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 Harmonization guidelines for the conduct of clinical trials. All trial site personnel received training in Good Clinical Practice (GCP).
The relevant ethics committees from each country approved the study and the details are listed in the attached supporting text document. Patients and their legal guardians (if they were minors) provided signed informed consent prior to being randomized to the different treatment arms. GCP-trained monitors recruited from all four participating countries regularly monitored the trial at all sites.
An open label, parallel-arm multi-centre individually randomized controlled trial.
Patients were enrolled from six clinical trials sites: Médecins Sans Frontières (MSF) Holland treatment centre, Um el Kher, Gedaref State, Sudan; Ministry of Health Hospital Kassab, Gedaref State, Sudan; Gondar University Hospital, Amhara State, Northern Ethiopia; Arba Minch Hospital, SNNPR state, Southern Ethiopia; Centre for Clinical Research, Kenya Medical Research Institute (KEMRI), Nairobi, Kenya; and Amudat Hospital, Nakapiripirit Region, Uganda.
Inclusion and exclusion criteria have been described previously 
. Briefly, patients aged 4–60 years with parasitologically confirmed VL were included, but patients with very severe VL or those with contraindications were excluded ( and ).
CONSORT Patient Flowchart – SSG vs. PM.
CONSORT Patient Flowchart – SSG vs. SSG&PM.
The three arms were SSG monotherapy (20 mg/kg/day for 30 days: reference arm), PM monotherapy (20 mg/kg/day for 21 days) and SSG & PM combination (SSG: 20 mg/kg/day, PM: 15 mg/kg/day both administered for 17 days).
Administration of PM (Gland Pharma, India) was intramuscular (IM). SSG (Albert David, India) was administered IM, or intravenous (IV) in Kenya. Patients requiring rescue medication were given liposomal amphotericin B, (manufactured as Ambisome®, Gilead, USA) according to national dosage guidelines of the participating countries. Patients were hospitalized for treatment and weekly monitoring of clinical and biological parameters. Follow-up visits were conducted 3 months and 6 months post end of treatment ( and ).
The primary efficacy endpoint was definitive cure, defined as parasite clearance from splenic, bone marrow or lymph node tissue aspirates 6 months after the end of treatment. Any patient who died from VL, received rescue medication during the trial, or had parasites detected at the 6-month assessment was considered a treatment failure. The secondary efficacy endpoint was parasite clearance from tissue aspirates at the end of treatment (SSG: day 31, PM: day 22, SSG & PM: day18). Treatment failure at the end of treatment was defined as death or receipt of rescue medication during initial hospitalization or presence of parasites at end of treatment necessitating rescue treatment. The presence of parasites at the end of the treatment, subsequently cleared without need for rescue treatment was considered a treatment success for primary outcome (definitive cure at 6 months follow-up), but a treatment failure for secondary outcome (cure at end of treatment). Slow responders were defined as patients with detectable parasites at end of treatment and parasite clearance at 6 months follow-up, without need for rescue treatment at any time. Parasitology was performed and reported according to an approved World Health Organization (WHO) method 
. The numbers of parasites in slide fields were counted under oil emersion at 100× magnification and counts recorded.
Other Data Collection
Safety was evaluated based on the occurrence of adverse events (AE), laboratory parameters (haematology and biochemistry), electrocardiogram (ECG) readings, and audiometry. AEs were classified according to the Medical Dictionary for Regulatory Activities (MedDRA) version 10 
. A treatment emergent AE (TEAE) was defined as an AE with onset between the first day of treatment and 30 days after end of treatment.
ECGs were performed at all sites using a portable self-reporting ECG machine (Cardiofax, Model ECG 9620, Nihon Kohden) with patients resting supine on their beds. Trial physicians reviewed tracings and reported any abnormality.
Post-kala-azar dermal Leishmaniasis (PKDL) was recorded actively as an adverse event during patient follow-up or reported directly by the patients in between follow-up dates.
Audiometric testing was performed at all trial sites except Um el Kher using Voyager 522 Portable Diagnostic Audiometer (Madsen, Taastrup, Denmark). In recruitment period 1, investigators reported audiometric data as normal, clinically insignificant or clinically significant 
. In period 2, hearing levels were recorded in detail for each ear at six frequencies. The following definitions were used to measure abnormalities; 1) disabling hearing impairment
(DHI): an average hearing level, over frequencies 500, 1000, 2000, 4000 Hz, of ≥31 dB in both ears for those <15 years and ≥41 dB for those aged ≥15 years; 2) audiometric shift:
a change in hearing level from baseline of ≥25 dB at ≥1 threshold frequency or ≥20 dB at ≥2 adjacent threshold frequencies.
All patients were offered counselling and HIV testing in accordance to national guidelines at screening.
Sample Size Determination
The trial was designed to have 90% power (β
0.1) to detect, at the 5% significance level (α
0.05), an absolute difference in efficacy of 15% between PM and SSG and 10% between SSG & PM and SSG regimens 
. An 85% efficacy was assumed in the reference arm and adjusting for 10% HIV co-infection and 10% loss to follow-up at 6 months post end of treatment, it was estimated that 404 and 195 patients per arm were required for the respective comparisons.
Being HIV-positive was not an exclusion criteria but the original protocol stated that there was to be a sufficient number of patients for a subgroup analysis excluding HIV patients (if deemed necessary).
As described at the end of the Introduction, recruitment and randomisation was carried out during two periods. In the first period, patients were randomised to SSG or SSG & PM combination arms, as part of a randomisation into three arms. Data from the third arm, a lower dosage regimen of PM found to be ineffective are not included here. In the second period, randomisation continued into one of three arms; SSG, SSG & PM arms as per period 1 and a PM monotherapy arm at a higher dosage regimen than previously (see Introduction and Interventions sections.)
In recruitment period 2 (using the higher 20 mg/kg dose of PM), randomization into 3 arms was continued until the desired sample size was reached for the PM versus SSG comparison. Randomization was then continued into one of two arms (SSG or SSG & PM) until reaching the sample size for the SSG versus SSG & PM comparison. Um el Kher site participated in period 1 only and Amudat site in period 2 only (during the two-arm randomization).
A computer-generated randomization list was produced with stratification by centre and block sizes of 15 until recruitment in the PM arm was completed, and block sizes of 10 thereafter. Allocation was concealed using opaque, sequentially numbered sealed envelopes. The randomization list and envelopes were prepared and stored securely at the LEAP Data Centre, based at the trial co-ordination centre in Nairobi.
Blinding of patients and investigators was not possible due to the different treatment durations and additional placebo injections were considered inappropriate.
Data were double-entered and validated in Epi-Info. Bespoke query generation programs were developed using Stata software, version 11 
. All statistical analyses were performed using Stata. Baseline data were summarized using mean and standard deviation (SD) or proportions where appropriate. Nutritional status was classified as normal, underweight, or severely underweight according to WHO Child Growth Standards in those <19 years and body mass index (BMI) in those ≥20 years 
For the SSG vs. PM comparison, patient data from randomisation during period 2 are included in this comparison. For the SSG vs SSG & PM comparison, patient data from randomisation into these arms in periods 1 and 2 are included in this comparison.
Efficacy data were analysed according to Intention-to-Treat (ITT) and Per-Protocol (PP). The PP population excluded those with pre-specified major protocol deviations (i.e. consent withdrawal after taking a dose of study medication, receipt of under 70% or over 130% of the expected treatment dosage, or receipt of alternative treatment to that of random allocation). Missing efficacy data were handled in two ways for each analysis population; complete-case analysis, where patients with missing data were excluded and worst-case analysis, where missing outcomes were considered treatment failures.
Efficacy is measured as the percentage of patients cured per arm. The treatment effect is the difference in efficacy between each test treatment (PM or SSG & PM) and the reference (SSG). Unadjusted treatment effects were calculated with exact binomial 95% confidence intervals (CI). Adjusted treatment effects were obtained using generalized linear models with a binomial distribution and identity link function. To assess possible effects of centre, age group (<18 years and ≥18 years) and recruitment period on efficacy after accounting for treatment allocation, regression models including treatment but with and without the covariate of interest were compared using the likelihood ratio test (LRT).
Treatment emergent adverse event (TEAE) rates were calculated as the number of TEAE, divided by the person-days at risk for each arm, and comparisons made using rate ratios. The treatment emergent period was defined as between day 1 of treatment and 30 days after the pre-defined treatment period, inclusive, therefore person-time at risk was as follows; SSG arm: 60 days, PM: 51 days, SSG & PM: 47 days. An adverse drug reaction was defined where an investigator recorded a probable, possible or unlikely relationship between the AE and study drug for VL.