Protective effect of Mediterranean-type glucose-6-phosphate dehydrogenase deficiency against Plasmodium vivax malaria
Abstract
X-linked glucose-6-phosphate dehydrogenase (G6PD) deficiency is the most common human enzymopathy. The severe Mediterranean variant (G6PD Med) found across Europe and Asia is thought to confer protection against malaria, but its effect is unclear. We fitted a Bayesian statistical model to observed G6PD Med allele frequencies in 999 Pashtun patients presenting with acute Plasmodium vivax malaria and 1408 population controls. G6PD Med was associated with reductions in symptomatic P. vivax malaria incidence of 76% (95% credible interval [CI], 58–88) in hemizygous males and homozygous females combined and 55% (95% CI, 38–68) in heterozygous females. Unless there is very large population stratification within the Pashtun (confounding these results), the G6PD Med genotype confers a very large and gene-dose proportional protective effect against acute vivax malaria. The proportion of patients with vivax malaria at risk of haemolysis following 8-aminoquinoline radical cure is substantially overestimated by studies measuring G6PD deficiency prevalence in healthy subjects.
Introduction
In red blood cells, glucose-6-phosphate dehydrogenase (G6PD; EC 1.1.1.49) is the only source of reduced nicotinamide adenine dinucleotide phosphate (NADPH) (Beutler, 1994; Luzzatto and Arese, 2018). G6PD deficiency reflects instability, not absence, of this enzyme, which is essential for normal cellular function. Mammalian red blood cells lack nuclei and the necessary protein synthetic pathway, and so, unlike nucleated cells, they cannot replenish degraded G6PD. In G6PD deficiency, the active enzyme content of erythrocytes declines markedly as they age. NADPH is essential for the maintenance of oxidant defences. Thus, as G6PD deficient red cells age, they become increasingly susceptible to oxidant haemolysis. G6PD deficiency is the most common enzyme abnormality in humans. It is found across the malaria-endemic world with mutant gene prevalences up to 35% (average 8–10%) (Howes et al., 2012). There are over 200 different polymorphic variants, most of which result in enzyme deficiency, but the degree of deficiency (from accelerated enzyme degradation) and thus vulnerability to oxidant haemolysis varies substantially among the different genotypes. G6PD deficiency is X-linked, so males are either normal or fully deficient. Women display these two phenotypes (normal or homozygous deficiency) as well as intermediate deficiency (heterozygotes). The heterozygote females are genetic mosaics as a result of early embryonic random X-chromosome inactivation (Lyonisation). Their blood contains a mixture of G6PD-normal and G6PD-deficient erythrocytes. Overall, at a population level, the proportion averages 50:50 of each cell type, but there is inter-individual variation, so in some heterozygotes, the large majority of erythrocytes are G6PD deficient. The high prevalences of G6PD deficiency in tropical areas, particularly in Africa, and in areas where malaria was once endemic, suggest that G6PD deficiency confers protection either against malaria or its adverse effects. But the mechanism of protection and its extent is unclear. This has been a subject of controversy and divergent opinion, with no clear conclusion. Claims have been made that there is no malaria protective effect provided by G6PD deficiency or that protective effects are seen in female heterozygotes only, or in male hemizygotes only, or in both (Mbanefo et al., 2017; Bienzle et al., 1972; Ruwende et al., 1995; Guindo et al., 2007; Mombo et al., 2003; Lopera-Mesa et al., 2015; Uyoga et al., 2015; Clarke et al., 2017). Most of the studies addressing this question have focussed on falciparum malaria in Africa where the majority of evidence supports a protective effect against severe malaria, particularly in female heterozygotes (Uyoga et al., 2015; Clarke et al., 2017). Whether male hemizygotes and female homozygotes are protected is unclear with evidence both for and against. A recent meta-analysis of 28 studies suggested a moderate protective effect against uncomplicated falciparum malaria (odds ratio [OR]: 0.77; 95% credible interval [CI], 0.59–1.02), but this estimate could be affected by publication biases (Mbanefo et al., 2017).
In 2002, Richard Carter and Kamini Mendis suggested that the evolutionary force selecting G6PD deficiency could have been either Plasmodium falciparum or Plasmodium vivax (Carter and Mendis, 2002). Historically, P. vivax had a wider geographic distribution, although it has now been eradicated from North America, Europe, and Russia. Elsewhere in the Americas, the horn of Africa, Asia, and Oceania, P. vivax has become the predominant cause of malaria in recent years. In general, the variants of G6PD deficiency that are prevalent in these areas where P. vivax infections occur, or once occurred, are more severe than the common (‘A−’) variant prevalent in the sub-Saharan African populations (in whom P. vivax malaria is rare and P. falciparum comprises the large majority of malaria infections), and in people with their genetic origin there. The most severe of the commonly found G6PD variants is the ‘Mediterranean’ variant (‘G6PD Med’). This results from a single C-T transition at nucleotide position 563, causing a serine phenylalanine replacement at amino acid position 188. G6PD Med is the predominant genotype in the Pashtun who live in Afghanistan, Pakistan, and India (Bouma et al., 1995; Leslie et al., 2010; Jamornthanyawat et al., 2014). An earlier retrospective study conducted in Afghan refugees in North-Western Pakistan (Leslie et al., 2010) suggested that G6PD Med hemizygotes were protected against vivax malaria, but there were too few observations to substantiate a trend to lower infection rates in female heterozygotes. To characterise the possible protective effects of G6PD Med against P. vivax malaria, we conducted a retrospective analysis of case–control data from clinical studies on vivax malaria, and epidemiological studies of G6PD deficiency that we have conducted in Afghanistan over the past 10 years. These data were then combined in a meta-analysis using all previously published data on G6PD deficiency in people of Pashtun ethnicity living in malaria-endemic areas.
Results
Retrospective case–control study
In total, 764 Pashtun patients presenting with acute vivax malaria (304 males, 460 females) and 699 Pashtun controls (342 males, 357 females) were studied; 236 healthy males came from the epidemiology study reported previously (Awab et al., 2017) and the remaining control subjects came from the same locations as the clinical malaria studies (Table 1; Figure 1). In the controls, the allele frequency of G6PD Med was estimated to be 7.8% (95% credible interval [CI], 6.3–9.5) under the assumption of Hardy–Weinberg equilibrium. There was no evidence of departure from Hardy–Weinberg in the controls (p=0.9). The proportions of G6PD Med male hemizygotes and female heterozygotes were substantially lower in patients with acute vivax malaria than in people incidentally visiting clinics or vaccination centres who did not have malaria (Table 1). Only 1.6% (5 of 304) of males with vivax malaria were hemizygotes (risk ratio [RR]; 95% CI, 0.12 [0.08–0.51]), while in the females with vivax malaria, 7.0% (32 of 460) (RR 0.61 [0.39–0.95]) and 0.7% (3 of 460) were heterozygotes and homozygotes, respectively. Under the Bayesian model, assuming Hardy–Weinberg equilibrium, these results suggest that G6PDd Med hemizygous males and homozygous females have 68% protection (95% CI, 39–85; i.e. relative reduction, given by 1-α in the model) against acute P. vivax malaria, and G6PDd Med heterozygous females have 51% protection (95% CI, 28–67, given by 1-β in the model).

Locations of the two vivax malaria clinical study sites in Eastern Afghanistan from the present study (red circles), and the approximate locations of the villages in the North-West frontier province of Pakistan where Afghan Pashtun refugees were enrolled in vivax malaria clinical trials and later included in case–control studies (Leslie et al., 2010) (blue circles).
Summary of all case–control data included in the meta-analysis.
*Data from this report; 23 of 236 male controls from the earlier epidemiological study (Jamornthanyawat et al., 2014) and 5 of 106 male controls from the later studies were hemizygotes.
General population (controls) | P. vivax malaria (cases) | ||||||||
---|---|---|---|---|---|---|---|---|---|
Awab et al* | Leslie et al., 2010 | Bouma et al., 1995 | Total | Awab et al* | Leslie et al., 2010 | Bouma et al., 1995 | Total | ||
Males | Hemizygous | 28 | 31 | 25 | 84 | 5 | 2 | 0 | 7 |
Normal | 314 | 285 | 214 | 813 | 299 | 155 | 0 | 454 | |
Females | Homozygous | 2 | 2 | 0 | 4 | 3 | 0 | 0 | 3 |
Heterozygous | 50 | 26 | 0 | 76 | 32 | 6 | 0 | 38 | |
Normal | 305 | 126 | 0 | 431 | 425 | 72 | 0 | 497 |
Geometric mean (range) P. vivax parasite densities were very similar in patients with wild-type G6PD (2099; 80–67,000 parasites/µL; N = 570) and female G6PD Med heterozygotes (2064; 310–26,110 parasites/µL; N = 25) and were slightly lower in the six patients with quantitative parasite counts who were G6PD Med hemizygotes or homozygotes (922; 150–3720 parasites/µL; p=0.14 from a logistic regression).
Effect of a haemoglobin exclusion criterion
In our earlier studies, patients with moderate or severe anaemia (haemoglobin <8 g/dL) were excluded, which could have biased our results. We therefore compared distributions of haemoglobin concentrations in the earlier clinical trials, in which this exclusion was applied, with those in the more recent studies in which there was no anaemia exclusion criterion (Figure 3). This indicated that 9.6% (29 of 302) patients would have been excluded if the earlier study threshold had been applied. This is predicted to have resulted in exclusion of a single female heterozygote. There was no significant difference in the haemoglobin concentrations at presentation comparing G6PD Med hemizygotes or homozygotes, heterozygotes, or wild type (Figure 3). This shows that exclusion of anaemic patients in the first part of the study did not affect the interpretation of the protective effect of G6PD deficiency on vivax malaria.
Meta-analysis
These data were then combined with data from the two previously reported studies on the prevalence of G6PD deficiency in the Pashtun ethnic group from Afghanistan. One was in healthy subjects only (Bouma et al., 1995), and the other included both vivax malaria patients and matched healthy control subjects (Leslie et al., 2010; Table 1). The meta-analysis of all three studies gave a slightly higher G6PD Med allele frequency of 8.8% (95% CI, 7.6–10.1). The overall protective effect in male hemizygotes and female homozygotes was estimated as 76% (95% CI, 58–88) and in female heterozygotes was 55% (95% CI, 38–68). The posterior distributions for these estimates from the meta-analysis are shown in Figure 2. The posterior probability was 0.98 that the protective effect observed in female heterozygotes was less than the protective effect observed in male hemizygotes and female homozygotes , suggesting that the protective effect is proportional to the gene dose.

Results from the meta-analysis assessing the protective effect of the Mediterranean variant of G6PD deficiency against Plasmodium vivax malaria.
The posterior distributions of 1-α (top: hemi/homozygotes) and 1-β (bottom: heterozygotes) are shown as percentages. These values represent the reduction in prevalence of clinical vivax malaria relative to G6PD normal individuals. The circles show the median estimates, with the 50% credible intervals shown by the thick blue lines and the 95% credible intervals shown by the thin blue lines.
Discussion
Mediterranean-type glucose-6-phosphate dehydrogenase deficiency (G6PD Med) prevalent in the Pashtun provided a strong and gene-dose-related protective effect against Plasmodium vivax malaria. This is a much greater protective effect than observed against P. falciparum malaria elsewhere (Mbanefo et al., 2017; Uyoga et al., 2015). It is probably explained by two factors. First, the degree of enzyme deficiency with G6PD Med is substantially greater than in the common African A− variant, which has been the main genotype studied previously in falciparum malaria studies. Second, P. vivax is generally more sensitive to oxidant effects than P. falciparum. Compared with P. falciparum, asexual stages of P. vivax are more sensitive to oxidant drugs (i.e. artemisinins and synthetic peroxides and 8-aminoquinolines) (Phyo et al., 2016; Pukrittayakamee et al., 2000). P. vivax may therefore be more sensitive to the oxidant stresses associated with G6PD deficiency. This large study from Afghanistan confirms earlier findings from a case–control study in Afghan Pashtun refugees, based mainly on phenotyping. This earlier study showed clear evidence of protection against vivax malaria in male hemizygotes, but in the smaller subgroup of genotyped female heterozygotes the uncertainty around the estimated effect was large (adjusted odds ratio [AOR]: 0.4, 95% CI, 0.16–1.02) (Leslie et al., 2010). These earlier studies, and a smaller series from Iran with phenotyping (Ebrahimipour et al., 2014), are all consistent with the present series. Combined together they show clearly that G6PD Med provides a substantial gene dose proportionate protection against vivax malaria. The inference of the gene-dose effect in females assumes that hemizygote males and homozygote females are phenotypically identical, and, as expected from the Hardy–Weinberg equilibrium, there were few female homozygotes, so the majority of the protective signal in this combined group is from the hemizygote males. In a survey conducted in periurban Manaus, in the Amazon region of Western Brazil, where P. vivax is now the predominant (90%) cause of malaria, there was also a very strong protective effect of G6PD Med against self-reported previous malaria (AOR: 0.010) (Santana et al., 2013). In comparison the protective effect of G6PD A– in the same location was an order of magnitude weaker (AOR: 0.119). This marked protective benefit of G6PD deficiency against P. vivax infections is critically important for the assessment of population haemolytic risk associated with giving 8-aminoquinoline antimalarials for the radical cure for vivax malaria. These are the only drugs providing radical cure of vivax malaria (prevention of relapse), but they are underused because G6PD deficiency testing is usually not available and prescribers are naturally concerned about precipitating dangerous haemolysis. The proportion of patients with vivax malaria at risk of serious haemolysis with G6PD Med in these studies is nearly four times lower than would be predicted from gene frequencies in the healthy population.
G6PD Med is among the most severe of the polymorphic genetic G6PD variants. It is found across the malaria-endemic regions of the world, having evolved independently on several occasions (Jamornthanyawat et al., 2014; Kurdi-Haidar et al., 1990). There is also evidence for protection against vivax malaria from the moderate severity G6PD Mahidol variant. A study from western Thailand found that P. vivax densities were lower in hemizygote males and also in heterozygote females presenting to clinics with vivax malaria, but there was no corresponding effect in falciparum malaria, and there was no apparent protective effect against malaria illness (Louicharoen et al., 2009). However, a case–control study from Northern Myanmar in a population with a high prevalence (25.2%) of the G6PD Mahidol variant gave adjusted odds ratios for having acute vivax malaria of 0.213 (95% CI, 0.093–0.487) for male 487A hemizygotes, and 0.248 (0.110–0.561) for female heterozygotes (Yi et al., 2019). In a multi-site survey of G6PD deficiency in malaria patients in Cambodia, where G6PD Viangchan predominates (WHO class two but quantitatively similar to G6PD Mahidol), phenotypic severe deficiency (i.e. <10% of population normal) provided stronger protection against P. vivax than P. falciparum infections (OR: O.45; 95% CI, 0.32–0.64) (Khim et al., 2013). Conversely, Dewasurendra et al. in Sri Lanka reported evidence of hemizygote protection from malaria and lower parasite densities for P. falciparum, but not P. vivax (Dewasurendra et al., 2015).
A recent meta-analysis of 28 studies (Mbanefo et al., 2017) addressing the question of whether G6PD deficiency protects against malaria showed a moderate protective effect against uncomplicated falciparum malaria (OR: 0.77; 95% CI, 0.59–1.02). The degree of protection was similar in female heterozygotes (OR: 0.7; 95% CI, 0.57–0.87) and in male hemizygotes and female homozygotes (OR: 0.7; 95% CI, 0.46–1.07) but did not reach statistical significance in the latter group. There was evidence for publication bias towards significant findings in the uncomplicated malaria comparisons. The same meta-analysis also concluded that there was no statistically significant protective effect in severe malaria or in P. vivax malaria, although there were limited data to assess the protective effects in P. vivax infections.
The degree of protection conferred by G6PD Med against P. vivax illness estimated in this study is large; it is similar in magnitude to the well-described protection conferred by Hb AS (sickle cell heterozygotes) against falciparum malaria (Williams et al., 2005).
It is possible that G6PD deficiency confers no significant protection against uncomplicated P. falciparum malaria, but protects only against life-threatening illness. This may reflect either inhibition of parasite multiplication or a different protective mechanism. One consistent clinical feature of G6PD deficiency is an increased risk of anaemia in acute infections as the deficient red cell haemolyse (Uyoga et al., 2015; Clarke et al., 2017). As severe anaemia (Hb < 5 g/dL) is one of the criteria for defining severe malaria, this results in a higher proportion of patients with G6PD deficiency presenting with severe malarial anaemia, and therefore being diagnosed as having severe malaria. This can bias genetic association studies (Watson et al., 2019). It has been suggested that G6PD deficiency may protect specifically against cerebral malaria, but a simpler explanation is that, in the context of severe falciparum malaria, a rapid onset of moderate anaemia protects against life-threatening complications such as cerebral malaria (White, 2018; Leopold et al., 2019). In most areas, P. falciparum and P. vivax coexist. The interaction between the malaria species is complex. P. falciparum was once prevalent in malaria-endemic regions of Afghanistan, but it has now been all but eliminated. There were no coinfections identified in this study, so P. falciparum was not a confounder.
There are several limitations to this study. It was not designed prospectively as a case–control study. It combines results from a prospective epidemiology study conducted 6 years ago and prospective clinical trials and sampling of controls from the same centres from 2018. For security reasons, careful matching of cases and controls (other than for location) was not possible. We do not have additional genetic marker data that could be used to adjust for population stratification. However, in the earlier study by Leslie et al., 2010, conducted in refugee camps in Pakistan, careful matching was done, particularly with reference to tribe (within the Pashtun group) and location, and that study’s findings are consistent with the present investigation. Nevertheless, it remains possible that uncharacterised variations between the different investigations and genetic heterogeneity within the Pashtun group are confounders. As many of the controls had non-malaria febrile illnesses, it is possible, although unlikely, that the prevalence of G6PD deficiency is higher in such patients to that in the general population. In the initial clinical studies reported here, and in the earlier studies in Pashtun refugees (Leslie et al., 2010), severe anaemia was an exclusion criterion. This could have reduced the proportion of G6PD-deficient patients, but there was no evidence for differences in presenting haemoglobin concentrations between G6PD-deficient and G6PD-normal patients (Figure 3). This study was confined to the G6PD Mediterranean genotype, so other G6PD polymorphisms were not studied, but these are unusual in the Pashtun. If there were undetected G6PD-deficient patients, then they would have diluted the apparent protective benefit. Potential confounders become progressively more relevant as effects become smaller.

Distribution of admission haemoglobin concentrations in P. vivax malaria cases (overall: top row; stratified by G6PD genotype: bottom row).
The top row shows the distribution of haemoglobin concentrations in the first part of the study (left panel, when there was a cut-off at 8 g/dL for inclusion) and in the second part of the study (right panel: after this inclusion criteria was relaxed). The bottom row shows these two distributions stratified by the patients’ G6PD genotypes.
Overall, this study shows that the G6PD Mediterranean genotype confers a very large and gene-dose proportional protective effect against vivax malaria. If G6PD Med is approximately at an equilibrium allele frequency in the Pashtun population (i.e. is under balancing selection), this implies that it must be harmful, i.e. malaria alone cannot create balancing selection. The main risks associated with G6PD deficiency are neonatal hyperbilirubinaemia and haemolysis following oxidant food and drugs, while the benefit is protection from malaria. Importantly for malaria treatment, the population risks associated with 8-aminoquinoline radical cure are greatly overestimated from G6PD prevalences in the healthy population.
Materials and methods
Study area and participants
Request a detailed protocolIn Afghanistan, malaria is confined to the northern plains, the Jalalabad basin and river valleys, which fringe the central mountains to the west and south (Figure 1). Malaria cases begin presenting in May, peak in July and August, and then decline with few cases seen after November. These studies were conducted over 10 years between 2009 and 2019.
The initial epidemiological study to assess the prevalence and genotypes of G6PD deficiency in Afghanistan was undertaken in one urban centre in each of nine provinces: Jalalabad (Nangarhar province), Mehtherlam (Laghman), Asadabad (Kunar), Maimana (Faryab), Taloqan (Takhar), Imamsahib (Kunduz), Pulikhumri (Baghlan), Pulialam (Lowgar), and the capital, Kabul (Kabul province) (Jamornthanyawat et al., 2014). Provinces were chosen on the basis of practical accessibility combined with having a high risk of vivax malaria (>1 case/1000 population/year), the exceptions being Kabul and Lowgar which have low malaria transmission. This analysis is restricted to the Pashtun ethnic group who live in the higher P. vivax transmission areas for two main reasons: the Pashtun comprises the majority of people in the Eastern provinces where malaria incidence is highest, and G6PD variants other than G6PD Med are rare in this ethnic group (Bouma et al., 1995; Leslie et al., 2010).
In the initial prevalence study in healthy subjects conducted in 2009, only males were studied (Jamornthanyawat et al., 2014). A convenience approach to sampling was used for reasons related to security. EDTA blood was obtained from healthy male adults and children attending outpatient medical laboratories with non-febrile illnesses or neighbouring vaccine administration who did not have malaria. Samples were transferred to filter-paper blood spots and stored in plastic zip-lock bags with silica gel at room temperature before transport to Bangkok for genotyping (Jamornthanyawat et al., 2014). In later prospective studies conducted in 2018 and 2019, subjects (both male and female) attending outpatients and those attending immunisation centres were sampled in the same way from the same sites in Nangarhar province as the clinical malaria studies (Figure 1).
Clinical studies
Request a detailed protocolPatients with symptomatic malaria presented to provincial malaria control centres adjacent to the Pakistan border. Jalalabad, the capital city of Nangarhar province, is a referral centre for the whole eastern region. Population movement takes place in both directions across this largely mountainous area bordering Pakistan. The malaria treatment protocols have been reported previously (Awab et al., 2017). Patients (aged >6 months) or older presenting with uncomplicated microscopy confirmed symptomatic vivax malaria were enrolled if they, or their carer, gave fully informed written informed consent. Inclusion criteria were monoinfection with P. vivax, axillary temperature ≥ 37.5°C or oral/rectal temperature ≥ 38°C, or history of fever in preceding 24 hr, and able to swallow oral medication and comply with study requirements. Exclusion criteria were any clinical or laboratory feature of severe malaria, significant comorbidity, known hypersensitivity to any study drugs, mixed species Plasmodium infection, and pregnancy or lactation. In the initial studies, patients with haemoglobin concentrations ≤ 8 g/dl were excluded, but in the later observational studies in 2019, anaemic patients were not excluded.
Laboratory procedures
Request a detailed protocolCapillary blood was collected for haemoglobin measurement (HemoCue), Giemsa staining of thick and thin blood films for parasite speciation and counts and G6PD testing. Before 2018 G6PD deficiency screening used the G-6-PD OSMMR 2000 (R and D Diagnostics) and NADPH fluorescent spot test kit (10 min incubation). From 2018, the CareStart rapid G6PD test was substituted. Dried blood spots were stored for later genotyping (Leslie et al., 2010; Awab et al., 2017). Genomic DNA was extracted using the QIAamp DNA Mini Kit (QIAGEN, Germany). Eluted genomic DNA samples were frozen at −20°C until PCR amplification. All samples were subjected to PCR-RFLP to assess the G6PD Mediterranean variant in exon 6, based on the protocol of Samilchuk et al., 1999 with modifications including use of primers F14948 and R15158 at 250 nM, a 2 µl volume of each genomic DNA template, and a final reaction volume of 30 µl containing 20 mM Tris–HCl (pH 8.4), 50 mM KCl, 1 mM MgCl2, 125 µM 4-deoxynucleotide triphosphate, and 0.05 units PlatinumTaq DNA polymerase (Invitrogen, Brazil). After pre-denaturation at 95°C for 5 min, 45 PCR cycles were performed involving denaturation at 94°C for 1 min, annealing at 55°C for 1 min, and extension at 72°C for 1 min, with post-extension at 72°C for 7 min. Ten microlitres of each PCR product was digested with 10 units of the restriction enzyme MboII (New England Biolabs Inc) at 37°C for 3 hr and visualised by 2% agarose gel electrophoresis. A product pattern of 104+98+28 bp indicates the G6PD Med mutation, whereas 202+28 bp indicates wild type. All PCR products of samples with mutant type and 10% of the wild-type samples were purified subsequently and the DNA was sequenced (Macrogen, Korea). Gene sequences of the PCR amplification products of g6pd (accession number: NM_001042351.2) were confirmed using the NCBI’s Blastn and Blastx programmes. Gene polymorphisms were assessed by comparison with the reference sequences using BioEdit v7.2.5.
Additional data included in the meta-analysis assessing the protective effect of G6PD deficiency
Request a detailed protocolA systematic literature review identified two other sources of relevant data in Pashtun subjects. Bouma et al., 1995 described a phenotyping survey in male Afghan and Pakistani Pashtun refugee schoolchildren (Safi and Mamund tribes, respectively). Leslie et al., 2010 described a P. vivax case–control study in which patients who had been enrolled in earlier clinical trials assessing vivax malaria treatments were matched to community controls. This was also an Afghan refugee population. The refugee villages were located close to Peshawar in the North-West frontier province of Pakistan (Figure 1), close to the border with Afghanistan. The majority of individuals were also from the Safi and Mamund tribes. This study relied mainly on phenotyping (except for heterozygous females). In the meta-analysis, the phenotyping of males was assumed to be 100% sensitive and specific for the G6PD Med genotype. Thus, for the study reported by Leslie et al., we included all data from males, but for females, we only included data from genotyped individuals (154 genotyped female controls and 78 genotyped female cases) as phenotyping identifies only a minority of heterozygotes. The inclusion and exclusion criteria for the vivax cases reported by Leslie et al. were very similar to those in our clinical studies in Eastern Afghanistan.
Statistical analysis
Request a detailed protocolTo estimate the degree of protection against vivax malaria and investigate a possible gene-dose effect, we analyzed all the available data under a Bayesian statistical model that estimates the relative risk reduction. We chose this model structure instead of a logistic regression with the allele as the predictive variable and the case–control status as the outcome (the usual model choice for genetic case–control studies) because we can assume that the protective effect in G6PD-deficient hemizygous males and homozygous females is the same (i.e. from a red cell perspective, they are identical). Log-odds estimated from a logistic regression model are not collapsible; therefore, this assumption cannot be input into a logistic regression model. Our formulation, in terms of relative reduction, allows for this assumption to be hard-wired into the model. It should be noted that this assumption applies to the marginal effect of G6PD deficiency, that is females could be more or less likely to have clinical vivax malaria relative to males, but the protective effect of hemizygote and homozygote G6PD deficiency is assumed identical, conditional on having clinical vivax malaria.
The model assumed that Hardy–Weinberg equilibrium held in the studied population(s). We tested this assumption by applying an exact test appropriate for X-linked biallelic markers in the control populations (p=0.9 for the data reported here; p=0.6 for the combined data in the meta-analysis, Graffelman and Weir, 2016). The unknown overall allele frequency of the Mediterranean variant of G6PD deficiency is denoted p. The total number of healthy males (controls) sampled is denoted ; the total number of P. vivax infected males (cases) is denoted ; within these groups, the number of hemizygote deficient males are denoted and , respectively. The same notation is used for the female subjects, except that the superscript HetD refers to heterozygous deficient, HomoD refers to homozygous deficient, and WT refers to G6PD normal.
The likelihood of the observed data for the non-malaria controls is given by:
Where θcontrols = .
The likelihood of the observed data for the vivax malaria cases is given by:
where θcases = .
The parameter α, such that 0≤α ≤1, denotes the vivax malaria protective effect of hemi- and homozygous deficiencies. A value of α = 1 implies no protective effect, and α = 0 implies a complete protective effect. Thus 1-α is the relative reduction in prevalence in hemi/homozygous G6PD-deficient P. vivax malaria cases compared to the controls. The parameter β, such that 0≤β≤1, denotes the protective effect of heterozygous G6PD deficiency with the same interpretation.
Weakly informative Bayesian priors were subjectively chosen as follows:
This model allows for an additive effect (i.e. the protective effect in heterozygous females is less than in homo/hemizygotes) or an increased effect in heterozygotes (expected under a model of balancing selection driven by heterozygous advantage). The Bayesian model was written in Stan (Stan Development Team, 2018), and a reproducible implementation of the results in rstan is provided in the supplemental materials. Four independent chains were run for 106 iterations, with the first half discarded for burn-in and second half thinned every 200 iterations, to give a total of 10,000 samples for the posterior distribution. Parameter estimates and uncertainty levels are reported as median posterior estimates with 95% CI (2.5% and 97.5% quantiles of the posterior distribution). Convergence of chains was checked visually with trace plots.
Ethical approval
Request a detailed protocolThe clinical studies were approved by the Institutional Review Board of the Afghan Public Health Institute, Ministry of Public Health, Afghanistan, the Ethics Committee of the Faculty of Tropical Medicine, Mahidol University, Thailand, and the Oxford Tropical Research Ethics Committee, Oxford University, UK. The clinical trial was registered with the clinical trials website http://www.clinicaltrials.gov/ct under the identifier NCT01178021.
Data availability
All data used in the analysis are available along with the code which is given in the supplementary materials.
References
-
Chloroquine-Primaquine versus chloroquine alone to treat vivax malaria in Afghanistan: an open randomized superiority trialThe American Journal of Tropical Medicine and Hygiene 97:1782–1787.https://doi.org/10.4269/ajtmh.17-0290
-
Prevalence and clinical presentation of glucose-6-phosphate dehydrogenase deficiency in Pakistani Pathan and Afghan refugee communities in Pakistan; implications for the use of primaquine in regional malaria control programmesTransactions of the Royal Society of Tropical Medicine and Hygiene 89:62–64.https://doi.org/10.1016/0035-9203(95)90661-4
-
Evolutionary and historical aspects of the burden of malariaClinical Microbiology Reviews 15:564–594.https://doi.org/10.1128/CMR.15.4.564-594.2002
-
The rate of Plasmodium vivax Infectivity within Glucose-6-Phosphate Dehydrogenase (G6PD) Deficient Individuals in Hormozgan Province, IranIranian Journal of Parasitology 9:402–406.
-
Origin and spread of the glucose-6-phosphate dehydrogenase variant (G6PD-Mediterranean) in the middle eastAmerican Journal of Human Genetics 47:1013–1019.
-
Effect of red blood cell variants on childhood malaria in Mali: a prospective cohort studyThe Lancet Haematology 2:e140–e149.https://doi.org/10.1016/S2352-3026(15)00043-5
-
Favism and Glucose-6-Phosphate dehydrogenase deficiencyNew England Journal of Medicine 378:60–71.https://doi.org/10.1056/NEJMra1708111
-
Human genetic polymorphisms and asymptomatic Plasmodium falciparum malaria in Gabonese schoolchildrenThe American Journal of Tropical Medicine and Hygiene 68:186–190.https://doi.org/10.4269/ajtmh.2003.68.186
-
Therapeutic responses to different antimalarial drugs in vivax malariaAntimicrobial Agents and Chemotherapy 44:1680–1685.https://doi.org/10.1128/AAC.44.6.1680-1685.2000
-
Glucose-6-phosphate dehydrogenase deficient variants are associated with reduced susceptibility to malaria in the Brazilian AmazonTransactions of the Royal Society of Tropical Medicine and Hygiene 107:301–306.https://doi.org/10.1093/trstmh/trt015
-
Sickle cell trait and the risk of Plasmodium falciparum malaria and other childhood diseasesThe Journal of Infectious Diseases 192:178–186.https://doi.org/10.1086/430744
Decision letter
-
Amy WesolowskiReviewing Editor; Johns Hopkins Bloomberg School of Public Health, United States
-
Eduardo FrancoSenior Editor; McGill University, Canada
-
Bridget S PenmanReviewer; School of Life Sciences, University of Warwick, United Kingdom
-
Anirban MondalReviewer; Case Western Reserve University, United States
In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.
Acceptance summary:
The exact relationship between G6PD deficiency and malaria protection remains uncertain. This study provides evidence that the G6PD Med mutation (563 C>T) protects against clinical Plasmodium vivax disease. It uses a Bayesian statistical approach which specifically elucidates the particular protection which female heterozygotes versus male hemizygotes (or female homozygotes) for the Med mutation may experience. This is an important contribution to our understanding of the relationship between G6PD deficiency and P. vivax.
Decision letter after peer review:
Thank you for submitting your article "Protective effect of Mediterranean type glucose-6-phosphate dehydrogenase deficiency against Plasmodium vivax malaria" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and a Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Bridget S Penman (Reviewer #1); Anirban Mondal (Reviewer #2).
The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.
Our expectation is that the authors will eventually carry out the additional experiments and report on how they affect the relevant conclusions either in a preprint on bioRxiv or medRxiv, or if appropriate, as a Research Advance in eLife, either of which would be linked to the original paper.
Summary
The exact relationship between G6PD deficiency and malaria protection remains uncertain. This study provides evidence that the G6PD Med mutation (563 C>T) protects against clinical Plasmodium vivax disease. It uses a Bayesian statistical approach which specifically elucidates the particular protection which female heterozygotes versus male hemizygotes (or female homozygotes) for the Med mutation may experience. This is an important contribution to our understanding of the relationship between G6PD deficiency and P. vivax.
Overall, the reviewers were positive about the work and its potential, but have some clear concerns that will require additional data, analyses, and interpretation. Below are the main points raised by the reviewers that would need to be addressed to for a revised manuscript.
1) The presence of mixed infections: although the work is focused on P. vivax, the majority (95%) of malaria in Afghanistan is caused by P. falciparum that means mixed species infections are likely high and P. falciparum infections may be obscuring P. vivax infections. It is not clear to what extent G6PD deficiency may impact the chance of being coinfected with both falciparum and vivax. Ideally, PCR verification of these samples would be performed to confirm the species for samples included in the analysis. Without this molecular data, the overall assessments of susceptibility to vivax malaria in association with G6PD Med is incomplete.
2) The analysis relies on a number of assumptions made about Pashtun population genetics (e.g. is it reasonable to assume the same frequency of the relevant mutation throughout all the tribes in the study, and should this be at Hardy Weinberg equilibrium?) and it is not clear to what extent these assumptions are justified since little evidence/support is provided. In particular, the assumptions about Hardy Weinberg equilibrium of G6PD Med within the Pashtun population need to be justified and supported since the analysis is highly reliant on this assumption.
3) The exclusion criteria does not appear to have been uniformly applied – in particular anemia was an exclusion criteria for only part of the data. This was not clear and may impact the overall significance of statistical results.
4) While the manuscript makes a number of conclusions about female homozygotes, these are not strongly supported by the evidence. In particular, the study is likely under-powered with regard to clinical associations among female homozygotes with G6PD Med, but this is not addressed and the stated conclusions are likely stronger than what can be supported by the data/analyses provided.
Reviewer #1:
There are aspects of the study which I would like to see explained further before I can be certain of the strength of the evidence.
1) The key result rests on the fact that the proportion of G6PD deficient individuals is lower among patients with vivax malaria than in controls without malaria. However, as the authors note: "In the initial clinical studies reported here severe anaemia was an exclusion criterion and this could have reduced the proportion of G6PD deficient patients [among malaria patients – my insertion]". To enable readers to judge for themselves how big an effect this could have had, it would be useful to indicate the numbers of patients affected -e.g. by providing a supplementary table which splits up the data in table 1 according to before and after the anaemia criterion was changed. It may also be useful to conduct a supplementary analysis among just the data where severe anaemia was not an exclusion criterion – but I do not want to request that without knowing the numbers involved.
2) The Leslie study [14], from which data was included in the meta-analysis had "372 cases and 743 controls", but from the numbers in Table 1 it seems that not all of these have been included in the present meta-analysis. The Materials and methods need to make clear which of the samples from the Leslie study were included in the meta-analysis, and why. I can imagine this may be something to do with the geographical origins of the refugee populations in the Leslie study, since this study focused on high malaria transmission regions only, but this needs to be explained.
3) The method used here implicitly assumes that there is little heterogeneity in G6PD Med frequencies between different Pashtun tribal groups currently living in malaria endemic regions. It is hard to know whether or not this is a reasonable assumption, because inevitably the evidence for G6PD Med frequencies in different tribal groups is based on small numbers (e.g. the Leslie study [14]). The authors do acknowledge this with the statement in the Discussion that "genetic heterogeneity within the Pashtun group" may be a confounder. However, I would like to see this potential limitation of the study discussed in more detail in a specific section of the Discussion or Introduction, including reasons why it is reasonable to assume a consistent Med frequency. Are there any other sources of evidence about genetic diversity within the Pashtun group, e.g. from studies at other loci?
4) The Materials and methods explain the case definition for P. vivax infection for the majority of the data (that which was obtained by the authors). However, it is not stated that the case definition for P. vivax infection was equivalent in the Leslie study, from which data was included in the meta-analysis.
5) The number of female homozygotes in this study is (inevitably) very small. I do not object to the authors including female homozygotes and male hemizygotes in the same category, I can see the biological logic behind it, and I like their entire approach to estimate the gene dose effect. However, on the other hand, surely it remains possible that, despite their similar red blood cells, a female homozygote for G6PD Med may not be exactly the same as a male hemizygote for G6PD Med with respect to their likelihood of displaying clinical vivax malaria. If the authors had fitted α, β and γ (for 3 different genotypes, with γ referring to the female homozygote), it seems likely to me that α and β would have been similar to the current estimates and γ would have had a huge credible interval because there is little data. If the authors concur, it would be worth acknowledging this explicitly in the Discussion – because to me this means that this study tells us quite a lot about male hemizygotes and female heterozygotes, but doesn't tell us much about female homozygotes.
Reviewer #2:
This review is mainly focussed on the statistical analysis part of the article.
The authors have used a well-established and standard Bayesian methodology for inference on the role of G6PD Med as a protective effect against vivax malaria. The Bayesian model is appropriate assuming that the Hardy-Weinberg equilibrium is held in the studied population(s). A Markov Chain Monte Carlo sampling based inference is used for the Bayesian inference.
I have the following concerns and/or comments which should be addressed.
1) The author states that weakly informative Bayesian priors were used in the model. But this is not true for the allele frequency parameter p. I believe Β(2,18) is a strongly informative prior for p. Some explanation for choosing such a strong prior for p should be included in the article. How are the values for hyperparameters (2 and 18) chosen? How sensitive are the results with respect to these hyperparameter values?
2) It is desired that the authors include some plots and statistics assessing the convergence of the MCMC method, e.g. trace plots of the posterior samples, the Gelman-Rubin statistics (Rhat), in the R-markdown document.
3) The Bayesian model assumed that the Hardy-Weinberg equilibrium is held in the studied population(s). Can the authors discuss why this assumption is reasonable in the studied population(s) and how the model can be modified if the equilibrium is disturbed by mutations, natural selection and non-random mating?
Reviewer #3:
The Short Report article suggests that the Mediterranean variant of G6PD deficiency confers a strong gene-dose proportional protective effect against symptomatic Plasmodium vivax malaria in the Pashtun ethnic group in Afghanistan. Results suggesting that G6PDd Med hemizygous males and homozygous females have 67% relative reduction against acute P. vivax malaria and G6PDd Med heterozygous females have 51% relative reduction is of particular interest for malaria and human genetics researchers. The authors present a very thoughtful discussion of P. vivax sensitivity to G6PD deficiency oxidant effects on parasite viability.
1) The authors assume that Hardy-Weinberg equilibrium exists in the populations studied but they provided no evidence that it exists. Please provide the evidence that Hardy-Weinberg equilibrium exists for the alleles/genotypes for G6PD. Please also discuss how a deviation from Hardy-Weinberg equilibrium would affect the results of this study.
2) Previous studies have shown that the G6PD Mediterranean allele is the predominant G6PD deficiency allele in the region. The G6PD gene is, however quite polymorphic, with over 180 mutations and at least two different G6PD Mediterranean allelic haplotypes. Thus the possibility that further genetic polymorphism is associated with G6PD deficiency and significant potential for population mixing (particularly in large cosmopolitan urban settings) may introduce genetic diversity in the G6PD gene. Please discuss how additional G6PD deficiency alleles could affect the results of this study.
3) Brief details were provided on malaria diagnosis by Giemsa-stained thick and thin blood films. P. falciparum accounts for 95% of malaria cases in Afghanistan; P. vivax 5% (http://www.emro.who.int/afg/programmes/malaria-leishmaniasis.html). Diagnosis of P. vivax and accurate assessment of mixed species infections in areas where these two species are co-endemic is known to confound malaria microscopy. Additionally, submicroscopic Plasmodium species infections (involving P. vivax in particular), revealed by parallel PCR methods, is now very well established. Please provide further details as to how Plasmodium species diagnosis was performed, how monoinfection by P. vivax was confirmed and how exclusion of other Plasmodium species and/or mixed species infections were ruled out. If the complexity of Plasmodium species infections has been confirmed by PCR, the authors should revise their Laboratory Procedures section and present their data in these Result section. If Plasmodium species infections have not been confirmed by PCR, the authors should perform this work.
[Editors' note: further revisions were suggested prior to acceptance, as described below.]
Thank you for resubmitting your work entitled "Protective effect of Mediterranean type glucose-6-phosphate dehydrogenase deficiency against Plasmodium vivax malaria" for further consideration by eLife. Your revised article has been evaluated by a Senior Editor and a Reviewing Editor.
The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:
1) Include information in the main text about co-infection. The Who data seems to suggest that co-infection may be possible, although unlikely, and this is not likely to impact the study's outcome. However, the authors should address this point with a few sentences in the manuscript since readers may be concerned by the absence of PCR confirmation.
2) Include a figure to highlight/provide more context/information about the data on the haemoglobin levels in the different genotypes. While the code provide is helpful, it would be much easier for the reader to interpret if it was summarized with a figure/legend. Also, all figures/etc. should be labeled and well references in the supplement.
3) We propose the following smaller edits (suggestions in capitals):
In the Abstract: "Thus, in THE absence of very large population stratification within the Pashtun ethnic group (i.e. confounding these results), the G6PD Med35 genotype confers a very large and gene dose proportional protective effect against acute vivax malaria. "
In the Discussion, when the protective effects of G6PD deficiency against falciparum are discussed, this sentence "It is possible that G6PD deficiency confers no significant protection against infection by P. falciparum, but protects only against life threatening illness." could be improved by replacing "infection by P. falciparum" with something like "mild falciparum disease" or "uncomplicated clinical falciparum malaria".… just because protecting against "infection" could also mean protecting against completely asymptomatic infection, whereas the bulk of the evidence that has been discussed relates to disease with clinical symptoms.
https://doi.org/10.7554/eLife.62448.sa1Author response
Overall, the reviewers were positive about the work and its potential, but have some clear concerns that will require additional data, analyses, and interpretation. Below are the main points raised by the reviewers that would need to be addressed to for a revised manuscript.
1) The presence of mixed infections: although the work is focused on P. vivax, the majority (95%) of malaria in Afghanistan is caused by P. falciparum that means mixed species infections are likely high and P. falciparum infections may be obscuring P. vivax infections. It is not clear to what extent G6PD deficiency may impact the chance of being coinfected with both falciparum and vivax. Ideally, PCR verification of these samples would be performed to confirm the species for samples included in the analysis. Without this molecular data, the overall assessments of susceptibility to vivax malaria in association with G6PD Med is incomplete.
We understand this reasonable concern but this description of the epidemiology in Afghanistan is wrong. It is based upon incorrect information provided (unfortunately) on the WHO EMRO site. The author of the WHO situation report may have got P. falciparum and P. vivax mixed up? The World Malaria report (2020) gives tells the true story. In fact, P. falciparum has now almost gone from Afghanistan, and was rare during the course of this study. It is certainly not a confounder.
2) The analysis relies on a number of assumptions made about Pashtun population genetics (e.g. is it reasonable to assume the same frequency of the relevant mutation throughout all the tribes in the study, and should this be at Hardy Weinberg equilibrium?) and it is not clear to what extent these assumptions are justified since little evidence/support is provided. In particular, the assumptions about Hardy Weinberg equilibrium of G6PD Med within the Pashtun population need to be justified and supported since the analysis is highly reliant on this assumption.
We agree that this is an important point and population stratification between cases and controls could bias the effect estimates. We cannot adjust for population stratification as we do not have other genetic marker data. We have added to the section in the Discussion outlining this limitation. However, the population stratification would have to be very large to account for such a large protective effect. The data from Leslie et al. (from a geographic location close to the current study sites, albeit with a refugee population – see map shown in Figure 1) do show that there are different allele frequencies between the different tribes comprising the Pasthtun ethnic group. We do not have data on case-control allele counts broken down by tribe, so we cannot adjust for this. However, we note that this does not make the model mis-specified per se (as the likelihood is still binomial for the aggregate data). The data from Leslie et al. are carefully matched so are no large differences between cases and controls. It is reassuring that the effect sizes between our data and the data from Leslie are very similar.
We have added a statistical test for deviations from Hardy –Weiberg equilibrium (H-W). This shows no evidence for a departure from H-W in the controls (i.e. data from the controls are entirely consistent with the H-W assumption). This is referred to in the Materials and methods and given in the Source code 1 file.
3) The exclusion criteria does not appear to have been uniformly applied – in particular anemia was an exclusion criteria for only part of the data. This was not clear and may impact the overall significance of statistical results.
We fully accept this criticism of our retrospective study, which we did acknowledge in the paper. We now present a sub-analysis to see if there are significant differences between the two groups (i.e. before versus after: when the exclusion criterion was applied versus afterwards when it was lifted) which might have affected our overall estimates. This shows that the 8g/dL criterion excludes about 10% of patients (and would have excluded 1 of the 10 G6PD deficient patients – a female heterozygote) if applied to the later series. There is no evidence that the exclusion criteria would have affected the effect estimates as there was no difference in the distribution of haemoglobin concentrations between G6PD deficient and G6PD normal patients (analysis now included).
4) While the manuscript makes a number of conclusions about female homozygotes, these are not strongly supported by the evidence. In particular, the study is likely under-powered with regard to clinical associations among female homozygotes with G6PD Med, but this is not addressed and the stated conclusions are likely stronger than what can be supported by the data/analyses provided.
This is correct-the number of female homozygotes is very small, as expected. We are unaware of any evidence that the red cells of hemizygotes and homozygotes are phenotypically different. The assumption that red cells of female homozygotes and male hemizygotes are identical is fully supported by the underlying biology (due to Lyonisation). The model only makes an assumption about relative effects (hemi or homo-zygous deficiency versus WT), so males and females could differ in their absolute risk but this would not affect the model estimates unless there was some other genetic variation that had a direct interaction with G6PD deficiency. Pooling of male hemizygotes and female homozygotes is now the norm for the G6PD deficiency (otherwise we could never say anything about female homozygotes, see for example the recent paper from Kenya: https://ashpublications.org/bloodadvances/article/4/23/5942/474381/Glucose-6-phosphate-dehydrogenase-deficiency-and ). It also the norm for general tests of association on the X-chromosome, see https://pubmed.ncbi.nlm.nih.gov/18441336/. Given the strong background biological knowledge, the rationale for not pooling hemizygotes and homozygotes is unclear.
Reviewer #1:
There are aspects of the study which I would like to see explained further before I can be certain of the strength of the evidence.
1) The key result rests on the fact that the proportion of G6PD deficient individuals is lower among patients with vivax malaria than in controls without malaria. However, as the authors note: "In the initial clinical studies reported here severe anaemia was an exclusion criterion and this could have reduced the proportion of G6PD deficient patients [among malaria patients – my insertion]". To enable readers to judge for themselves how big an effect this could have had, it would be useful to indicate the numbers of patients affected -e.g. by providing a supplementary table which splits up the data in Table 1 according to before and after the anaemia criterion was changed. It may also be useful to conduct a supplementary analysis among just the data where severe anaemia was not an exclusion criterion – but I do not want to request that without knowing the numbers involved.
We present a supplementary analysis split as requested.
2) The Leslie study [14], from which data was included in the meta-analysis had "372 cases and 743 controls", but from the numbers in Table 1 it seems that not all of these have been included in the present meta-analysis. The Materials and methods need to make clear which of the samples from the Leslie study were included in the meta-analysis, and why. I can imagine this may be something to do with the geographical origins of the refugee populations in the Leslie study, since this study focused on high malaria transmission regions only, but this needs to be explained.
This is partly a mistake on our part and partly a lack of clarity. We had only included data from Leslie et al. for the individuals who had been genotyped (418 samples). However, genotyping data are only necessary for females (where the phenotypic test cannot determine heterozygosity). Therefore, we can augment the total number of samples in the meta-analysis to 316 male controls and 157 male cases (total of 703 cases and controls). We can only use data from the genotyped females reported in Leslie for the meta-analysis (124 controls and 78 cases). We have added one line to the Materials and methods to clarify this.
3) The method used here implicitly assumes that there is little heterogeneity in G6PD Med frequencies between different Pashtun tribal groups currently living in malaria endemic regions. It is hard to know whether or not this is a reasonable assumption, because inevitably the evidence for G6PD Med frequencies in different tribal groups is based on small numbers (e.g. the Leslie study [14]). The authors do acknowledge this with the statement in the Discussion that "genetic heterogeneity within the Pashtun group" may be a confounder. However, I would like to see this potential limitation of the study discussed in more detail in a specific section of the Discussion or Introduction, including reasons why it is reasonable to assume a consistent Med frequency. Are there any other sources of evidence about genetic diversity within the Pashtun group, e.g. from studies at other loci?
We accept that this is a limitation. As pointed out above the populations studied in this investigation, and those reported earlier by Leslie et al. were located in a similar geographic area divided by the Afghanistan-Pakistan border. There are few population genetic studies of the Pashtun ethnic group, but there is evidence that it is a diverse group (https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0076748). Having additional genetic marker data would allow for adjustment for population stratification but this is not available.
4) The Materials and methods explain the case definition for P. vivax infection for the majority of the data (that which was obtained by the authors). However, it is not stated that the case definition for P. vivax infection was equivalent in the Leslie study, from which data was included in the meta-analysis.
Inclusion criteria in the Leslie et al. clinical trial (which contributed cases to their retrospective case-control evaluation) were microscopically confirmed P vivax malaria, age 3 years or older, written or witnessed verbal consent (by parents in the case of minors), available for the duration of follow-up, and willingness to be tested for G6PD deficiency at admission. Exclusion criteria were general condition requiring hospital admission, evidence of any concomitant infection or disease likely to mask treatment response, known allergy to any classes of the study drugs, known methemoglobin reductase deficiency, treatment within the past 7 days with any drug with known antimalarial properties or with an investigational drug within 30 days, severe anaemia (haemoglobin <7 g/dL), mixed infection (P vivax and P falciparum), and pregnancy (confirmed by testing), lactation, or both. Thus, the criteria, case definition and the patient populations were all very similar to the current studies by Awab et al. with which they were pooled.
5) The number of female homozygotes in this study is (inevitably) very small. I do not object to the authors including female homozygotes and male hemizygotes in the same category, I can see the biological logic behind it, and I like their entire approach to estimate the gene dose effect. However, on the other hand, surely it remains possible that, despite their similar red blood cells, a female homozygote for G6PD Med may not be exactly the same as a male hemizygote for G6PD Med with respect to their likelihood of displaying clinical vivax malaria. If the authors had fitted α, β and γ (for 3 different genotypes, with γ referring to the female homozygote), it seems likely to me that α and β would have been similar to the current estimates and γ would have had a huge credible interval because there is little data. If the authors concur, it would be worth acknowledging this explicitly in the Discussion – because to me this means that this study tells us quite a lot about male hemizygotes and female heterozygotes, but doesn't tell us much about female homozygotes.
We agree that the study conclusions regarding the fully deficient state are almost entirely based on males hemizygotes. However, the model assumption is about a marginal relative effect: it could be the case that female homozygotes and male hemizygotes do not have the same absolute probability of having clinical vivax (for reasons unrelated to G6PD deficiency). However, the model assumption is that, conditional on having clinical malaria, the protective effect of the hemizygote and homozygotes states are equal. There is no evidence that this would not be the case but, as mentioned, the data are too sparse to support this claim (the biological prior takes precedence over the sparse information in the data). We have one sentence in Discussion making this assumption explicit.
Reviewer #2:
This review is mainly focussed on the statistical analysis part of the article.
The authors have used a well-established and standard Bayesian methodology for inference on the role of G6PD Med as a protective effect against vivax malaria. The Bayesian model is appropriate assuming that the Hardy-Weinberg equilibrium is held in the studied population(s). A Markov Chain Monte Carlo sampling based inference is used for the Bayesian inference.
I have the following concerns and/or comments which should be addressed.
1) The author states that weakly informative Bayesian priors were used in the model. But this is not true for the allele frequency parameter p. I believe Β(2,18) is a strongly informative prior for p. Some explanation for choosing such a strong prior for p should be included in the article. How are the values for hyperparameters (2 and 18) chosen? How sensitive are the results with respect to these hyperparameter values?
We disagree that this is a strong prior. The Β(2,18) distribution has 80% of its mass between 2.8% and 19% (median value is 8.6%) Given our prior knowledge (i.e. G6PDd Med in Afghanistan has around 5-10% allele frequency), this fits the standard usage of the “weakly informative”. A plot is shown for illustration in Author response image 1.
2) It is desired that the authors include some plots and statistics assessing the convergence of the MCMC method, e.g. trace plots of the posterior samples, the Gelman-Rubin statistics (Rhat), in the R-markdown document.
The model converges very well given its simplicity (unimodal distribution). Traceplots are now included in the RMarkdown and referred to in the Materials and methods.
3) The Bayesian model assumed that the Hardy-Weinberg equilibrium is held in the studied population(s). Can the authors discuss why this assumption is reasonable in the studied population(s) and how the model can be modified if the equilibrium is disturbed by mutations, natural selection and non-random mating?
We now include standard tests for H-W equilibrium in the RMarkdown (reported in the main text as well). There is no evidence in the controls for departure from H-W.
Reviewer #3:
The Short Report article suggests that the Mediterranean variant of G6PD deficiency confers a strong gene-dose proportional protective effect against symptomatic Plasmodium vivax malaria in the Pashtun ethnic group in Afghanistan. Results suggesting that G6PDd Med hemizygous males and homozygous females have 67% relative reduction against acute P. vivax malaria and G6PDd Med heterozygous females have 51% relative reduction is of particular interest for malaria and human genetics researchers. The authors present a very thoughtful discussion of P. vivax sensitivity to G6PD deficiency oxidant effects on parasite viability.
1) The authors assume that Hardy-Weinberg equilibrium exists in the populations studied but they provided no evidence that it exists. Please provide the evidence that Hardy-Weinberg equilibrium exists for the alleles/genotypes for G6PD. Please also discuss how a deviation from Hardy-Weinberg equilibrium would affect the results of this study.
As above. There is no evidence for large departures from H-W.
2) Previous studies have shown that the G6PD Mediterranean allele is the predominant G6PD deficiency allele in the region. The G6PD gene is, however quite polymorphic, with over 180 mutations and at least two different G6PD Mediterranean allelic haplotypes. Thus the possibility that further genetic polymorphism is associated with G6PD deficiency and significant potential for population mixing (particularly in large cosmopolitan urban settings) may introduce genetic diversity in the G6PD gene. Please discuss how additional G6PD deficiency alleles could affect the results of this study.
Previous studies have only found the G6PD Mediterranean variant in the Pashtun, but these have not been exhaustive. If additional G6PD deficiency variants did occur in this study they would have diluted the protective effect estimate. We have added a sentence in the Discussion.
3) Brief details were provided on malaria diagnosis by Giemsa-stained thick and thin blood films. P. falciparum accounts for 95% of malaria cases in Afghanistan; P. vivax 5% (http://www.emro.who.int/afg/programmes/malaria-leishmaniasis.html). Diagnosis of P. vivax and accurate assessment of mixed species infections in areas where these two species are co-endemic is known to confound malaria microscopy. Additionally, submicroscopic Plasmodium species infections (involving P. vivax in particular), revealed by parallel PCR methods, is now very well established. Please provide further details as to how Plasmodium species diagnosis was performed, how monoinfection by P. vivax was confirmed and how exclusion of other Plasmodium species and/or mixed species infections were ruled out. If the complexity of Plasmodium species infections has been confirmed by PCR, the authors should revise their Laboratory Procedures section and present their data in these Result section. If Plasmodium species infections have not been confirmed by PCR, the authors should perform this work.
This is an unfortunate mistake by WHO EMRO.
https://doi.org/10.7554/eLife.62448.sa2Article and author information
Author details
Funding
Wellcome Trust (Principle Fellowship of NJ White 093956/Z/10/Z)
- Ghulam R Awab
Wellcome Trust (Major Overseas Programme-Thailand Unit Core Grant 093956/Z/10/Z)
- Kanokon Suwannasin
Wellcome Trust (Training Fellowship 107548Z/15/Z)
- Ghulam R Awab
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Acknowledgements
The investigators are grateful to the doctors, laboratory technicians, and malaria workers in the sites in Afghanistan, the Afghanistan National Malaria and Leishmania Control Program (NMLCP), the two provincial health directorates, and the senior management of the Ministry of Public Health and WHO-Afghanistan. We are also very grateful to the Molecular Tropical Medicine and Genetics laboratory staff in the Faculty of Tropical Medicine, Mahidol University for their help with genotyping.
Ethics
Clinical trial registration Clinicaltrials.gov NCT01178021.
Human subjects: The clinical studies were approved by the Institutional Review Board of the Afghan Public Health Institute, Ministry of Public Health, Afghanistan, the Ethics Committee of the Faculty of Tropical Medicine, Mahidol University, Thailand, and the Oxford Tropical Research Ethics Committee, Oxford University, UK.
Senior Editor
- Eduardo Franco, McGill University, Canada
Reviewing Editor
- Amy Wesolowski, Johns Hopkins Bloomberg School of Public Health, United States
Reviewers
- Bridget S Penman, School of Life Sciences, University of Warwick, United Kingdom
- Anirban Mondal, Case Western Reserve University, United States
Version history
- Received: August 25, 2020
- Accepted: February 3, 2021
- Accepted Manuscript published: February 5, 2021 (version 1)
- Version of Record published: February 15, 2021 (version 2)
Copyright
© 2021, Awab et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 1,961
- Page views
-
- 190
- Downloads
-
- 14
- Citations
Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading
-
- Epidemiology and Global Health
Background:
Machine learning (ML) techniques improve disease prediction by identifying the most relevant features in multidimensional data. We compared the accuracy of ML algorithms for predicting incident diabetic kidney disease (DKD).
Methods:
We utilized longitudinal data from 1365 Chinese, Malay, and Indian participants aged 40–80 y with diabetes but free of DKD who participated in the baseline and 6-year follow-up visit of the Singapore Epidemiology of Eye Diseases Study (2004–2017). Incident DKD (11.9%) was defined as an estimated glomerular filtration rate (eGFR) <60 mL/min/1.73 m2 with at least 25% decrease in eGFR at follow-up from baseline. A total of 339 features, including participant characteristics, retinal imaging, and genetic and blood metabolites, were used as predictors. Performances of several ML models were compared to each other and to logistic regression (LR) model based on established features of DKD (age, sex, ethnicity, duration of diabetes, systolic blood pressure, HbA1c, and body mass index) using area under the receiver operating characteristic curve (AUC).
Results:
ML model Elastic Net (EN) had the best AUC (95% CI) of 0.851 (0.847–0.856), which was 7.0% relatively higher than by LR 0.795 (0.790–0.801). Sensitivity and specificity of EN were 88.2 and 65.9% vs. 73.0 and 72.8% by LR. The top 15 predictors included age, ethnicity, antidiabetic medication, hypertension, diabetic retinopathy, systolic blood pressure, HbA1c, eGFR, and metabolites related to lipids, lipoproteins, fatty acids, and ketone bodies.
Conclusions:
Our results showed that ML, together with feature selection, improves prediction accuracy of DKD risk in an asymptomatic stable population and identifies novel risk factors, including metabolites.
Funding:
This study was supported by the National Medical Research Council, NMRC/OFLCG/001/2017 and NMRC/HCSAINV/MOH-001019-00. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
-
- Epidemiology and Global Health
The NIH-funded RECOVER study is collecting clinical data on patients who experience a SARS-CoV-2 infection. As patient representatives of the RECOVER Initiative’s Mechanistic Pathways task force, we offer our perspectives on patient motivations for partnering with researchers to obtain results from mechanistic studies. We emphasize the challenges of balancing urgency with scientific rigor. We recognize the importance of such partnerships in addressing post-acute sequelae of SARS-CoV-2 infection (PASC), which includes ‘long COVID,’ through contrasting objective and subjective narratives. Long COVID’s prevalence served as a call to action for patients like us to become actively involved in efforts to understand our condition. Patient-centered and patient-partnered research informs the balance between urgency and robust mechanistic research. Results from collaborating on protocol design, diverse patient inclusion, and awareness of community concerns establish a new precedent in biomedical research study design. With a public health matter as pressing as the long-term complications that can emerge after SARS-CoV-2 infection, considerate and equitable stakeholder involvement is essential to guiding seminal research. Discussions in the RECOVER Mechanistic Pathways task force gave rise to this commentary as well as other review articles on the current scientific understanding of PASC mechanisms.