Articles Pregnancy outcomes after first-trimester treatment with artemisinin derivatives versus non-artemisinin antimalarials: a systematic review and individual patient data meta-analysis Makoto Saito, Rose McGready, Halidou Tinto, Toussaint Rouamba, Dominic Mosha, Stephen Rulisa, Simon Kariuki, Meghna Desai, Christine Manyando, Eric M Njunju, Esperanca Sevene, Anifa Vala, Orvalho Augusto, Christine Clerk, Edwin Were, Sigilbert Mrema, William Kisinza, Josaphat Byamugisha, Mike Kagawa, Jan Singlovic, Mackensie Yore, Anna Maria van Eijk, Ushma Mehta, Andy Stergachis, Jenny Hill, Kasia Stepniewska, Melba Gomes, Philippe J Guérin, Francois Nosten, Feiko O ter Kuile, Stephanie Dellicour Summary Lancet 2023; 401: 118–30 Background Malaria in the first trimester of pregnancy is associated with adverse pregnancy outcomes. Artemisinin- Published Online based combination therapies (ACTs) are a highly effective, first-line treatment for uncomplicated Plasmodium falciparum November 25, 2022 malaria, except in the first trimester of pregnancy, when quinine with clindamycin is recommended due to concerns https://doi.org/10.1016/ about the potential embryotoxicity of artemisinins. We compared adverse pregnancy outcomes after artemisinin- S0140-6736(22)01881-5 based treatment (ABT) versus non-ABTs in the first trimester of pregnancy. See Comment page 81 WorldWide Antimalarial Resistance Network, Oxford, Methods For this systematic review and individual patient data (IPD) meta-analysis, we searched MEDLINE, Embase, UK (M Saito MD, and the Malaria in Pregnancy Library for prospective cohort studies published between Nov 1, 2015, and Dec 21, 2021, K Stepniewska PhD, containing data on outcomes of pregnancies exposed to ABT and non-ABT in the first trimester. The results of this Prof P J Guérin PhD, search were added to those of a previous systematic review that included publications published up until Prof F O ter Kuile PhD); Infectious Diseases Data November, 2015. We included pregnancies enrolled before the pregnancy outcome was known. We excluded Observatory, Oxford, UK pregnancies with missing estimated gestational age or exposure information, multiple gestation pregnancies, and if (M Saito, K Stepniewska, the fetus was confirmed to be unviable before antimalarial treatment. The primary endpoint was adverse pregnancy Prof P J Guérin, outcome, defined as a composite of either miscarriage, stillbirth, or major congenital anomalies. A one-stage IPD Prof F O ter Kuile); Centre for Tropical Medicine and Global meta-analysis was done by use of shared-frailty Cox models. This study is registered with PROSPERO, Health, Nuffield Department of number CRD42015032371. Medicine, University of Oxford, Oxford, UK (M Saito, Findings We identified seven eligible studies that included 12 cohorts. All 12 cohorts contributed IPD, including Prof R McGready MBBS, K Stepniewska, Prof P J Guérin, 34 178 pregnancies, 737 with confirmed first-trimester exposure to ABTs and 1076 with confirmed first-trimester Prof F Nosten PhD); Division of exposure to non-ABTs. Adverse pregnancy outcomes occurred in 42 (5·7%) of 736 ABT-exposed pregnancies Infectious Diseases, Advanced compared with 96 (8·9%) of 1074 non-ABT-exposed pregnancies in the first trimester (adjusted hazard ratio Clinical Research Center, [aHR] 0·71, 95% CI 0·49–1·03). Similar results were seen for the individual components of miscarriage (aHR=0·74, Institute of Medical Science, University of Tokyo, Tokyo, 0·47–1·17), stillbirth (aHR=0·71, 0·32–1·57), and major congenital anomalies (aHR=0·60, 0·13–2·87). The risk of Japan (M Saito); Shoklo Malaria adverse pregnancy outcomes was lower with artemether–lumefantrine than with oral quinine in the first trimester of Research Unit, Faculty of pregnancy (25 [4·8%] of 524 vs 84 [9·2%] of 915; aHR 0·58, 0·36–0·92). Tropical Medicine, Mahidol University, Mae Sot, Thailand (Prof R McGready, Interpretation We found no evidence of embryotoxicity or teratogenicity based on the risk of miscarriage, stillbirth, or Prof F Nosten); Clinical Research major congenital anomalies associated with ABT during the first trimester of pregnancy. Given that treatment with Unit of Nanoro, Institut de artemether–lumefantrine was associated with fewer adverse pregnancy outcomes than quinine, and because of the Recherche en Sciences de la known superior tolerability and antimalarial effectiveness of ACTs, artemether–lumefantrine should be considered Santé, Nanoro, Burkina Faso (Prof H Tinto PhD, the preferred treatment for uncomplicated P falciparum malaria in the first trimester. If artemether–lumefantrine is T Rouamba PhD); Ifakara Health unavailable, other ACTs (except artesunate–sulfadoxine–pyrimethamine) should be preferred to quinine. Continued Institute, Rufiji, Tanzania active pharmacovigilance is warranted. (D Mosha PhD, S Mrema BSc); School of Medicine and Pharmacy, University Teaching Funding Medicines for Malaria Venture, WHO, and the Worldwide Antimalarial Resistance Network funded by the Hospital of Kigali, University of Bill & Melinda Gates Foundation. Rwanda, Kigali, Rwanda (Prof S Rulisa PhD); Kenya Medical Research Institute, Copyright © 2022 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Centre for Global Health Research, Kisumu, Kenya Introduction anaemia, pregnancy loss, preterm birth, intrauterine (S Kariuki PhD); Centers for Plasmodium falciparum infection in early pregnancy growth restriction, and low birthweight.2,4–10 In areas with Disease Control and Prevention, Atlanta, GA, USA impairs placental vasculogenesis and angiogenesis, 1–4 and stable malaria transmission, more than 60% of malaria is associated with gestational hypertension, maternal infections during pregnancy are estimated to occur in the 118 www.thelancet.com Vol 401 January 14, 2023 Articles (M Desai PhD); Tropical Diseases Research in context Research Centre, Ndola, Zambia (C Manyando PhD); Department Evidence before this study exposed to artemisinin-based treatments (ABTs) and non-ABTs of Basic Sciences, Copperbelt Until late 2022, WHO treatment guidelines recommend 7 days in the first trimester were similar (adjusted hazard ratio University, Ndola, Zambia of quinine (with clindamycin, if available) for treating [aHR] 0·71, 95% CI 0·49–1·03). Similar results were seen for the (E M Njunju MSc); Faculty of Medicine, uncomplicated Plasmodium falciparum malaria in the first individual components of miscarriage (aHR=0·74, 0·47–1·17), Eduardo Mondlane University, trimester of pregnancy, despite its poor tolerability, adherence, stillbirth (aHR=0·71, 0·32–1·57), and major congenital Maputo, Mozambique and effectiveness. In 2017, a systematic literature review and anomalies (aHR=0·60, 0·13–2·87). First-trimester treatment (E Sevene PhD); Centro de meta-analysis assessed the safety of artemisinin derivatives in with ACTs was associated with a lower risk of adverse Investigação em Saúde de Manhiça, Manhiça, the first trimester of human pregnancies, showing no difference pregnancy outcomes (0·59, 0·39–0·89) than oral non-ABTs. Mozambique (E Sevene, in the risk of miscarriage, stillbirth, and congenital anomalies Similarly, artemether–lumefantrine was associated with a lower A Vala MVD, O Augusto MPH); between artemisinin-exposed and quinine-exposed risk of adverse pregnancy outcomes than the standard of care School of Public Health, pregnancies. In 2021, WHO invited us to update the evidence with oral quinine (0·58, 0·36–0·92). University of Ghana, Dodowa, Ghana (C Clerk* PhD); on the safety of antimalarials in the first trimester of pregnancy Implications of all the available evidence Department of Reproductive to establish whether artemisinin-based combination therapy Malaria in the first trimester can have severe consequences for Health, Moi University, Eldoret, (ACT) could be reconsidered for the treatment of malaria in the Kenya (Prof E Were MD); the pregnancy. The currently available safety data (ie, no first trimester. No relevant meta-analyses on the clinical safety National Institute of Medical evidence of embryotoxicity and teratogenicity associated with Research, Amani Medical of artemisinin derivatives in the first trimester were identified first-trimester malaria treatment), together with the superior Research Centre, Muheza, through our literature search, except the one published in 2017. tolerability, higher efficacy, longer duration of post-treatment Tanzania (W Kisinza PhD); Department of Obstetrics and Added value of this study prophylaxis, and wide availability of ACTs, suggest that Gynaecology, Makerere This Article, by use of a one-stage, individual patient data (IPD), artemether–lumefantrine, with the most safety evidence University, Kampala, Uganda meta-analysis approach, presents current information on the available, should replace quinine-based regimens as the (J Byamugisha PhD, M Kagawa PhD); Icon, Prague, safety of artemisinin compounds in the first trimester of preferred treatment for uncomplicated P falciparum malaria in Czech Republic (J Singlovic BSc); pregnancy. This analysis includes data from two additional the first trimester of pregnancy. Other ACTs, except those with VA Los Angeles and University studies (five cohorts), one updated dataset from a previously a first-trimester contraindicated partner drug (eg, artesunate– of California, Los Angeles included study, and the four studies included in the previous sulfadoxine–pyrimethamine), should be considered if National Clinician Scholars Program, VA Greater systematic review in 2017. Compared with the previous meta- artemether–lumefantrine is unavailable. Implementing new Los Angeles Healthcare System analysis, in which only aggregated data were available from the strategies to ensure the creation of robust evidence on the Health Services Research and largest contributing site and sites with no events did not benefit–risk profile of antimalarials for treatment in the first Development Service contribute to the meta-analysis, this update includes IPD from trimester of pregnancy will be crucial in ensuring that Center of Innovation, Los Angeles, CA, USA all eligible studies and analysed all eligible data. We show that individuals at hig risk (ie, people who are pregnant in the first (M Yore MD); Department of the risks of adverse pregnancy outcomes in pregnancies trimester) can access the best treatments. Clinical Sciences, Liverpool School of Tropical Medicine, Liverpool, UK (A M van Eijk PhD, J Hill PhD, Prof F O ter Kuile, first trimester.11–14 During this period, most pregnancies reported in animal studies.19,20 The effects included embryo S Dellicour PhD); Centre for are not protected by insecticide-treated nets, which are resorption, pregnancy loss, and cong enital anomalies Infectious Disease provided at the first antenatal care visit that typically including skeletal defects (eg, shortened or bent long Epidemiology and Research, University of Cape Town, happens in the second or third trimester,15 or by bones and scapulae, misshapen ribs, cleft sternebrae, and Cape Town, South Africa intermittent preventive treatment in pregnancy (IPTp) incompletely ossified pelvic bones) and heart defects (U Mehta PhD); Department of with the antifolate sulfadoxine–pyrimethamine, which is (eg, ventricular septal and great vessel defects).21 The most Pharmacy, School of Pharmacy, and Department of Global contraindicated in the first trimester. Therefore, effective sensitive period is assumed to be the 7 weeks between Health, School of Public Health, recognition and prompt, safe, and effective treatment of 6 and 12 weeks of gestation when primitive erythroblasts, University of Washington, malaria is a priority in the first trimester of pregnancy to the suggested primary target of embryotoxicity, Seattle, WA, USA protect the parent and the fetus. predominantly circulate in human embryos.22 (Prof A Stergachis PhD); UNICEF/ UNDP/World Bank/WHO Until late 2022, WHO treatment guidelines for In 2017, WHO reviewed the evidence on the safety of Special Programme for uncomplicated P falciparum malaria in the first trimester artemisinin-based treatment (ABT) in the first trimester of Research and Training in recommend a 7-day course with quinine combined with pregnancy compared with non-artemisinin-based Tropical Diseases, Geneva, clindamycin (if available),16 which is poorly tolerated, treatment (non-ABT).23,24 This review, which included Switzerland (M Gomes PhD); School of Public Health and poorly adhered to, and less effective than artemisinin- 30 618 pregnancies with 717 confirmed ABT first-trimester Community Medicine, based combination therapies (ACTs).17,18 ACTs have been exposures from five studies, suggested no increase in the Institute of Medicine, the recommended first-line treatment in the second and risk of miscarriage, stillbirth, or major congenital anomalies University of Gothenburg, third trimesters of pregnancy and in all other patient with ABTs compared with quinine in the first trimester. Gothenburg, Sweden (M Gomes) groups since 2006.16 However, ACTs have not been However, it did not change WHO treatment guidelines, *Dr Clerk died in March, 2018 recommended in the first trimester of pregnancy due to which still recommend quinine-based regimens.25,26 concerns about the potential embryotoxicity and In 2021, WHO commissioned an update to the previous teratogenicity of the artemisinin class of compounds meta-analysis to assess whether ACTs should be www.thelancet.com Vol 401 January 14, 2023 119 Articles Correspondence to: reconsidered for the treatment of uncomplicated smoking status, drinking habits, literacy, education, and Dr Stephanie Dellicour, P falciparum malaria in the first trimester of pregnancy. IPTp doses received) and calendar year were extracted at Department of Clinical Sciences, Liverpool School of Tropical This Article expands on the previous meta-analysis 23 by the individual level. Country, and calendar year if year Medicine, Liverpool L3 5QA, UK including all studies published or completed since 2015 was not available at the individual level, was extracted at stephanie.dellicour@lstmed. and by doing an individual patient data (IPD) meta-analysis. the study level. ac.uk Pregnancies were categorised as being exposed to ABT, Methods exposed to non-ABT, or neither (ie, unexposed). Only Search strategy and acquisition of IPD exposures confirmed by at least one written medical record This systematic review and IPD meta-analysis was or multiple data sources were included. Pregnancies with done according to a registed protocol (PROSPERO unconfirmed exposures were excluded to avoid potential CRD42015032371) and is reported according to the misclassification of exposure. Two exposure risk periods PRISMA-IPD statement.27 The results of the previous were defined: the entire first trimester (EGA from ≥2 weeks systematic review, including studies published up to and 0 days to <14 weeks and 0 days) or the putative embryo- November, 2015, are reported elsewhere.23 We did an sensitive period (EGA from ≥6 weeks and 0 days to updated literature search for studies published between <13 weeks and 0 days). Ethics approval was not required Nov 1, 2015, and Dec 21, 2021, by use of MEDLINE, for this analysis. Embase, and the Malaria in Pregnancy Library. The newly identified studies were combined with those Outcomes identified in the previous literature search, which The primary outcome was a composite of either miscarriage included studies published before 2015. A full list of (spontaneous fetal loss before EGA 28 weeks), stillbirth See Online for appendix search terms used is provided in the appendix (p 2). In (fetal loss at or after EGA 28 weeks),28 or major congenital addition, malaria researchers were contacted for any anomalies. Secondary outcomes included miscarriage, other potential data sources, inlcuding unpublished stillbirth, fetal loss (ie, miscarriage or stillbirth), and major studies. congenital anomalies. The rationale for the use of a Eligible studies included prospective cohorts enrolling composite primary outcome was to assess the overall participants before pregnancy outcomes were known; had embryotoxicity and fetal toxicity of the artemisinins shown documented data on ABT exposure in the first trimester; in animals, which resulted in either pregnancy loss or and included comparator groups (women exposed to non- major congenital anomalies among livebirths, and to ABT or women unexposed to any antimalarials). There account for the competing nature of these outcomes. Major were no language restrictions. The Malaria in Pregnancy congenital anomalies were defined as any structural Library includes grey literature (eg, conference abstracts anomaly deemed to be of surgical, medical, or cosmetic and PhD theses). Retrospective studies, case series, case importance at birth, detected by surface examination of reports, and studies not reporting pregnancy outcomes livebirths by trained birth attendants (appendix p 2). were excluded. After preliminary screening of article titles, Congenital anomalies with a suspected genetic cause abstracts were assessed for eligibility by AMvE and SD. (identified with the International Classification of Diseases, AMvE did the search and documentation of identified and 10th edition, code Q90–99) were excluded. selected studies, and conflicts were resolved through discussion between reviewers. Finally, investigators of the Statistical analysis eligible studies were invited to provide fully anonymised The primary analyses compared the risk of adverse IPD for the meta-analysis. pregnancy outcomes in ABT-exposed women with that in non-ABT-exposed pregnancies (the reference group) for Participant eligibility, definition of exposure, and the two risk periods by using adjusted hazard ratio (aHR) comparison groups with 95% CI. In addition, we compared the risks of Women enrolled in included studies during pregnancy pregnancies exposed to ABT or non-ABT with the risks in (ie, before pregnancy outcome was known) were included pregnancies unexposed to any antimalarials during the in the analysis. Pregnancies were excluded if data on exposure risk period. We also analysed comparisons the estimated gestational age (EGA) were missing, the between ACTs (eg, oral ABT excluding artesunate fetus was confirmed to be unviable at enrolment, or monotherapy and artesunate–clindamycin) and oral exposure information was incomplete. Multiple gestation non-ABTs, and between artemether–lumefantrine and pregnancies (eg, twins) were also excluded as they have a oral quinine-based treatments. Crude prevalences of higher risk of adverse pregnancy outcomes than major congenital anomalies among livebirths are singleton pregnancies. Women were included on the presented with 95% CIs by Wilson’s method because of basis of their exposure to antimalarials regardless of small percentages. confirmation of malaria infection, malaria species, or A one-stage, random-effects, IPD meta-analysis was disease severity. The baseline characteristics (ie, age, done for each outcome based on the exposure status in the gravidity, parity, previous history of miscarriage or two different risk periods by use of Cox models with shared stillbirth, height, bodyweight, marital status, HIV status, frailty to account for within-cohort clustering. The time to 120 www.thelancet.com Vol 401 January 14, 2023 Articles outcome was based on EGA (in weeks), with the observation time starting from the EGA at enrolment 630 potentially eligible studies identified through accounting for left truncation. The risk period was from an database search EGA of 2 weeks (or 28 weeks for stillbirth) until pregnancy outcome. The risk period in women who had not had an 4 potentially eligible studies identified through adverse pregnancy outcome was extended to 50 weeks (or other sources 28 weeks for miscarriage) to distinguish women with missing outcomes who were censored at the last 634 screened observation. The number of women included in the IPD meta-analysis is shown as the denominator regardless of the availability of the outcome, unless otherwise stated. 585 excluded 528 did not report antimalarial treatment in Exposure status was treated as a time-dependent first trimester of pregnancy variable. The time after enrolment but before exposure 57 were reviews was treated as unexposed in the analysis. Women exposed to sulfadoxine–pyrimethamine in the first trimester of 49 study abstracts screened for eligibility pregnancy, to both ABT and non-ABT during the risk period, or to unconfirmed antimalarials were censored at the time of these exposures. Pregnancies exposed to the 35 excluded 19 no report of artemisinin by trimester same class of antimalarials more than once during the 16 data overlap with data from included studies exposure risk period were included. Exposures outside each risk period (first trimester or embryo-sensitive period) were not considered. The Newcastle-Ottawa scale 7 with data on ACT exposure in first trimester excluded was used to assess the risk of bias. 2 retrospective design Only three potential confounders identified a priori 4 no comparator 1 treatment allocation bias (age, gravidity, and calendar year) were available across all cohorts. We did sensitivity analyses using multiple imputat ions for missing potential confounders 7 cohort studies were asked to provide IPD (appendix p 2). HIV status was missing in 75% (25 699 of 34 178) of women. Thus, a sensitivity analysis 7 cohort studies provided IPD (n=38 996 pregnancies) excluding confirmed women who were HIV-positive was done. E-values were used to quantify the effects of unmeasured confounders. These e-values represent the 4818 participants excluded* minimum risk ratio at which an unmeasured confounder 194 sulfadoxine–pyrimethamine taken in first trimester before enrolment needs to be associated with both the exposure and the 2689 fetus confirmed not viable at enrolment outcome to make the lower limit of 95% CI exceed unity.29 89 incomplete information on antimalarial Other sensitivity analyses were also done, excluding exposures in first trimester 1190 missing EGA pregnancies with non-P falciparum malaria, excluding 364 multiple gestation pregnancies with multiple exposures to the same class of 41 congenital or genetic atypicality 364 exposure status not confirmed before antimalarials, and including unconfirmed exposures. An outcome additional exploratory analysis was done to measure the 565 exposed to both ABT and non-ABT before enrolment effect of each exposure week on pregnancy outcomes 609 exposed to unconfirmed antimalarials (appendix p 2). The proportional hazard assumption for before enrolment the exposure groups was tested by use of Schoenfeld residuals. For sensitivity analyses for handling clustering 7 studies with contemporaneous comparison of within cohorts, stratified and fixed-effects Cox models IPD included (12 cohorts) were fitted to the primary outcome. Statistical 34 178 pregnancies with exposure status in first trimester or embryo-sensitive period heterogeneity was not assessed quantitatively as not all 32 365 (32 947) unexposed† cohorts had events in both ABT-exposed and non-ABT- 737 (584) exposed to ABT† exposed groups. However, analyses were repeated by 1076 (823) exposed to non-ABT† removing one cohort at a time to assess whether there were any influential cohorts. Analyses were done with Figure 1: Study selection ABT=artemisinin-based treatment. ACT=artemisinin-based combination therapies. EGA=estimated gestational Stata 16.1 MP and R 4.1.2. age. IPD=individual patient data. *Multiple reasons for exclusion allowed. †Data in parentheses are the number of women for whom data are available during the embryo-sensitive period. Role of the funding source The funders of the study had no role in study design, Results data collection, data analysis, data interpretation, or The search identified 634 studies for screening, which writing of the report. included 371 additional studies since the previous review www.thelancet.com Vol 401 January 14, 2023 121 Articles in 2017 (figure 1).23 Overall, 14 studies assessed one was an updated dataset from a previously published ABT exposure in the first trimester of pregnancy. cohort,34,35 one was an unpublished pregnancy cohort Seven of these studies were excluded because they did study in Burkina Faso,36 and one was an unpublished not meet the eligibility criteria (appendix pp 12–14). The multicountry study with four cohorts from the WHO remaining seven were included in our meta-analysis: Special Programme for Research and Training in Tropical four were also included in the previous publication,30,31,32,33 Diseases Pregnancy Exposure Registry.37 The study Unexposed (n=32 365) ABT-exposed* (n=737) Non-ABT-exposed† (n=1076) n Mean (SD) or n Mean (SD) or n Mean (SD) or n (%) n (%) n (%) EGA at exposure, weeks NA NA 737 8·8 (3·0) 1076 8·9 (3·0) Duration of follow-up, weeks 32 365 22·4 (9·9) 737 21·0 (9·3) 1076 23·1 (10·5) Pregnancy outcome available 32 365 28 698 (88·7%) 737 669 (90·8%) 1076 769 (71·5%) Age, years 32 268 26·1 (6·5) 737 25·6 (6·3) 1076 25·1 (6·6) Gravidity .. ·· .. ·· .. ·· 1 32 226 9453 (29·3%) 736 314 (42·7%) 1074 336 (31·3%) 2 32 226 6691 (20·8%) 736 128 (17·4%) 1074 214 (19·9%) ≥3 32 226 16 082 (49·9%) 736 294 (39·9%) 1074 524 (48·8%) Parity .. ·· .. ·· .. ·· 0 27 414 8175 (29·8%) 267 102 (38·2%) 999 354 (35·4%) 1 27 414 6258 (22·8%) 267 60 (22·5%) 999 199 (19·9%) ≥2 27 414 12 981 (47·4%) 267 105 (39·3%) 999 446 (44·6%) Previous miscarriage 28 298 6078 (21·5%) 541 91 (16·8%) 1030 274 (26·6%) Previous stillbirth 27 728 759 (2·7%) 535 10 (1·9%) 978 39 (4·0%) Height (m) 1415 1·6 (0·1) 83 1·6 (0·1) 7 1·6 (0·1) Bodyweight (kg) 1441 59·7 (9·5) 232 50·0 (9·5) 772 46·3 (6·7) Married 24 767 23 978 (96·8%) 461 412 (89·4%) 907 898 (99·0%) HIV positive 7777 643 (8·3%) 508 36 (7·1%) 194 5 (2·6%) Current smoker 22 895 4805 (21·0%) 305 51 (16·7%) 827 305 (36·9%) Any alcohol consumed during pregnancy 2058 295 (14·3%) 113 14 (12·4%) 28 9 (32·1%) Literate 8467 5300 (62·6%) 65 30 (46·2%) 113 51 (45·1%) Education .. ·· .. ·· .. ·· No education 4635 907 (19·6%) 437 65 (14·9%) 102 6 (5·9%) Primary education 4635 2846 (61·4%) 437 255 (58·4%) 102 74 (72·5%) Secondary education or higher 4635 882 (19·0%) 437 117 (26·8%) 102 22 (21·6%) IPTp doses .. ·· ·· ·· ·· ·· 0 6902 1504 (21·8%) 135 56 (41·5%) 183 39 (21·3%) 1 6902 2202 (31·9%) 135 29 (21·5%) 183 60 (32·8%) 2 6902 2692 (39·0%) 135 26 (19·3%) 183 74 (40·4%) 3 6902 390 (5·7%) 135 18 (13·3%) 183 9 (4·9%) 4 6902 114 (1·7%) 135 6 (4·4%) 183 1 (0·5%) NA 20 905 ·· 194 ·· 811 ·· Gestational age measured by ultrasound 28 831 20 069 (69·6%) 497 248 (49·9%) 1000 621 (62·1%) Location .. ·· .. ·· .. ·· Burkina Faso 32 365 4980 (15·4%) 737 43 (5·8%) 1076 173 (16·1%) Ghana 32 365 246 (0·8%) 737 5 (0·7%) 1076 4 (0·4%) Kenya 32 365 1305 (4·0%) 737 74 (10·0%) 1076 5 (0·5%) Tanzania 32 365 1714 (5·3%) 737 156 (21·2%) 1076 69 (6·4%) Uganda 32 365 171 (0·5%) 737 3 (0·4%) 1076 3 (0·3%) Mozambique 32 365 710 (2·2%) 737 19 (2·6%) 1076 5 (0·5%) Rwanda 32 365 1571 (4·9%) 737 77 (10·4%) 1076 0 Zambia 32 365 763 (2·4%) 737 166 (22·5%) 1076 6 (0·6%) Thailand–Myanmar border 32 365 20 905 (64·6%) 737 194 (26·3%) 1076 811 (75·4%) (Table 1 continues on next page) 122 www.thelancet.com Vol 401 January 14, 2023 Articles Unexposed (n=32 365) ABT-exposed* (n=737) Non-ABT-exposed† (n=1076) n Mean (SD) or n (%) n Mean (SD) or n Mean (SD) or n (%) n (%) (Continued from previous page) Study year .. ·· .. ·· .. ·· 2000–04 32 365 5243 (16·2%) 737 55 (7·5%) 1076 374 (34·8%) 2005–09 32 365 9986 (30·9%) 737 318 (43·1%) 1076 364 (33·8%) 2010–17 32 365 17 136 (52·9%) 737 364 (49·4%) 1076 338 (31·4%) Women are categorised according to the first exposure in the first trimester. Unexposed women represent pregnancies with no antimalarial exposure according to any sources during the first trimester. Pregnancies exposed to antimalarials in the first trimester contributed person-time to the unexposed group until they were exposed (with either confirmed or unconfirmed exposures). ABT=artemisinin-based treatment. EGA=estimated gestational age. IPTp=intermittent preventive treatment in pregnancy. n=number of women. NA=not applicable. *Including 637 pregnancies exposed to artemisinin-based combination therapies (525 exposed to artemether–lumefantrine, 32 to artesunate–amodiaquine, 58 to artesunate–mefloquine, 19 to artenimol–piperaquine, and three to artesunate–atovaquone–proguanil), 95 to artesunate with or without clindamycin, and five to parenteral artesunate, on the basis of first exposure in the first trimester. †Including 917 pregnancies exposed to oral quinine (715 exposed to quinine monotherapy and 202 to quinine plus clindamycin), nine to parenteral quinine, 147 to chloroquine, one to mefloquine, one to atovaquone–proguanil, and one to quinine plus mefloquine, on the basis of first exposure in the first trimester. Table 1: Characteristics of women unexposed to antimalarials and women with confirmed exposure to ABT or non-ABT in the first trimester of pregnancy design and quality of the included and excluded studies adverse outcomes in participants followed up until birth are summarised in the appendix (pp 3–13). All studies or other end of pregnancy was 2531 (8·4%) of 30 136. After were assessed as low risk of bias. excluding 165 women (one in the ABT group, two in the All seven eligible studies contributed IPD and involved non-ABT group, and 162 in the unexposed group) whose 34 178 pregnancies (appendix p 15). They were done covariate information was missing, 34 013 women between 2000 and 2017 in nine countries: 11 cohorts in contributed to the covariate-adjusted analysis. The aHR of sub-Saharan Africa (n=12 268 pregnancies [5196 from adverse pregnancy outcomes in ABT-exposed pregnancies Burkina Faso, 255 from Ghana, 1384 from Kenya, (42 [5·7%] of 736) compared with non-ABT-exposed 734 from Mozambique, 1648 from Rwanda, 1939 from pregnancies (96 [8·9%] of 1074) was 0·71 (95% CI Tanzania, 177 from Uganda, and 935 from Zambia]) and 0·49–1·03) in the first trimester. The risk was similar one in Asia on the Thailand–Myanmar border (n=21 910). between the exposure groups in the embryo-sensitive Compared with the previous review published in 2017,23 period (37 [6·3%] of 584 in the ABT group vs 8126 additional pregnancies and 60 additional artemisinin 60 [7·3%] of 822 in the non-ABT group; aHR 0·95, exposures were included in this meta-analysis that were 0·63–1·45; figure 2A). not included in the previous publication, either because An analysis restricted to ACT versus oral non-ABT they are new data or due to differences in the eligibility exposure in the first trimester showed that the risk of criteria for statistical analyses. adverse pregnancy outcomes in pregnancies exposed to The mean age was 26·1 years (SD 6·5), and 10 103 ACTs was lower than in oral non-ABT exposed pregnancies (29·7%) of 34 036 were primigravida (table 1). Pregnancy (aHR 0·59, 95% CI 0·39–0·89; appendix p 24). Most of outcomes were available for 30 136 (88·2%) of all these exposures were to artemether–lumefantrine and 34 178 pregnancies. Of the 34 178 pregnancies, oral quinine. The risk of adverse pregnancy outcomes in 1813 (5·3%) had confirmed exposure to any antimalarial pregnancies exposed to artemether–lumefantrine in the in the first trimester (1569 once, 200 twice, 40 three first trimester was lower than in pregnancies exposed to times, and four pregnancies were exposed four oral quinine in the first trimester (25 [4·8%] of 524 in the times), including 737 (2·2%) confirmed ABT-exposed artemether–lumefantrine group vs 84 [9·2%] of 915 in the pregnancies and 1076 (3·1%) confirmed non-ABT- quinine group; aHR 0·58, 0·36–0·92) but not significantly exposed pregnancies. The 737 ABT-exposed pregnancies different in the embryo-sensitive period (22 [4·9%] of 445 included 637 exposed to ACTs (of which 525 were vs 51 [7·5%] of 684; aHR 0·71, 0·43–1·20; figure 2B). The exposed to artemether–lumefantrine; appendix numbers for the other ACTs were too small to do further pp 18–19). The 1076 non-ABT-exposed pregnancies in- analyses for specific ACTs. clu ded 917 exposed to oral quinine. The remaining The risk of adverse pregnancy outcomes did not differ 32 365 pregnancies were not exposed to any antimalarial between ABT-exposed pregnancies and unexposed in the first trimester. The analysis that was restricted to pregnancies (aHR 0·92, 95% CI 0·67–1·26). By contrast, the embryo-sensitive period included 584 confirmed the risk was higher for non-ABT-exposed pregnancies ABT-exposed pregnancies and 823 non-ABT-exposed than for unexposed pregnancies (aHR 1·30, 1·06–1·60; pregnancies (appendix pp 16–17). appendix p 20). 2531 (7·4%) of all 34 178 pregnancies had an adverse Further sensitivity analyses came to similar conc lusions. pregnancy outcome. The proportion of pregnancies with The results of analyses restricted to pregnancies with only www.thelancet.com Vol 401 January 14, 2023 123 Articles A ABT Non-ABT HR (95% CI) aHR (95% CI) p value e-value Composite outcome First trimester 42/737 96/1076 0·70 (0·49–1·02) 0·71 (0·49–1·03) 0·071 3·54 Embryo-sensitive period 37/584 60/823 1·00 (0·66–1·53) 0·95 (0·63–1·45) 0·83 2·60 Miscarriage First trimester 27/670 76/1072 0·73 (0·46–1·15) 0·74 (0·47–1·17) 0·19 3·71 Embryo-sensitive period 23/533 46/819 1·09 (0·65–1·84) 1·02 (0·61–1·70) 0·95 2·72 Stillbirth First trimester 13/646 12/745 0·74 (0·33–1·64) 0·71 (0·32–1·57) 0·39 5·79 Embryo-sensitive period 12/518 6/609 1·22 (0·45–3·28) 1·18 (0·44–3·18) 0·75 4·05 Fetal loss First trimester 40/737 88/1076 0·69 (0·47–1·01) 0·70 (0·47–1·02) 0·065 3·68 Embryo-sensitive period 35/584 52/823 1·01 (0·65–1·57) 0·96 (0·62–1·49) 0·86 2·64 Major congenital anomaly First trimester 2/737 8/1076 0·60 (0·12–2·85) 0·60 (0·13–2·87) 0·52 15·64 Embryo-sensitive period 2/584 8/823 0·77 (0·16–3·70) 0·72 (0·15–3·49) 0·69 12·89 0 0·5 1·0 1·5 2·0 Favours ABT Favours non-ABT B Artemether– Quinine HR (95% CI) aHR (95% CI) p value e-value lumefantrine Composite outcome First trimester 25/525 84/917 0·57 (0·36–0·90) 0·58 (0·36–0·92) 0·021 4·98 Embryo-sensitive period 22/445 51/685 0·72 (0·43–1·21) 0·71 (0·43–1·20) 0·20 4·17 Miscarriage First trimester 15/465 68/915 0·65 (0·36–1·18) 0·67 (0·37–1·23) 0·20 4·89 Embryo-sensitive period 12/398 40/683 0·78 (0·39–1·53) 0·77 (0·39–1·52) 0·45 4·65 Stillbirth First trimester 10/488 12/592 0·55 (0·23–1·29) 0·53 (0·22–1·24) 0·14 8·47 Embryo-sensitive period 10/415 6/470 0·93 (0·33–2·59) 0·90 (0·32–2·51) 0·84 5·71 Fetal loss First trimester 25/525 80/917 0·55 (0·34–0·88) 0·56 (0·35–0·90) 0·016 5·18 Embryo-sensitive period 22/445 46/685 0·72 (0·43–1·22) 0·72 (0·42–1·21) 0·21 4·20 Major congenital anomaly First trimester 0/525 4/917 No events No events NA NA Embryo-sensitive period 0/445 5/685 No events No events NA NA 0 0·5 1·0 1·5 2·0 Favours artemether–lumefantrine Favours quinine Figure 2: Adverse pregnancy outcomes in the first trimester and during the embryo-sensitive period in women exposed to antimalarials (A) Compares women treated with ABT with women treated with a non-ABT antimalarial. (B) Compares women treated with artemether–lumefantrine with women treated with an oral quinine-based treatment. The composite primary outcome includes miscarriage, stillbirth, or major congenital anomalies; fetal loss includes miscarriage or stillbirth. Adjusted by age group (<20 years, 20–29 years, 30–39 years, or ≥40 years), gravidity (1, 2, or ≥3 number of pregnancies, including the current pregnancy), and study year (2000–04, 2005–09, or 2010–17). A shared frailty Cox model was fitted to adjust for within-study clustering. The numbers in the ABT, non-ABT, artemether-lumefantrine, and quinine columns represent the pregnancies included in the unadjusted analysis. In the adjusted analysis, three women (one exposed to artemether–lumefantrine and two exposed to quinine) with a missing covariate (gravidity) were not included. ABT=artemisinin-based treatment. aHR=adjusted hazard ratio. HR=hazard ratio. NA=not available. one exposure in the first trimester or pregnancies with trimester and exceeded 2·5 for the embryo-sensitive P falciparum malaria were similar to those of the primary period, even after accounting for measured potential analysis (appendix pp 25–26). The results of multiple confounders. imputation models accounting for age group, gravidity, In addition, a sensitivity analysis excluding pregnancies marital status, smoking status, previous history of with known positive HIV status came to conclusions miscarriage and stillbirth, and calendar year were similar similar to the primary analyses (appendix p 31). A to the primary model (appendix pp 28–29). All e-values sensitivity analysis that included all unconfirmed exceeded 3·5 for the exposure analysis in the first exposures showed that the aHRs for ABT-exposed women 124 www.thelancet.com Vol 401 January 14, 2023 Articles compared with non-ABT-exposed women were similar to First trimester Embryo-sensitive period those obtained from the primary analysis that only included confirmed exposures (appendix pp 43–47). No ABT Non-ABT Unexposed ABT Non-ABT (n=623) (n=681) (n=26 270) (n=503) (n=558) single study changed the conclusion or the direction of the effect estimates when excluded from the analysis Any major congenital anomaly* 2 (0·32%)† 8 (1·17%)‡ 182 (0·69%) 2 (0·40%)† 8 (1·43%)‡ (appendix pp 39–42). The results were similar when only Multiple congenital anomalies 0 1 (0·15%) 36 (0·14%) 0 1 (0·18%) sub-Saharan African countries were included (appendix Nervous system 0 0 27 (0·10%) 0 0 pp 39–42) and were not affected by different statistical Eye 0 0 8 (0·03%) 0 0 models for handling clustering within cohorts Ear, face, and neck 0 0 13 (0·05%) 0 0 (appendix p 30). Congenital heart defects§ 0 1 (0·15%) 15 (0·07%) 0 1 (0·18%) In an exploratory analysis assessing the effect by Orofacial clefts 1 (0·16%) 2 (0·29%) 30 (0·12%) 0 2 (0·36%) gestational week of exposure, there was no clear Digestive system 0 0 20 (0·08%) 1 (0·20%) 0 difference in aHRs in ABT-exposed pregnancies than in Abdominal wall defects 0 0 10 (0·04%) 0 0 non-ABT-exposed pregnancies or unexposed pregnancies Urinary 0 0 4 (0·02%) 0 0 in a specific week (appendix p 27). Genital 0 0 8 (0·03%) 0 0 1910 miscarriages (6·0%) occurred among Limb 1 (0·16%) 5 (0·73%) 46 (0·18%) 1 (0·20%) 5 (0·90%) 32 042 pregnancies enrolled in or before the 28th week of Other anomalies or syndromes 0 1 (0·15%) 24 (0·10%) 0 1 (0·18%) gestation. The aHR of miscarriage in ABT-exposed Anomalies excluded from 0 0 3 (0·01%) 0 0 pregnancies compared with non-ABT-exposed pregnan- EUROCAT subgroups¶ cies in the first trimester was 0·74 (95% CI 0·47–1·17) Data are n (%). ABT=artemisinin-based treatment. *Some cases of congenital anomalies appear in multiple subgroups; and in the embryo-sensitive period was 1·02 (0·61–1·70; therefore, the total in each column might not add up. †The ABT in utero exposed cases of congenital anomalies were: figure 2A). A similar pattern was observed for the risk in (first trimester) one case of cleft lip and palate and one case of bilateral syndactyly; (embryo-sensitive period) one case of bilateral syndactyly and one case of imperforated anus. The case of imperforated anus was not counted in the first artemether–lumefantrine-exposed pregnancies compared trimester analysis as the pregnancy was censored at 4 weeks of gestation due to exposure to chloroquine; because 4 weeks with quinine-exposed pregnancies in the first trimester is not within the embryo-sensitive period, the pregnancy was not censored within this period and thus was included in the (0·67, 0·37–1·23) and in the embryo-sensitive period analysis of the embryo-sensitive period. ‡The eight non-ABT in utero exposed cases of congenital anomalies (first trimester and embryo-sensitive period) were: two cases of cleft lip and palate; two cases of unilateral talipes; one case of (0·77, 0·39–1·52; figure 2B). The risk of misc arriage was syndactyly (both hands and feet) and bilateral talipes; one case of congenital heart defect; one case of amniotic banding on significantly higher in first-trimester pregnancies treated right hands, polydactyly, and one foot with missing toe; and one case of bilateral brachysyndactyly. §None of the studies with non-ABT than in pregnancies not exposed to were designed to systematically screen and detect congenital heart defects. 13 cases of congenital heart defects antimalarials in the first trimester, but this increased risk (12 unexposed and one non-ABT) were detected on the Thailand–Myanmar border, the only site systematically screening for heart murmurs. There were two cases of congenital heart defects detected in Kenya and one case detected in Tanzania was not observed with ABTs (appendix p 20). (all unexposed). The cases of congenital heart defects included one fatal case at 5 months (non-ABT); three cases of heart 429 (1·5%) stillbirths occurred among 29 338 preg- murmur and other major anomalies; five of murmur and cyanosis; and seven confirmed diagnoses (marked nancies followed up after 28 weeks of gestation. The aHR cardiomegaly with increased pulmonary vasculature on chest x-ray, dysplastic pulmonary valve, suspected congenital rubella with heart murmur, tetralogy of Fallot, congenital atrioventricular block with heart murmur, pulmonary artery of stillbirths in ABT-exposed pregnancies compared with atresia, hypoplastic left heart syndrome, and an ectopia cardis). ¶Two cases of inguinal hernia, and one case of fetal non-ABT-exposed pregnancies was 0·71 (95% CI hydrops. 0·32–1·57) in the first trimester and 1·18 (0·44–3·18) in Table 2: Summary of the distribution of major congenital anomalies by EUROCAT subgroups38 the embryo-sensitive period (figure 2A). A similar pattern was observed in artemether–lumefantrine-exposed pregnancies compared with quinine-exposed pregnancies one cohort34,35 and other studies did not systematically in the first trimester (0·53, 0·22–1·24) and in the embryo- screen for heart defects. The limb anomaly reported in sensitive period (0·90, 0·32–2·51; figure 2B). the ABT-exposed group was a bilateral syndactyly, 2339 (6·8%) fetal losses (miscarriages and stillbirths) whereas rodent studies reported limb deformities, occurred in 34 178 pregnancies. The aHR of fetal loss including bent or shortened long bones. The aHR of in ABT-exposed compared with non-ABT-exposed major congenital anomalies in ABT-exposed compared with pregnancies was 0·70 (95% CI 0·47–1·02) in the first non-ABT-exposed pregnancies was 0·60 (95% CI trimester and 0·96 (0·62–1·49) in the embryo-sensitive 0·13–2·87) in the first trimester and 0·72 (0·15–3·49) in period (figure 2A). The risk was lower in artemether– the embryo-sensitive period (figure 2A). No major lumefantrine-exposed pregnancies than in quinine- congenital anomalies were detected in the 482 livebirths exposed pregnancies in the first trimester (0·56, from pregnancies exposed to artemether–lumefantrine 0·35–0·90), but was not significantly different in the in the first trimester (none [0%, 95% CI 0·00–0·79] embryo-sensitive period (0·72, 0·42–1·21; figure 2B). of 482). The prevalence of major congenital anomalies in 192 (0·6%) major congenital anomalies were detected the quinine-exposed group was four (0·74%, 0·29–1·88) in 34 178 enrolled pregnancies, and 192 (0·7%, 95% CI of 543. 0·60–0·80) of 27 574 livebirths (table 2; appendix pp 21–22). Compared with unexposed pregnancies, the risk of Neither limb deformities or congenital heart defects, major congenital anomalies did not differ in both which were reported in animals, were observed in ABT-exposed pregnancies (aHR 0·99, 95% CI ABT-exposed pregnancies, although cardiac auscultation 0·24–4·03) and non-ABT pregnancies (1·65, 0·81–3·36; of neonates was systematically assessed only in appendix p 20). www.thelancet.com Vol 401 January 14, 2023 125 Articles Discussion when these primitive erythroblasts are predominant in In this meta-analysis we found no evidence of the circulation, substantial depletion of primitive embryotoxicity or teratogenicity based on the assessment erythroblasts can occur, leading to fatal consequences of miscarriage, stillbirth, or major congenital anomalies for fetal development. In humans, however, primitive associated with ABT exposure during the first trimester erythropoiesis occurs for 6 weeks, and transient of pregnancy. Furthermore, the risk of the composite reductions of erythroblasts by short-term exposure (if outcome in women exposed to ABTs in the first trimester any) could thus be replenished by newly produced was probably lower than in women exposed to non-ABT cells.22 In monkeys, toxicity was only observed when (aHR 0·71, 95% CI 0·49–1·03). This finding was robust artesunate was administered at 12 mg/kg per day for as all sensitivity analyses showed results in the same 12 days or longer,43 suggesting that treatment courses direction and e-values were high. In analyses restricted to shorter than 12 days are insufficient to cause substantial comparing ACTs with oral non-ABTs, ACTs were depletion of embryonic erythroblasts. Unlike in animal associated with a significantly lower risk of adverse models, a short treatment course of 3–7 days with pregnancy outcomes (aHR 0·59, 95% CI 0·39–0·89). artesunate at a 2–4 mg/kg per day target dose in humans Most of these ACTs were artemether–lumefantrine might not be sufficient to result in embryotoxicity or treatments, which were associated with a significantly have a clinically significant effect. Our analysis showed lower risk of adverse pregnancy outcomes than no increased risk of embryotoxicity during EGA pregnancies treated with oral quinine in the first 6–12 weeks when primitive erythroblasts predominantly trimester (aHR 0·58, 0·36–0·92). circulate in human embryos.22 Human exposure data The adverse effects of malaria in the first trimester of provide the greatest degree of confidence on pregnancy need to be considered when interpreting embryotoxicity,44 and this Article shows that although antimalarial safety risks. Malaria in pregnancy is the results of animal toxicology studies help to identify associated with a 33% increase in pregnancy loss,39 and potential teratogens, results need to be interpreted with this increase can be as high as 60% in the first trimester.34 caution as they might not always be directly applicable Furthermore, malaria infection in the first trimester to humans. impairs placental villous and vascular development,1–4 In this Article, no major congenital anomalies were leading to fetal growth restriction, preterm birth, and observed in the artemether–lumefantrine-exposed pregnancy loss.40 This Article showed that the risk of group, and the 95% CI estimates suggest that the miscarriage was significantly higher in first-trimester prevalence of major congenital anomalies were between pregnancies treated with non-ABT than in pregnancies 0·00% and 0·79%. This upper confidence limit is similar not exposed to antimalarials in the first trimester, which to the 0·69% (95% CI 0·60–0·80) background rate of is expected because of the effect of malaria itself on major congenital anomalies detected at birth by surface adverse pregnancy outcomes. By contrast, an increased examination in the group unexposed to antimalarials risk was not observed when ABTs were used to treat and the rate of 0·74% (0·29–1·88) in the quinine- malaria in the first trimester of pregnancy. This finding exposed group. The prevalence of major congenital suggests that prompt treatment with effective anomalies at birth observed in our study was lower antimalarials can counteract some of the adverse effects than the 2% reported in high-income countries.45 This of malaria infection in early pregnancy. lower observed prevalence partly reflects the exclusion of In animal models, including in rodents and monkeys, defects of suspected genetic cause and that the artemisinin (as a class) was reported to have embryotoxic assessment was restricted to a surface examination just effects. We did not observe any increased risks of at birth. Only approximately 60% of anomalies are miscarriage, fetal loss, or the composite adverse generally detected at birth.46 In the USA, the prevalence pregnancy outcome after ABT compared with non-ABT of major congenital anomalies detectable by surface exposures in either the first trimester or embryo-sensitive examination within the first week after birth was period. Possible explanations for the observed differences approximately 1·3%.47 Furthermore, congenital heart between animals and humans have been proposed.21,41 defects, the most common major congenital anomalies Infection with P falciparum might protect or reduce with a prevalence of approximately 1% in the USA and potential embryotoxicity in pregnant humans on ABT Europe,48 were not adequately assessed in the included because artemisinins concentrate in infected red blood cohorts. Only one cohort (on the Thailand–Myanmar cells, which reduces the availability of free artemisinin border)34,35 used chest auscultation and assessment for and its derivatives. severe cyanosis to screen for potential heart defects at Furthermore, the length and dose of exposure could birth. This site reported a prevalence of congenital heart be crucial. In some animal models (eg, rats), the defects of 0·1% (16 of 15 974) in the unexposed group, nucleated primitive erythroblasts, which are the primary showing that additional detection methods like target of embryotoxicity, are produced in just a few days echocardiography might be needed, but paediatric (about 3 days in rats, between days 10 and 13 after echocardiography was not available in any of the conception).21,22,42 If rats are treated with artemisinin included cohorts. Only congenital anomalies in liveborn 126 www.thelancet.com Vol 401 January 14, 2023 Articles babies were included in our analysis as autopsies were (>3·5 in the first trimester and >2·5 in the embryo- unavailable. Therefore, the composite adverse outcome, sensitive period) are reassuring. This means that the of which we did not observe any increased risk in minimum strength of the association an unmeasured ABT-exposed women, is a better measurement to assess confounder would need to have is a value of >2·5 for the overall embryotoxicity and fetal toxicity than the both the treatment and the outcome to conceal potential individual components of the composite outcome and increased risks of adverse pregnancy outcomes. accounts for the competing nature of these outcomes. Confounders with e-values of more than 2 are uncommon The high rate (ie, >50%) of embryolethal and teratogenic in clinical research.29 outcomes in rats, rabbits, and monkeys was not observed Second, measurement error could have happened in in the human data.20,21 specifying exposure status and estimating gestational The benefits of 3-day ACTs for treating uncomplicated age. Measurement error in specifying exposure status malaria in the first trimester of pregnancy compared was why we only included confirmed exposures in the with the 7-day, administered every 8 h courses of oral primary analysis. Gestational age might not be accurate quinine include much better efficacy, tolerability, and and precise enough to assess the risks of adverse adherence.17,18,49A 2020 systematic review of trials in the outcomes in short periods (eg, 1 week), although second and third trimester of pregnancy reported that gestational age was estimated with ultrasound in malaria treatment failure with quinine was six times 69% of pregnancies in this analysis. higher than with artemether–lumefantrine (aHR 6·11, Third, information on adherence and dosage was 95% CI 2·57–14·54).17 The increased duration of post- largely unavailable. Fourth, heterogeneity between treatment prophylaxis conferred by ACTs is another studies was not quantified, although no specific cohorts important benefit in pregnancy as they prevent new unduly influenced the direction of the effect estimates. infections for several weeks, whereas quinine has no Fifth, we did not adjust for multiple comparisons. In post-treatment prophylactic effect due to its short addition to the composite primary outcome, we also half-life.50,51 Furthermore, adherence to quinine is low presented the results of each individual component of because it is associated with cinchonism, nausea, and the primary outcome and did several subgroup analyses. hypoglycaemia.17,52 Finally, except for artemether–lumefantrine, we could Additionally, harmonising the first-line treatment of not assess the drug-specific effects of ACTs because of uncomplicated malaria in the first trimester with that in the few first-trimester antimalarial exposures to each other trimesters and the rest of the population would ACT and the low incidence of outcomes. However, simplify case-management practices, service delivery, preclinical studies suggest that the safety concerns with and supply-chain management. As first-trimester artemisinins in early pregnancy are a class effect that pregnancy is the only indication for oral quinine, the would involve all artemisinin derivatives. The partner supply of quinine and clindamycin is problematic in drugs of included ACTs are either considered likely to be many countries; in some parts of sub-Saharan Africa, safe (eg, lumefantrine or 4-aminoquinolines, including quinine is rarely available in public facilities and most amodiaquine and piperaquine) or are already approved first-trimester malaria is already treated with first-line (eg, mefloquine) for use in the first trimester of ACTs.53–57 pregnancy in some countries, such as the UK and the Although a cost-effectiveness analysis is beyond the USA, although the possibility of synergistic toxicity scope of this paper, our results suggest that artemether– might remain.20,58 Sulfadoxine–pyrimethamine is an lumefantrine, and possibly other ACTs, are likely to be ant ifolate, and so the combination of artesunate–sulfa- more cost-effective than quinine–clindamycin because of doxine–pyrimethamine is contraindicated in the first the disability-adjusted life-year associated with poorly trimester of pregnancy. The safety data of pyronaridine treated malaria, the simple antimalarial supply in any trimester of pregnancy are scarce, although no management, and the case management of women of safety signals were reported with pyronaridine in childbearing age that does not require screening for preclinical studies.58 pregnancy before treatment. A previous analysis from a single study done on the This study has some limitations, partly due to the Thailand–Myanmar border, which was included in the observational nature and scope of the existing data.23 current meta-analysis, implied a potential increased risk Artemisinin safety data are available largely from of adverse pregnancy outcomes with the combination of observational studies because ACTs are not yet artesunate–mefloquine in the embryo-sensitive period.34 recommended in the first trimester by WHO. Although No further studies reported on this combination. The addressed in this meta-analysis, risks of bias, authors of that study suggested that this finding might confounding, and heterogeneity (both clinical and reflect the fact that women treated with artesunate– methodological) are intrinsically higher in observational mefloquine presented with fever in early pregnancy studies than in randomised trials. First, the range of (EGA <10 weeks), when the risk of miscarriage is highest. potential confounding factors available across the Furthermore, fetal viability was not confirmed before datasets was small. However, the relatively high e-values treatment because artesunate–mefloquine was more www.thelancet.com Vol 401 January 14, 2023 127 Articles commonly prescribed at outpatient clinics before UM, MG, and AS did individual studies and contributed to acquisition individuals became aware of their pregnancy, or in the of data. SD and MS drafted the manuscript. SD, MS, and KS have early 2000s when gestational ultrasound was not available directly accessed and verified the underlying data. All authors had permission to access all the data in the study if they wished. All authors in that cohort.34 had final responsibility for the decision to submit for publication A major strength of this Article was the contribution of (except for CC). All authors contributed to the review and editing of IPD from all eligible studies and the updated methods, this Article (except for CC). which could include all available data. As a result, the Declaration of interests statistical power of the study was increased and we were We declare no competing interests. able to standardise the definitions of exposure and Data sharing outcomes and apply the same statistical models De-identified data are available from the Worldwide Antimalarial accounting for left truncation and the time-dependent Resistance Network data repository (http://www.wwarn.org/working-together/sharing-data/accessing-data), which can be requested via the nature of the exposure. To avoid recall bias and ensure Worldwide Antimalarial Resistance Network Data Access Committee appropriate comparisons, only studies with internal (https://www.wwarn.org/about-us/governance-people/data-access- comparators and studies in which exposures were committee). documented before the pregnancy outcome was known Acknowledgments were eligible. We thank all study participants and study teams. In addition, we thank the WHO International Birth Defect Panel for their review of all cases of This Article highlights the challenges in obtaining congenital anomalies from the WHO TDR Pilot Pregnancy Registry, the quality data on the safety of antimalarials in the first Pregnancy Registry in Burkina Faso, and the Assessment of the Safety of trimester of pregnancy. Generating robust evidence on Antimalarial Drug Use During Early Pregnancy studies. We are grateful the benefits and risks of antimalarial drugs in the first to Lewis Holmes (Medical Genetics and Metabolism Unit, MassGeneral Hospital for Children, Boston, MA, USA) and Asher Ornoy (Hebrew trimester of pregnancy is time-consuming, resource- University Hadassah Medical School and Adelson School of Medicine, intensive, and challenging.23 Despite the large number Ariel University, Israel) for their valuable feedback on the manuscript. of pregnancies exposed to ACTs in the first trimester, it This meta-analysis was partly funded by the Medicines for Malaria has taken more than 20 years to accumulate more Venture and WHO. The Worldwide Antimalarial Resistance Network, which supported the curation of the data for this Article, is supported by than 700 well documented pregnancies exposed to The Bill & Melinda Gates Foundation. The findings and conclusions in artemisinins in the first trimester. Pregnancy registries this Article are those of the authors and do not necessarily represent the provide reliable data on the safety of specific medicines official position of the US Centers for Disease Control and Prevention. in pregnancy in the postmarketing phase, but these References registries are time-consuming and resource-intensive. 1 Moeller SL, Nyengaard JR, Larsen LG, et al. Malaria in early pregnancy and the development of the placental vasculature. Therefore, complementary approaches, including active J Infect Dis 2019; 220: 1425–34. registries and interventional studies (ie, trials in the first 2 Griffin JB, Lokomba V, Landis SH, et al. Plasmodium falciparum trimester), should be considered to increase the creation parasitaemia in the first half of pregnancy, uterine and umbilical artery blood flow, and foetal growth: a longitudinal Doppler of data needed for adequate benefit–risk assessment and ultrasound study. Malar J 2012; 11: 319. to ensure the best antimalarials are available in a timely 3 McClure EM, Meshnick SR, Lazebnik N, et al. A cohort study of manner for the treatment of malaria in the first trimester Plasmodium falciparum malaria in pregnancy and associations with of pregnancy. uteroplacental blood flow and fetal anthropometrics in Kenya. Int J Gynaecol Obstet 2014; 126: 78–82. In conclusion, our findings indicate a favourable risk– 4 Schmiegelow C, Matondo S, Minja DTR, et al. Plasmodium benefit profile for use of artemether–lumefantrine in the falciparum infection early in pregnancy has profound consequences first trimester of pregnancy. This Article supports the for fetal growth. J Infect Dis 2017; 216: 1601–10. 5 Huynh BT, Cottrell G, Cot M, Briand V. Burden of malaria in early recommendation by WHO to include artemether– pregnancy: a neglected problem? Clin Infect Dis 2015; 60: 598–604. lumefantrine as the preferred treatment option for 6 Harrington WE, Moore KA, Min AM, et al. Falciparum but not vivax uncomplicated P falciparum malaria in the first trimester malaria increases the risk of hypertensive disorders of pregnancy in 59 women followed prospectively from the first trimester. BMC Med of pregnancy. Other ACTs, except those with partner 2021; 19: 98. drugs contraindicated in the first trimester (eg, 7 Accrombessi M, Yovo E, Fievet N, et al. Effects of malaria in the first artesunate–sulfadoxine–pyrimethamine), are preferred to trimester of pregnancy on poor maternal and birth outcomes in quinine if artemether–lumefantrine is unavailable. Benin. Clin Infect Dis 2019; 69: 1385–93.8 Moore KA, Simpson JA, Wiladphaingern J, et al. Influence of the However, as with all drugs in pregnancy, a possible risk of number and timing of malaria episodes during pregnancy on rare adverse events or embryotoxicity cannot be excluded. prematurity and small-for-gestational-age in an area of low Continued active pharmacovigilance should be transmission. BMC Med 2017; 15: 117. 9 Valea I, Tinto H, Drabo MK, et al. An analysis of timing and encouraged and supported, particularly for ACTs other frequency of malaria infection during pregnancy in relation to the than artemether–lumefantrine and for new antimalarial risk of low birth weight, anaemia and perinatal mortality in Burkina drugs being developed. Faso. Malar J 2012; 11: 71. 10 Cottrell G, Mary JY, Barro D, Cot M. The importance of the period Contributors of malarial infection during pregnancy on birth weight in tropical SD, FOtK, JH, ES, FN, AS, and RM contributed to the study Africa. Am J Trop Med Hyg 2007; 76: 849–54. conception. AMvE and SD did the literature review. MS, SD, KS, and JS 11 Walker PGT, ter Kuile FO, Garske T, Menendez C, Ghani AC. curated the data. MS, SD, KS, FN, RM, PJG, and FOtK contributed to Estimated risk of placental infection and low birthweight the method. MS and KS did statistical analyses. RM, HT, TR, DM, SR, attributable to Plasmodium falciparum malaria in Africa in 2010: SD, SK, MD, CM, EMN, ES, AV, OA, CC, EW, SM, WK, JB, MK, MY, a modelling study. Lancet Glob Health 2014; 2: e460–67. 128 www.thelancet.com Vol 401 January 14, 2023 Articles 12 Accrombessi M, Fievet N, Yovo E, et al. Prevalence and associated 33 Mosha D, Mazuguni F, Mrema S, Sevene E, Abdulla S, Genton B. risk factors of malaria in the first trimester of pregnancy: Safety of artemether-lumefantrine exposure in first trimester of a preconceptional cohort study in Benin. J Infect Dis 2018; pregnancy: an observational cohort. Malar J 2014; 13: 197. 217: 1309–17. 34 Moore KA, Simpson JA, Paw MK, et al. Safety of artemisinins in 13 Hounkonnou CPA, Briand V, Fievet N, et al. Dynamics of first trimester of prospectively followed pregnancies: submicroscopic Plasmodium falciparum infections throughout an observational study. Lancet Infect Dis 2016; 16: 576–83. pregnancy: a preconception cohort study in Benin. Clin Infect Dis 35 Saito M, Carrara VI, Gilder ME, et al. A randomized controlled trial 2020; 71: 166–74. of dihydroartemisinin-piperaquine, artesunate-mefloquine and 14 Tuikue Ndam N, Tornyigah B, Dossou AY, et al. Persistent extended artemether-lumefantrine treatments for malaria in Plasmodium falciparum infection in women with an intent to pregnancy on the Thailand-Myanmar border. BMC Med 2021; become pregnant as a risk factor for pregnancy-associated malaria. 19: 132. Clin Infect Dis 2018; 67: 1890–96. 36 Rouamba T, Valea I, Bognini JD, et al. Safety profile of drug use 15 Moller AB, Petzold M, Chou D, Say L. Early antenatal care visit: during pregnancy at peripheral health centres in Burkina Faso: a systematic analysis of regional and global levels and trends of a prospective observational cohort study. Drugs Real World Outcomes coverage from 1990 to 2013. Lancet Glob Health 2017; 5: e977–83. 2018; 5: 193–206. 16 WHO. WHO guidelines for malaria. 2022. https://www.who.int/ 37 Mehta U, Clerk C, Allen E, et al. Protocol for a drugs exposure publications/i/item/guidelines-for-malaria (accessed pregnancy registry for implementation in resource-limited settings. June 30, 2022). BMC Pregnancy Childbirth 2012; 12: 89. 17 Saito M, Mansoor R, Kennon K, et al. Efficacy and tolerability of 38 European Concerted Action on Congenital Anomalies and Twins. artemisinin-based and quinine-based treatments for uncomplicated EUROCAT congenital anomalies prevalence charts and tables. falciparum malaria in pregnancy: a systematic review and 2022. https://eu-rd-platform.jrc.ec.europa.eu/eurocat/eurocat-data/ individual patient data meta-analysis. Lancet Infect Dis 2020; prevalence_en (accessed Feb 23, 2022). 20: 943–52. 39 Gamble C, Ekwaru JP, ter Kuile FO. Insecticide-treated nets for 18 Burger RJ, van Eijk AM, Bussink M, Hill J, ter Kuile FO. preventing malaria in pregnancy. Cochrane Database Syst Rev 2006; Artemisinin-based combination therapy versus quinine or other 2: CD003755. combinations for treatment of uncomplicated 40 Moore KA, Simpson JA, Scoullar MJL, McGready R, Fowkes FJI. Plasmodium falciparum malaria in the second and third trimester of Quantification of the association between malaria in pregnancy and pregnancy: a systematic review and meta-analysis. stillbirth: a systematic review and meta-analysis. Lancet Glob Health Open Forum Infect Dis 2015; 3: ofv170. 2017; 5: e1101–12. 19 González R, Pons-Duran C, Bardají A, Leke RGF, Clark R, 41 D’Alessandro S, Menegola E, Parapini S, Taramelli D, Basilico N. Menendez C. Systematic review of artemisinin embryotoxicity in Safety of artemisinin derivatives in the first trimester of pregnancy: animals: implications for malaria control in human pregnancy. a controversial story. Molecules 2020; 25: E3505. Toxicol Appl Pharmacol 2020; 402: 115127. 42 Clark RL. Teratogen update: malaria in pregnancy and the use of 20 Clark RL. Safety of treating malaria with artemisinin-based antimalarial drugs in the first trimester. Birth Defects Res 2020; combination therapy in the first trimester of pregnancy. 112: 1403–49. Reprod Toxicol 2022; 111: 204–10. 43 Clark RL, Arima A, Makori N, et al. Artesunate: developmental 21 Clark RL. Embryotoxicity of the artemisinin antimalarials and toxicity and toxicokinetics in monkeys. potential consequences for use in women in the first trimester. Birth Defects Res B Dev Reprod Toxicol 2008; 83: 418–34. Reprod Toxicol 2009; 28: 285–96. 44 Brent RL. Utilization of animal studies to determine the effects and 22 Gomes C, Boareto AC, Dalsenter PR. Clinical and non-clinical human risks of environmental toxicants (drugs, chemicals, safety of artemisinin derivatives in pregnancy. Reprod Toxicol 2016; and physical agents). Pediatrics 2004; 113 (suppl): 984–95. 65: 194–203. 45 Dolk H, Loane M, Garne E. The prevalence of congenital anomalies 23 Dellicour S, Sevene E, McGready R, et al. First-trimester in Europe. Adv Exp Med Biol 2010; 686: 349–64. artemisinin derivatives and quinine treatments and the risk of 46 Bower C, Rudy E, Callaghan A, Quick J, Nassar N. Age at diagnosis adverse pregnancy outcomes in Africa and Asia: a meta-analysis of of birth defects. Birth Defects Res A Clin Mol Teratol 2010; observational studies. PLoS Med 2017; 14: e1002290. 88: 251–55. 24 WHO Malaria Policy Advisory Committee and Secretariat. Malaria 47 Holmes LB, Nasri HZ, Hunt AT, Zash R, Shapiro RL. Limited Policy Advisory Committee to the WHO: conclusions and surface examination to evaluate potential teratogens in a resource- recommendations of eighth biannual meeting (September 2015). limited setting. Birth Defects Res 2021; 113: 702–07. Malar J 2016; 15: 117. 48 Hoffman JI, Kaplan S. The incidence of congenital heart disease. 25 McGready R, Nosten F, Barnes KI, Mokuolu O, White NJ. Why is J Am Coll Cardiol 2002; 39: 1890–900. WHO failing women with falciparum malaria in the first trimester of pregnancy? Lancet 2020; 395: 779. 49 Saito M, Gilder ME, Nosten F, McGready R, Guérin PJ. Systematic literature review and meta-analysis of the efficacy of artemisinin- 26 Rao VB, Jensen TO, Jimenez BC, et al. Malaria in pregnancy: a call based and quinine-based treatments for uncomplicated falciparum for a safe, efficient, and patient-centred approach to first-trimester malaria in pregnancy: methodological challenges. Malar J 2017; treatment. Lancet Glob Health 2018; 6: e607–08. 16: 488. 27 Stewart LA, Clarke M, Rovers M, et al. Preferred reporting items for 50 WHO. Guidelines for the treatment of malaria, 3rd edn. Geneva: systematic review and meta-analyses of individual participant data: WHO Press, 2015. the PRISMA-IPD statement. JAMA 2015; 313: 1657–65. 51 Bretscher MT, Dahal P, Griffin J, et al. The duration of 28 WHO. Neonatal and perinatal mortality country, regional and global chemoprophylaxis against malaria after treatment with artesunate- estimates. Geneva: World Health Organization, 2006. amodiaquine and artemether-lumefantrine and the effects of 29 VanderWeele TJ, Ding P. Sensitivity analysis in observational PFMDR1 86Y and PFCRT 76T: a meta-analysis of individual patient research: introducing the e-value. Ann Intern Med 2017; data. BMC Med 2020; 18: 47. 167: 268–74. 52 Achan J, Talisuna AO, Erhart A, et al. Quinine, an old anti-malarial 30 Tinto H, Sevene E, Dellicour S, et al. Assessment of the safety of drug in a modern world: role in the treatment of malaria. Malar J antimalarial drug use during early pregnancy (ASAP): protocol for a 2011; 10: 144. multicenter prospective cohort study in Burkina Faso, Kenya and 53 Hill J, D’Mello-Guyett L, Hoyt J, van Eijk AM, ter Kuile FO, Mozambique. Reprod Health 2015; 12: 112. Webster J. Women’s access and provider practices for the case 31 Manyando C, Mkandawire R, Puma L, et al. Safety of artemether- management of malaria during pregnancy: a systematic review and lumefantrine in pregnant women with malaria: results of a meta-analysis. PLoS Med 2014; 11: e1001688. prospective cohort study in Zambia. Malar J 2010; 9: 249. 54 Dellicour S, Desai M, Aol G, et al. Risks of miscarriage and 32 Rulisa S, Kaligirwa N, Agaba S, Karema C, Mens PF, de Vries PJ. inadvertent exposure to artemisinin derivatives in the first trimester Pharmacovigilance of artemether-lumefantrine in pregnant women of pregnancy: a prospective cohort study in western Kenya. Malar J followed until delivery in Rwanda. Malar J 2012; 11: 225. 2015; 14: 461. www.thelancet.com Vol 401 January 14, 2023 129 Articles 55 Mosha D, Mazuguni F, Mrema S, Sevene E, Abdulla S, Genton B. 58 Clark RL. Animal embryotoxicity studies of key non-artemisinin Safety of artemether-lumefantrine exposure in first trimester of antimalarials and use in women in the first trimester. pregnancy: an observational cohort. Malar J 2014; 13: 197. Birth Defects Res 2017; 109: 1075–126. 56 Webster J, Hoyt J, Diarra S, et al. Adoption of evidence-based global 59 WHO. 22nd meeting of the Malaria Policy Advisory Group policies at the national level: intermittent preventive treatment for (MPAG). 2022. https://www.who.int/news-room/events/ malaria in pregnancy and first trimester treatment in Kenya, detail/2022/10/11/default-calendar/22nd-meeting-of-the-malaria- Malawi, Mali and The Gambia. Health Policy Plan 2021; 35: 1364–75. policy-advisory-group (accessed Nov 16, 2022). 57 Sangaré LR, Weiss NS, Brentlinger PE, et al. Patterns of anti- malarial drug treatment among pregnant women in Uganda. Malar J 2011; 10: 152. 130 www.thelancet.com Vol 401 January 14, 2023