray makes a prevalence of HIV similar to or higher than the national level likely.

Over the last decade visceral leishmaniasis has emerged as a major opportunistic infection associated with HIV in at least 28 countries, mainly in Europe (where L. infantum is the causative organism) (WHO & UNAIDS 1998; Desjeux 1999; Pintado & Lópéz-Vélez 2001). Research in Europe has found that individuals co-infected can present with vague and atypical symptoms of visceral leishmaniasis; absence of hepatosplenomegaly, non-specific gastro- intestinal problems, with parasites often found outside the reticulo-endothelial system and negative serology (Davidson 1997; Desjeux 1999; Pintado et al. 2001). Co-infected patients have a much poorer prognosis, with up to 80% relapsing, and a reduced life-expectancy (WHO & UNAIDS 1998; Desjeux 1999). To date there have been 271 reported cases of co-infection of HIV and visceral leishmaniasis in Ethiopia (Wolday et al. 2001). However, there is very little information on co-infection regarding estimation of true numbers, clinical presentation or outcome in Africa, where the causative organism is L. donovani (WHO & UNAIDS 1998; Wolday et al. 2001).

This study documents the outcomes of a large cohort of individuals treated for visceral leishmaniasis, identifies factors that predict mortality and describes the clinical picture of visceral leishmaniasis in HIV co-infected patients.

Material and methods

The study was a retrospective review of treatment charts from patients admitted from 1 December 1998 to 1 May 2000. Diagnosis of visceral leishmaniasis was made on clinical presentation (case definition: fever for more than two weeks, with splenomegaly and/or wasting) (WHO 1990; Veeken 1998) and was confirmed by either (i) positive serology using direct agglutination test (DAT) titre ≥1:6 400 or by (ii) demonstration of parasites from spleen or lymph node aspiration. Individuals previously treated for visceral leishmaniasis were excluded from the study. The outcome of interest was death during treatment. Patients who completed and survived the full treatment programme (of 30 injections stibogluconate 20 mg/kg per day) and were discharged with recorded clinical improvement (with or without test of cure, i.e. clearance of parasites confirmed by splenic or lymphatic aspiration) were considered cured. Follow-up at six months to establish relapse was not undertaken in this study.

HIV testing was performed during two distinct periods in the study. From December 1998 to April 1999 two rapid tests were used (Capillus and HIV spot) for all patients and those who had two positive results were considered HIV-positive. During this period a comparative drug trial was implemented at the treatment site and HIV tests were readily available. From October 1999 to May 2000 testing re-started, but due to scarcity of HIV tests, only one rapid test (either Capillus and HIV spot) was used for individuals fitting the WHO acquired immunodeficiency syndrome (AIDS) definition (WHO & UNAIDS 1998). Those with a positive test result were considered HIV-positive. HIV testing was voluntary, and pre- and post-test counselling was provided.

Laboratory confirmation of visceral leishmaniasis, HIV or other conditions was not always available due to proximity to the frontline with Eritrea, inconsistency of supplies, inaccessibility during the rainy season and lack of electricity.

After pilot testing, 41 variables were collected from the charts, comprising data on demographics, clinical presentation on admission, laboratory tests, symptoms, co-morbidity and outcome. Only patient numbers were recorded so no individual can be identified from the study. Statistical analysis was carried out using JMP statistical package (SAS Institute 2000). Pearson Chi-square test, Fisher's exact test and crude odds ratios were used to compare proportions in independent groups of categorical data. Multiple logistic regression models were developed to determine which variables best predicted mortality. Exact 95% confidence intervals were calculated for proportions of binomial variables and for adjusted odds ratios. Denominators vary for each variable because not all sections of the charts were completed.

The study was approved by Trinity College Public Health Research Ethics Committee.


Between 1 December 1998 and 1 May 2000, 839 people were admitted to the programme and all had a documented outcome (i.e. survived/cured, died, defaulted or transferred). Of these 839, 48 were excluded from the study: 14 relapsed cases who had been treated previously, six cases with post-visceral leishmaniasis dermal leishmaniasis, 21 defaulters and seven transfers. Therefore 791 individuals with their first diagnosis of visceral leishmaniasis were included in the study analysis, of whom 645 were cured and 146 died. Three hundred and seventy-eight (47.8%) patients were diagnosed with a positive aspirate and 321 (40.6%) with positive serology. The remaining 92 (11.6%) individuals had either borderline/negative results in serology or aspirates (53) or no diagnostic test (39) and were all treated based on clinical signs and symptoms.

The median age of the 791 cases of visceral leishmaniasis was 25 years (10 months to 77 years). More than three-quarters of the cohort were aged between 15 and 45 years. The male to female ratio was 11:1; females were younger than males, with 42% of the females (27/64) younger than 15 compared to 9% of the males (68/722, P < 0.001). Children under the age of 15 had a lower mortality rate (3/95, 3%, 95% CI: 0.7–9%) than those aged 15 and over (142/692, 21%, 95% CI: 16–21.6%).

There was no significant difference in mortality between those who had been diagnosed by laboratory tests (DAT or aspirate) and those treated on clinical grounds only (χ2 = 1.0, d.f. = 1, P = 0.3). The cohort experienced the classical signs and symptoms associated with visceral leishmaniasis (Table 1). The overall death rate was 18.5% (95% CI: 15.8–21.3%).

Table 1.  Signs and symptoms of Ethiopian visceral leishmaniasis patients (n = 791)
Signs and symptoms%No. diagnosed/ denominator
  1. * Proxy for severe debilitation.

  2. † Adults aged over 18, with no recorded oedema.

Carried by stretcher*13 91/701
BMI <16†30116/390

Of the 213 individuals tested for HIV antibodies, 49 (23%) were HIV (antibody) positive. One hundred and twenty-seven of the 213 tested individuals had two tests, with 24 testing positive. Eighty-six of the 213 tested for HIV underwent only one test (fitting the WHO AIDS definition), and 25 of them tested positive. The median age (range) of this group was 29 (18–55), all were over the age of 18 years and 45 (92%) were male. Of those who tested positive for HIV, three-quarters had one or more episodes of diarrhoea compared to half of those who tested negative for HIV antibodies (χ2 = 9.5, d.f. = 1, P = 0.002). There was no significant difference in mortality between those who had two tests (14, 10%) and those who had one test (10, 12.7%), (χ2 = 0.3, d.f. = 1, P = 0.6).

Of the 41 variables recorded, bivariate analysis identified 11 variables significantly associated with death (Table 2). In order to identify independent patient characteristics or risk factors that predicted mortality, logistic regression models were constructed to control for confounding and describe statistical interaction. Gender and the variables that were significant using bivariate analysis were entered in the models and only statistically significant variables were retained in the final models. Two separate models were created: model A excludes all those tested for HIV and model B includes only those tested for HIV.

Table 2.  Characteristics associated with death from visceral leishmaniasis in Ethiopian patients despite treatment
CharacteristicCured n (%)Died n (%)Crude odd ratios (CI)
  1. * Other sites of oedema: feet, lower limbs, scrotum, face.

  2. † Coma, convulsions, psychiatric disturbance or mono-neuropathy.

Age group (n)n = 642n = 145 
 0–14 years (95)92 (96.8)3 (3.2)1
 15–45 years (659)529 (80.3)130 (19.7)7.5 (2.3–30.3)
 >45 years (33)21 (63.6)12 (36.4)17.5 (4.1–86.8)
Gender (n)n = 640n = 146 
 Male (722)585 (81.0)137 (19.0)1
 Female (64)55 (85.9)9 (14.1)1.4 (0.7–3.2)
Mobility on admission (n)n = 572n = 129 
 Walking (413)367 (88.9)46 (11.1)1
 Supported by a stick (197)154 (78.2)43 (21.8)2.2 (1.4–3.6)
 Carried by stretcher (91)51 (56.0)40 (44.0)6.3 (3.6–10.8)
Duration of illness before admission (n)n = 580n = 124 
 0–8 weeks (474)407 (85.9)67 (14.1)1
 9–16 weeks (183)126 (68.9)37 (20.2)1.8 (1.1–2.9)
 17–24 weeks (51)35 (68.6)16 (31.4)2.7 (1.4–5.5)
 25–48 weeks (16)12 (75.0)4 (25.0)1.4 (0.37–4.8)
BMI (n)n = 314n = 76 
 ≥16 (274)234 (85.4)40 (14.6)1
 <16 (116)80 (69.0)36 (31.0)2.6 (1.5–4.6)
Haemorrhage (n)n = 599n = 133 
 At least one recorded episode (161)100 (62.1)61 (45.9)1
 No recorded episodes (571)499 (87.4)72 (12.6)4.2 (2.8–6.5)
Diarrhoea (n)604133 
 At least one recorded episode (381)285 (74.8)96 (25.2)1
 No recorded episodes (356)319 (89.6)37 (10.4)2.9 (1.9–4.5)
Vomiting (n)n = 601n = 132 
 At least one recorded episode (206)147 (71.4)59 (28.6)1
 No recorded episodes (527)454 (86.1)73 (13.9)2.5 (1.7–3.8)
Site of oedema (n)n = 599n = 135 
 None recorded (617)506 (82.0)111 (18.0)1
 Ascites (25)12 (48.0)13 (52.0)4.9 (2.1–11.9)
 Other* (92)81 (88.0)11 (12.0)0.6 (0.3–1.24)
Neurological complications (n)n = 599n = 136 
 No complications recorded (723)596 (82.4)127 (17.6)1
 One or more complication † (12)3 (25.0)9 (75.0)14.1 (3.4–66.5)
Jaundice (n)n = 599n = 136 
 No jaundice recorded (730)599 (82.1)131 (17.9)1
 Jaundice (6)1 (16.7)5 (83.3)22.9 (2.6–521.5)
HIV status (n)n = 189n = 24 
 Negative (164)153 (93.3)11 (6.7)1
 Positive (49)36 (73.5)13 (26.5)5.0 (1.9–13.3)

Model A (Table 3) shows that individuals admitted on a stretcher were 11 times more likely to die during treatment for visceral leishmaniasis than those who walked unaided or with the support of a stick. Individuals with a body mass index (BMI) of less than 16 were four times more likely to die than those with a BMI of 16 or more. Those who vomited one or more times during treatment were more than three times more likely die than those who did not vomit, while those who had a recorded haemorrhage during treatment were also more than three times more likely to die than those who did not. Individuals who experienced at least one episode of diarrhoea were more than twice as likely to die than those who did not have diarrhoea.

Table 3.  Model A: logistic regression model* to identify the characteristics that predict death from visceral leishmaniasis despite appropriate treatment in the subcohort not tested for HIV
 Total no.Died no. (%)Adjusted odds ratioP-value
  1. Whole model χ2 = 70.62, R2 = 0.28, P = <0.0001.

  2. * Initial model also included gender, duration of illness before admission, spleen and liver size on admission, oedema, neurological complications and clinical signs of HIV.

 Walked/stick43672 (16.5)1 
 Carried by stretcher6033 (55.0)11.1 (3.5–43.2)0.0001
 >1620739 (18.8)1 
 <169436 (38.3)3.8 (1.8–8.3)0.0006
Haemorrhaged at least once
 No36160 (14.3)1 
 Yes9849 (50.0)3.1 (1.3–7.2)0.008
Vomited at least once
 No38562 (16.1)1 
 Yes6246 (34.1)3.5 (1.6–7.5)0.002
Diarrhoea at least once
 No26332 (12.2)1 
 Yes26177 (29.5)2.6 (1.2–6.0)0.02

In model B (Table 4), logistic regression was carried out on the subcohort of individuals who were tested for HIV. Three variables remained significant in this model: vomiting and haemorrhage, and, the most important predictor of death during treatment for visceral leishmaniasis, testing positive for HIV antibodies. HIV-infected people were more than four times more likely to die than those who tested negative for HIV.

Table 4.  Model B: logistic regression model* to identify the characteristics that predict death from visceral leishmaniasis in the subcohort of 213 tested for HIV
 Of n = 213Of deaths n = 24Adjusted odds ratio (95% CI)P-value
  1. Whole model χ2 = 20.77, R2 = 0.14, P = <0.0001.
    * Initial model also included gender, mobility on admission, duration of illness before admission, BMI, spleen and liver size on admission, diarrhoea, oedema, neurological complications and clinical signs of HIV.

HIV status
 Positive4913(26.5)4.5 (1.8–11.4)0.001
Haemorrhaged at least once
 Yes6812(17.4)2.7 (1.1–6.7)0.04
Vomited at least once
 Yes7113(18.3)2.6 (1.05–6.6)0.04


The study population displayed the classical signs and symptoms of visceral leishmaniasis, such as fever, cough and splenomegaly (Maru 1979; WHO 1990; Bryceson 1996). The death rate (18.5%, 95% CI: 15.8–21.3%) was higher than the expected 10% or less (WHO 1990; Bryceson 1996). It compares unfavourably to the death rates in other MSF treatment programmes, 5.7% in North Sudan and 11% in South Sudan, which were observed in severely debilitated patients (Seaman et al. 1996; Veeken et al. 2000). Although worldwide, more males than females are infected with visceral leishmaniasis (WHO 1990), the very high male to female ratio found in this study, 11:1, reflects the high numbers of single males who make up the migrant worker population in the Tigray region and who are at risk of contracting visceral leishmaniasis.

The primary outcome of interest in the study was mortality during treatment for visceral leishmaniasis. Splenic aspirations to prove cure, at discharge, were not routinely done, consequently discharge was based on clinical improvement (feeling of well-being, absence of fever, weight gain) (Berhe et al. 2001). Follow-up of individuals six months after discharge is recommended to monitor episodes of relapse, but was beyond the scope of this study.

The retrospective nature of the study meant it was limited to the information already recorded, information that had been originally collected to diagnose and treat cases rather than for research purposes, and values were missing for all of the variables. This limited the scope of the study to explore in more detail the predictors of death.

The inaccessibility of the project site, because of conflict and weather conditions, meant that routine laboratory confirmation of visceral leishmaniasis, HIV and other conditions was not always possible due to lack of equipment and materials. Diagnosis of visceral leishmaniasis on clinical presentation was made for 92 (11.6%) cases. The proportion of cases who died was similar among those who had a laboratory test to confirm their diagnoses to those who were diagnosed using clinical criteria.

HIV status was assessed for 213 individuals, in some cases based on the results of one rapid test. These tests have a sensitivity 98.8% or above and were used following WHO guidelines for surveillance. The individuals selected for HIV testing based on the WHO AIDS definition (WHO & UNAIDS 1998) and who subsequently had one rapid test were possibly at a more advanced stage of HIV illness, because they had been identified based on physical signs and symptoms rather than a guideline to test all newly admitted cases. Routine testing of all new cases was the policy only during the comparative drug trial period. Therefore, those selected for testing using the WHO definition were, in theory, more likely to die and therefore this analysis could overestimate the magnitude of the effect of HIV. The use of two regimes limits the generalisability of the results. But there was no significant difference in mortality between those who had two tests and those who had one test. There was no significant difference in mortality rates with regard to the comparative drug trial which ran for a period in the treatment programme (Veeken et al. 2000).

The clinical presentation of visceral leishmaniasis can be largely indistinguishable from AIDS-related illnesses, for example, weight loss, fever and splenomegaly (WHO & UNAIDS 1998; Desjeux 1999). Also, very little is known as yet about the symptoms of visceral leishmaniasis and HIV co-infection in African populations (WHO & UNAIDS 1998; Wolday et al. 2001). There was no evidence of atypical presentation of visceral leishmaniasis and HIV co-infection in this study population. Apart from age, the only statistically significant clinical difference between HIV-negative and HIV-positive visceral leishmaniasis patients was diarrhoea (P = 0.005).

Multiple regression was carried out on two separate subcohorts, those not tested for HIV antibodies (model A) and those tested (model B). In model A, five independent factors that predicted death among those treated for visceral leishmaniasis were identified. These were: admitted by stretcher (proxy for debilitation), BMI less than 16 (proxy for wasting), haemorrhage, vomiting and diarrhoea. In model B, the subcohort of individuals who were tested for HIV antibodies, vomiting and haemorrhage were again identified as independent risk factors that predicted death among those treated for visceral leishmaniasis. In addition, in this group HIV-positive status was the most important predictor of death among those treated for visceral leishmaniasis.

Haemorrhage is frequently associated with visceral leishmaniasis both as a symptom and risk factor for death (WHO 1990; Bryceson 1996), but this is the first study to show it as an independent predictor of death. However due to lack of data, it could not be established if site of bleeding or severity of bleeding were factors for mortality. Using bivariate analysis, a study in South Sudan had already found an association between vomiting and increased likelihood of death (Seaman et al. 1996) while we found that vomiting at least once during treatment was an independent risk factor for death.

The difference in the other significant variables between the two models is intriguing. Seaman's study in South Sudan found that a BMI of less than 12 was an independent predictor of death (Seaman et al. 1996). Being carried by a stretcher is obviously an indicator of severe illness, as more than half (55%) of these individuals died. The symptom diarrhoea was examined for interaction with other variables and none was observed. Using bivariate analysis, experiencing one or more episodes of diarrhoea was more common among HIV-positive than HIV-negative individuals.

One must also consider the independent effect of HIV co-infection on the population. Although model B has a smaller number of subjects than model A, it showed that those who were HIV-positive were also five times more likely to die than those who were not. The effect on mortality of undiagnosed HIV in the untested group (model A) cannot be assessed and this is another limitation of the study. Although in this group debilitation, wasting and diarrhoea were independent risk factors for death during treatment, they are also possible signs and symptoms of the opportunistic conditions associated with HIV/AIDS.

What is certain is that with the spread of HIV infection to rural areas and the large-scale migrant movements, these two diseases will continue to collide in visceral leishmaniasis endemic areas. Currently as there is no affordable treatment available for HIV/AIDS in the African context, preventive measures must be taken to reduce the incidence of both HIV and visceral leishmaniasis. Community education on both visceral leishmaniasis and HIV prevention, intensive treatment of sexually transmitted diseases, vector control, and use of impregnated bed nets to reduce transmission are suggested interventions to halt the spread of these infections (WHO & UNAIDS 1998; Desjeux 1999). Other measures to be considered for those with visceral leishmaniasis are active case-finding, rigorous follow up of treated cases, treatment of relapsed cases (who may be a reservoir for visceral leishmaniasis) and routine testing for HIV in all visceral leishmaniasis patients infections (WHO & UNAIDS 1998; Desjeux 1999; Wolday et al. 2001). Finally, research to identify the unknown reservoir and new treatment protocols should also be actively pursued, in conjunction with improved epidemiological surveillance in affected countries.


We thank the staff at MSF for their support and assistance, in particular Dr Michael Marlett and Dr Yousef Melaku. Thanks to the staff at the Department of Community Health and General Practice, TCD also, for their support during this project.