Comparison of artemether-lumefantrine and chloroquine with and without primaquine for the treatment of Plasmodium vivax infection in Ethiopia: A randomized controlled trial

Background Recent efforts in malaria control have resulted in great gains in reducing the burden of Plasmodium falciparum, but P. vivax has been more refractory. Its ability to form dormant liver stages confounds control and elimination efforts. To compare the efficacy and safety of primaquine regimens for radical cure, we undertook a randomized controlled trial in Ethiopia. Methods and findings Patients with normal glucose-6-phosphate dehydrogenase status with symptomatic P. vivax mono-infection were enrolled and randomly assigned to receive either chloroquine (CQ) or artemether-lumefantrine (AL), alone or in combination with 14 d of semi-supervised primaquine (PQ) (3.5 mg/kg total). A total of 398 patients (n = 104 in the CQ arm, n = 100 in the AL arm, n = 102 in the CQ+PQ arm, and n = 92 in the AL+PQ arm) were followed for 1 y, and recurrent episodes were treated with the same treatment allocated at enrolment. The primary endpoints were the risk of P. vivax recurrence at day 28 and at day 42. The risk of recurrent P. vivax infection at day 28 was 4.0% (95% CI 1.5%–10.4%) after CQ treatment and 0% (95% CI 0%–4.0%) after CQ+PQ. The corresponding risks were 12.0% (95% CI 6.8%–20.6%) following AL alone and 2.3% (95% CI 0.6%–9.0%) following AL+PQ. On day 42, the risk was 18.7% (95% CI 12.2%–28.0%) after CQ, 1.2% (95% CI 0.2%–8.0%) after CQ+PQ, 29.9% (95% CI 21.6%–40.5%) after AL, and 5.9% (95% CI 2.4%–13.5%) after AL+PQ (overall p < 0.001). In those not prescribed PQ, the risk of recurrence by day 42 appeared greater following AL treatment than CQ treatment (HR = 1.8 [95% CI 1.0–3.2]; p = 0.059). At the end of follow-up, the incidence rate of P. vivax was 2.2 episodes/person-year for patients treated with CQ compared to 0.4 for patients treated with CQ+PQ (rate ratio: 5.1 [95% CI 2.9–9.1]; p < 0.001) and 2.3 episodes/person-year for AL compared to 0.5 for AL+PQ (rate ratio: 6.4 [95% CI 3.6–11.3]; p < 0.001). There was no difference in the occurrence of adverse events between treatment arms. The main limitations of the study were the early termination of the trial and the omission of haemoglobin measurement after day 42, resulting in an inability to estimate the cumulative risk of anaemia. Conclusions Despite evidence of CQ-resistant P. vivax, the risk of recurrence in this study was greater following treatment with AL unless it was combined with a supervised course of PQ. PQ combined with either CQ or AL was well tolerated and reduced recurrence of vivax malaria by 5-fold at 1 y. Trial registration ClinicalTrials.gov NCT01680406


Introduction
This Statistical Analysis Plan (SAP) provides a detailed and comprehensive description for the analysis of the antimalarial efficacy study entitled "Evaluating the efficacy of artemether-lumefantrine alone compared to artemether-lumefantrine plus primaquine and chloroquine alone compared to chloroquine plus primaquine for Plasmodium vivax infection". The purpose of the study is to assess the therapeutic efficacy of arthemether-Lumefantrine (AL) compared to artemether-lumefantrine + primaquine (AL+PQ) and chloroquine (CQ) compared to chloroquine + primaquine (CQ+PQ) for P. vivax infection and to determine the number of recurrent vivax episodes in patients receiving radical cure compared to those who did not . The study design is described in a Protocol titled "Ethiopia antimalarial in vivo efficacy study 2012: Evaluating the efficacy of artemether-lumefantrine alone compared to artemether-lumefantrine plus primaquine and chloroquine alone compared to chloroquine plus primaquine for Plasmodium vivax infection" and the trial registered with ClinicalTrials.gov (NCT01680406).
The analysis will be presented in a report, which will be used as the basis of the primary research publications according to the study publication plan. This SAP describes the statistical methods for the primary and secondary outcomes of the study as defined in the protocol as well as additional pharmacokinetic analyses.

Study design
The aim of the study is to improve the radical cure of P. vivax. This requires the elimination of blood stage parasites with schizontocidal drugs, followed by the removal of the dormant liver stages with a hypnozonticidal agent; the trial design aims to define these two antimalarial modes of action.
The trial is a randomised, open label evaluation of clinical and parasitological responses to directly observed treatment for uncomplicated P. vivax malaria. Patients with uncomplicated P. vivax monoinfection who meet the study inclusion criteria are enrolled, randomised, and treated with either AL or CQ with or without PQ, and monitored for 12 months. Patients with recurrent episodes of malaria are treated again with the same regimen and continue to be followed. The first treatment episode is partially supervised while subsequent treatment with primaquine is unsupervised. Patients failing in any arm within 14 days of treatment are treated with rescue treatment (Quinine for 7 days). Patients presenting with vivax parasitaemia > 14 days following treatment will receive the same radical therapy as at enrollment.
The recurrence of P. vivax parasitaemia during follow up can result from either recrudescence of the same parasite, reinfection with a new parasite, or relapse from a liver stage parasite which can be the same or different from the initially observed parasite 1,2 .
Failure of schizontocidal treatment (CQ and AL) can result in recrudescence of the same parasite early during follow up (usually before 42 days), whereas failure from antirelapse treatment can occur from approximately 16 days to over a year. The earliest time of recurrence is dependent upon the pharmacokinetic profile of the antimalarial treatment regimen which defines the post treatment prophylaxis.

Sample size
The design of the clinical study is to compare four treatment arms AL or CQ alone with or without primaquine. From a previous efficacy study, CQ monotherapy had an efficacy of 68% at day 42 (null hypothesis). It is assumed that the addition of PQ will decrease the chance of relapse by at least 60% resulting in a success in 87% of patients. Group sample sizes of 97 in each group were set to achieve 90% power to detect a difference between group proportions of 19% using the two-sided Z-test with pooled variance test statistic. The significance level of the test was targeted at 0.05. Adjusting the alpha level to 0.025 for multiple comparisons will increase the sample size to 118 in each group. The AL arm showed even a lower efficacy at day 42 in our previous study of 58%. Using this lower success rate of 58% seen with AL at day 42, resulted in a smaller required sample size of 68. Taking into account 10-20% loss to follow-up at 42 days, a sample size of 120 subjects for each treatment was determined to be optimal.

Study objectives
The primary objective is to assess the day 28/42 therapeutic efficacy of AL compared to AL + PQ and CQ compared to CQ + PQ against P. vivax infection. The secondary objective to compare AL vs CQ and AL+PQ vs CQ+PQ and also the overall number of Phase 1 of the study evaluates the clinical, parasitological, and haematological parameters for P. vivax infection over a 42-day follow-up period, to compare the schizontocidal drug efficacy of CQ

Study Question Endpoints for Comparison
A) Does the addition of 14 day course of supervised primaquine reduce the risk of treatment failure by 28/42 days Comparison of efficacy of CQ versus CQ+PQ and AL versus AL+PQ at day 28/42. First episode of malaria only. B) What is the most efficacious schizontocidal treatment for P. vivax?
Comparison of CQ versus AL efficacy at day 28 and day 42. First episode of malaria only. C) What is the most effective treatment for early parasite clearance?
Comparison of parasitaemia day 0-3 and fever clearance CQ vs CQ+PQ and AL vs AL+PQ and AL vs CQ D) What is the efficacy of low dose primaquine for radical cure?
Comparison of the cumulative risk and incidence rate of P. vivax recurrences at 12 months for CQ vs CQ+PQ and AL vs AL+PQ. E) What is the most efficacious radical cure for P. vivax?
Comparison of the cumulative risk of recurrence and incidence rate of P. vivax at 12 months for CQ+PQ vs AL+PQ F) What is the effectiveness of unsupervised primaquine Comparison of the cumulative risk and incidence rate of P. vivax recurrences at 12 months for CQ vs CQ+PQ and AL vs AL+PQ, using the second episodes of malaria (unsupervised) as starting point for the analyses and AL. Phase 2 continues the monthly follow-up of these patients for one year to assess the frequency of recurring vivax infections and thus the overall radical cure of CQ+PQ and AL+PQ over 12 months. The comparison of each drug (CQ and AL) with and without primaquine defines the antirelapse efficacy of the currently recommended dose of primaquine for P. vivax in Ethiopia.

Demographic and baseline characteristics
Once enrolled, all patients are randomized into four treatment groups. Patients in each group will be described with respect to baseline characteristics. These will include means (standard deviations) or medians (interquartile ranges) for non-normally distributed data. Categorical characteristics will be presented as counts and percentages.  The incidence rate of any parasitaemia at 1 year following treatment (includes any species of infection), comparing CQ with CQ+PQ, AL with AL+PQ and CQ+PQ with AL+PQ (Qu. D &E)

"Effectiveness" endpoints:
 The cumulative risk (time to first event) of recurrent P. vivax parasitaemia at 6 months following treatment of the second episode of malaria (unsupervised treatment), comparing with the enrollment episode (Qu. F). Detection of non-vivax parasitaemia before day 28 or 42 of follow up will result in readmission of antimalarial treatment, and thus these patients will be censored for the schizontocidal efficacy analyses (Qu A and B).

PCR adjustment:
Parasite genotyping can determine whether a recurrent parasitaemia is homologous to the initial infection. Recurrence of P. vivax genetically identical to the pre-treatment isolate can occur from either a true recrudescence of the initial infection or a relapse from hypnozoites generated from the prior blood stage infection; current molecular methods are unable to distinguish between these alternatives. However PCR adjusted efficacy can reduce the confounding effects of recurrent parasitaemias arising from new infection or relapse from a different strain. For this purpose recurrent samples are defined as homologous (recrudescence/relapse) if at least one allele is shared with the day 0 sample at each locus investigated and heterologous (re-infection/relapse) if no alleles are shared with the day 0 samples at 1 or more loci. Recurrence outcomes are defined as indeterminate if a sample pair exhibits no informative data at all loci investigated. Informative data from a minimum of 3 loci is required to call a homologous recurrence event and one locus to call a heterologous event. For calculation of the PCR-adjusted cure rate, homologous recurrences are classified as recrudescences (although these may include homologous relapses) and heterologous recurrences are classified as re-infections or heterologous relapses and are being censored. .
Patient data with samples negative for P. vivax by PCR at enrollment will be excluded in this analyses.

Adjudication of month 12 (Phase 2) treatment outcome assessment
When the number of scheduled visits after day 42 is incomplete the following rules will be applied: Deviation Effect More than 45 days without blood film examination after day 42 -In the survival analysis lost to follow up on day of last visit -In the incidence rate analysis the period of missing observation will be deducted from the total period of observation.

Adjudication for endpoint on parasite prevalence on day 0-3
For missed blood films (microscopy) the following rules will be applied:

Adjudication for endpoint on fever clearance on day 0-3
For missed temperature records, the same rules as for missing blood slides will apply.

Adjudication of hemoglobin outcomes
For missing haemoglobin measurements during follow up the following rules apply:

Outcome assessment Deviation Action
Fractional change in Hb between baseline and day 3 Missing Hb measurement on day 3 Exclude from analysis

Handling of missing data on drug course
Patients can have an incomplete course of treatment or data on drug administration may be missing.
No imputation of treatment course will be made for patients with missing data. Subgroup analyses may be performed for patients with full versus partial treatment courses if appropriate.

Handling of missing data for Adverse Events
For patients with missing data on adverse events the most conservative approach will be used: For the efficacy analysis, an intention-to-treat (ITT) and modified Intention to treat (mITT) approach will be adopted, with the ITT analysis being the primary approach, to provide comparison of the different drug treatments.

Intention to treat
To provide a pragmatic comparison of the different drug treatments, the principle of intention-totreat, will be the main strategy of analysis adopted for the primary and secondary endpoints. These analyses will be conducted on all patients assigned to the treatment groups as randomized, regardless of the study treatment received.

modified Intention to Treat
A modified intention to treat approach will be used as secondary analyses. Patients with protocol violations (e.g. wrong treatment or no treatment information available) will be censored form the respective episode onwards.

Safety Population
For the analysis of safety outcomes, all patients who effectively received any drug (i.e. at least one treatment dose) are included in the safety analysis in the treatment group they actually received.

Additional analyses (CQ drug concentrations) population
For the analyses of CQ drug levels only patients who failed before day 28 in the CQ arm and who had blood collected for drug levels will be included.
6 Statistical methods

Demographic and Baseline
Details of patients screened, those who meet the study inclusion criteria, those who are eligible and randomized, those who are eligible but not randomized (e.g. G6PD deficient), those who withdraw from the study after randomization and those who are lost to follow-up will be summarized in CONSORT flow diagrams (one until day 42 endpoint and one until end of study).
The number of patients discontinuing from the study will be tabulated by reason for study discontinuation. The number (%) of patients attending scheduled follow-up visits by study day (days 1 to 28 and 42, monthly visits until 1 year) will be reported.
The baseline value is defined as the last available value before randomization.

Demographic characteristics
 Gender: male / female  Median Age (years)  Age in classes: 12 months up to 5 years, 5 years up to 15 years, ≥15 years  Median Weight

Prior or concomitant medications
All medications taken within fourteen days before randomization and until the end of the study are reported in the case report form pages.
• Prior medications are those the patient used prior (seven days before) to first study drug intake. Prior medications can be discontinued before first administration or can be ongoing during treatment phase. • Concomitant medications are those used during the treatment with the study drug and during the first phase of follow up until day 42. Concomitant medications can be ongoing prior medication and/or continue into post treatment medication. • Post treatment medications are those the patient took in the period running from the day after day 42 up to the end of the study.

Efficacy analyses
In this study, patients with repeated episodes are treated with the same regimen they received for their initial episode. However only the first episode is mostly directly observed (all schizonticidal treatment is observed but primaquine administration is partially observed), whereas for all consecutive episodes patients are advised to take their treatment at home (unsupervised). The analyses for schizontocidal efficacy therefore only takes the first episode into consideration (enrollment episode). For the "effectiveness" analyses only episodes following the first recurrences (ie second and greater episodes, using the second episode as starting point) will be included.

Analyses of primary efficacy endpoints
6.3.1 Cumulative risk of P. vivax treatment failure at day 28/42 following treatment with AL compared to AL+PQ (Qu. A) The response to treatment will be assessed using survival analyses (Kaplan Meier). Comparisons between treatment arms will be presented as Hazard Ratios using a Cox regression model. The proportional hazards assumption will be assessed by comparing visually the log(cumulative hazard) by time of follow-up curves for each covariable category and subsequently by fitting and comparing models with and without time of follow-up interaction terms.
In the survival analyses patients with a recurrent vivax parasitaemia before day 28/42 are considered failures, patients lost to follow up or with other a non-vivax parasitaemia before day 28/42 are censored at the respective day. Patients with no recurrence until day 28/42 are considered cured for the purpose of this analysis. Incidence risk of P. vivax treatment failures at day 28/42 following treatment with CQ compared to CQ+PQ (Qu. A) 6.4 Analyses of secondary efficacy endpoints 6.4.1 Schizontocidal Endpoints: All analyses for schizontocidal endpoints refer to the first malaria episode (enrollment episode).

Cumulative risk of P. vivax treatment failures at day 28 and day 42 following treatment
with CQ compared to AL (Qu. B) The incidence risk of recurrent P. vivax over 28 and 42 days will be assessed using survival analyses (Kaplan Meier), similar to 6.3.1.
6.4.1.2 PCR-adjusted cumulative risk of P. vivax at day 28 and 42 following treatment with AL compared to CQ (Qu. B) The incidence risk (cumulative risk) of recurrent P. vivax over 28 and 42 days will be assessed by survival analysis (Kaplan Meier), similar to 6.3.1, but parasites with different genotypes will be censored from the analysis. If in all of the markers used at least one allele is matching in the enrollment sample and in the sample taken at the day of recurrence, the recurrence will be considered the same infection. Otherwise, it will be considered a different infection. In the event that molecular data is missing from the day of recurrence, the patient will be censored at this day without a recorded endpoint outcome (i.e., that patient will not contribute any influence on the risk estimate).

Cumulative risk of any parasitaemia at day 28 and 42 following treatment with AL and CQ (includes any species of infection) (Qu. B)
If there are enough non-vivax episodes, the incidence risk of any recurrent parasitaemia over 28 and 42 days will be assessed by survival analysis (Kaplan Meier). In this analysis patients with any recurrent parasitaemia before day 28/42 are considered failures, patients lost to follow up before day 28/42 are censored at the respective day. Patients with no recurrence until day 28/42 are considered cured for the purpose of this analysis. The comparison of risks between CQ vs AL will inform overall schizontocidal efficacy and post treatment prophylaxis.
6.4.1.4 Proportion of patients with P. vivax parasitaemia on day 1, 2 and 3 after treatment with AL, CQ, AL+PQ and CQ+AL (Qu. C) Parasite clearance will be presented using the proportions of patients who remain parasitaemic on each day (days 1 to 3).

Proportion of patients with fever on day 1, 2 and 3 after treatment with AL , CQ, AL+PQ
and CQ+PQ (Qu. C) Fever clearance time will be presented as the proportion of patients who were febrile or had a history of fever on day 0 and who became febrile (<37.5°C) day 1, 2 and 3 with no subsequent measured fever or history of fever on subsequent daily within the next 48 hours. Similar rules as for the parasite clearance will be used to handle missing visits.

Incidence rate of all recurrent episodes of P. vivax parasitaemia over one year (Qu. D &E)
The total number of P. vivax episodes occurring during the 1 year follow up will be presented as incidence estimate in person year of observation (PYO) per treatment arm.
Incidence rates will be calculated by dividing the number of P. vivax episodes by the number of person-years of observation (PYO) in the study population. On the individual patient level the start date for PYO is the day of enrolment into the study, stop date is the last visit performed (either completed study at 1 year or any last visit before lost to follow up and/or censoring). The period between start and stop dates for each patient will be calculated in days and divided by 365 to determine PYO, which will then be summed for all participants.
Patients receiving antimalarial treatment will be assumed to have a period of 28 days of post treatment prophylaxis in the chloroquine arm and 14 days in the AL arm following each antimalarial treatment and this period will thus be subtracted from their total period of follow up (time at risk). Patients who received Quinine will have 7 days subtracted.
For this analysis an intention to treat approach is used as the primary approach. Patients who received a treatment other than allocated at enrollment for other episodes with P. vivax monoinfections are excluded from this time point in the mITT analyses. Comparison between incidence rates will be presented as Incidence Rate Ratios (IRR) calculated using a negative binomial regression model.
The comparison of risks between CQ vs CQ+PQ and AL vs AL+PQ will inform primaquine efficacy (Qu. D) and the comparison of risks between CQ+PQ vs AL+PQ will inform overall efficacy for radical cure (Qu. E).
In a sensitivity analysis the incidence will be recalculated without subtracting the period of post treatment prophylaxis, to quantify the total incidence rate of all recurrences.

The cumulative risk of recurrent P. vivax parasitaemia over one year (Qu. D &E)
The incidence risk of recurrent P. vivax over total follow up (up to 12 months) will be assessed by survival analysis (Kaplan Meier). In this analysis patients with a recurrent vivax parasitaemia before month 12 are considered failures, patients lost to follow up or with other a non-vivax parasitaemia before month 12 are censored at the respective day. Patients with no recurrence until month 12 are considered cured for the purpose of this analysis.
The comparison of risks between CQ vs CQ+PQ and AL vs AL+PQ will inform primaquine efficacy (Qu D) and the comparison of risks between CQ+PQ vs AL+PQ will inform overall efficacy for radical cure (Qu E).

Incidence rate of any parasitaemia at 1 year following treatment (includes any species of infection) (Qu. D & E)
If there are enough non-vivax episodes the analyses will be performed similar as outlined in 6.4.2.1 but for any parasitaemia irrespective of species. In this analysis, patients with any recurrent parasitaemia before month 12 are considered failures, patients lost to follow up before month 12 are censored at the respective day. Patients with no recurrence until month 12 are considered cured for the purpose of this analysis.

Effectiveness endpoints
For the effectivness analysis the enrollment episode (semi-supervised) ) will be compared to the first recurrence (unuspervised) This analysis will only be conducted if a sufficient number of cases with recurrences are available. The incidence risk of recurrent P. vivax parasitaemia at 12 months following unsupervised treatment (Qu. F) The cumulative risk of recurrent P. vivax after unsupervised treatment at 1 year will be assessed by survival analysis (Kaplan Meier). The starting point of the survival analysis will be the day of the first vivax recurrence. Only patients with a recurrence will therefore be included in this analysis. Patients with a recurrent vivax parasitaemia thereafter and before month 12 are considered failures, patients lost to follow up or with another non-vivax parasitaemia before month 12 are censored at the respective day. Patients with no recurrence until month 12 are considered cured for the purpose of this analysis. Comparison will be made within treatment arms with risk of recurrence after supervised treatment (6.4.2.2).
A similar subgroup analyses will be performed to compare patients who received a full dose and those who received an incomplete dose due to missed visit or not adherence on the days between supervised treatment days.

Analyses of safety data
The summary statistics (including number, mean, median, interquartile range, standard deviation, minimum and maximum) of all vital signs variables (values and changes from baseline) will be calculated for each visit by treatment group.