Safety of primaquine given to people with G6PD deficiency: systematic review of prospective studies

Background Haemolysis risk with single dose or short course primaquine was evaluated in glucose-6-phosphate dehydrogenase (G6PD) deficient people. Methods Major electronic databases (to August 2016) were searched for single or short course 8-aminoquinolines (8-AQ) in (1) randomized comparisons against placebo in G6PD deficient people; and (2) observational comparisons in G6PD deficient compared to replete people. Two authors independently assessed eligibility, risk-of-bias, and extracted data. Results Five randomized controlled trials and four controlled observational cohorts were included. In G6PD deficient individuals, high-dose (0.75 mg/kg) PQ resulted in lower average haemoglobin levels at 7 days (mean difference [MD] −1.45 g/dl, 95% CI −2.17 to −0.74, 2 trials) and larger percentage fall from baseline to day 7 (MD −10.31%, 95% CI −17.69 to −2.92, 3 trials) compared to placebo. In G6PD deficient compared to replete people, average haemoglobin was lower at 7 days (MD −1.19 g/dl, 95% CI −1.94 to −0.44, 2 trials) and haemoglobin change from baseline to day 7 was greater (MD −9.10%, 95% CI −12.55 to −5.65, 5 trials). One small trial evaluated mid-range PQ dose (0.4–0.5 mg/kg) in G6PD deficient people, with no difference detected in average haemoglobin at day 7 compared to placebo. In one cohort comparing G6PD deficient and replete people there was a greater fall with G6PD deficiency (MD −4.99%, 95% CI −9.96 to −0.02). For low-dose PQ (0.1–0.25 mg/kg) in G6PD deficient people, haemoglobin change from baseline was similar to the placebo group (MD 1.72%, 95% CI −1.89 to 5.34, 2 trials). Comparing low dose PQ in G6PD deficient with replete people, the average haemoglobin was lower in the G6PD deficient group at 7 days (−0.57 g (95% CI −0.97 to −0.17, 1 trial)); although change from baseline was similar (MD −1.45%, 95% CI −5.69 to 2.78, 3 trials). Conclusions Falls in average haemoglobin are less marked with the 0.1 to 0.25 mg/kg PQ than with the 0.75 mg/kg dose, and severe haemolytic events are not common. However, data were limited and the evidence GRADE was low or very low certainty. Electronic supplementary material The online version of this article (doi:10.1186/s12936-017-1989-3) contains supplementary material, which is available to authorized users.

and classification of the genotype and phenotype of G6PD deficiency is complicated and understanding is continually developing ( Fig. 1; [3]). An estimated 400 million people worldwide carry G6PD gene mutations [4][5][6], and the relatively high prevalence of between 5 and 33% in the malaria endemic countries of sub-Saharan Africa and Asia [6] makes the drug potentially unsafe within these populations.
In 2010, the World Health Organization (WHO) reaffirmed a recommendation made many years previously for a single dose of 0.75 mg/kg of PQ to be administered with primary treatment for falciparum malaria to reduce transmission [7]. In reconsidering the risks, the WHO lowered the recommended single dose for gametocytocidal effect to 0.25 mg/kg in 2012, and limited PQ's use to areas threatened by artemisinin resistance or areas of low endemicity approaching elimination [8]. Then in 2015, the policy recommendation was changed to recommend use in 'low transmission areas' and it was explicitly stated that G6PD testing was not required before single dose use [9].
The change in policy from 0.75 to 0.25 mg/kg was based partly on a WHO Evidence Review Group report [10,11] summarizing published literature narratively, but did not use standard synthesis approaches to critically appraise the evidence. In light of the importance of the recommendation when implemented at scale, it could mean millions of people exposed to PQ or a related drug for the first time [12]. In this systematic review, a narrower objective was considered: to assess the risk and extent of haemolysis in people with G6PD deficiency given PQ or other 8-AQ, from existing research.

Inclusion criteria
Study designs Comparisons of single or short course 8-AQ versus placebo given to G6PD deficient people in randomized trials and controlled cohort designs were sought. This was supplemented with comparisons of 8-AQ given to G6PD deficient or G6PD replete people.
Participants Adults or children tested for G6PD deficiency.
Intervention Single dose or short course 8-AQ alone or added to malaria treatment. Since doses and schedules vary by country, any single dose regimens were included. The short course was defined as any regimen lasting up to 7 days. The results are expressed within two comparisons (PQ vs placebo in G6PD deficient people; and PQ in G6PD deficient people compared with G6PD replete people).
Control Comparison 1: placebo or no intervention. Comparison 2: PQ in G6PD replete individuals.
Outcomes Haemoglobin values at baseline and day 7. Absolute and percentage decline in haemoglobin.
Information sources and search strategy The following major electronic databases were searched using the criteria in Additional file 1 to August 2016: the Cochrane Infectious Diseases Group Specialized Register, the Cochrane Central Register of Controlled Trials (CEN-TRAL), MEDLINE, Scopus, Web of Science; EMBASE and LILACS. All relevant trials, regardless of language or publication status (published, unpublished, in press and in progress) were identified. The MIM Pan-African Malaria Conferences and the American Society of Tropical Medicine and Hygiene (ASTMH) (from 2004 to 2016) were also searched for relevant abstracts of unpublished trials, and researchers and other experts in the field of malaria chemotherapy were contacted. The reference lists of all retrieved articles were also reviewed [11].
Study selection and data collection RS and OU independently screened the results of the search strategy and retrieved the full text of all relevant citations. The inclusion criteria were independently applied using a spreadsheet and any discrepancies were discussed or referred to a third reviewer (PAG or PMG). RS and OU independently extracted data on study methods, participants, interventions and outcomes, and attempted to contact authors where data were missing.
Risk of bias RS and OU independently assessed the risk of bias for each study using predefined standards based on the Cochrane risk of bias tool for randomized controlled trials, and the EPOC criteria [15] for assessing risk of bias of non-randomized studies (Additional file 2). Each domain was categorized as 'low' risk of bias, 'high' risk of bias or 'unclear' . Where the judgment was that the risk of bias was unclear, attempts were made to contact the authors for clarification. The quality of evidence was assessed using the GRADE approach [16].
Percentage change in haemoglobin (Hb) concentration from baseline was calculated as: This percentage change was categorized as: greater than −5%, greater than −10% and greater than −20%. Moderate and severe anaemia were defined as haemoglobin less than 8 g/dl and less than 5 g/dl, respectively [17][18][19][20][21]. Where the data were not provided, all recent study authors were contacted. The author of two trials [22,23] provided additional individual patient data, which were included in the analysis.
The values were reported as: • Mean difference in average haemoglobin values at day 7. • Mean difference in percent haemoglobin value change between baseline and day 7. • Relative risk of ≥5, ≥10 or ≥20% decline in haemoglobin concentration by day 7. • Relative risk of moderate (≤8 g/dl) and severe (≤5 g/ dl) anaemia at day 7.
The dichotomous outcomes were reported as risk ratios (RRs) and continuous outcomes as mean differences. Both measures are presented with 95% confidence intervals (CIs). When standard deviations were not reported [24], we imputed these values from comparatively sized trials [22].
Synthesis of results All eligible studies of primaquine were summarized and analysed using Review Manager 5.3. Data were presented in tables and meta-analyses performed when appropriate. In the absence of statistical heterogeneity, a fixed-effects model was used. A randomeffects model was used where moderate heterogeneity was detected but it was deemed still reasonable to combine trials.
This systematic review was reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines [25,26]. PRISMA checklist is provided in the Additional files 3. The review protocol was published a priori [3].

Study selection and study characteristics
1904 unique citations were identified from the search (Fig. 2). 1797 were excluded after abstract screening and 98 after applying the inclusion criteria to the full text. We excluded 98 studies for the reasons given in the characteristics of excluded studies Additional File 4. Five randomized controlled trials [22-24, 27, 28] and four controlled observational cohorts [29][30][31][32], all published between 2004 and 2016, met the inclusion criteria (see Table 1). Eight studied primaquine, and one study available as a conference abstract studied tafenoquine. G6PD deficiency detection methods, phenotypes and genotypes are summarized in Table 2. Four studies reported that they used the fluorescent spot test (FST) for screening [27,[29][30][31], and one study used the rapid test rather than FST [28]. Studies also reported enzyme quantification and/or polymerase chain reaction to detect G202A/ A376G mutation [22,23,27,28] or Mahidol mutation   [29,30,32]. The reported genotypes varied across studies, and G6PD activity was not always measured and not consistently reported, so no stratified analysis was possible. PQ compared to placebo in G6PD deficient individuals Four randomized controlled trials reported haemoglobin levels [22,23,27,28]. Three out of the four studies recruited only children and adolescents [22,23,27] and one recruited participants of all ages [28]. Three studies were conducted in individuals with confirmed malaria (P. falciparum [22,27,28]).
Risk of bias assessment The risk of bias of the nine included studies is summarized in Fig. 3.
Baseline characteristics Six studies were classed as low risk of bias [22,[27][28][29][30][31]; in two it was unclear; and in one there were baseline differences that were likely to influence the results [22].
Blinding of outcome assessors One study was low risk (an RCT) [22]; six studies were classed as unknown risk of bias; two studies did not blind outcome assessors [24,32].
Attrition and incomplete outcome data All studies were deemed to be at low risk of attrition bias.
Reporting bias and selective reporting One study was at high risk of reporting bias as stated outcomes were not reported within the results [24]. Of the five trials described as randomized [22-24, 27, 28], risk of bias due to random sequence generation and allocation concealment was low in two studies [22,28] and unclear in the remaining three studies [23,24,27].

Certainty of evidence
Our assessment of the quality of evidence using the GRADE approach is presented in summary of findings tables (Additional files 5,6,7,8,9,10).

High dose (0.75 mg/kg) PQ
PQ compared to placebo in G6PD deficient people Three randomized trials [22,23,27] report 0.75 mg/kg PQ versus placebo in people with G6PD deficiency, all part of larger trials. Numbers with G6PD deficiency were small, and one trial excluded those with severe deficiency [27]. Two trials were considered low risk of bias [22,27], while one was at high risk of bias due to large baseline differences in haemoglobin concentration between groups [23].
Risk of anaemia Two trials reported this [22,23], with an 11-fold increase in the proportion of individuals with ≥20% fall (low certainty evidence, Fig. 4.3).
Moderate and severe anaemia at day 7 Two trials reported this [22,23], with an increased risk of anaemia defined by ≤8 g/dl by day 7, with one individual with Hb ≤5 g/dl (low certainty evidence, Fig. 4.4).
PQ in G6PD deficient compared to G6PD replete people Five controlled observational cohorts evaluated changes in haemoglobin when 0.75 mg/kg PQ was given to people with and without G6PD deficiency [22,23,27,30,31].
Risk of anaemia Two trials reported this [22,23] and there was a 3.4-fold increase in risk for ≥20% fall (very low certainty evidence, Fig. 5.3).
Moderate and severe anaemia at day 7 Two trials reported this [22,23] and there was increased risk of anaemia defined as ≤ 8 g/dl by day 7, with one individual with Hb ≤5 g/dl (Fig. 5.4).

Mid-range dose (0.4-0.5 mg/kg)
PQ compared to placebo in G6PD deficient people One trial [27] evaluated 0.4 mg/kg PQ. This trial did not report haemoglobin at day 7 or risk of anaemia.
PQ in G6PD deficient compared to G6PD replete people Two controlled cohort studies evaluated changes in haemoglobin with a mid-range dose of PQ given to G6PD deficient and G6PD replete people [24,27]. The doses used in these studies ranged from 0.4 to 0.5 mg/kg. The two trials did not report risk of anaemia.
Percent haemoglobin value change between baseline and day 7 One trial [27] reported this and found greater % change in G6PD deficient people (MD −4.99%, 95% CI −9.96 to −0.02, 113 participants, very low certainty evidence, Fig. 7.2). (0.1-0.25 mg/kg) PQ compared to placebo in G6PD deficient people Two trials [27,28] provide data for this, using a single dose of PQ versus placebo in people with G6PD deficiency. The doses in these studies ranged from 0.1 to 0.25 mg/kg. These trials did not report mean haemoglobin at followup or risk of anaemia.

Tafenoquine (Table 3)
Tafenoquine in G6PD deficient compared to G6PD replete people One controlled observational cohort compared average maximum fall in haemoglobin values in G6PD   deficient and G6PD replete people given tafenoquine. 300 mg (5 mg/kg) of tafenoquine produced a greater average maximum decline in haemoglobin over an average of 9 days in G6PD deficient than in G6PD replete individuals ( Table 3). The average maximal decline in haemoglobin following 100 mg (1.67 mg/kg) and 200 mg (3.33 mg/kg) of TQ did not demonstrate a difference.

Discussion
In this review, comparative information about effects of PQ (and tafenoquine) on haemoglobin in G6PD deficient people was summarized. The aim was to quantify the risk of adverse effects in those given single dose primaquine to reduce infectiousness of P. falciparum. Few trials to date have addressed this issue directly, so indirect evidence from trials that had the main aim of assessing impact of 8-AQ on P. falciparum gametocytes was sought. Whilst these trials generate some useful information on haemoglobin, they were not necessarily powered for the outcomes studied here. The nine trials in this review were conducted with varying levels of methodological rigor. Nevertheless, with the 0.75 mg/kg single dose in G6PD deficient people there is an average 1.45 g/dl greater fall in haemoglobin by day 7 compared to placebo, and risks of anaemia (20% drop in haemoglobin) are increased 11-fold. For the mid-range dose of PQ in G6PD deficient compared to G6PD replete people, one trial reported a greater fall in haemoglobin of 3.92 g/dl by day 7.
For the currently recommended low dose of 0.25 mg/ kg, there was no difference in the percentage fall in haemoglobin by day 7 in G6PD deficient people given low dose PQ compared to placebo. However, in one trial of PQ given to G6PD deficient compared with G6PD replete people, the average haemoglobin value at day 7 was 0.57 g/dl lower in the deficient group. No difference was seen in percent change in haemoglobin between G6PD deficient and replete people. These data are somewhat reassuring, but the total number of the trials and people in the low dose (0.1 to 0.25 mg/kg) is few, and large falls in a few individuals cannot be excluded.
The lack of sufficient data prevented any analysis stratified by G6PD type and enzyme activities. The reported genotypes varied across studies. It should be noted that one trial contributing to both comparisons [27,33] included only G6PD 'normal' participants based on the FST test, but further genotyping showed wide variation in G6PD genotypes. Therefore, this study likely underestimated the effect in G6PD deficient participants. Individual patient data meta-analysis is needed to explore variation due to phenotypic G6PD activity, genotype or applied diagnostic method, which is beyond the scope of the present analysis.
The enduring uncertainty surrounding safety, shown clearly by low or very low certainty of the evidence, should prompt further research on the effect of lower doses of PQ in people with G6PD deficiency, including examining effects on the most prevalent genotypes. Only this additional information will enable policies to be based on an adequate evidence base and reassure national and local programme managers that the drug is safe to be used without prior G6PD testing.
Several studies on the safety of the currently recommended single low dose of 0.25 mg/kg PQ in combination with artemisinin combination therapy in G6PD deficient individuals are ongoing (clinicaltrials.gov NCT02174900; NCT02654730; NCT02535767). These studies will help refine our understanding of the safety of PQ in G6PD deficient individuals. Enhanced and continuing pharmacovigilance is needed to fully characterize the safety of PQ in the range of G6PD deficiency variants [34,35].

Conclusions
The evidence assessed for this systematic review suggests that the lower single dose (0.1 to 0.25 mg/kg) of PQ given with the goal of reducing transmission of P. falciparum is less likely to cause haemolytic effects in people with G6PD deficiency than the previous 0.75 mg/kg dose, and that severe haemolytic events are not very common. However, the evidence is based on few trials and a small number of people. The roll-out and scale-up of the 0.25 mg/kg dose in countries without prior experience would greatly benefit from more reliable data demonstrating the safety of this dose.