Bulletin of the World Health Organization

Cost–effectiveness of artemisinin combination therapy for uncomplicated malaria in children: data from Papua New Guinea

Wendy A Davis a, Philip M Clarke b, Peter M Siba c, Harin A Karunajeewa a, Carol Davy d, Ivo Mueller c & Timothy ME Davis a

a. University of Western Australia, Fremantle Hospital, PO Box 480, Fremantle 6959, Western Australia, Australia.
b. University of Sydney, School of Public Health, Sydney, Australia.
c. Papua New Guinea Institute of Medical Research, Madang, Papua New Guinea.
d. University of Adelaide, School of Population Health, Adelaide, Australia.

Correspondence to Timothy ME Davis (e-mail: tdavis@cyllene.uwa.edu.au).

(Submitted: 01 November 2010 – Revised version received: 21 December 2010 – Accepted: 04 January 2011 – Published online: 01 February 2011.)

Bulletin of the World Health Organization 2011;89:211-220. doi: 10.2471/BLT.10.084103


Recent estimates show that Plasmodium falciparum causes 225 to 500 million cases of malaria and nearly one million malaria-related deaths worldwide each year.13 Of these deaths, 96% take place in low-income countries and 87% in children younger than 5 years of age. An additional 70 to 390 million people develop P. vivax malaria annually.3 Although P. vivax is generally thought to cause benign disease, it can lead to severe illness and even death.4 Malaria generates an annual loss of about 34 million disability-adjusted-life-years, is the twelfth leading cause of morbidity in the world,2 and undermines society and the economy in many ways.5 In places with intense disease transmission, its long-term detrimental repercussions on economic growth and development transcend the additive costs of individual cases.5

Fortunately, the burden of malaria can be reduced through effective case management and vector control measures, including the use of insecticide-treated bed nets and indoor residual insecticide spraying.5 Although choosing the right antimalarial treatment is crucial for effective management of uncomplicated infections, cost–effectiveness studies have been few. All the studies performed since the 1990s, when artemisinin combination therapy (ACT) became available, have focused on sub-Saharan Africa and India and on P. falciparum malaria610 despite increasing recognition of the pathogenic importance of P. vivax.11 Wide differences in the cost and efficacy of ACT regimens make comparative cost–effectiveness studies indispensable.

Both P. vivax and P. falciparum exist in Oceania and parts of Asia, where malaria is hyper- or holoendemic. A particularly complex epidemiological situation exists in low-lying areas of Papua New Guinea,12,13 where we recently conducted a trial comparing three ACTs with conventional therapy for uncomplicated paediatric malaria.14 In this paper we assess the relative cost–effectiveness of these four treatment regimens.


An open-label, randomized, parallel-group study (Australian New Zealand Clinical Trials Registry ACTRN12605000550606) was conducted between April 2005 and July 2007 in two northern coastal provinces of Papua New Guinea (Madang and East Sepik) among children 6 to 60 months of age with uncomplicated P. falciparum or P. vivax malaria who presented to two rural clinics. All children were randomly allocated to one of the following: (i) conventional treatment with chloroquine (CQ) plus sulfadoxine (S) plus pyrimethamine (P), henceforth referred to as CQ+S+P; (ii) artesunate (ARTS) plus S plus P, henceforth ARTS+S+P; (iii) dihydroartemisinin (DHA) plus piperaquine (PQ), henceforth DHA+PQ; and (iv) artemether (A) plus lumefantrine (L), henceforth A+L. Recruitment, informed consent and other study procedures have been described in detail elsewhere.14 The present comparative assessment of cost–effectiveness is based on data from 656 children (482 with P. falciparum malaria and 195 with P. vivax malaria, since 21 children had both pathogens) who were recruited based on eligibility and followed according to the study protocol (Fig. 1).

Fig. 1. Number of patients in trial of antimalarial treatment regimens conducted in 2005-2007 in Papua New Guinea who were eligible for participation and were followed from randomization through day 42. Cost-effectiveness analyses were done using 2007–2008 cost data
Fig. 1. Number of patients in trial of antimalarial treatment regimens conducted in 2005-2007 in Papua New Guinea who were eligible for participation and were followed from randomization through day 42. Cost-effectiveness analyses were done using 2007–2008 cost data
PCR, polymerase chain reaction.a “PCR-corrected” denotes correction for re-infections identified by means of polymerase chain reaction genotyping of polymorphic Plasmodium falciparum loci.

Drugs and dosage

The drugs used in the trial were from the following laboratories and were administered in the following combinations and doses:

  • CQ+S+P: CQ (Aspen Health, St Leonards, Australia), 10 milligrams (mg) of base per kilogram (kg) of body weight daily for 3 days, plus S and P (Roche, Basel, Switzerland) as 25 mg per kg of S plus 1.25 mg per kg of P with the first dose of CQ;
  • ARTS+S+P: ARTS (Sanofi-Aventis, Paris, France), 4 mg per kg daily for 3 days, plus S and P as 25 mg per kg of S plus 1.25 mg per kg of P with the first dose of ARTS;
  • DHA+PQ (Beijing Holley-Cotec, Beijing, China): DHA, 2.5 mg per kg, plus PQ phosphate, 20 mg per kg, daily for 3 days;
  • A+L (Novartis Pharma AG, Basel, Switzerland): A, 1.7 mg per kg, plus L, 10 mg per kg, twice daily for 3 days.

All drugs were administered with water except for A+L, which was given with milk, as instructed by the manufacturer, to increase bioavailability. All doses were supervised except for the evening doses of A+L, which were taken at home.

Clinical follow-up

Axillary temperature measurements and blood film microscopy were scheduled on days 1, 2, 3, 7, 14, 28 and 42. Asymptomatic children who developed uncomplicated or severe malaria during follow-up were given rescue antimalarial therapy consisting of quinine (7–20 mg of base per kg twice daily for 7 days) intramuscularly if unable to tolerate oral therapy, or otherwise orally plus a single dose of S+P. Efficacy was assessed in accordance with World Health Organization (WHO) definitions,15 namely, (i) early treatment failure: signs of severe disease16 or an inadequate parasitological response by day 3; (ii) late parasitological failure: parasitaemia between days 4 and 42 (after a favourable response on days 1–3); (iii) late clinical failure: late parasitological failure plus an axillary temperature > 37.5 °C, and (iv) adequate parasitological and clinical response: none of the conditions in i–iii. In cases of P. falciparum malaria, these outcomes were corrected for re-infection by genotyping the disease pathogen through polymerase chain reaction (PCR).14 Since no equivalent protocols have been established for re-emergent P. vivax,11 treatment outcomes among patients with P. vivax malaria were not corrected.

Economic analyses

Our economic analyses focused primarily on societal costs. We estimated the cost of travelling between home and the health clinic plus the direct costs of health care (i.e. conventional treatment and ACTs, visits to health clinics, clinical tests and rescue antimalarial therapy, when indicated). We compared the net costs and net effectiveness of the three ACTs with those of conventional therapy with CQ+S+P and expressed the results as cost–effectiveness ratios. All analyses and comparisons were performed separately for P. falciparum and P. vivax on both a per protocol and a modified intention-to-treat basis.14 Per protocol analyses included children with complete follow-up or confirmed treatment failure and excluded those who were treated for malaria without confirmatory microscopy or who defaulted from follow-up. These excluded patients were retained in the modified intention-to-treat group, to which we applied: (i) a worst-case approach (early treatment failure assumed for those excluded on or before day 3 and late parasitological failure or late clinical failure assumed for all others) and (ii) a best-case approach (all missing follow-up blood films were assumed to be parasite-negative). In a secondary analysis, we extrapolated outcomes to estimate the increase in life expectancy obtained with the most effective treatment based on the estimated mortality associated with P. falciparum and the remaining life expectancy.

For each patient, standardized data were collected at each scheduled clinic visit and at any extra unscheduled visits on days when fever or other symptoms were present (sick days). These data included the doses of all the drugs being used to treat the malaria and its symptoms or complications (e.g. trial medication, rescue quinine, paracetamol, iron and folate supplements, etc.). Unit costs were obtained from the Papua New Guinea National Department of Health under the Medium-Term Expenditure Framework,17 participating clinics, Novartis Pharma AG and local suppliers (Table 1) and were combined with resource volumes to obtain a net cost per patient during follow-up. Mean net costs and associated 95% confidence intervals (CIs) were calculated for each treatment arm. Costs are reported in 2007–2008 United States dollars (US$) and undiscounted due to the relative brevity of the trial.

All study participants were scheduled for eight clinic visits (including day 0 and excluding sick days). However, since usual-care (non-trial) visits are different and less frequent, we conducted a complementary analysis with the costs of conventional therapy or ACT treatment based on a single clinic visit comprising both diagnosis and treatment. We assumed the same number of subsequent sick-day visits for all children except those with early treatment failure, whose clinic visit scheduled for day 1–3 was replaced by a sick-day visit. For each patient, the actual cost of trial visits was replaced by the estimated cost of standard practice, which depended on treatment allocation. We assumed that no-cost forms of fat, such as in breast milk or fat-containing foods, were always consumed with A+L except during the initial clinic visit (day 0).

The primary endpoint of the trial was treatment failure by day 42.14 A secondary analysis of the lifetime benefits of using A+L to treat P. falciparum malaria was performed by using lifetables to estimate the potential life years saved. All results are reported as means and standard deviations (SDs) or mean differences and 95% CIs. The CIs for the key cost–effectiveness ratios were calculated by Fieller's method.18 The effect of assumptions on the main results was examined by sensitivity analyses involving a cost–effectiveness analysis using best- and worst-case scenarios and modified intention-to-treat assumptions. All data were analysed with SPSS 15.0 (SPSS Inc., Chicago, United States of America) and Microsoft Excel 97 (Redmond, USA).



Table 2 shows the mean cost per patient for CQ+S+P and the three ACTs and the mean cost difference between these regimens over the duration of the study, by type of cost and allocation for P. falciparum and P. vivax in both trial and usual care settings. For P. falciparum malaria, ARTS+S+P, DHA+PQ and A+L cost between US$ 0.26 and US$ 0.48 more per patient, on average, than CQ+S+P. The A+L group had the additional cost of milk. Transportation costs made clinic visits the most expensive component. Other costs – quinine rescue treatment, treatment for anaemia, paracetamol or clinic visits – did not differ significantly between conventional therapy and ACT regimens. The total costs of usual care were significantly higher in the ARTS+S+P and A+L groups than in the conventional therapy group.

For P. vivax malaria, ARTS+S+P, DHA+PQ and A+L cost between US$ 0.22 and US$ 0.36 more per patient, on average, than CQ+S+P (Table 2). Again, the A+L group had the additional cost of milk. Other costs did not differ significantly between conventional therapy and ACT regimens. The total costs of usual care were significantly higher in the A+L group than in the group that received CQ+S+P.


In the per protocol analysis of the 388 children with P. falciparum malaria who completed the trial, 341 (87.9%) had an adequate parasitological and clinical response, with the highest rate of success (95.2%) in the A+L group, followed by the DHA+PQ group (88.0%), the ARTS+S+P group (85.4%) and the CQ+S+P group (81.5%).14 Among the 154 children with P. vivax who attended the clinic on day 42, 69.4% had an adequate parasitological and clinical response in the DHA+PQ group compared with 13.0%, 33.3% and 30.3% in the CQ+S+P, ARTS+S+P and A+L groups, respectively.14 Table 3 documents the proportion of treatment successes and costs for each pathogen species and type of analysis (per protocol versus modified intention-to-treat) in a usual care setting. The incremental number of successes and costs, together with the incremental cost–effectiveness ratios, are also shown for each ACT compared with CQ+S+P.


The primary measure of cost–effectiveness was the incremental cost–effectiveness ratio, defined as the incremental cost per incremental treatment success of ACT relative to the comparator, CQ+S+P (see cost–effectiveness planes in Fig. 2). In a usual care setting, DHA+P+Q and A+L are cost-effective alternatives to CQ+S+P. For DHA+PQ the average cost per treatment success was US$ 2.95, with an 87% probability of the cost–effectiveness ratio being < US$ 50.00. For A+L the average cost per treatment success was US$ 6.97, with a 99% probability of this being < US$ 50.00. The incremental cost–effectiveness ratio of A+L relative to DHA+PQ, the next best alternative, was US$ 10.60 per success, with a 92% probability of this ratio being < US$ 50.00. For P. vivax (Fig. 2), DHA+PQ was the most effective treatment, with an average cost saving of US$ 0.18 per success in a usual care setting when compared with CQ+S+P, and a > 99% probability of the cost per success being < $1.00.

Fig. 2. Planes showing the cost–effectiveness of each intervention relative to conventional treatment with chloroquine (CQ) plus sulfadoxine (S) plus pyrimethamine (P) for children with (a) Plasmodium falciparum malaria and (b) P. vivax malaria, in a usual care setting, Papua New Guinea, 2007–2008
Fig. 2. Planes showing the cost–effectiveness of each intervention relative to conventional treatment with chloroquine (CQ) plus sulfadoxine (S) plus pyrimethamine (P) for children with (a) <em>Plasmodium falciparum</em> malaria and (b) <em>P. vivax</em> malaria, in a usual care setting, Papua New Guinea, 2007–2008
ACT, artemisinin combination therapy.

Cost per life year saved by artemether plus lumefantrine

In 2008, a child aged between 1 and 4 years in Papua New Guinea could expect to live another 64.9 years.19 Mortality from P. falciparum malaria treatment failure has been estimated at 0.185%, or about 1.9 deaths per 1000 patients.20 In our study, the incremental success of A+L over CQ+S+P in children 6–60 months old was 13.7% (95.2% versus 81.5% success). Thus, for every 1000 patients with P. falciparum malaria who are treated with A+L instead of CQ+S+P, an additional 0.253 life is saved (0.00185 × 1000 × 0.137 = 0.253). The increase in average life expectancy per 1000 cases of P. falciparum malaria treated with A+L instead of CQ+S+P is approximately 0.253 × 64.9 years = 16.4 years (or 12.8 years when discounted at 3%, as recommended by Gold et al.21). The extra cost associated with A+L versus CQ+S+P treatment was US$ 955 per 1000 cases treated (per protocol analysis). Therefore, the cost of A+L per life year saved was US$ 955/16.4 = US$ 58.23 (or US$ 74.6 when benefits are discounted at 3%).


This is the first economic analysis of a range of contemporary treatment options for children living in an area of intense malaria transmission caused by several species of Plasmodium. According to the findings, in Papua New Guinea the most cost-effective treatment for paediatric P. falciparum malaria differs from that for paediatric P. vivax malaria for clinical and financial reasons. In a usual care setting, three ACTs proved more effective against P. falciparum malaria but also more costly than CQ+S+P. Both A+L and DHA+PQ were found to be cost-effective alternatives to CQ+S+P, with A+L being the more effective of the two drugs but also the most expensive. The availability of donated or subsidized drugs and willingness to pay can determine the choice of regimen. For P. vivax malaria, DHA+PQ was found to be the most effective treatment and led to savings in a usual care setting when compared with CQ+S+P. A+L was the least cost-effective ACT regimen against P. vivax malaria among children in Papua New Guinea.

Our data for P. falciparum malaria are comparable to those from other studies. In a 2005 prospective observational study in Zambia,6 the incremental cost–effectiveness ratio for A+L versus S+P was estimated at US$ 4.10 per case successfully treated from the health care provider’s perspective. This is similar to the US$ 6.97 per success found in the present analyses. In a study from KwaZulu Natal Province in South Africa, A+L proved more cost-effective than S+P at the population level despite its greater cost, mainly owing to a reduction in local transmission.9 A predictive modelling study concluded that ACTs are extremely likely to be cost-effective under most conditions, unless resistance to S+P is very low.7 The findings of an Indian study comparing CQ monotherapy with mefloquine and A+L were difficult to interpret because the cost of A+L was unavailable and many patients on CQ experienced early treatment failure.8 In studies from two sub-Saharan regions, high ACT coverage was found to be the most cost-effective strategy for malaria control (9–12 international dollars per disability-adjusted life year averted) compared with insecticide-treated bed nets, indoor residual spraying and intermittent preventive treatment in pregnancy.22

Some population data from Papua New Guinea are detailed enough to estimate the benefits of ACT regimens. In 2004, 1 660 645 malaria patients (28.7% of the total population, assuming no double counting) were treated in Papua New Guinea.20 Of these patients, 92 956 required further treatment as outpatients and 29 406 as inpatients.19 These figures suggest a substantial need for secondary treatment or for treatment for severe malaria. Our trial data showed that nearly 70% of uncomplicated malaria cases were caused by P. falciparum and 30% by P. vivax. If the findings of the trial’s per protocol analysis are projected to the general population of Papua New Guinea (5 789 796), US$ 49 321 per year could be saved and 280 981 more cases could be successfully treated every year by using DHA+PQ instead of CQ+S+P to treat children with P. vivax malaria. If A+L were used to treat P. falciparum malaria, an extra US$ 1 110 141 would have to be expended to attain an additional 159 256 treatment successes every year. These estimates may have to be revised, however, if in future the use of insecticide-treated bed nets and other integrated malaria control measures increases and the malaria burden decreases in Papua New Guinea.23

At a cost per life year saved of US$ 58, A+L appears to be a highly cost-effective and affordable regimen for uncomplicated paediatric malaria in Papua New Guinea, where the gross national income per capita in 2008 was US$ 1040.24 In a large multicentre trial that compared ARTS with quinine for the treatment of severe malaria in south-east Asia, the incremental cost per death averted by the use of ARTS was US$ 140 from the provider’s perspective,25 and it was concluded that substituting quinine with ARTS provided a return on investment that few health interventions could match in terms of immediate health gains and minimal additional cost.

Treatment compliance has not been considered in the present economic analysis. All treatments were given over three days, but A+L requires two doses per day and the medication has to be taken with a fatty meal, so that compliance may be lower than with other therapies. In a cost–effectiveness study of ACTs in sub-Saharan Africa, estimated compliance with ACTs ranged from 30% to 60% compared with 85% to 95% for S+P.7 Opportunity costs, such as the time parents took off work to look after their sick children and take them to the health clinic, were not factored into our analysis. Prompt recovery after effective treatment might allow parents to return to work more quickly. The intervention trial was not designed to investigate the impact of mixed infections.14 The fact that the most effective novel treatment for P. falciparum malaria proved to be the least effective for P. vivax malaria may have implications in terms of cost–effectiveness, but the present data suggest that DHA+PQ would be the regimen of choice for mixed infections.

As conventionally recommended in cost–effectiveness analyses, we based our estimates on the primary trial endpoints. In countries such as Papua New Guinea, PCR is not performed as part of usual care to diagnose malaria and monitor treatment effectiveness. Worth noting in this regard is the absence of between-group differences in adequate clinical and parasitological response rates on day 42, without PCR correction, among children with P. falciparum malaria (67.9%, 63.9%, 62.6% and 64.2% for CQ+S+P, ARTS+S+P, DHA+PQ and A+L, respectively).14 In the absence of PCR correction, CQ+S+P is substantially more cost-effective than the other three regimens because it is the least expensive. However, most early treatment failures in the intervention trial were observed in the CQ+S+P group. Thus, continued use of CQ+S+P in the holoendemic setting where the trial was conducted is likely to result in increased parasite resistance and a greater disease burden.14 For this reason, WHO has proscribed this type of therapeutic procrastination.15

In areas of Papua New Guinea and of other resource-poor tropical countries, microscopy and/or rapid diagnostic tests may not be available or reliable.1 Empirical treatment may be administered based on symptoms and on the most likely infecting Plasmodium species. Although an economic analysis of such practices was beyond the scope of the present study, a progressive reduction in the cost of rapid diagnostic tests and a resulting increase in their availability may improve the accuracy of malaria diagnosis in countries such as Papua New Guinea, and this, in turn, would reduce the cost of unnecessary treatment. We did not consider the cost implications of continuing to carry Plasmodium gametocytes after treatment. Although this occurred much more often after CQ+S+P than after any of the ACTs,14 there is emerging evidence that S+P impairs both P. falciparum gametocyte infectivity and Anopheles mosquito survival.26

Our intervention trial14 has led to changes in the national malaria treatment guidelines; A+L has replaced CQ+S+P as the recommended first-line therapy for uncomplicated malaria and is being distributed throughout the country with the help of the Global Fund to Fight AIDS, Tuberculosis and Malaria, and other entities.1 DHA+PQ has been recommended as an alternative regimen because it is more effective than A+L against P. vivax. Our cost–effectiveness data support these recommendations, but further research is needed to determine if they are broadly applicable to parts of Oceania and Asia where CQ+S+P is losing its effectiveness and ACTs are available.


The main trial was sponsored by WHO Western Pacific Region, Rotary Against Malaria (Papua New Guinea) and the National Health and Medical Research Council of Australia (grant 353663). TMED was supported by an NHMRC Practitioner Fellowship.

Competing interests:

None declared.