Primaquine is the only widely available antimalarial drug that kills mature Plasmodium falciparum gametocytes, the stage of the parasite lifecycle responsible for the transmission of malaria from humans to mosquitoes.1 Since 2012, WHO has recommended the addition of a single low dose of primaquine (0·25 mg/kg) to artemisinin-based combination therapy (ACT) in areas with artemisinin partial resistance (ART-R), to limit the spread of resistance strains, and in areas of low transmission, to reduce onward transmission of P falciparum malaria and to advance malaria elimination.²

Despite malaria-control efforts, including pivotal integrated vector control, young children (aged <5 years) remain highly susceptible to malaria infection and have the highest burden of clinical malaria.³ Because these infections are often associated with gametocytaemia, children contribute substantially to the infectious malaria reservoir.^4–6 Furthermore, although the effectiveness of primaquine depends on achieving high coverage rates at a population level, no paediatric primaquine formulation has yet been marketed.⁷

At high total doses, primaquine can cause severe drug-induced haemolysis in individuals with glucose-6-phosphate dehydrogenase (G6PD) deficiency.⁸ However, WHO recommends that single low-dose primaquine can be given safely without G6PD testing. Previously, we reported the safety and efficacy of single low-dose primaquine in all age groups.^9,10 Young children seemed to be at the highest risk of anaemia, although insufficient data were available to enable a robust comparison between young children and older children (aged 5 years to <15 years) and adults (aged ≥15 years). Additional individual-participant data have since become available, allowing us to reassess the safety and efficacy of single primaquine doses up to 0·75 mg/kg for blocking P falciparum transmission—in young children compared with older children and adults and in areas of low transmission compared with areas of moderate-to-high transmission.

Of 1062 studies screened for full-text review, 30 were eligible for inclusion, of which individual patient data were available from 23 studies involving 6056 patients from 1G countries: 1171 (19·3%) were younger than 5 years, 2827 (46·7%) were aged between 5 years and <15 years, and 2058 (34·0%) were aged ≥15 years (figure 1). 19 studies were RCTs, two were open-label clinical trials, and two were prospective observational studies. 16 studies included a no-primaquine group and 16 studies were conducted in Africa. Apart from one study¹⁵ that used age-based primaquine dosing, all studies used primaquine target doses, which ranged from 0·0625 mg/kg to 0·75 mg/kg, although most studies included 0·25 mg/kg (14 studies) and 0·75 mg/kg (nine studies; appendix p 7).

14 studies from ten countries measured gametocytaemia by molecular methods, with data for 2449 (93·5%) of 2619 patients available for inclusion in the efficacy analysis (figure 1, appendix p 7). Seven studies used quantitative nucleic acid-based amplification and the remaining seven used quantitative RT-PCR, with target transcripts including Pfs25, Pfs230p, and Pfg377 mRNA.

At enrolment, gametocytes were detected in 1764 (72·0%) of 2449 patients. Baseline gametocytaemia was more prevalent in young children (421 [82·9%] of 508) than in older children (1039 [74·0%] of 1404; p<0·001) and adults (304 [56·6%] of 537; p<0·001), and was also more prevalent in moderate-to-high transmission settings (1009 [90·0%] of 1121) than in low-transmission settings (755 [56·9%] of 1328; p<0·001; appendix p 11).

Across all age groups and transmission settings, patients given primaquine were less likely to have gametocytes at day 7 and day 14 than patients given ACT alone (tables 1, 2; appendix pp 12–15). No difference was observed in the effect of single low-dose primaquine on day 7 gametocyte positivity in young children compared with older children (adjusted odds ratio [aOR] 1·08, 95% CI 0·52–2·23; p=0·84) and adults (0·50, 0·20–1·25; p=0·14), and between low-transmission and moderate-to-high transmission settings (1·07, 0·46–2·52; p=0·86; appendix p 16). At day 7 and day 14, patients with higher baseline gametocyte densities were more likely to have gametocytes, whereas patients with baseline hyperparasitaemia and higher haemoglobin concentrations were less likely to have gametocytes (table 1).

Notably, the rate of decline in gametocyte carriage associated with primaquine dose differed between patients treated with artemether–lumefantrine and those treated with dihydroartemisinin–piperaquine (figure 2). Of patients given less than 0·2 mg/kg primaquine, those treated with artemether–lumefantrine were less likely to have gametocytaemia at day 7 and day 14 than those treated with dihydroartemisinin–piperaquine. However, for patients who received a single dose of at least 0·25 mg/kg primaquine with artemether–lumefantrine or dihydroartemisinin–piperaquine, gametocyte carriage was similarly low on day 7 and day 14 (figure 2, table 2).

Adding single low-dose primaquine to ACT probably decreases malaria transmission, as inferred from gametocyte carriage by day 7 (table 3).

Four studies reported mosquito-feeding experiments, three provided data on patient infectiousness (ie, infected at least one mosquito) and the proportion of mosquitoes infected between day 0 and day 14, and one study reported patient infectiousness at day 7. A single 0·25 mg/kg dose of primaquine was tested in all four studies, a no-primaquine group was included in three studies, and other primaquine doses were tested in two studies. Patients were treated with either artemether–lumefantrine (two studies) or dihydroartemisinin–piperaquine (two studies). Two studies included children younger than 15 years and two included patients older than 5 years. Two studies were conducted in low-transmission areas and two in moderate-to-high transmission areas.

Greater reduction in mosquito infectivity over time was observed with a primaquine target dose of 0·25 mg/kg than with doses of 0·0625–0·125 mg/kg; higher primaquine doses had similar effects to 0·25 mg/kg (appendix p 18). The risk of mosquitoes becoming infected with the use of single low-dose primaquine was not different in young children than in older children (aOR 1·36, 95% CI 0·07–27·71; p=0·84) and adults (0·31, 0·01–8·84; p=0·50), and between low-transmission and moderate-to-high transmission settings (0·18, 0·01–2·95; p=0·23; table 4). The risk of a patient infecting at least one mosquito and the percentage of mosquitoes infected with primaquine use were strongly associated with gametocyte density at the time of mosquito feeding (appendix p 18). Mosquito infectivity decreased over time with artemether–lumefantrine alone but not with dihydroartemisinin–piperaquine alone. However, with a primaquine dose of 0·25 mg/kg, the risk of infecting at least one mosquito was zero by day 3 regardless of the type of ACT (data not shown).

Whether adding single low-dose primaquine to ACT decreases malaria transmission is uncertain, as inferred from the risk of a patient infecting at least one mosquito (table 3).

The haematological safety analysis included data from 5772 patients in 20 studies spanning 14 countries (figure 1, appendix p 19). Of 1169 young children included, 55 (0·95%) were infants (aged 6 months to <1 year). G6PD status was assessed in 16 studies (appendix pp 7–8) with results known for 4243 (73·5%) of 5772 patients; 131 (12·1%) of 1085 young children, 202 (9·0%) of 2246 older children, and 50 (5·5%) of 912 adults were G6PD-deficient. All young children deficient in G6PD and all children (<15 years) with intermediate G6PD activity were from study sites in Africa.

A single primaquine dose was given to 4066 (70·4%) of 5772 patients. Except for one large study15 that used age-based dosing (median actual dose 0·21 mg/kg, range 0·07–0·41), all study groups used target weight-based doses ranging from 0·0625 mg/kg to 0·75 mg/kg. Patient characteristics at enrolment were generally similar between age groups and between primaquine and no-primaquine groups for each age category. However, young children were more commonly hyperparasitaemic and anaemic than were older children and adults and, in adult males, hyperparasitaemia was more common in the primaquine group than in the no-primaquine group (appendix p 19). We observed no difference in severe anaemia at day 7 between the primaquine and no-primaquine groups for each age category (appendix p 22).

An increase in primaquine dose of 0·1 mg/kg was not associated with a decrease in haemoglobin concentration of more than 25% alongside moderate-to-severe anaemia (haemoglobin <10 g/dL) by day 7 across age groups, transmission settings, and in patients with normal or unknown G6PD activity (table 5, appendix pp 24–26). Primaquine at a dose of 0·25 mg/kg might not increase symptoms of anaemia in the first week after starting treatment (table 3).

However, the odds of a decrease in haemoglobin concentration of more than 25% by day 7 with moderate-to-severe anaemia were approximately eight times higher (aOR 7·93, 95% CI 2·08–30·33) in young children than in adult males (table 5). We found no association between primaquine dose and a decrease in haemoglobin concentration by day 7 in different transmission settings or across age groups, except in older children, in whom a 0·1 mg/kg increase in primaquine dose was associated with a very small (0·02 [95% CI 0·0001–0·04] g/dL) decrease in haemoglobin concentration (appendix p 26). For G6PD-deficient patients (n=341; four studies), each 0·1 mg/kg increase in primaquine dose was associated with a 0·18 (0·09–0·26) g/dL decrease in haemoglobin concentration by day 7 (appendix p 26). Severe anaemia by day 7 was associated with lower baseline haemoglobin concentration, hyperparasitaemia, and low transmission intensity but not with age group, G6PD status, or primaquine dose (appendix p 26).

Of 3750 patients with haemoglobin measurements on days 0, 2 or 3, and 7, 2159 (57·6%) had a nadir of haemoglobin concentration on day 2 or day 3, regardless of age group, primaquine use, or G6PD status. The mean haemoglobin concentration increased similarly in both primaquine and no-primaquine groups by day 21 or day 28, across all age groups, and regardless of G6PD status (appendix pp 27–28).

13 studies with 3755 patients were included in the adverse event analysis (figure 1). 1860 (49·5%) of 3755 patients had at least one adverse event and 64 (1·7%) patients had at least one serious adverse event within 28 days. Nine studies (including 2282 patients) were RCTs with a no-primaquine group and four studies (1473 patients) did not have a no-primaquine group (two studies with randomisation of ACT and two observational studies).

In the nine RCTs, 1020 (44·7%) of 2282 patients had at least one adverse event, with adverse events occurring in 326 (64·7%) of 504 young children, 554 (40·9%) of 1356 older children, and 140 (33·2%) of 422 adults (p<0·001). In all age groups, the risks of adverse events of grade 2 or higher, serious adverse events, and adverse events of special interest (vomiting and gastrointestinal disturbances) were generally similar in patients who received primaquine and in those who did not (appendix pp 29–30).

34 (1·5%) of 2282 patients reported 35 serious adverse events within 28 days in the RCTs with a no-primaquine group (appendix pp 31–34). Of these 35 events, eight were considered possibly related to primaquine (seven decreases in haemoglobin concentration and one report of vomiting), 12 were considered unrelated or unlikely to be related to primaquine, and six were from one study that did not report relatedness. 22 of the serious adverse events were haematological, and all 22 were reported in three studies and within 7 days of ACT initiation. These haematological serious adverse events were based on a protocol-defined criterion of a decrease in haemoglobin concentration of more than 25% in two studies, and on the definition of serious adverse events according to the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use guidelines¹⁶ in one study. In one study, ten patients (of whom five received placebo and five received primaquine) had blood transfusions within the first week; two of these events were considered related to primaquine by the authors of the primary study. All serious adverse events were transient, and all patients recovered fully.

Little to no difference in the risk of vomiting, any other adverse event, and serious adverse events was found between the primaquine and no-primaquine groups (table 3).

In the four studies without no-primaquine groups, 727 (49·4%) of 1473 patients reported adverse events and 18 (1·6%) of 1111 patients reported serious adverse events within 7 days of ACT initiation, with no serious adverse events reported in young children. Overall, three deaths were reported; all were considered unrelated to primaquine by the authors of the primary study, with one considered possibly related to dihydroartemisinin–piperaquine. One patient had a pre-existing immune-compromised condition and died on admission day 2; one patient died 10 days after being incorrectly enrolled and treated with artemether–lumefantrine and single low-dose primaquine despite a negative malaria test; and one patient died after developing a rash and respiratory distress after the first dose of dihydroartemisinin–piperaquine.

Across all age groups, the risk of haemoglobinuria was similar for patients in the primaquine and no-primaquine groups, and was also independent of the dose of primaquine (appendix pp 35–37).

None of the included studies were primarily designed to compare the safety and efficacy of primaquine between young children, older children, and adults (appendix pp 38–39); of the eight studies that were eligible but not included in the analysis (either because individual patient data were not shared [n=7] or because the study was ongoing [n=1]), all reported adverse events but only one reported membrane-feeding experiments and haematology safety data (appendix p 40).

This systematic review and individual patient data meta-analysis, which included more than 6000 patients from 23 studies in 16 countries, provides robust support for the very good safety and efficacy of single low-dose primaquine for P falciparum transmission blocking, in all age groups and in areas of low and moderate-to-high transmission intensity.

At enrolment, gametocytaemia was more prevalent in moderate-to-high transmission areas than in low transmission areas, and more prevalent in young children than in older children or adults. The efficacy of single low-dose primaquine in reducing day 7 and day 14 gametocytaemia was similar in young children and in older children and adults. This finding highlights the urgent need for a paediatric primaquine formulation for P falciparum transmission blocking, particularly given the higher prevalence of gametocytes in young children and the risks associated with administering inappropriately high doses when giving adult formulations to young children.

Patients treated with dihydroartemisinin–piperaquine were eight times more likely to have gametocytaemia on day 14 than those given artemether–lumefantrine, when the ACT was administered either alone or with a very low dose of primaquine. However, the difference was not significant with a primaquine dose greater than 0·2 mg/kg. A significant reduction in mosquito infectivity was observed with artemether–lumefantrine but not with dihydroartemisinin–piperaquine. Previous studies^17–19 also showed that artemether–lumefantrine has a better gametocyte clearing effect than dihydroartemisinin–piperaquine, but less than that of artesunate–mefloquine. Although the long half-life of dihydroartemisinin–piperaquine is considered advantageous, adding single low-dose primaquine should improve its effectiveness in advancing malaria elimination efforts and limiting the spread of resistant parasites.

ART-R has emerged and is now spreading in several African countries with moderate-to-high transmission.^17,18,20 ART-R results in delayed parasite clearance, with potentially increased prevalence and density of gametocytes even without subsequent recrudescence.¹ Such effects would facilitate greater transmission of resistant than sensitive parasites,¹⁹ as seen previously with other antimalarials.²¹ Artemisinin drug pressure exacerbates the spread of partially resistant parasites, as more sensitive than resistant parasites are killed in treated infections.¹⁹ This individual patient data meta-analysis shows that, among patients seeking treatment, baseline gametocytaemia was much more prevalent in moderate-to-high transmission settings (90·0%) than in low-transmission settings (56·9%; p<0·001). Our findings support that the potent efficacy of primaquine is consistent regardless of malaria transmission intensity. As such, even in moderate-to-high transmission areas threatened by ART-R, adding single low-dose primaquine has the potential to reduce the further selection of resistant parasites by artemisinin-based treatments and counteract the greater transmissibility of artemisinin-resistant parasites.²² This finding is consistent with the WHO 2012 recommendation to add a single low dose of primaquine (0·25 mg/kg) to ACT in areas threatened by ART-R and in malaria elimination areas.²

WHO revised this guidance in 2015²³ to recommend single low-dose primaquine only in low-transmission areas; this revision was based on the concern that reducing malaria transmission requires a high coverage of primaquine, which is hindered by a large reservoir of asymptomatic parasite carriers in areas of moderate-to-high transmission. However, control of asymptomatic infections in areas threatened by ART-R is best achieved by strengthening integrated vector control and preventive treatment strategies that are not artemisinin-based, such as vaccines and seasonal malaria chemoprevention, to decrease the prevalence of asymptomatic malaria without selecting for artemisinin-resistant parasites.²⁴

Malaria frequently results in a decrease in haemoglobin concentration through the destruction of red blood cells (both those containing and not containing parasites) and bone marrow dyserythropoiesis.^25,26 Young children are more prone to develop anaemia after malaria infection, anaemia-associated nutritional deficiencies, and intestinal helminthic infections than are older children and adults.^27,28 Drugs that can result in iatrogenic haemolysis could increase the severity of anaemia. However, our analysis showed that a single primaquine dose did not result in changes in haemoglobin concentration in patients with normal or intermediate G6PD activity, with similar effects seen in young children as in older children or adults. In patients with G6PD deficiency, all of whom were from Africa, each 0·1 mg/kg increase in primaquine dose was associated with a small (0·18 g/dL) decrease in haemoglobin concentration and a 33% increase in moderate-to-severe anaemia, which is reassuringly transient. These data support the safety of the WHO-recommended single target dose of primaquine (0·25 mg/kg, range 0·15–0·38 mg/kg). Although young children had more adverse events and serious adverse events than older children and adults, there was no association with primaquine use in controlled studies. Similarly, no differences in vomiting, any gastrointestinal disturbance, and haemoglobinuria were observed with primaquine use in any age category.

Although, to our knowledge, this is the largest analysis to date on the safety and efficacy of single low-dose primaquine for P falciparum malaria transmission blocking, our study has several limitations. No data were available on the safety and tolerability of repeated single low-dose primaquine use for recurrent P falciparum infection, which remains a knowledge gap that is particularly relevant to areas of high transmission intensity. Nevertheless, the findings from vivax malaria research on repeated primaquine dosing offer reassuring evidence.²⁹ Gametocytes are the sexual stages of the malaria parasites capable of infecting mosquitoes, and are the most widely used predictor of P falciparum transmission blocking efficacy. However, assessing gametocyte carriage—even when using sensitive molecular methods—underestimates the sterilising effects of primaquine, as many infections with low gametocyte density are not transmitted.³⁰ Transmission-blocking effects are therefore often found to be greater in membrane-feeding experiments and infectivity studies than in gametocyte clearance studies, yet the relatively small number of such infectivity studies published testifies to their difficulty, particularly in young children. Only two of the four studies on membrane feeding included young children; more data in individuals from this age group would therefore strengthen the evidence of efficacy in this key population. Even in this large individual patient data meta-analysis, we had little data on infants (aged 6 months to <1 year; n=55). Few (n=8) young children deficient in G6PD received primaquine doses greater than 0·25 mg/kg, so we were not able to estimate the effect of higher primaquine doses on haemoglobin concentration in this vulnerable population.

This systematic review and individual patient data meta-analysis corroborated that adding the WHO-recommended target dose of 0·25 mg/kg primaquine to ACTs is safe and effective as a P falciparum malaria transmission-blocking tool, in all age groups and in areas of low and moderate-to-high transmission intensity. When given without primaquine or with a very low dose of primaquine, artemether–lumefantrine has better gametocytocide and transmission-blocking effects than dihydroartemisinin–piperaquine. The higher prevalence of gametocytaemia in young children than in adults highlights the importance of developing a user-friendly, affordable paediatric primaquine formulation for young children for the reduction of P falciparum transmission in areas threatened by ART-R, including sub-Saharan Africa, and in pre-elimination areas.