Search

Plasmodium falciparum malaria remains one of the greatest causes of mortality and morbidity globally. Estimates of treatment efficacy provide critical information for optimising treatment guidelines. The aim of this thesis was to quantify the magnitude of variation in the derived estimates attributable to study design and analytical approaches in uncomplicated P. falciparum in patients treated with artemisinin combination therapies (ACTs). This thesis utilised individual patient data collated from 92 clinical trials, including 31,507 patients treated with ACTs extracted from WorldWide Antimalarial Resistance Network (WWARN) repository. This thesis presents a number of novel findings relevant to clinical researchers and policy makers. The current recommended follow-up period of 28 days with artemether-lumefantrine regimen missed an estimated 58% [95% Confidence Interval (CI): 47% - 74%] of treatment failures in Africa and 32% [95% CI: 15% - 45%] in Asia. For dihydroartemisinin-piperaquine (DP) regimen, a study with 42 day follow-up was estimated to miss 47% [95% CI: 19 - 90%] of the treatment failures in those who were adequately dosed. These results confirmed that studies adhering to the current follow-up recommendations provide an overestimate of the drug efficacy and longer follow-up is warranted for these regimens to adequately capture treatment failures. The derived estimates were further vulnerable to analytical approaches adopted. Kaplan-Meier (K-M) approach, the currently recommended method, overestimated treatment failure by up to 3.6% compared to a competing risk survival analysis approach. Simulation studies demonstrated that the magnitude of this overestimation increased with declining drug efficacy in high transmission areas, where a substantial fraction of patients develop new infections. Similarly, excluding patients with indeterminant outcomes (i.e. whether parasite recurrence is a treatment failure or new infection cannot be determined) was associated with an upwards bias in the derived estimate of efficacy by up to 1.7% compared to using the sophisticated methods of multiple imputation and inverse probability weighting for dealing with the indeterminant outcomes. This thesis highlighted that the choice of analytical approaches led to variability around the derived estimates of drug efficacy. The findings of this thesis are discussed with regards to current World Health Organization guidance to change treatment policy.