The protocol for this trial and supporting CONSORT checklist are available as supporting information; see Checklist S1
and Protocol S1
Study design and patients
This prospective, randomized, open-label, three-arm trial was carried out at five centres in three countries of East Africa: MSF Holland treatment centre, Um el Kher, Gedaref State, Sudan; Ministry of Health Hospital, Kassab, Gedaref State, Sudan; Leishmania Research and Treatment Centre, Gondar University Hospital, Amhara Regional State, northern Ethiopia; Leishmania Research and Treatment Centre, Arba Minch Hospital, Gama Gofa, Southern Nations, Nationalities and Peoples (SNNP) Regional State, southern Ethiopia; and Centre for Clinical Research (CCR), Kenya Medical Research Institute (KEMRI), Nairobi, Kenya.
Eligible patients were 4–60 years old and had to have clinical symptoms (fever and splenomegaly) and a diagnosis of VL, confirmed by microscopic verification of parasites in spleen, lymph node or bone marrow tissue aspirates. Patients were excluded if they: (1) had taken any antileishmanial drug in the preceding 6 months; (2) showed severe protein or caloric malnutrition (Kwashiokor or marasmus); (3) had previous hypersensitivity reaction to SSG or aminoglycosides; (4) had a concomitant severe infection (except HIV) or any other serious underlying (cardiac, renal, hepatic) disease; (5) had conditions associated with splenomegaly, such as schistosomiasis; (6) had a history of cardiac arrhythmia or an abnormal electrocardiogram (ECG); (7) were pregnant or lactating; (8) had any relevant outliers of safety laboratory parameters (hemoglobin <5 g/dL, white blood cells <1×103/mm3, platelets <40,000/mm3, liver function test values >3 times higher than upper limit of normal, serum creatinine values above upper limit of normal); or (9) had clinical hearing loss.
Eligible patients were identified during regular field trips to villages in the endemic areas, or presented spontaneously at the hospital trial sites. At presentation, the trial and its purpose were described to patients in their local language. If patients agreed, they were transported to the trial sites for further investigation of eligibility. If eligible, patients were enrolled after written informed consent was signed either by themselves or, in case of minors, by their parent(s) or guardian. The study protocol was approved in each country by the relevant Institutional National Scientific Ethics Committees and the Ethics Committee of the London School of Hygiene and Tropical Medicine. The trial was registered with ClinicalTrials.gov (number NCT00255567).
The trial was conducted in accordance with the Declaration of Helsinki (2002 version) relating to the conduct of research on human subjects and followed the International Committee on Harmonisation (ICH) guidelines for the conduct of clinical trials. All trial site personnel received relevant training in Good Clinical Practice (GCP). The WHO-TDR handbooks were used for training and for reference before and during the trial.
A Data Safety Monitoring Board (DSMB) was appointed at trial start and met regularly throughout the trial. Members were drawn from each of the participating countries. The trial was regularly monitored at all sites by GCP-trained monitors recruited from Sudan, Ethiopia, Kenya and Uganda. Monitors were allocated to trial sites not within their own country to ensure independence from investigators and to ease communications when dealing with issues arising at the trial sites. Wherever possible during monitoring visits, data entered into the trial case report forms (CRFs) were checked against source data (e.g. laboratory log books, analyzer printouts, nursing records) for verification.
Allocation to treatment was by means of sequentially numbered, sealed envelopes, generated from a computerized randomization list. Each centre received a box of uniquely numbered sealed envelopes from the LEAP Trial Coordination Centre in Nairobi, where centralized randomization and envelope preparation were carried out in blocks of 15 to maintain randomization balance within centers. For each enrolled patient, the investigator took the next lowest numbered envelope and allocated the patient to the treatment given on the card inside.
Patients received either PM alone (manufactured by Gland Pharma, India) at a daily dose of 15 mg/kg body weight given intramuscularly for 21 days, or SSG alone (manufactured by Albert David, India) at a daily dose of 20 mg/kg body weight given intramuscularly or intravenously (KEMRI trial site) for 30 days, or the combination of both at the same daily doses for 17 days. For the duration of treatment, patients were treated as inpatients, thus ensuring high compliance. Patients were followed up at 3 (optional) and 6 months after treatment. If patients did not attend follow-up, health workers visited the patient's village to try to establish if he/she was alive and well, had died or had moved away. Patients for whom trial medication had to be stopped due to lack of response or a serious adverse event (SAE) were given rescue medication (liposomal amphotericin B, manufactured as Ambisome® by Gilead, USA) according to national VL guidelines from participating countries.
Baseline, efficacy and safety measurements
Baseline information was obtained from all patients through a standard clinical assessment. Patients were offered counseling and testing for HIV in accordance to national guidelines, though this was not a prerequisite to inclusion. Formal assessments were carried out at baseline, days 7, 14, 21, end of treatment, and 3 and 6 months after treatment. This included a clinical assessment (clinical symptoms, vital signs, weight, height, spleen and liver size), hemoglobin, white cell count (not performed in Um el Kher site), platelets, urea, creatinine, liver function tests (bilirubin, aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase), urinalysis, ECG and audiometry (not performed in Um le Kher site). Amylase test was not routinely done in all sites but was performed by investigators if clinically indicated. ECG printouts with corrected QT intervals were interpreted by the site physician and were stored with source documents. Audiometry was performed using a standardized procedure by site investigators who were trained by a qualified audiometrist and recorded as hearing levels in dB at 0.25, 0.5, 1, 2, 4 and 8 kHz frequencies 
. All reported abnormal audiometric readings were reviewed by the audiometrist. Abnormal audiometry was defined as any patient with audiometric readings in either of the following categories: mild 26–40 dBHL (decibel hearing level); moderate 41–55 dBHL; moderately severe 56–70 dBHL; severe 71–90 dBHL; or profound >90 dBHL. Clinically significant hearing loss or clinically abnormal hearing was defined as any patient who reports to have a hearing loss.
Parasitology was done at baseline, end of treatment and 6 months after treatment. A spleen, lymph node or bone marrow aspirate was performed at diagnosis and for assessment of cure. This varied between sites according to local guidelines. During follow-up, bone marrow aspirates were done in the event of spleen or lymph node regression.
The primary endpoint was cure (ie, efficacy) at 6 months after treatment, based on the absence of leishmania parasites in tissue aspirates. Any patient who died from VL, received rescue medication during the treatment period or had parasites visualized on microscopy at the 6-month assessment was regarded as a treatment failure in primary efficacy analyses.
The secondary endpoint was initial cure at the end of treatment (i.e. at day 22 for PM, at day 31 for SSG, and at day 18 for the combination). Any patient who died from VL, received rescue medication during the period of the trial or had parasites visualized on microscopy at the end of treatment was regarded as a treatment failure in the secondary efficacy analyses. Clinical and biological parameters such as temperature, weight, spleen size and hemoglobin were used by clinicians to decide if rescue medication was indicated.
A slow responder was defined as a patient who was parasite positive at end of treatment but was deemed not to require rescue medication due to improvements in clinical and biological parameters and who went on to clear parasites at the primary endpoint (6-month assessment). Therefore, a treatment failure at initial cure could be a treatment success at 6-month follow-up as a slow responder to treatment, provided that no rescue medication was administered.
The evaluation of safety was based on adverse event (AE) monitoring, ECGs, safety laboratory analyses (hematology and biochemistry) and audiometry. AEs were classified according to MedDRA, version 10, and defined as treatment emergent (TEAE) if onset was between the first day of treatment and 30 days after treatment.
Data were analyzed with Stata software, version 9. At baseline, continuous data were summarized using mean and standard deviation (SD) and binary data using proportions. Primary and secondary efficacy analyses were by intention to treat (ITT). Where missing parasitological data at the primary endpoint (6-month follow-up) occurred due to loss to follow-up or death unrelated to VL in patients not receiving rescue medication before loss, two analysis approaches were taken within the ITT framework: complete-case analysis, where patients with missing data were excluded, and worst-case analysis, where the missing outcome was assumed to be treatment failure.
The number and percentage cured is presented by arm and centre. Fisher's exact test was used to assess evidence of significant differences in cure.
TEAE rate per arm was calculated as the number of events per arm divided by the person-time at risk. For the rate analysis, patient-time at risk was calculated as the number of patients per arm multiplied by the maximum patient-time at risk for TEAE (treatment period plus 30 days, i.e. 60 days for SSG and 51 days for PM). Poisson regression was used to obtain a rate ratio comparing PM to SSG, adjusted for centre.
At the start of the trial, it was estimated that 217 patients would be required per arm to detect a 10% difference in efficacy among HIV-uninfected patients at the 5% level, assuming 90% power, 95% efficacy in the reference (SSG) arm, and a 20% drop-out rate from the analysis (10% due to HIV co-infection and 10% due to loss to follow-up between end of treatment and 6-month assessment). HIV infection was not an exclusion criterion. Whilst co-infection was expected to be low, efficacy and toxicity may vary in those with and without HIV. A comparison of PM to SSG is presented here following recruitment of 135 patients per arm after which point recruitment into the 15 mg/kg/day PM arm was stopped.