Efficacy and tolerability of artemisinin-based and quinine-based treatments for uncomplicated falciparum malaria in pregnancy: a systematic review and individual patient data meta-analysis

Summary Background Malaria in pregnancy affects both the mother and the fetus. However, evidence supporting treatment guidelines for uncomplicated (including asymptomatic) falciparum malaria in pregnant women is scarce and assessed in varied ways. We did a systematic literature review and individual patient data (IPD) meta-analysis to compare the efficacy and tolerability of different artemisinin-based or quinine-based treatments for malaria in pregnant women. Methods We did a systematic review of interventional or observational cohort studies assessing the efficacy of artemisinin-based or quinine-based treatments in pregnancy. Seven databases (MEDLINE, Embase, Global Health, Cochrane Library, Scopus, Web of Science, and Literatura Latino Americana em Ciencias da Saude) and two clinical trial registries (International Clinical Trials Registry Platform and ClinicalTrials.gov) were searched. The final search was done on April 26, 2019. Studies that assessed PCR-corrected treatment efficacy in pregnancy with follow-up of 28 days or more were included. Investigators of identified studies were invited to share data from individual patients. The outcomes assessed included PCR-corrected efficacy, PCR-uncorrected efficacy, parasite clearance, fever clearance, gametocyte development, and acute adverse events. One-stage IPD meta-analysis using Cox and logistic regression with random-effects was done to estimate the risk factors associated with PCR-corrected treatment failure, using artemether-lumefantrine as the reference. This study is registered with PROSPERO, CRD42018104013. Findings Of the 30 studies assessed, 19 were included, representing 92% of patients in the literature (4968 of 5360 episodes). Risk of PCR-corrected treatment failure was higher for the quinine monotherapy (n=244, adjusted hazard ratio [aHR] 6·11, 95% CI 2·57–14·54, p<0·0001) but lower for artesunate-amodiaquine (n=840, 0·27, 95% 0·14–0·52, p<0·0001), artesunate-mefloquine (n=1028, 0·56, 95% 0·34–0·94, p=0·03), and dihydroartemisinin-piperaquine (n=872, 0·35, 95% CI 0·18–0·68, p=0·002) than artemether-lumefantrine (n=1278) after adjustment for baseline asexual parasitaemia and parity. The risk of gametocyte carriage on day 7 was higher after quinine-based therapy than artemisinin-based treatment (adjusted odds ratio [OR] 7·38, 95% CI 2·29–23·82). Interpretation Efficacy and tolerability of artemisinin-based combination therapies (ACTs) in pregnant women are better than quinine. The lower efficacy of artemether-lumefantrine compared with other ACTs might require dose optimisation. Funding The Bill & Melinda Gates Foundation, ExxonMobil Foundation, and the University of Oxford Clarendon Fund.


Appendix 1. Sensitivity analyses of risk factors for PCR-corrected treatment failure
Treatment effect was assessed by the piece-wise Cox model, which split the observation period into 0-28 days and >28 days, in order to satisfy the proportional hazards assumption.
Another pre-planned sensitivity analysis was conducted with P. vivax intercalated infection (n=233) being censored on the day of P. vivax appearance (following the current WHO guidelines for non-pregnant populations). The results were similar to the results of the other models (data not shown).
A post-hoc sensitivity analysis was conducted regarding the indeterminate PCR results (Appendix Table 1-2).
One model regarded indeterminate PCR results as recrudescence (the worst-case scenario) and the other model regarded indeterminate PCR results as reinfection (the best case-scenario). The results were similar to our main analysis model. Finally, to assess the difference of study designs, only randomised control trials (4571 episodes in eight studies) were included. The results were similar to the results of the other models (Appendix Table 1-3). Excluding one study site at a time also did not change the results (data not shown). 0.90 ⱡNumber of women enrolled on day 0 or followed up on day 29. * Adjusted by parity, parasitaemia, previous antimalarial exposure and interaction between treatment and time. AAP: artesunate with atovaquone-proguanil, AC: artesunate with clindamycin, AL: artemetherlumefantrine, AS: artesunate monotherapy, ASAQ: artesunate-amodiaquine, ASMQ: artesunate-mefloquine, ASSP: artesunate-sulfadoxinepyrimethamine, CI: confidence interval, DP: dihydroartemisinin-piperaquine, HR: hazard ratio, Q: quinine monotherapy, QC: quinine with clindamycin.

Appendix 2. Dose response of AL
The dose-response was analysed only for AL, which was associated with a higher risk of treatment failure than other ACTs. The dose-response of quinine could not be investigated as quinine was administered based on body weight (i.e. patients received a very similar dose in mg/kg of body weight). Dose-response of AL (lumefantrine dose in mg per body weight in kg) was analysed by including women who completed the standard six-dose regimen over three days.
The standard dose (total 2880 mg of lumefantrine) was given to 1262 women regardless of body weight, following the study protocols of each study (11 study sites in eight studies). Fatty food was co-administered fully in four studies and partially in three studies but was not specified in one study. Twelve women did not have information on body weight and were thus excluded from the analyses. Overall, there was no apparent dose-response (aHR 0·99 per 1 mg/kg of lumefantrine, 95% CI 0·96 to 1·02, p=0·59) (Appendix Table 5).
However, when the dose effect was assessed for each malaria transmission intensity area (p-value for interaction 0·09), a linear dose-response was found in the low transmission, but not in moderate or high transmission areas (Appendix Table 6). In low transmission areas, women receiving a higher dose (i.e. women with lower body weight) had a lower risk of PCR-corrected treatment failure (aHR 0·94 per 1 mg/kg of lumefantrine, 95% CI 0·88 to 1·00, p=0·048). Co-administration of fat was not associated with treatment outcome.

Appendix 5. Acute adverse events
There was six maternal deaths, one of which was due to severe malaria which developed after treatment with ASSP. 9 The deaths of the other five women were considered not to be associated with malaria.
The symptoms that developed within a week after starting treatment were analysed by mixed-effects logistic regression on nine commonly assessed symptoms. The pooled analysis below was conducted by including only studies in which each symptom was actively assessed (Appendix Table 5-1).
A total of 4623 women were actively assessed for the presence of headache. The risk of headache was higher in ASAQ (aOR 1·56, 95% CI 1·14 to 2·14, p=0·01) than AL (Appendix Table 5 -6). No other baseline characteristics (except the presence of headache on day 0) were associated with the risk of headache.