Latest research explores new statistical approaches for estimating antimalarial drug efficacy

WWARN’s Clinical Scientific Group have published research in BMC Medical Research methodology and Malaria Journal comparing different statistical approaches for deriving cumulative estimates of drug efficacy from clinical studies. Results indicate that the Cumulative Incidence Function (CIF) approach should be considered as an alternative to the widely used Kaplan-Meier method for calculating efficacy estimates in high transmission areas.


In clinical studies looking at the efficacy of antimalarial treatments for uncomplicated Plasmodium falciparum malaria, the primary measurement in determining efficacy is the level of recurrence of parasites which are genetically identical to parasites in the initial infection i.e. recrudescence. Recurrent infection can however occur as a result of a new infection with P. falciparum or another species of malaria parasite. Recrudescent infection can be potentially pre-empted when the number of parasites of the newly acquired infection outnumbers those of the existing infection, or if the new infection is due to a more resistant parasite strain. This gives rise to a scenario where the new infection masks or outcompetes the recrudescent infection. Such events are considered as competing risk events in statistical literature.

The Kaplan-Meier (K-M) method is currently recommended by the World Health Organization (WHO) for estimating the efficacy of antimalarial treatments but does not account for competing risk events. The Cumulative Incidence Function (CIF) provides an alternative approach for estimating efficacy by accounting for competing risk events.

Authors set out to comprehensively investigate how the choice of analytical method can impact the derived estimates and interpretation of drug efficacy data in antimalarial clinical studies. In the first instance, simulation studies were carried out for different scenarios of existing drug efficacy in  areas of high and low malaria transmission intensity.

The simulation study showed substantial differences in the derived estimates of drug efficacy between the two methods in areas of high malaria transmission. The Kaplan-Meier approach overestimated failure (underestimated the efficacy) and the degree of overestimation in treatment failure reached as high as 3% when the drug efficacy fell to 90%.

In the next study, comparison of the two methods was applied to real datasets. Data were used from studies collected in the WWARN data repository, this included 92 studies conducted between 1996 and 2014, accounting for 31,379 patients. Similar to the previous study, the K-M method overestimated recrudescence in high transmission settings, becoming increasingly pronounced as antimalarial efficacy decreased. 

Just over a third of treatments in the studies had an observed proportion of new infections greater than 10% (a threshold considered to make studies vulnerable to competing risk bias), suggesting that situations with competing risk events are the rule rather than the exception in high transmission settings.

First author of the studies, Prabin Dahal, commented: "Competing risk analysis has not gathered much attention in antimalarial literature. We undertook these studies to compare the methods both in a simulation study and with real data. We explored when and where does the competing risk approach prove to be a useful alternative to the currently used Kaplan-Meier approach for measuring antimalarial drug efficacy. We were able to demonstrate that the choice of analytical approach can have implications on the derived estimate, especially in the areas of high transmission settings. Hence, it is important to utilise the current statistical tools at our disposal to maximise the information obtained from the clinical trials. It is also equally important to take into consideration the biological plausibility of a new infection outcompeting an existing recrudescent infection, which is very difficult to discern with data collected during a clinical trial. We suggest interpreting the results of competing risk analysis by taking this limitation into account."

These results have important clinical consequences, as for new drugs to be eligible as first line treatments the estimates of their efficacy have to exceed certain thresholds (e.g. 95% is the desired efficacy for introducing a new drug as a first line therapy and 90% efficacy is required for existing drugs). Both studies demonstrate that such a threshold based approach for judging the current status of antimalarial efficacy is vulnerable to the choice of analytical approach used, especially when the derived estimates are at the cusp of these thresholds of drugs being accepted or rejected. Authors suggest that in high transmission settings, where large proportions of recurrence are attributed to new infections, competing risk survival analysis provides an alternative approach for estimating drug efficacy.

Find out more about the WWARN Clinical Scientific Group

Read the papers:

Evaluating antimalarial efficacy in single-armed and comparative drug trials using competing risk survival analysis: a simulation study. BMC Medical Research Methodology. 2019; 19:107.

Competing risk events in antimalarial drug trials in uncomplicated Plasmodium falciparum malaria: a WorldWide Antimalarial Resistance Network individual participant data meta-analysis. Malaria Journal. 2019. 18:225.