Cost-effectiveness of artesunate for the treatment of severe malaria
Corresponding Author Y. Lubell, London Schol of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK. E-mail: email@example.com
Objective To explore the cost-effectiveness of artesunate against quinine based principally on the findings of a large multi-centre trial carried out in Southeast Asia.
Methods Trial data were used to compare mortality of patients with severe malaria, treated with either artesunate or quinine. This was combined with retrospectively collected cost data to estimate the incremental cost per death averted with the use of artesunate instead of quinine.
Results The incremental cost per death averted using artesunate was approximately 140 USD. Artesunate maintained this high level of cost-effectiveness also when allowing for the uncertainty surrounding the cost and effectiveness assessments.
Conclusion This analysis confirms the vast superiority of artesunate for treatment of severe malaria from an economic as well as a clinical perspective.
Objectif: L’artésunate a été démontré supérieur à la quinine dans la prise en charge des cas sévères de malaria dans les essais cliniques menés en Asie. L’étude ci-présente explore le rapport efficacité-coût de l’artésunate, en se basant principalement sur les résultats d’une large étude multicentrique menée en Asie du sud-est.
Méthodes: Les données des essais ont été utilisées pour comparer la mortalité chez les patients atteints de malaria sévère, traités soit avec l’artésunate soit avec la quinine. Cela a été combiné avec des données collectées rétrospectivement sur le coût afin d’estimer l’incrément de coût par décès évité avec l’utilisation de l’artésunate plutôt de la quinine.
Résultats: L’incrément de coût par décès évité en utilisant l’artésunate était d’environ 140 USD. L’artésunate maintenait aussi ce haut niveau de rapport efficacité-coût lorsque l’incertitude entourant les évaluations de coût et d’efficacitéétait prise en compte.
Conclusion: Cette analyse confirme la supérioritéélevée de l’artésunate pour le traitement de la malaria sévère du point de vue économique ainsi que d’une perspective clinique.
Objetivo: En ensayos clínicos realizados en Asia se ha demostrado que el artesunato es superior a la quinina para el manejo de la malaria severa. Este estudio explora la costo-efectividad del artesunato, basándose principalmente en los hallazgos de un ensayo multicéntrico realizado en el sudeste asiático.
Métodos: Los datos del ensayo se utilizaron para comparar la mortalidad entre pacientes con malaria severa, tratados con artesunato o quinina. Estos datos se combinaron con datos de coste recolectados retrospectivamente, con el fin de estimar el coste incremental por muerte evitada con el uso de artesunato en vez de quinina.
Resultados: El coste incremental por muerte evitada utilizando artesunato fue de aproximadamente 140 USD. El artesunato mantuvo este alto nivel de costo-efectividad también cuando se permitía la incertidumbre asociada a las evaluaciones del coste y la efectividad.
Conclusión: Este análisis confirma la gran superioridad del artesunato como tratamiento de la malaria severa tanto desde un punto de vista económico como clínico.
Artemisinin combination therapies (ACT) are now recommended as first-line treatment of uncomplicated falciparum malaria in almost all malaria endemic countries (WHO 2007). They have been shown repeatedly to be more effective and cost-effective than their predecessors (Honrado et al. 1999; Agnamey et al. 2005; Wiseman et al. 2006; Yeung 2006), require only once a day dosing and are associated with few adverse effects. For severe malaria, quinine has been the traditional standard treatment in both developed and developing countries. Quinine is an effective antimalarial, but it is not simple to administer, and it has a narrow therapeutic ratio. It is associated with a risk of local toxicity following intramuscular injection, and significant risks of systemic toxicity (hypoglycaemia, hypotension if administered rapidly). Quinine must be given three times daily either by rate-controlled intravenous infusion or intramuscular injection to the anterior thigh (Anstey et al. 2006). A growing body of evidence summarised in recent reviews demonstrates the considerable superiority of artesunate relative to quinine in terms of mortality rates without increasing rates of neurological deficit (Cochrane review estimate: RR 0.62, 95% CI 0.51 to 0.75) (Jones et al. 2007). The studies so far have included mostly adults in Asia, although of the 1 461 patients enrolled into the large multi-centre SEAQUAMAT trial, 202 were children, and benefits were similar in both age groups (Dondorp et al. 2005).
The SEAQUAMAT study was conducted across 10 sites in four South East Asian countries. A large multi-centre trial of similar design is underway to estimate the comparative effectiveness of artesunate in children in Sub-Saharan Africa. In the SEAQUAMAT trial, mortality in patients treated with artesunate was 35% lower than in quinine recipients. The implication was that, on average, for every 13 patients treated with artesunate instead of quinine, one death would be averted.
Despite these promising results, and endorsement by the WHO treatment guidelines, many local guidelines in malaria endemic countries continue to recommend quinine as the drug of choice for severe malaria (WHO 2006). The second most frequently recommended treatment for severe malaria is artemether (WHO 2007), even though its advantage over quinine in terms of mortality is limited (Hien et al. 1996; Pittler & Ernst 1999). Artesunate has only recently been added to the policy guidelines of a limited number of countries in Asia (WHO 2007), and its cost-effectiveness has yet to be assessed. The aim of this paper is therefore to examine the costs and consequences of switching from quinine to artesunate from an economic perspective.
A cost-effectiveness analysis framework was used to determine the cost per death averted by switching from quinine to artesunate for inpatients with severe malaria. The interventions being considered were quinine and artesunate for treatment of severe malaria. The drugs were given intravenously. Once patients had recovered sufficiently to take tablets, they continued with the same antimalarial taken orally to complete a course of 7 days.
The perspective taken was that of the provider, so only costs incurred by the hospitals were accounted for, as this was considered of most immediate relevance for decision making purposes by ministries of health considering policy change.
The SEAQUAMAT study was a multi-centre trial carried out between 2003 and 2005 in one site each in Bangladesh, India, Indonesia and seven sites in Myanmar and has been described in detail elsewhere (Dondorp et al. 2005). Relevant patient-specific data from the trial for this analysis include mortality for each of the drugs, dosages used, and the length of stay in hospital as inpatients. The incidence of significant neurological sequelae are also summarised although as their incidence in the prospective studies was very low, they were not incorporated in the final measure of outcome.
The costs included are the provider costs resulting from a switch from quinine to artesunate. Costs for artesunate were obtained from the producer and include shipment costs. Quinine costs and those for i.v. sets and syringes to administer the drugs were obtained from the International Drug Price Indicator Guide (Pittler & Ernst 1999). The cost of i.v. sets and syringes used to administer the drugs are presented for general comparison, but they are not included in the calculation of the cost per death averted due to the variability in routine administration practices of the treatments. Drug costs were increased by 15% to account for taxes and an extra 10% for wastage (Gold et al. 1996). Standard inpatient care costs for each country were obtained from the WHO-CHOICE database and are specific to the level of hospital at each site. These costs include ‘hotel’ expenditures – those for personnel, capital and food and exclude drug costs (WHO 2008a). Both drug costs and inpatient-care costs were calculated for each patient individually based on their length of stay and drug dosage used.
We assumed that, apart from the cost of trial drugs, the inpatient costs per day were the same for both treatment arms. Labour costs were also assumed to be equal, despite the fact that artesunate is considerably easier and simpler to administer. Costs were converted from local units to US dollars at the relevant year, adjusted for inflation using the consumer price index, and reported in 2008 USD. Table 1 shows the unit costs used.
Table 1. Costs for treatment, equipment and inpatient care used in the analysis
|Quinine vial (300 mg salt)||$0.19 (1)||International Drug Price Indicator Guide (accessed 2 June 2008)|
|Quinine tab (300 mg)||$0.04 (1)|
|Artesunate vial (60 mg)||$1.2 (1)||Quote from the producer, Guilin Pharma, Personal communication , 16 May 2008|
|Artesunate tab (50 mg)||$0.17 (1)|
|Artesunate administration equipment||$0.3 (1 × 5 ml syringe and 2 × needles)|| |
|Quinine administration equipment||$1.2 (1 × 5 ml syringe, 2 × needles, 1 × infusion set, 1 IV solution)|| |
|Cost per inpatient day||Bangladesh $5.7|| WHO-CHOICE (2008) (accessed 23 May 2008)|
|Myanmar $2.1 (mean)|
The number needed to treat (NNT) to avert one death was calculated from the difference in mortality between the two arms. This was multiplied by the difference in average cost of treating a patient with each of the drugs, providing the incremental cost per death averted using artesunate. This figure is equivalent to the incremental cost-effectiveness ratio (ICER) calculated using standard methods (Drummond et al. 2005). The primary outcome is the pooled estimate for cost per death averted, merging cost and effectiveness data from all sites. As the trial was a multi-centre study, results are also reported stratified by country. The immediacy of the costs and benefits (deaths averted) meant that no discounting was needed.
Uncertainty surrounding both outcomes and costs was explored using probabilistic sensitivity analysis. The treatment outcome, a binary variable with values representing outcomes of either dead or alive, was assigned a beta distribution using the mortality frequency to define the distribution parameters. This allows for the greater uncertainty in sites that recruited fewer patients. Probability distributions were fitted to the cost of antimalarials given to all patients (gamma distributions; these are skewed to the right reflecting the tendency of cost data to be positively skewed (Nixon & Thompson 2004). Using the @Risk Excel plug-in (Palisade corp., Newfield, NY, USA), a Monte Carlo simulation was carried out to observe the impact these uncertainties had on the ICER.
A threshold analysis was carried out to estimate the cost at which artesunate ceases to be cost-effective. This was done using decision thresholds of $575 and £3450 per death averted. These values were obtained using WHO’s thresholds for the willingness to pay to avert the loss of a disability adjusted life year (Jamison et al. 1993; WHO 2008a), multiplied by the average remaining life expectancy for a patient that survives their illness, based on life expectancy tables for the relevant countries (WHO 2008b), and discounted at 3% (Gold et al. 1996).
Health outcomes are shown in Table 2. The mortality amongst patients treated with artesunate was considerably lower in all countries; the pooled estimate for mortality amongst the artesunate group was 34.7% lower than that for quinine (95% CI 18.5–47.6%; P = 0.0002). There were very few instances of neurological sequelae, three in the quinine arm and seven in the artesunate arm. The difference between these was not statistically significant (P = 0.2).
Table 2. Patient outcomes by country and pooled
| Mortality|| 32.5% (75/231)|| 26.4% (19/72)|| 12.3% (18/146)|| 18.4% (52/282)|| 22.4% (164/731)|
| Neurological sequelae||0||0||1||2||3|
| Mortality|| 23.4% (52/222)|| 21.4% (15/70)|| 6.3% (9/143)|| 10.5% (31/295)|| 14.7% (107/730)|
| Neurological sequelae||3||1||1||2||7|
The mean costs for each country and the pooled estimate are shown in Table 3. The variation in costs is due to both differences in unit costs, and in the average length of stay in each hospital. Despite this variation, overall the difference in average cost between the treatment groups was fairly consistent, costs being slightly higher for recipients of artesunate.
Table 3. Treatment cost per patient by country
| Drug cost||$5.0||$5.5||$4.6||$4.8||$4.8|
| Drug administration cost||$8.2||$9.3||$9.0||$8.5||$8.7|
| Inpatient care cost||$21.8||$45.6||$50.2||$12.7||$27.6|
| Total|| $35.0 || $60.4 || $63.8 || $26.0 || $41.1 |
| Drug cost||$15.8||$21.6||$13.4||$13.0||$14.7|
| Drug administration cost||$1.1||$1.4||$1.0||$1.0||$1.1|
| Inpatient care cost||$27.0||$47.3||$52.6||$12.9||$28.3|
| Total|| $44.0 || $70.3 || $67.0 || $26.9 || $44.1 |
Costs and outcomes are combined in Table 4. Given the number of assumptions involved, the costs in this table exclude the equipment needed to administer the drugs, making artesunate appear even more costly than quinine compared with the totals in Table 3. As previously reported (Dondorp et al. 2005), treatment with artesunate was associated with a relative risk of 0.65. The bottom row specifies the incremental cost per death averted. The pooled ICER is $135.6, with a range of values from $104 in Myanmar where the largest number of patients were recruited, to $361 in India where fewest patients were recruited.
Table 4. Costs for each of the treatments, combined with the relative risk to produce the numbers needed to treat and the incremental cost per death averted
|Mean cost for patients treated with quinine||$26.8||$51.1||$54.8||$17.5||$32.4|
|Mean cost for patients treated with artesunate||$42.8||$68.9||$66.0||$25.8||$43.0|
|Relative risk for treatment with artesunate||0.72||0.81||0.51||0.57||0.65|
|Numbers needed to treat to avert a death||11||20||17||13||13|
| Incremental cost per death averted || $177.2 || $358.9 || $185.7 || $104.8 || $135.6 |
By assigning a probability distribution to all hospital inpatient-care costs and treatment outcomes, the uncertainty surrounding these parameters is carried through to the ICER for artesunate. Using a Monte Carlo simulation, the mean cost per death averted was found to be $140.2, approximating the deterministic calculation. By removing the highest and lowest 2.5% of observations, a 95% interval was created, ranging from −$120 to $455. The negative values indicate those instances where artesunate is more effective and less costly, providing hospitals with savings in costs of treating severe malaria1.
Use of the threshold analysis to estimate the point at which artesunate ceases to be cost-effective showed that if decision makers are willing to pay $575 to avert the loss of a life, artesunate would be cost-effective up to a cost of $4.2 per vial, over three times its current selling price. For a higher threshold of $3450 the cost could be $24.4, over 20 times its current selling price, before the drug ceases to be cost-effective.
This economic evaluation confirms the considerable superiority of artesunate over quinine for the treatment of severe malaria, at least for adults in South East Asia. The pooled estimate suggests that the incremental cost of averting a death using artesunate is approximately $140. The variation in results, both by country and in the sensitivity analysis do not diminish the strength of these results. In fact the sensitivity analysis shows that the use of artesunate can be cost saving, in addition to being clinically superior.
Cost-effectiveness analyses require a comparison of results to decision thresholds to determine whether an intervention can be considered a good investment. Gross domestic product (GDP) per capita is increasingly considered a benchmark for determining when an intervention is considered cost effective, with GDP per capita being compared with the cost of averting the loss of a life year in full health (Sachs 2002). As the mean cost per death averted is well below the GDP per capita of even the poorest countries, there is no doubt that the use of artesunate to treat severe malaria represents an extremely good investment. This is further supported by the fact that these cost estimates suggest a much better return on investment than those for other well accepted malaria related interventions, such as $858 per death averted from implementing an environmental control programme (Utzinger et al. 2001), or a range of $254 to $3437 per death averted for insecticide treated mosquito nets (Mulligan et al. 2005).
A number of simplifying assumptions have been made that reduce the advantage of artesunate. First no attempt has been made to quantify and cost the labour requirements of administering the drugs. Quinine is administered three times per day and infusions need careful monitoring whereas artesunate is given by as a daily bolus injection and therefore requires no special nursing attention. Second, the costs for equipment used to administer the drugs, shown to be higher for quinine, were not included in the calculation of cost per death averted. This was done as quinine can be given by the intramuscular as well as the intravenous route, and patients with severe illness will often be given intravenous fluids irrespective of the route of drug administration. Although not used in the calculation of cost per death averted, the estimate of the different costs of administering the drugs is provided in table 3 and clearly favours artesunate. Third, any potential differences in the treatment of adverse events such as hypoglycaemia, a specific adverse effect of quinine (White 1983), have also been ignored but would obviously favour artesunate. In the trial hypoglycaemia following treatment was indeed less frequent with artesunate (Mantel–Haentzel stratified odds ratio and 95%CI from the SEAQUAMAT study; 0.31 [0.12–0.78]), although monitoring for hypoglycaemia would still be required for both treatments (Anstey et al. 2006).
Despite these assumptions, that all reduce the potential cost-effectiveness of artesunate, the cost per death averted demonstrates that it remains a highly attractive intervention. The production of a good manufacturing practise (GMP) artesunate is currently underway2http://clinicaltrials.gov/ct2/show/NCT00298610 (accessed 11 November 2008) and as the sensitivity analysis shows, artesunate would continue to be cost-effective for the treatment of severe malaria even if its price was significantly higher.
Neurological sequelae were reported but not accounted for as these were found not to differ significantly and occurred in less than 1% of cases, although importantly there was no evidence of a significant increase in risk in the artesunate group (seven of 730) compared with the quinine group (three of 731). Reporting the results as cost per Disability Adjusted Life Year (DALY) averted would therefore add little to the analysis.
The perspective taken in this analysis is that of the provider. Ideally, economic evaluations would include a broader range of costs and benefits, including costs incurred by patients, and how these differ between interventions. In this instance there was no reason to expect major differences in costs, and as the data for this were not readily available, this was excluded from the analysis.
Multi-centre trials can pose a number of challenges in analysing and interpreting results. Pooling data is not always a valid procedure whilst stratifying results by site may result in significant loss of power and fail to make full use of available data (Grieve et al. 2007). Pooling cost data raises concerns around how prices are standardised, as unit costs can vary widely across countries. In this analysis prices were standardised using official exchange rates and inflating their value from the year in which they were reported to 2008 USD. An alternative approach is the use of international dollars that adjust for purchasing power parity, however this has not yet entered mainstream use and can cause confusion for those not familiar with purchasing power adjustment.
Some variation in effectiveness between sites was observed, but as there was no prior reason to expect significant differences in treatment effects, and the observed differences were not statistically significant, pooling the data was considered justifiable.
With over a million annual deaths due to malaria, it is imperative that the most effective and cost-effective treatments be used for patients with severe illness. Artesunate is considerably superior to quinine for the treatment of severe malaria in Asia and has been incorporated into the national guidelines in a limited number of countries including China, Vietnam, India, Indonesia, Thailand, Lao PDR and PNG. However, the guidelines in many countries in the region continue to recommend quinine or artemether. This study has demonstrated that from a cost-effectiveness perspective, substituting quinine with artesunate would provide a return on investment that few health interventions could match in terms of both immediate health gains and minimal, if any, additional cost. There seems no reason to deny patients the best available treatment.
Although a negative value could also imply the inverse (artesunate associated with higher mortality and lower costs than quinine), in this case artesunate was found to be superior in all iterations of the simulation.
Safety and Efficacy Study of IV Artesunate to treat malaria.
We thank the directors and staff of the trial hospitals, and the doctors, research nurses, and research assistants for their help; Prof Kyaw Myint, Minister of Health, Myanmar, for his support; Kyaw Win, for his advice and help in initiating this study; and Marja Dondorp-Schilstra for designing the database and Prof Richard Peto for advice. The trial was funded by a grant from the Wellcome Trust, and was coordinated as part of the Wellcome Trust-Mahidol University-Oxford Tropical Medicine Research Programme funded by the Wellcome Trust of Great Britain. We thank Richard Grieve and Borislava Mihaylova for their comments on the manuscript.