After upload to the WWARN Data Repository, WWARN will standardise data sets according to the WWARN Clinical Data Management and Statistical Analysis Plan and pool all primary data sets into a single database of quality-assured individual patient data.
Confounders will be selected using a causal diagram framework.
a) The proportion of patients with mild or moderate anaemia at day 90, 180 and 360 will be determined and described stratified by the use and dose of primaquine, number of recurrences, timing of recurrences, and antimalarial elimination half-life.
b) Linear mixed effects modelling will be used to assess the impact of primaquine on haemoglobin over time, with estimation of the effect of primaquine on haemoglobin at the outcome day (day 90, 180 and 360) derived from the model
c) A Poisson regression model to assess incidence rate of moderate anaemia over 12 months will be undertaken on the subset of studies that followed patients actively and measured haemoglobin at least monthly.
d) Cox regression analysis for the time to first episode of moderate and moderately severe anaemia during follow-up (90, 180 and 360 days) will be performed separately, with shared frailty for study-site.
e) The effect of the total number of malaria episodes prior to the outcome day and timing of recurrences prior to the haemoglobin on the outcome day at day 90, 180 and 360 will be assessed using separate multivariable linear mixed-effects modelling.
f) The effect of antimalarial half-life on haemoglobin at the outcome day (day 60, 90, 180 and 360) will be assessed using multivariable linear mixed-effects modelling.
g) The analyses described in d) to f) above will be repeated using the clinical threshold of anaemia as the outcome using a multivariable logistic mixed effects model.
A detailed statistical analysis plan has been developed.