Background: Prevention of malaria infection during pregnancy in HIV-negative women currently relies on the use of long-lasting insecticidal nets together with intermittent preventive treatment in pregnancy with sulfadoxine–pyrimethamine (IPTp-SP). Increasing sulfadoxine–pyrimethamine resistance in Africa threatens current prevention of malaria during pregnancy. Thus, a replacement for IPTp-SP is urgently needed, especially for locations with high sulfadoxine–pyrimethamine resistance. Dihydroartemisinin–piperaquine is a promising candidate. We aimed to estimate the cost-effectiveness of intermittent preventive treatment in pregnancy with dihydroartemisinin–piperaquine (IPTp-DP) versus IPTp-SP to prevent clinical malaria infection (and its sequelae) during pregnancy. Methods: We did a cost-effectiveness analysis using meta-analysis and individual trial results from three clinical trials done in Kenya and Uganda. We calculated disability-adjusted life-years (DALYs) arising from stillbirths, neonatal death, low birthweight, mild and moderate maternal anaemia, and clinical malaria infection, associated with malaria during pregnancy. Cost estimates were obtained from data collected in observational studies, health-facility costings, and from international drug procurement databases. The cost-effectiveness analyses were done from a health-care provider perspective using a decision tree model with a lifetime horizon. Deterministic and probabilistic sensitivity analyses using appropriate parameter ranges and distributions were also done. Results are presented as the incremental cost per DALY averted and the likelihood that an intervention is cost-effective for different cost-effectiveness thresholds. Findings: Compared with three doses of sulfadoxine–pyrimethamine, three doses of dihydroartemisinin–piperaquine, delivered to a hypothetical cohort of 1000 pregnant women, averted 892 DALYs (95% credibility interval 274 to 1517) at an incremental cost of US$7051 (2653 to 13 038) generating an incremental cost-effectiveness ratio (ICER) of $8 (2 to 29) per DALY averted. Compared with monthly doses of sulfadoxine–pyrimethamine, monthly doses of dihydroartemisinin–piperaquine averted 534 DALYS (−141 to 1233) at a cost of $13 427 (4994 to 22 895), resulting in an ICER of $25 (−151 to 224) per DALY averted. Both results were highly robust to most or all variations in the deterministic sensitivity analysis. Interpretation: Our findings suggest that among HIV-negative pregnant women with high uptake of long-lasting insecticidal nets, IPTp-DP is cost-effective in areas with high malaria transmission and high sulfadoxine–pyrimethamine resistance. These data provide a comprehensive overview of the current evidence on the cost-effectiveness of IPTp-DP. Nevertheless, before a policy change is advocated, we recommend further research into the effectiveness and costs of different regimens of IPTp-DP in settings with different underlying sulfadoxine–pyrimethamine resistance. Funding: Malaria in Pregnancy Consortium, which is funded through a grant from the Bill & Melinda Gates Foundation to the Liverpool School of Hygiene and Tropical Medicine.
We used outcome data from three clinical trials in Siaya County in western Kenya, and Tororo and Busia District in eastern Uganda, which enrolled HIV-negative pregnant women of all gravidities who had a high uptake of long-lasting insecticidal nets.10, 11, 12 To our knowledge, there are no other similar trials published. Although located in two different countries, the three trial sites are only a maximum of 110 km apart. The sites share a high intensity of malaria transmission and the effectiveness of IPTp-SP is compromised by widespread parasite resistance to sulfadoxine–pyrimethamine as a result of single-nucleotide polymorphisms in the P falciparum dihydrofolate reductase (dhfr) and dihydropteroate synthetase (dhps) genes. The prevalence of sulfadoxine–pyrimethamine quintuple-mutant parasites (a biomarker of resistance to sulfadoxine–pyrimethamine) is high, but the prevalence of the sextuple mutant is less than 6%.13 The Kenya trial randomly assigned 1031 women to receive either IPTp-SP3 (n=515) or IPTp-DP3 (n=516).12 A third group were assigned to receive intermittent screening followed by treatment with dihydroartemisinin–piperaquine, which was found to be unsuitable to replace IPTp-SP because of a lack of clinical efficacy, and this group was excluded from this cost-effectiveness analysis. The Uganda-I trial in Tororo, Uganda, randomly assigned 300 women to one of three groups, IPTp-SP3 (n=106), IPTp-DP3 (n=94), or IPTp-DPmonthly (n=100).11 The Uganda-II trial, in Busia District, Uganda, randomly assigned 782 women to either IPTp-DPmonthly (n=391) or IPTp-SPmonthly (n=391).10 Baseline characteristics were generally similar between the trials. In the Kenya trial, the gestational age of women at enrolment was 16–32 weeks (mean 22·8–23·0 weeks depending on group),12 while in the Uganda trials it was 12–20 weeks (mean 15·2–15·5 weeks in the Uganda-I trial11 and 15·0–15·4 weeks in the Uganda-II trial10). Women in the Uganda-I trial were on average younger than in the other two trials (mean 22·2 years vs 23·4 years in the Kenya trial and 23·0 years in the Uganda-II trial). Women in the Kenya trial received on average 2·7 doses of treatment in both groups; women in the Uganda-I trial received 2·8 (IPTp-SP3), 2·9 (IPTp-DP3), and 5·9 doses (IPTp-DPmonthly); and women in the Uganda-II trial received 6·0 doses in both groups. The mean haemoglobin at baseline was lower in women in the Kenya trial compared with women in the Uganda trials (105 g/L in the Kenya trial vs 119 g/L in the Uganda-I trial and 115 g/L in the Uganda-II trial). Details of the study sites, trial design, participant characteristics, key outcomes, and sample size calculations have been previously published,10, 11, 12 and an overview is provided in the appendix (pp 3–4). Model outcomes were selected on the basis of clinical and economic relevance and availability of disability weights to calculate DALYs. Both child and maternal outcomes were included, namely neonatal death, stillbirth, low birthweight (<2·5 kg), maternal anaemia (haemoglobin <110 g/L for Kenya and Uganda-I and <100 g/L for Uganda-II), and clinical malaria. Separate decision tree models were developed for all child outcomes (neonatal death, stillbirth, and low birthweight), maternal anaemia, and clinical malaria (appendix pp 5–8), because disability weights for concurrent events (ie, maternal anaemia and clinical malaria) are unknown. In the decision tree for child outcomes, neonatal death branches off before low birthweight, as birthweight was not recorded for all babies and priority was given to the most severe outcome. Prevalence and incidence data in the model incorporated the individual trial results. Pooled effect estimates combining the effect of IPTp-DP3 versus IPTp-SP3 from the Kenya and Uganda-I trials were obtained using fixed-effects meta-analyses. DALYs were calculated by adding up the DALYs from all outcomes included in the model. Other serious adverse events and tolerance measures documented in the trial were excluded from the model (for both effect and cost), as no study drug-related difference in their incidence was recorded. All three trials concluded that IPTp-DP was efficacious; however, the main outcomes driving their conclusions differed among the trials. Both the Kenya trial and the Uganda-II trial found a lower number of stillbirths and neonatal deaths in the IPTp-DP group, and the Uganda-I trial identified no difference in stillbirths, but a small, non-significant increase in neonatal deaths, probably occurring by chance because of the small sample size of the trial. The number of low birthweight babies in the IPTp-DP group increased in the Kenya trial, but was lower in both of the Uganda trials. Only the Uganda-II trial found a decrease in maternal anaemia in the IPTp-DP group, and in the other trials the level of anaemia was relatively similar between IPTp-SP and ITPp-DP groups. Finally, in all three trials, episodes of clinical malaria were substantially lower in all IPTp-DP groups. We adopted a health-care provider perspective to estimate the incremental fixed and variable costs of delivering the interventions and their cost savings. The costs of nurses' time, drugs, and of treating consequential health outcomes of malaria infection in pregnancy, comprising maternal anaemia, maternal clinical malaria, and post-delivery hospitalisation costs for low birthweight, were included. Health-care provider costs of a neonatal death or stillbirth (such as mortuary costs) were not collected because we did not anticipate these outcomes would drive the trials results, and hence these costs are excluded from the cost estimation. Cost data were collected by doing an observational study of trial participants in Kenya to measure the average administration time for intermittent preventive treatment (n=44) which was multiplied by the mean cost of nurses' time to estimate the total cost of nurses' time per administration; health-care facility costings (n=4) to estimate the costs of malaria infection during pregnancy and its sequelae; and an analysis of international drug procurement databases to calculate the drug prices per administration.14 The cost of nurses' time was estimated by averaging salaries and benefits provided to nurses obtained from Ministries of Health in Kenya and Uganda for 2018. For all costs, economic costs were calculated and valued in constant 2018 US$ using the local consumer price index and average 2018 exchange rates.15, 16 No health-care facility costing was done in Uganda. Therefore, the costs of malaria infection during pregnancy (and its consequences) from Kenya were used for Uganda, adjusting costs by the ratio of average nurse salaries in the two countries, which was 0·30 (ie, nurses' salaries in Uganda are 30% of those in Kenya). Ex-factory drug prices from procurement databases were adjusted for insurance and freight (10%), in-country transport (10%), and wastage (5%). Assumptions used in calculating costs of maternal anaemia and malaria infection during pregnancy are published elsewhere.14 All cost and observational data collection was approved by the ethics committees of the London School of Hygiene & Tropical Medicine, US Centers for Disease Control and Prevention, and the Kenya Medical Research Institute. Verbal consent was obtained from each participant being observed. Outcome data for IPTp-SP3 and IPTp-DP3 from the Uganda-I and Kenya trials were pooled using both fixed-effects and random-effects meta-analysis models, with the effect of the random-effects model explored in the sensitivity analysis. Because we followed the structure of the decision trees, the results are slightly different from the results reported by Desai and colleagues8 in a review and meta-analysis. The cost-effectiveness of IPTp-DP3 versus IPTp-SP3 (pooled estimate); IPTp-DPmonthly versus IPTp-DP3 and IPTp-SP3 (Uganda-I); and IPTp-DPmonthly versus IPTp-SPmonthly (Uganda-II), was analysed based on the decision trees from the health-care provider perspective using a lifetime horizon to reflect the lifelong mortality and morbidity effects of the adverse health outcomes of malaria infection during pregnancy. DALYs were calculated and then summed for all child and maternal outcomes using disability weights from Global Burden of Disease studies,17, 18, 19 case fatality rates from secondary literature (except for low birthweight, which is captured in the mortality estimates of the trials),20, 21 and local life expectancies, with no age weighting, and a discount rate of 3%.22 The results were expressed as an incremental cost-effectiveness ratio (ICER) for a hypothetical cohort of 1000 pregnant women, calculated by dividing the incremental costs of the intervention by the DALYs averted—ie, The robustness of our results was tested using deterministic and probabilistic sensitivity analysis. The probabilistic sensitivity analysis included 10 000 iterations, producing a point estimate for each iteration (a simulation point), which was plotted in the cost-effectiveness plane. Subsequently, using all 10 000 iterations, the mean, median, and credibility intervals (CrIs; a 95% CI based on percentiles) of the differences in costs, effects, and ICER were calculated. Appropriate distributions were assigned to each parameter following cost-effectiveness guidelines (Table 1, Table 2).22 For zero events (neonatal death in IPTp-SP3 group and clinical malaria in the IPTp-DPmonthly group in the Uganda-I trial), a transformation of 0 plus 0·1 was used in the probabilistic sensitivity analysis to be able to model uncertainty of these parameters using distributions. The probabilistic sensitivity analysis results were compared with multiple country-specific cost-effectiveness thresholds and, for each threshold, the likelihood of being cost-effective was calculated. The cost-effectiveness thresholds were calculated by using estimates of country-level thresholds by Woods and colleagues27 and Ochalek and colleagues,28 adjusted for inflation. These thresholds ranged from $79·8 to $1273·1 for Kenya and from $30·6 to $793·0 for Uganda. Cost-effectiveness thresholds should be used as a guide for interpreting the results, rather than as an actual cutoff for cost-effectiveness. The Uganda-I trial was relatively small compared to the other trials (300 participants vs 1031 participants in the Kenya trial and 782 participants in the Uganda-II trial). The Uganda-I trial reported one stillbirth in each group, and zero neonatal deaths in 98 participants in the IPTp-SP3 group, two (2%) in 88 participants in the IPTp-DP3 group, and three (3%) in 97 participants in the IPTp-DPmonthly group. All five neonatal deaths in the Uganda-I trial occurred in the first 3 days of life and the causes of death were prematurity related complications. These differences in number of neonatal deaths were not significant; however, when these numbers are extrapolated to a hypothetical cohort of 1000 women in our models, the IPTp-DPmonthly group can never be more cost-effective than the other treatment regimens, because of the higher number of neonatal deaths in this group. The numbers of stillbirths and neonatal deaths in the three groups in the Uganda-I trial are very small; therefore the differences among the groups cannot be estimated precisely and might have occurred by chance (table 2). Therefore, although we used the trial results in the base case analysis, we also modelled mortality in the deterministic sensitivity analysis for the cost-effectiveness analysis of IPTp-DPmonthly versus IPTp-DP3 and IPTp-SP3. Cost and DALY input parameters Mild anaemia was haemoglobin between 90 and <110 g/L, moderate anaemia was haemoglobin <90 g/L. Low birthweight was <2·5 kg. DALY=disability-adjusted life-year. IPTp-SP3=intermittent preventive treatment in pregnancy with three doses of sulfadoxine–pyrimethamine. IPTp-DP3=intermittent preventive treatment in pregnancy with three doses of dihydroartemisinin–piperaquine. IPTp-DPmonthly=intermittent preventive treatment in pregnancy with monthly doses of dihydroartemisinin–piperaquine. IPTp-SPmonthly=intermittent preventive treatment in pregnancy with monthly doses of sulfadoxine–pyrimethamine. α=shape parameter. β=rate parameter of the distribution. NA=not applicable. GBD=Global Burden of Disease study. Measures of effect Data are shown as events per 1000 women following the decision tree, as well as α (events) and β (no events). Low birthweight was <2·5 kg. IPTp-SP3=intermittent preventive treatment in pregnancy with three doses of sulfadoxine–pyrimethamine. NA=not applicable. IPTp-DP3=intermittent preventive treatment in pregnancy with three doses of dihydroartemisinin–piperaquine. IPTp-SPmonthly=intermittent preventive treatment in pregnancy with monthly doses of sulfadoxine–pyrimethamine. IPTp-DPmonthly=intermittent preventive treatment in pregnancy with monthly doses of dihydroartemisinin–piperaquine. Hb=haemoglobin. Analysis of the observations of nurses' time to administer the intervention and the meta-analysis of outcomes was done using STATA version 15, and the international procurement data analysis, health-care facility costing studies, and the cost-effectiveness modelling were done in Microsoft Excel (using Visual Basic for Applications for the probabilistic sensitivity analysis). Our analytical and reporting methodology was guided by the Gates Reference Case.22 The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. SF and KH had full access to all the data in the study and had final responsibility for the decision to submit for publication.