Glucose-6-Phosphate Dehydrogenase Deficiency

Literature and regulatory information updated: 6 October 2026.

0. Initial Reference Table

1. Introduction

Glucose-6-phosphate dehydrogenase deficiency is the commonest human enzymopathy, yet remains among the most preventable causes of serious iatrogenic harm because clinical expression depends almost entirely on exposures under medical control: medicines, foods and timely recognition of acute haemolysis. The condition results from variants in G6PD, located at the end of the X chromosome's long arm, reducing activity of the first pentose phosphate pathway enzyme and thus the mature erythrocyte's capacity to regenerate reduced glutathione [5, 12].

The foundational description emerged from the US antimalarial research programme in the 1950s. After observing that certain patients, particularly African Americans, developed haemolytic anaemia when treated with primaquine, Alving, Carson, Flanagan and Ickes demonstrated in a 1956 Science communication that primaquine-sensitive erythrocytes had a measurable enzyme deficiency, establishing for the first time a link between a medicine, an adverse reaction and an inherited biochemical red cell defect [13]. The term “primaquine sensitivity” predated gene identification and remained in use for years; its history explains why G6PD deficiency pharmacogenetics remains a teaching paradigm for the discipline [2, 14].

Favism, a haemolytic episode following fava bean ingestion, was recognised long before modern enzymology and remains an important cause of severe acute anaemia in children in Mediterranean, Middle Eastern and Asian regions; Vicia faba beans contain vicine and convicine, whose metabolites divicine and isouramil generate free radicals [2, 5]. The World Health Organization recognised its public health significance early: it convened a study group in December 1966, published a variant classification scheme based on residual activity and enzyme biochemical properties in 1971, and consolidated it in a working group document in 1989 [3, 15].

The chronology of discovery helps to explain terminology that is still in circulation. In 1932, Otto Warburg and Walter Christian identified in yeast and in erythrocytes an enzyme that oxidised glucose-6-phosphate with NADP as an obligatory intermediate and named it Zwischenferment; only later was it recognised as corresponding to G6PD [5]. In 1957, Gennaro Sansone and Giuseppe Segni, in Genoa, found enzyme deficiency in every patient with a history of favism; in 1958, Arieh Szeinberg and colleagues, in Tel Aviv, showed that transmission was compatible with a gene on the X chromosome; in 1961, Franco Panizon, in Sassari, and Spyros Doxiadis, in Athens, identified the deficiency as a cause of severe neonatal jaundice [5]. In 1962, Ernest Beutler demonstrated that inactivation of one of the two X chromosomes creates mosaicism in the blood cells of heterozygous women, and in 1965 Lucio Luzzatto and Norman Allan, in Ibadan, showed that the A− variant is structurally distinct from the A variant and not merely a smaller amount of the same protein [5]. Philip Cohen and Michael Rosemeyer established in 1969 that the active form of the human enzyme is a dimer or a tetramer of identical subunits; Barbara Migeon mapped the gene to Xq28 in 1980; Maria Grazia Persico cloned the complementary DNA from 1981 onwards, and by 1986 the complete sequence and the genomic structure were available; and the three-dimensional structure of the human tetramer was solved by X-ray crystallography in 1999 [5].

This five-class scheme remained in use for approximately half a century but accumulated two practical problems recognised by the technical panel convened by the Global Malaria Programme in January 2022: for several variants, interindividual enzyme activity variability crossed the 10% threshold separating class II from class III, and both classes conferred susceptibility to the same manifestations, namely neonatal jaundice, favism and drug-induced acute haemolytic anaemia [3]. A further impetus came from tafenoquine authorisation in 2018 in the United States and Australia: its long half-life and inability to be stopped once administered made reliable pretreatment G6PD activity measurement mandatory [3, 16].

The revised classification, endorsed with minor changes by the WHO Malaria Policy Advisory Group in March 2022 and published in the Bulletin of the World Health Organization in 2024, combines former classes II and III into a single class B, retains class A for variants associated with chronic haemolysis, creates class C for variants without haemolysis and class U for those of uncertain clinical significance [3]. In parallel, the Clinical Pharmacogenetics Implementation Consortium published an expanded guideline in 2023, replacing its previous rasburicase-only version and systematically reassessing 48 medicines for acute haemolytic anaemia risk, concluding that many historical regulatory warnings lack support in peer-reviewed literature [2]. These two documents and the malaria and neonatal hyperbilirubinaemia guidelines provide the basis for practical management [6, 7].

2. Molecular Genetics and Pathophysiology

Glucose-6-phosphate dehydrogenase converts glucose-6-phosphate to 6-phosphogluconolactone in the first irreversible step of the pentose phosphate pathway, reducing nicotinamide adenine dinucleotide phosphate (NADP) to NADPH [2, 12]. The enzyme is ubiquitous, but its clinical importance centres on mature erythrocytes, which lack nuclei and mitochondria and depend on this pathway and 6-phosphogluconate dehydrogenase as their sole NADPH source [2]. NADPH donates electrons to glutathione reductase and thioredoxin reductase: reduced glutathione and peroxiredoxins constitute the central red cell antioxidant system, neutralising hydrogen peroxide and other reactive oxygen species [17].

The magnitude of the endogenous oxidative load explains why this dependence is critical. The haemoglobin concentration in the erythrocyte is not below 5 mM and reversible binding of oxygen to haem is imperfect: conformational fluctuations in the haem pocket allow the entry of water or of a small anion, with transfer of an electron from iron to oxygen, which generates methaemoglobin and superoxide radicals; this autoxidation process affects 2% to 3% of total haemoglobin per day [5]. Methaemoglobin is reduced by NADH-cytochrome b5 reductase 3, while superoxide is converted by superoxide dismutase into hydrogen peroxide, which the erythrocyte eliminates through three enzymatic mechanisms — catalase, glutathione peroxidase coupled to glutathione reductase, and peroxiredoxin 2 regenerated by thioredoxin reductase — all dependent on NADPH [5, 17]. The relative quantitative contribution of the three has not been established, but acatalasaemia does not run with haemolytic anaemia, whereas glutathione reductase deficiency is associated with favism, which gives particular weight to the glutathione pathway [5]. Because the mature erythrocyte does not synthesise proteins, G6PD activity declines with cell age: it is high in reticulocytes and about one tenth remains in the oldest erythrocytes, an amount that is still sufficient when the enzyme is normal [5].

Haemolysis therefore occurs indirectly. Implicated medicines are not enzyme substrates; deficiency deprives cells of the capacity to buffer the oxidative load generated by these medicines, fava beans or infection [2]. When glutathione is depleted, haemoglobin oxidises and precipitates as Heinz bodies, membranes are damaged and cells lyse; haem iron oxidation may generate methaemoglobin, so methaemoglobinaemia accompanies anaemia with some frequency [2, 17]. Saudubray and colleagues' textbook emphasises that, within this family of pentose phosphate pathway disorders, G6PD deficiency alone has exclusively haematological manifestations, unlike ribose-5-phosphate isomerase, transaldolase and transketolase deficiencies [12].

The sequence of events in a cell exposed to an oxidant is well characterised. The agent — primaquine, another risk medicine, or divicine and isouramil from fava beans — generates reactive oxygen species and hydrogen peroxide; the normal erythrocyte responds by accelerating NADPH recycling, the deficient erythrocyte cannot do so, and ferryl haemoglobin and hemichromes are formed, while lipids and the thiol groups of cytoplasmic and membrane proteins become oxidised [5]. Binding of hemichromes to the cytoskeleton gives rise to Heinz bodies, and aggregation of membrane proteins produces rigid hemighosts; reduced glutathione has a dual role here, as a substrate of glutathione peroxidase and as a regenerator of thiol groups previously oxidised to disulphides [5]. The most damaged cells undergo intravascular haemolysis, with haemoglobinuria, and free plasma haemoglobin binds nitric oxide, causing smooth muscle dystonia, which explains the abdominal pain; in other cells, aggregates of oxidised band 3 bind immunoglobulin G and complement fragment C3c, and opsonisation leads to erythrophagocytosis, that is, extravascular haemolysis, which explains the jaundice and the splenomegaly [5]. Because the gradient of enzyme activity with cell age also exists in deficient cells, the oldest erythrocytes are the first to be destroyed and reticulocytes are relatively resistant, which in a severe episode may correspond to destruction of up to two thirds of circulating erythrocytes [5].

G6PD spans approximately 18 kb at Xq28 and contains 13 exons and 12 introns, exon 1 being non-coding; the polypeptide chain has 514 amino acids and the enzymatically active forms are the dimer and the tetramer, since the monomer has no activity [5, 17]. More than 230 variants have been described at the molecular level and 1,341 variants have already been catalogued in population and clinical databases, of which 268 were interpreted using American College of Medical Genetics and Genomics criteria in a 2023 study, including 186 that were absent from ClinVar or listed there as of uncertain significance [3, 18]. The overwhelming majority are missense variants producing single amino acid substitutions; a few in-frame deletions also occur [2]. The absence of large deletions and of frameshift variants is consistent with the observation that complete abolition of enzyme activity is lethal in utero, demonstrated in gene-knockout mice [2].

The tally of the 230 variants with a known mutation is informative: 101 belong to the former class I, 99 to the former classes II and III, 5 to class IV and 25 remain without a defined class; 198 result from substitution of a single amino acid, 18 from two or three substitutions and 11 from in-frame deletions [5]. The only variant arising from a nonsense mutation described, G6PD Georgia, was found in a heterozygous woman, and there are no frameshift variants, which reinforces the inference that complete loss of activity is incompatible with life [5]. Some variants result from two mutations in cis: the best known example is the A− allele, in which substitution at codon 126 (p.Asn126Asp), innocuous on its own, coexists with one of three other substitutions — p.Val68Met, p.Arg227Leu or p.Leu323Pro — and only the combination produces the deficiency [5]. Transcription of the gene depends on a promoter region with seven GC boxes, two of which are essential because they bind the Sp1 and AP-2 factors, and expression requires a specific chromatin conformation [5].

Most low-activity variants impair protein stability; the most severe predominantly affect the dimer interface encoded by exon 10, affecting dimerisation and substrate binding [2]. Variants causing chronic non-spherocytic haemolytic anaemia cluster specifically in this exon: reported examples include G6PD Calvo Mackenna, G6PD Riley, G6PD Wisconsin, G6PD Tennessee, G6PD Veracruz and G6PD Yucatan [17]. The direct relationship between residual activity and substrate affinity provides a biochemical mechanism for the severity gradient between variants [19].

Two biochemical parameters separate the phenotypes in practice. Because the intraerythrocytic concentrations of glucose-6-phosphate and of NADP are well below saturation, a high Michaelis constant for glucose-6-phosphate is an adverse factor, whereas a low constant is advantageous in the steady state, although less so under oxidative stress: this is the explanation proposed for the Mediterranean variant, with very low residual activity, remaining in class B instead of causing chronic haemolysis [5]. Thermal instability is the most frequent determinant, and its comparison between pairs of variants affecting the same amino acid is illuminating: G6PD Plymouth (p.Gly163Asp, chronic haemolysis) differs from G6PD Mahidol (p.Gly163Ser, triggered acute haemolysis) only by its lower thermostability, and the same contrast exists between G6PD Santiago (p.Arg198Pro) and G6PD Nilgiri (p.Arg198His); in both cases, the substitution associated with the chronic form alters the charge of the residue and the other does not [5]. The catalytic constant is moderately reduced in many variants and profoundly reduced in G6PD Nilgiri and G6PD Santiago, and the residual activity of variants associated with chronic haemolysis is, as a rule, 1% of normal or less, at times undetectable, with partial overlap with the former class II [5]. From an evolutionary standpoint, deficient variants cluster at residues of intermediate conservation, on the assumption that mutations at highly conserved residues are lethal and that those at non-conserved residues do not cause deficiency [5].

One methodological consequence follows for variant interpretation: in the case of G6PD, pathogenicity prediction algorithms are of limited value, because demonstration of enzyme deficiency is the better test — any variant with reduced enzyme activity has the potential to cause clinical manifestations [5].

Nomenclature requires care because historical literature names variants after places and uses different transcript numbering. Normalisation performed for this article using VariantValidator on MANE Select transcript NM_001360016.2 confirms the following correspondences [1].

Numbering in NM_001360016.2 matches NM_001042351.3, the transcript used in most articles; NM_000402.4, however, has a different reading frame, and the same genomic coordinate c.202G>A corresponds in that transcript to p.(Gly68Ser), so mixing numbering systems causes interpretation errors [1]. The A− allele is actually a haplotype: it combines c.202G>A with polymorphism c.376A>G, which does not cause deficiency alone; complete linkage disequilibrium between them (D' = 1.0) has been documented in northern Brazilian populations [20].

The current variant classification, replacing that of 1985/1989, is shown below [3, 15].

Two footnotes in the original table have decisive clinical implications: a variant with activity below 20% belongs to class A only if associated with chronic haemolysis; association with triggered acute haemolysis places it in class B, and unknown manifestations in class U; class U is a temporary assignment [3]. To date, no identified variants have median activity in hemizygous males or homozygous females between 45% and 60%; any such variant should initially be placed in class U [3]. Class V, reserved for increased activity, was discontinued because it rested on a single case, G6PD Hektoen, whose causal allele was never identified [2, 3].

The new classification applies to genetic variants and not to individuals: at any given moment, a person may have activity above or below the median for their variant [3]. Class A differs from class B not by the level of activity but by coexisting chronic haemolysis [3]. Within class B, the severity of episodes is not uniform: episodes tend to be more frequent and more severe with variants of lower mean activity, such as G6PD Coimbra and G6PD Mediterranean, than with variants of higher mean activity, such as G6PD Seattle and G6PD Kalyan-Kerala [3].

Variants associated with chronic haemolysis, corresponding to the former class 1, have been described in every population and account for more than 70 identified causal mutations; the family history is inconstant, because a proportion of the mutations arise de novo, which means the hypothesis cannot be excluded simply because no other cases are known in the family [11]. In women carrying a variant of this group, enzyme assay is as a rule normal, because erythrocyte selection is strongly biased in favour of the non-deficient allele: cells that inactivate the normal allele have greatly reduced competitive survival and disappear from the circulation [11]. In some carriers, progressive weakening of this selection has been described, with the emergence of a haemolytic tendency after the age of 50, which justifies molecular study of the gene in the mothers of affected children and repetition of the assay in that age range [11].

Systematic quantification confirms this heterogeneity. In a literature review focused on polymorphic variants, with activity values from 2,291 males, mean residual activity estimates for 16 common variants ranged from 1.9% to 33%, and for most variants the majority of deficient males had activity below 30% of normal [19]. A systematic review using individual data from 4,320 people in 17 studies across 10 countries, including 1,738 carriers of one of 24 confirmed variants, found medians of 29.0% for A− (range 1.7%–76.6%), 16.9% for Mediterranean (3.3%–21.3%), 10.2% for Mahidol, 9.0% for Vanua Lava and 7.5% for Viangchan in hemizygous males and homozygous females [21].

Random X-chromosome inactivation determines the phenotype in heterozygous females. Because inactivation occurs very early in embryogenesis, deficient erythrocyte proportions range from nearly 0% to nearly 100%, potentially outweighing differences between variants: a G6PD Seattle heterozygote with 80% deficient erythrocytes may experience a more severe episode than a G6PD Coimbra heterozygote with 20% [3]. Consequently, genotype alone cannot assign a phenotype to people with more than one X chromosome, including women and people with Klinefelter syndrome [2].

For this reason, the designation “X-linked recessive”, still found in textbooks, is incorrect: the deficiency is frequently expressed in heterozygous women, both biochemically and clinically [5]. In any population, heterozygous women are more frequent than deficient hemizygous males, and homozygous women are much rarer; from this follows a recurrent epidemiological pitfall, since surveys reporting the frequency of the deficiency in a sample with both sexes pooled underestimate the allele frequency, because only a proportion of heterozygous women is classified as deficient — the frequency of the deficient allele is given accurately by the frequency of deficient males [5]. Stating that the deficiency “is more frequent” or “more expressed” in males is also incorrect: what is observed is that expression is consistent in hemizygous males and homozygous females and highly variable in heterozygous females [5].

The geographical distribution of deficient variants closely tracks current and historical malaria endemicity, and G6PD deficiency is the classical example of a human X-linked polymorphism maintained by protection of heterozygous females against malaria mortality [2, 5]. This selection explains frequencies of 30% or more in certain populations in endemic regions [2].

3. Clinical Spectrum and Manifestations by System

Most people with G6PD deficiency are asymptomatic between episodes, and many never discover their deficiency [2, 3]. There are three distinct clinical presentations with partly different pathophysiology: neonatal jaundice, triggered acute haemolytic anaemia, and chronic non-spherocytic haemolytic anaemia [3].

Acute haemolytic anaemia is the characteristic manifestation. It develops hours to days after oxidant exposure, with progressive pallor, jaundice and dark urine, accompanied by non-specific features such as abdominal pain, hepatosplenomegaly, tachypnoea, tachycardia and hypotension; haemoglobinuria should prompt consideration of acute kidney injury [22]. Laboratory findings include normocytic anaemia with reticulocytosis and leukocytosis with reactive neutrophilia, unconjugated hyperbilirubinaemia and moderately elevated transaminases, with negative direct and indirect Coombs tests; blood films may show red cell fragments [22]. Heinz bodies, formed from denatured haemoglobin, are the morphological correlate [2, 17].

The typical picture of a favism crisis adds details that are useful at the bedside. The patient is usually a boy between 2 and 10 years of age, pale, jaundiced and prostrate, with abdominal pain and at times fever, a spleen that may be palpable and dark urine described as “red wine” or “cola”; the blood film shows moderate to very severe anaemia with marked morphological changes, and supravital staining with methyl violet reveals Heinz bodies [5]. There is often neutrophilic leucocytosis with a left shift, with platelets usually normal, raised unconjugated bilirubin and lactate dehydrogenase, and low or undetectable haptoglobin [5]. Girls and adults are not excluded from this picture [5].

Haemolysis related to the deficiency is characteristically dose-dependent, for fava beans as for medicines, and the cases that reach the emergency department are the visible part of the phenomenon: when the anaemia is mild, the patient may never be assessed in hospital and may remain undiagnosed, and in many cases there is no anaemia at all, only compensated haemolysis [5]. When the trigger is a medicine, the clinical picture is superimposable on that of favism, with the difference that exposed patients are often already anaemic from their underlying disease, which makes comorbidity decisive in assessing severity [5].

The chronology of the episode has immediate practical value. Haemolysis sets in between 24 hours and 3 days after the oxidant agent is taken, with appreciable variability even within the same person; the anaemia continues to worsen until 7 to 8 days after withdrawal of the agent and begins to recover from the eighth to tenth day, so that a stable haemoglobin value in the first hours does not permit discharge without reassessment [11]. Haemolysis triggered by fava beans is accompanied by earlier and more severe signs than that triggered by medicines, and the anaemia may be very acute in children, requiring urgent transfusion; haemolysis occurring during an infection is as a rule of more moderate intensity [11]. The most feared complication is acute kidney injury, above all in adults, usually reversible but at times requiring a short period of renal replacement therapy [11]. Ketoacidotic decompensation of diabetes and acute myocardial infarction have been described as triggering circumstances, with the caveat that in those series there was frequently concomitant drug exposure or infection, which precludes attributing the episode to the metabolic or ischaemic event alone [11].

The spectrum of infections documented as triggers is wide but erratic. Bacterial infections must be severe to precipitate haemolysis, and episodes have been described in association with pneumonia, brucellosis, rickettsiosis, maxillary abscesses due to streptococci or staphylococci and even Clostridioides difficile infection; there are likewise reports during hepatitis A, B and E, cytomegalovirus infection and one case of dengue [5]. In viral hepatitis, hyperbilirubinaemia may be excessive even with little or no evidence of haemolysis, and acute renal failure is a possible complication [5]. When haemolysis regresses with control of the infection, it is reasonably well established that the infection was the trigger — a distinction that matters, because episodes attributed to the antibiotic were due to the infection being treated, and the converse also occurs [5].

The course, unlike that of autoimmune haemolytic anaemia, is typically one of rapid recovery, and the most striking feature is the speed of the reticulocyte response, which appears in less than 48 hours, too quickly to be explained by the hypoxia-erythropoietin axis alone and attributable to accelerated marrow release of reticulocytes already formed [5].

Methaemoglobinaemia may accompany anaemia because haem iron oxidation proceeds unchecked and NADPH scarcity also impairs methaemoglobin reduction [2, 22]. This association has immediate therapeutic implications because usual methaemoglobinaemia treatment, methylene blue, is a high-risk medicine in this setting [2, 22].

Episode severity depends on several simultaneous factors: principally the drug dose, but also baseline haemoglobin, the variant involved and other individual factors affecting erythrocyte physiology [2, 3]. Favism can be fatal, particularly in children [2]. Alongside fava beans and medicines, infection is a recognised trigger, and when a medicine is administered for infection it may be difficult to determine which caused haemolysis [2]. Hyperuricaemia and sepsis, which generate reactive oxygen species, also confer risk [2]. Episodes triggered by household insecticidal naphthalene exposure through contact or inhalation have been reported, including severe jaundice and kernicterus in a 4-day-old infant [17].

Neonatal jaundice is the second commonest presentation and carries the greatest public health burden. The 2022 US hyperbilirubinaemia guideline for newborns of at least 35 weeks describes G6PD deficiency as one of the leading causes of hazardous hyperbilirubinaemia resulting in kernicterus in the United States and worldwide, and lists it as a neurotoxicity risk factor alongside gestational age below 38 weeks, albumin below 3.0 g/dL, isoimmune haemolytic disease, sepsis and significant clinical instability during the preceding 24 hours [6]. Affected neonates more frequently receive phototherapy before discharge, probably because of increased bilirubin production and reduced conjugation, and have greater readmission and retreatment risks [6]. A clinically troubling feature is that total serum bilirubin may rise suddenly and extremely, making prediction or prevention difficult, and even after an apparently haemolytic event there may be little or no laboratory evidence of haemolysis [6].

The pathophysiology of neonatal jaundice is not superimposable on that of favism: the 2022 review on the subject proposes that, rather than membrane destruction with Heinz bodies, what predominates is a disturbance of the oxidant-antioxidant balance with impaired recycling of peroxiredoxin 2, affecting bilirubin clearance [17]. Coexisting Gilbert syndrome increases the risk: prolonged neonatal jaundice occurs in neonates heterozygous or homozygous for the Gilbert-type promoter insertion who additionally have G6PD deficiency, Coombs-negative ABO disease, spherocytosis or prolonged breastfeeding [23]. The US guideline refers to the UGT1A128* allele along the same line of reasoning, through the CPIC review [2]. The 2020 review adds that, in studies from different regions of the world, including Africa, the frequency of neonatal jaundice was consistently higher in deficient neonates than in normal controls, that coexistence of a UGT1A1 promoter variant raises that frequency further, and that the peak incidence of hyperbilirubinaemia occurs on the second and third days of life [5]. This timing has a direct organisational consequence: mother and neonate may already have been discharged by then, which leads to delay in or failure of a diagnosis that calls for urgent phototherapy or exchange transfusion [5].

The French national diagnosis and care protocol (PNDS) describes precisely the departure from physiological jaundice and quantifies the risk. Physiological jaundice, due to unconjugated bilirubin, appears between the second and third day, progresses until the fifth to seventh and resolves before the fifteenth; in the deficient neonate, induced haemolysis can make bilirubin accumulate above the reference values for postnatal age in hours within hours [11]. The risk of neurological complications is described above 340 µmol/L of circulating bilirubin, but three quarters of published cases of kernicterus followed values above 520 µmol/L [11]. G6PD deficiency is the second commonest cause of severe neonatal hyperbilirubinaemia worldwide and accounts for a quarter to a third of the aetiologies identified in cohorts of severe hyperbilirubinaemia or of indication for exchange transfusion [11].

Two patterns are recognised as indicative of the deficiency, both confined to the first two weeks of life: an acute, severe hyperbilirubinaemia, corresponding to haemolysis between the fourth and tenth day, with very high values and neurotoxic risk; and an early hyperbilirubinaemia, in the first 24 to 36 hours, prolonged through the first week, without immunological cause and without overt haemolysis initially [11]. Unlike later in life, neonatal haemolysis is not generally triggered by an oxidant exposure, but the history should always look for one, in particular through breast milk [11]. Enzyme assay is requested in the face of any severe or persistent non-immunological jaundice, that is, with negative direct antiglobulin test and elution, in neonates of either sex whose parents come from areas of high prevalence; reference values for activity at birth are higher than after 6 months of age, which conditions interpretation [11]. The main differential diagnosis, among severe hyperbilirubinaemias due to unconjugated bilirubin, is ABO fetomaternal incompatibility, usually of earlier onset, between the first and third day, followed by the other constitutional haemolyses, such as erythrocyte membrane disorders and pyruvate kinase deficiency [11].

Chronic non-spherocytic haemolytic anaemia defines the most severe phenotype and is rare, with an estimated frequency below 10 per million [2]. These patients have haemolysis without any trigger, usually with enzyme activity below 10% of normal, and are at risk of neonatal jaundice with possible kernicterus, gallstones, reticulocytosis and splenomegaly, with increased susceptibility to oxidants [2, 4].

The picture approaches that of hereditary spherocytosis, with jaundice, gallstones and highly variable severity, from mild forms diagnosed incidentally in adulthood to forms dependent on regular transfusion in a minority; the difference of greatest semiological value is that a history of neonatal jaundice, often severe, is the rule in this form and not in spherocytosis [5]. Any agent capable of triggering acute haemolysis in a deficient person provokes in these patients an acute haemolysis superimposed on the chronic one [5]. In transfusion-dependent patients, splenectomy has proved highly effective: in the documented case of a patient with the class A variant G6PD Harilaou (p.Phe216Leu), who presented with severe neonatal jaundice requiring exchange transfusion and continued to have moderate to severe anaemia with frequent transfusions, splenectomy performed at 9 years of age rendered him transfusion-independent, with resolution of the persistent reticulocytosis [5].

The clinical heterogeneity of this form also depends on non-erythrocyte factors. A heterozygous woman with a de novo mutation and skewed X-inactivation was described with the G6PD Volendam variant (p.Pro172Ser), with neonatal jaundice and mild anaemia with exacerbations associated with infections, which illustrates that the chronic form requires neither hemizygosity nor homozygosity [5].

Outside the erythrocyte, the enzyme takes part in the neutrophil oxidative burst, and G6PD deficiency belongs in the differential diagnosis of chronic granulomatous disease precisely because it impairs this function and increases susceptibility to infection [24]. Extra-erythrocytic manifestations are, taken together, not prominent, probably because the molecular basis of the deficiency is as a rule protein instability and nucleated cells can compensate for it by increasing enzyme synthesis [5]. Granulocyte impairment is best characterised in patients with chronic haemolysis and very low residual activity, such as carriers of G6PD Barcelona, who suffered severe bacterial infections through inability to generate enough NADPH for an effective oxidative burst — a resemblance that in the past led to G6PD deficiency being confused with X-linked chronic granulomatous disease [5]. In the same direction, a higher frequency of severe sepsis after major trauma was described in deficient patients, an observation made in carriers of the common A− variant [5]. Isolated cases of acute rhabdomyolysis with myoglobinuria and, at times, acute renal failure have been published, a very rare occurrence when compared with acute haemolytic anaemia and usually, though not always, associated with it [5]. The deficiency has also been linked to cataract in the context of disturbances of lens redox status, although the 2020 review considers that the controversy on this point has been settled in the negative, that is, that the deficiency does not favour cataract formation [5, 25]. As to oncology and cardiovascular risk, one study found a lower frequency of adenomatous polyps and colon cancer in deficient people, a result that the authors of the review consider to require confirmation before any mechanistic interpretation, and a study in US military personnel described a slightly higher risk of cardiovascular disease in deficient men than in controls (odds ratio 1.39; 95% confidence interval 1.04-1.87) [5]. Numerous studies have investigated associations between G6PD activity and diverse diseases, but critical analysis of these associations was explicitly left outside the scope of the pharmacogenetic guideline, which warrants caution before converting them into clinical recommendations [2].

One laboratory consequence with an impact beyond haematology deserves note: in G6PD deficiency, glycated haemoglobin is underestimated, through reduction of the pool of older erythrocytes, so that its value in the diagnosis and monitoring of diabetes is compromised and should be set against blood glucose and, where available, other indicators of metabolic control [11].

4. Clinical Recognition and Warning Signs

Recognition centres on three situations: acute haemolytic episodes, jaundiced neonates and imminent prescription of a high-risk medicine.

With acute anaemia, jaundice, dark urine and negative Coombs tests, particularly after fava bean ingestion, fever, intercurrent infection or recent medicine introduction, G6PD deficiency should be considered immediately [2, 22]. Haemolysis associated with methaemoglobinaemia, presenting clinically as cyanosis and low peripheral oxygen saturation poorly responsive to oxygen therapy, should strongly suggest this condition and influences treatment selection [22]. Family history is frequently negative, and parental geographical origin—sub-Saharan Africa, Middle East, Mediterranean, Arabian Peninsula and Southeast Asia—is relevant history [6, 23].

In neonates, the US guideline identifies severe hyperbilirubinaemia and atypical trajectories, such as elevated bilirubin in a formula-fed infant or late-onset jaundice, as warning signs [6]. Listed hyperbilirubinaemia risk factors, beyond genetic ancestry suggesting inherited red cell disease, include bilirubin rise exceeding 0.3 mg/dL per hour in the first 24 hours or 0.2 mg/dL per hour thereafter, phototherapy before discharge, a sibling or parent requiring phototherapy or exchange transfusion, exclusive breastfeeding with suboptimal intake, scalp haematoma or significant bruising, Down syndrome and a macrosomic newborn of a diabetic mother [6].

The French protocol adds two operational points to recognition in the neonatal period. The first is the inadequacy of visual assessment: the severity of hyperbilirubinaemia is easily underestimated by observing the intensity of the jaundice, above all in dark skin, so that quantification is always necessary, by blood assay or by transcutaneous reflectometry, with the reading compared against reference values for postnatal age in hours; the clinical factors that make such quantification mandatory are jaundice before 24 hours of life, gestational age below 38 weeks, African, Caribbean or Asian family origin and ineffective breastfeeding [11]. The second is the set of signs that turn the situation into a therapeutic emergency and that should be taught to parents and to the professionals providing follow-up: a neonate who is inconsolable or, conversely, drowsy, high-pitched crying and hyperextension of the neck with the head thrown backwards [11]. Because the deficiency becomes symptomatic above all on the fourth and fifth day, that is, after discharge from the maternity unit, the protocol makes recognition depend on written transmission of the information about the jaundice in the child health record, on coordination between the maternity unit and the structures supporting the return home, and on postponement of discharge when surveillance cannot be ensured in early, severe jaundice [11].

At prescription, the warning sign is the medicine itself. Rasburicase, indicated for the initial management of hyperuricaemia in patients with leukaemia, lymphoma or solid tumours receiving anticancer treatment with a risk of tumour lysis, carries a boxed warning and a formal contraindication in G6PD deficiency, with an explicit instruction to screen beforehand those patients at increased risk of deficiency, citing African or Mediterranean ancestry as examples [26]. The 2020 review goes further on this point and recommends that the test always be performed before rasburicase is given, on the grounds that restricting screening to ethnic groups in which the deficiency is common is unsafe: ethnic group is an imprecise concept and ancestry may be mixed or only partly known [5]. This is a more demanding recommendation than the wording of the US label, which limits prior screening to patients at increased risk [5, 26]. Tafenoquine requires testing before prescription, without exception, in both US products [16, 27].

The converse is also clinically significant: avoidance lists given to patients and pharmacy services traditionally include, through extrapolation across drug classes, medicines whose risk lacks supporting evidence; a comprehensive review of 30 medicines listed as contraindicated in manuals and other sources found supporting literature for only seven [2, 28]. Indiscriminate lists can lead to refusal of useful, safe therapies [2].

This classification by evidence is not the only one in use, and the difference has practical consequences for those practising in Europe. The French protocol refers to the “medicines and G6PD deficiency” list of the French national medicines agency, published in 2008 and revised in 2014, built from the literature and from pharmacovigilance data and organised into three levels of precaution: contraindicated medicines, medicines to be avoided, and medicines whose maximum dose must be respected; it recommends that the patient take the list to every consultation and stresses that no test is able to establish, in vivo or in vitro, the haemolytic character of a molecule [11]. The medicines formally contraindicated in that list are nalidixic acid, methylene blue by injection or for surgical use, dapsone, metamizole sodium, nitrofurantoin, primaquine, rasburicase, oral sulfadiazine, sulfafurazole, sulfaguanidine, sulfamethoxazole and trimethoprim in the co-trimoxazole combination, and sulfasalazine [11].

The divergence from the 2023 pharmacogenetic guideline is explicit and concerns several of these medicines: nalidixic acid, sulfamethoxazole, sulfadiazine, sulfisoxazole and sulfasalazine are classified as of low to no risk by the systematic review of primary evidence, and nitrofurantoin as of medium risk with an optional recommendation, whereas the French list keeps them as contraindicated [2, 11]. The resolution is not symmetrical: the 2023 guideline is the more recent document, was based on an explicit reappraisal of the peer-reviewed literature on 48 medicines and distinguishes G6PD status-dependent drug-induced anaemia from drug-induced anaemia independent of that status, and so is the reference for therapeutic decisions; the French list, more restrictive, also rests on national pharmacovigilance and remains the enforceable document in France, with mandatory reporting of haemolytic accidents to the regional pharmacovigilance centre, even when the medicine is already on the list [2, 11]. In Portugal and in Brazil there is no equivalent national authority list identified in the sources consulted, so the recommended practice is to document G6PD status, follow the risk classification of the 2023 guideline and reserve the French list for recognising labels and national warnings that reflect it.

Two particular cases have their own recommendation in the French protocol. The lidocaine and prilocaine combination as a plaster or cream should not be used in infants aged 0 to 3 months with known or suspected deficiency, and above 3 months there is an increased risk of methaemoglobinaemia with repeated, closely spaced applications, to be weighed before each prescription [11]. As to vitamin K1, there are no pharmacovigilance or experimental data to justify contraindicating it in the prophylaxis of haemorrhagic disease of the newborn or in the prophylaxis of deficiency in adults and children, at the recommended doses, the more so as there is no therapeutic alternative; rare cases of haemolysis have been described with rapid administration at curative doses, without formal incrimination of the medicine, so it is recommended that the doses and modes of administration be respected, by slow intravenous or intramuscular injection, and that any haemolytic risk be monitored [11]. This position agrees with the classification of vitamin K as of low to no risk by the pharmacogenetic guideline, adding to it the precaution about the rate of administration [2, 11].

5. Diagnosis: Criteria, Tests and Algorithms

Diagnosis is biochemical in most settings and genetic in a growing minority. Most diagnoses use enzyme activity rather than genotype for three reasons: unpredictable phenotype in heterozygous females, slow uptake and turnaround of genetic testing, and limited allele coverage in some assays [2].

The reference method is ultraviolet spectrophotometry of G6PD activity in whole blood, usually expressed in units per gram of haemoglobin and converted to a percentage of normal [29]. An individual-participant meta-analysis of 15,811 participants aged 0 to 86 years found median activity in normal control samples of 10.0 U/g Hb using the Trinity assay (range 6.3–14.0) and 8.3 U/g Hb using the Randox assay (6.8–15.6); the adjusted male median—the standardised 100% activity metric—ranged from 5.7 to 12.6 U/g Hb across the 13 studies that used the Trinity assay, and mean coefficients of variation for replicate measurements ranged from 1.6% to 14.9% between laboratories [29]. The authors derived a universal 100% activity threshold of 9.4 U/g Hb, yielding diagnostic thresholds of 6.6 U/g Hb for 70% and 2.8 U/g Hb for 30% activity [29]. The practical implication is that an absolute value is uninterpretable without knowing the assay and the local reference value. The 2020 review states the same idea in operational terms: in most laboratories the normal range of activity lies between 7 and 11 units per gram of haemoglobin, reference ranges vary between laboratories and some of those published are too wide, so that a result below 80% of the lower limit of normal should be considered deficient [5].

The timing of sampling determines the value of the result. Measurement during or shortly after a haemolytic episode gives a falsely normal result, because the older deficient erythrocytes have already lysed and the surviving reticulocytes and young erythrocytes have a higher enzyme content; a recent transfusion produces the same effect, as the transfused cells are normal [2]. The American hyperbilirubinaemia guideline states, on the same logic, that if suspicion is strong and the measured activity is normal or near normal, measurement should be repeated at least three months later [6]. Other limitations apply: newborns have higher activity than older children and adults; activity correlates positively with leucocyte and platelet counts, and there are no standardised corrections for haematological abnormalities, so that measurements are particularly unreliable in patients with anaemia or acute leukaemia—precisely those in whom reliable assignment of G6PD status matters most before rasburicase [2]. Enzyme activity alone also does not distinguish deficiency with chronic non-spherocytic haemolytic anaemia from deficiency without it [2]. When measurement made during or immediately after the crisis falls within the normal range, the alternatives are to repeat the assay some weeks later or to search directly by DNA analysis for a mutation known in the patient's population [5].

Collection and transport conditions are part of the result. The assay requires blood in anticoagulant, preferably acid citrate dextrose or, alternatively, citrate; because most deficient variants correspond to unstable proteins, the tube should go immediately to a competent laboratory and, if the interval before assay exceeds 24 hours, the sample should be refrigerated and transported between +4 °C and +6 °C [11]. The minimum interval after a transfusion is quantified: at least three months [11].

There are two ways of unmasking the two commonest biases in interpretation. For the reticulocyte bias, the laboratory is asked to measure, in the same sample, another erythrocyte enzyme activity—hexokinase or pyruvate kinase—which acts as an internal control of the mean age of the erythrocytes studied: apparently normal G6PD activity accompanied by hexokinase activity 2 to 3 times above normal, in a reticulocyte-rich sample, points to deficiency [11]. For the opposite bias, it is important to know that a reduced haemoglobin content per erythrocyte, that is, the hypochromia of thalassaemia or of iron deficiency, artificially increases the measured activity and may mask the deficiency, especially in a heterozygous woman [11]. The reticulocyte count should therefore always accompany the request.

As to repetition, the French protocol is more restrictive than the general rule of repeating after three months: a diagnosis made under good conditions does not need confirmation, except when the first assay was performed in the neonatal period or in the first months of life, in which case a check after 9 to 12 months of age is proposed, when the reference values are already those of the older child [11]. Semi-quantitative cellular methods, based on detection of NADPH or on formazan reduction, serve mass screening, but perform less well than the reference assay in the diagnosis of heterozygous women [11].

Between episodes, in individuals deficient for class B variants, the usual markers of haemolysis—haptoglobin, reticulocytes, lactate dehydrogenase and total and indirect bilirubin—are normal, and so is the full blood count; chronic regenerative anaemia appears only with the rare class A variants [11]. During the crisis, besides anaemia with reticulocytosis, the blood film may suggest the diagnosis through the presence of Heinz bodies, stained with cresyl blue or methyl violet, and through suggestive cytological changes such as bite cells and ghost cells [11].

Transfusion poses an additional problem in regions of high prevalence, where donor blood may be deficient. Blood from deficient donors can generally be used safely and universal donor screening is not feasible, but blood with normal G6PD is preferable in three situations: transfusion of a deficient patient in acute haemolytic anaemia, exchange transfusion for neonatal jaundice, and patients with regular transfusion requirements [5].

The classical semiquantitative screening method is the fluorescent spot test, used in neonatal programmes because of low cost but missing a significant proportion of moderate deficiency and partially deficient heterozygous females [17]. In malaria settings, quantitative and semiquantitative point-of-care devices are available, including biosensors integrating enzyme activity and haemoglobin concentration measurements [14, 17].

The lineage of these tests explains what each measures and why some persist. The methaemoglobin reduction test, described by Brewer, rests on the fact that the return of methaemoglobin to haemoglobin in intact erythrocytes, after addition of sodium nitrite, depends on intracellular G6PD activity: normal samples revert to red and deficient ones remain brownish [5]. The brilliant cresyl blue test, of Motulsky and Campbell-Kraut, uses the decolourisation time of the dye in haemolysates as an approximate measure of activity [5]. Both allowed large numbers of samples to be processed and formed the basis of the first population surveys; from them, in 1965 and 1968, two cytochemical tests were developed that allow normal and deficient erythrocytes to be counted separately on the same film, which makes them particularly suited to the study of heterozygous women, technically easier in the version using tetrazolium dyes [5]. The fluorescent spot test, developed in Beutler's laboratory between 1966 and 1968, requires only 10 microlitres of blood and a filter paper with a reagent mixture, read under an ultraviolet lamp after 15 minutes, and has retained its popularity for its reliability, low cost and simplicity [5]. The impetus for quantitative point-of-care tests came from the return of primaquine and the introduction of tafenoquine, whose label makes testing mandatory, and some of the commercial devices allow identification not only of hemizygous males and homozygous females but also of heterozygous women at risk [5].

Opting for genetic testing has its own advantages and limits: it can be performed on stored DNA and avoids the misclassification caused by reticulocytosis, but it requires knowledge of the mutations prevalent in the patient's population, may fail to detect an unusual or novel allele and, in heterozygous women, does not predict the degree of deficiency, so that in them phenotypic testing is superior to genotyping [5].

The indications for molecular study are delimited in operational terms in the French protocol. Genotyping is indicated in all cases of chronic haemolytic anaemia, that is, in class A variants, where it confirms the diagnosis, establishes carrier status in female relatives—whose assay is usually normal—and is a prerequisite for any prenatal diagnosis [11]. Outside that group, it is justified when the assay does not allow a conclusion: a patient transfused less than three months earlier, a markedly reticulocytic patient, or the need to confirm heterozygosity in a woman [11]. Routine molecular distinction between the former classes 2 and 3 is considered of little relevance, because the preventive advice is identical and because severe events have been described with the A− variant; an exception is made when a drug that is advised against is formally indicated and the potential severity of the haemolytic risk needs to be better estimated [11]. In patients with another cause of chronic haemolysis, such as sickle cell disease, and particularly in cohort follow-up, the study may be more systematic, because co-inheritance of a deficient trait may modify the clinical course [11]. The protocol also warns prescribers that, in out-of-hospital settings, the cost of genotyping variants of the former classes 2 and 3 may be charged to the patient [11].

The set of investigations for initial assessment is likewise specified. In the symptomatic patient, with haemolysis or jaundice, the request comprises full blood count with reticulocyte count, examination of the blood film, total and indirect bilirubin, renal function, G6PD assay under the conditions described, blood group with Rh and Kell phenotype and screening for irregular antibodies; extended red cell phenotyping for the Jk, Fy and MNS antigens is requested at a second stage in the rare cases with chronic anaemia [11]. Serology for human immunodeficiency virus and hepatitis B and C viruses forms part of the initial assessment, although it is no longer systematically indicated in transfusion follow-up [11]. In class A variants, a baseline abdominal ultrasound scan and parvovirus B19 serology are added [11]. In the asymptomatic individual, whose assay is requested as part of family screening or before administration of a contraindicated medicine, full blood count, reticulocytes and G6PD assay are sufficient [11].

Genotype interpretation follows the pharmacogenetic guideline's predicted phenotype table, reproduced in condensed form below [2].

Genotype can establish deficiency in hemizygous, homozygous or compound heterozygous individuals, but not heterozygous females, in whom random X-inactivation creates a mosaic of normal and deficient erythrocytes varying between individuals and over time, so overall activity may cover the entire normal-to-deficient range and drug-induced haemolysis may be indistinguishable from that of a hemizygous male [2]. The number of alleles reported by a pharmacogenetic test indicates X chromosome number: a two-allele result in a person identifying as male may reveal Klinefelter syndrome, occurring in approximately 1 in 600 people assigned male at birth; services routinely testing G6PD genotype should plan genetic counselling involvement and diagnostic confirmation [2].

Genetic test limitations should be included in reports. Several commercial assays screen only some common alleles, so patients may have rare, different or novel alleles and a “negative” result does not exclude deficiency; ancestry helps assess whether the panel captures relevant alleles [2]. Reports should specify gene, versioned transcript (NM_001360016.2 as MANE Select), HGVS nomenclature, zygosity, American College of Medical Genetics and Genomics classification and method limitations [1, 18]. In a 2023 cataloguing study, 186 of 268 interpreted variants were absent from ClinVar or listed as uncertain, illustrating the classification problem's scale [18]. A variant of uncertain significance does not become pathogenic because a compatible clinical response occurs. Recent data indicate reliable genetic results, especially in people with one X chromosome, and greater information from combining both approaches than either alone [2, 30].

In a study sequencing G6PD in 2,139 unselected, ethnically diverse US children with leukaemia, 85% of 48 class II–III deficient alleles were A− or Asahi, both class III [2, 31]. This supports targeted panels in specific populations while illustrating their inadequacy in others.

For malaria, the WHO defines operational thresholds and test types. Version 10.0 of its malaria guidelines, published on 10 September 2026, includes a good practice statement requiring G6PD testing to guide Plasmodium vivax relapse prevention and strong, moderate-certainty recommendations on qualitative and semiquantitative point-of-care tests [7]. In males and females, activity below 30% is deficient; qualitative tests are considered highly accurate in distinguishing activity above or below this threshold but do not identify heterozygous females with intermediate 30% to 70% activity, who are classified as normal [7]. Semiquantitative tests with fixed 30% and 70% thresholds are considered highly accurate at 30% and accurate at 70% for distinguishing deficient and intermediate individuals [7]. Women with intermediate activity also risk haemolysis, albeit less [7].

Tafenoquine safety has been established only for enzyme activity of at least 70% of normal, the drug trial inclusion criterion [2, 32].

Universal neonatal screening using fluorescent spot or quantitative testing is established or proposed in high-incidence areas such as Asia, Africa and the Middle East [2]. The WHO recommends G6PD activity screening in all neonates in high-prevalence countries [17]. In Malaysia, for example, neonatal screening has been mandatory since 1986, although the recommended fluorescent spot test misses deficient heterozygous females and individuals undergoing acute haemolysis [17]. The 2022 US guideline does not institute universal screening: key action statement 14 requires G6PD activity measurement in any neonate with jaundice of unknown cause whose total serum bilirubin rises despite intensive phototherapy, rises suddenly, rises again after initial decline, or requires escalation of care [6]. Identifying G6PD deficiency, hereditary spherocytosis or other membrane defects helps recognise infants at risk of recurrent haemolysis and informs families about future pregnancy risk [6].

The introduction of systematic newborn screening for the deficiency remains a matter of debate in many countries and, where it exists, it is organised mainly at regional or local level [11]. France has no systematic screening, either of newborns, blood donors or military personnel; when the deficiency is known, it is a contraindication to blood donation [11]. Outside the neonatal context, targeted screening continues, with assay offered to the siblings and maternal uncles of the male index case [11].

6. Differential Diagnosis

Acute haemolytic anaemia with a negative Coombs test in a previously healthy patient suggests a limited set of diagnoses, which sequential investigations can distinguish promptly.

A key conceptual point is that acute haemolytic anaemia during drug treatment does not itself prove G6PD deficiency: the pharmacogenetic guideline was specifically developed to separate G6PD-dependent from G6PD-independent drug-induced anaemia, noting that most literature linking medicines to haemolysis relied on older case reports without non-deficient comparators and sometimes insufficient investigation of other anaemia causes [2]. Conversely, normal enzyme activity measured during an episode does not exclude the diagnosis [2, 6].

Acquired reduction of enzyme activity and a coexisting second condition should also be considered, as in the published case of chronic parvovirus infection with G6PD deficiency mimicking Diamond–Blackfan anaemia [38].

7. Treatments and Risk Management Strategies

No treatment corrects the enzyme deficiency. The clinical value of knowing a person's G6PD status lies entirely in avoidance and anticipation [2].

7.1. Treatment of acute episodes

Management of the crisis is supportive: remove the trigger, hydrate, monitor haemoglobin and renal function, and transfuse when necessary [22]. The grading is simple: if the anaemia is severe, transfusion may be urgent; if it is not, fluid support and analgesia are generally sufficient [5]. As most patients were previously asymptomatic, it is the episode itself that prompts the request for enzyme assay, which should be quantitative [5]. Associated methaemoglobinaemia requires particular care, because methylene blue, the usual treatment in that situation, is classified as a high-risk drug in G6PD deficiency and may aggravate the haemolysis; the first measure is always to remove the triggering agent [2, 22]. In neonatal jaundice, phototherapy follows the thresholds of the American guideline, which are higher than those of 2004 because the committee concluded that bilirubin neurotoxicity occurs only at concentrations well above the previous exchange transfusion thresholds, with distinct thresholds for newborns who have risk factors for neurotoxicity, among them G6PD deficiency [6]. Exchange transfusion remains for severe cases [6].

The minimum assessment in the emergency department is specified: full blood count, reticulocyte count, blood group, screening for irregular antibodies and serum creatinine, with red cell concentrate transfusion decided by clinical tolerance and the degree of anaemia, bearing in mind that haemoglobinuria indicates that haemolysis is still active [11]. The patient and carers should recognise the signs that require immediate review—port wine-coloured urine, pallor, asthenia and breathlessness—and should flag the deficiency at every emergency admission, whatever the reason, presenting the care card; the same notification is due at any anaesthetic consultation, particularly when surgery carries a risk of infection or hypoxia [11]. Specific anaesthetic recommendations for G6PD deficiency exist, published by the orphananesthesia project in March 2015 and cited by the French protocol [11].

7.2. Medicines classified by risk

The 2023 Clinical Pharmacogenetics Implementation Consortium guideline classified 48 medicines into three categories, considering primary evidence strength, usage frequency, regulatory warnings and plausible mechanisms generating reactive oxygen species [2].

Medicines without published articles linking them to increased risk received no recommendation, corresponding to methodology level C and including aspirin above 1 g/day, chlorpropamide, dabrafenib, gliclazide, glimepiride, glipizide, mepacrine, mesalazine, moxifloxacin, nicorandil, nitrofural, probenecid, sodium nitrite, sulfacetamide, tolazamide and trametinib, among others [2].

Recommendations depend on phenotype. Normal-phenotype patients need not avoid any category based on G6PD status; deficient patients should avoid high-risk medicines and use medium-risk medicines cautiously; patients with deficiency and chronic non-spherocytic haemolytic anaemia should avoid both high- and medium-risk medicines and use all others cautiously with monitoring, given the high acute exacerbation risk and lack of data; variable-phenotype patients, corresponding to heterozygous females, require enzyme activity measurement and subsequent recommendations according to the measured phenotype [2]. For deficient patients, closer monitoring is justified with higher doses or longer regimens, infection or another oxidative state, and combinations of several medium- and low-risk medicines [2]. There is no reason for recommendations to differ between children and adults [2].

Two entries deserve note because they do not appear in the usual pharmacological lists. The 2020 review includes the cosmetic use of henna among exposures with a documented association with haemolysis in deficient individuals, which matters in populations where cutaneous application is common, including in children [5]. Conversely, it confirms that the acetylsalicylic acid regimen of 75 to 100 mg daily, widely used in the prophylaxis of cardiovascular events, is safe in deficient individuals, reinforcing the rule that drug-induced haemolysis is dose-dependent [2, 5].

Nitrofurantoin illustrates the tension between evidence and regulatory warnings. Despite more than 70 years of worldwide use, few acute haemolytic anaemia cases following administration have been published; adverse-reaction database analysis estimated approximately 1 per 100,000 prescriptions, considered too low to be compatible with deficiency-conferred predisposition, although G6PD was not tested in most cases [2, 39]. Conversely, one study found acute haemolytic anaemia in 3 of 6 deficient patients given nitrofurantoin, and another demonstrated shortening of deficient erythrocyte half-life comparable to primaquine [2]. Medium-risk classification with an optional recommendation reflects this uncertainty [2].

Dapsone provides the opposite example, with well-documented quantified risk. Two trials of antimalarial preparations containing dapsone at 2.5 mg/kg once daily for 3 days included 95 deficient hemizygous boys, 24 deficient homozygous girls and 200 heterozygous girls; in the first two groups, mean maximum haemoglobin fall was −2.64 g/dL (range −6.70 to +0.30), significantly greater than in the artemether–lumefantrine comparator (adjusted difference −1.46 g/dL; 95% confidence interval −1.76 to −1.15); haemoglobin fell by at least 40% from baseline in 24 of 119 deficient children (20.2%), and 13 of 119 (10.9%) required transfusion [40]. Among heterozygous girls, mean maximum fall was −1.83 g/dL (range +0.90 to −5.20), and 1 in 200 (0.5%) required transfusion; all children recovered, and all deficient children had A−, formerly class III [40]. These figures are central to merging classes II and III: an allele historically labelled as causing “moderate” deficiency produced acute haemolytic anaemia in 98% of deficient boys in one series, with 11% requiring transfusion [2].

7.3. Primaquine and tafenoquine in malaria

Primaquine is a special case because alternative regimens have been studied in deficient patients and reflected in WHO and Centers for Disease Control and Prevention guidelines [2]. Standard relapse-prevention dosing of 0.25–0.5 mg/kg daily for 14 days for P. vivax or P. ovale should be avoided in deficient patients, with two exceptions following established guidelines: for radical treatment of hepatic forms, 0.75 mg/kg once weekly for 8 weeks under WHO guidance, or 45 mg once weekly for 8 weeks in adults under CDC guidance, with close haemolysis monitoring; and for gametocytocidal P. falciparum treatment, a single 0.25 mg/kg dose considered safe and effective without monitoring [2]. No primaquine dose is considered safe with chronic non-spherocytic haemolytic anaemia [2].

There is a dose divergence to record. The 2020 review describes radical cure of the hepatic forms of P. vivax as a 14-day course of primaquine at 0.75 mg/kg daily, which regularly causes haemolysis in a deficient individual, including in most heterozygous women, and gives the much lower single dose of 0.25 mg/kg for eliminating gametocytes after artemisinin-based combination therapy, which is safe in deficient individuals [5]. The reference dose for radical cure is that of the guidelines, namely 0.25 to 0.5 mg/kg daily for 14 days or 1 mg/kg daily for 7 days, with the short regimen reserved for activity of 70% or above, and it is this that should be prescribed; in deficient patients the accepted regimen is 0.75 mg/kg once weekly for 8 weeks, under close medical supervision, and not a daily course [2, 7].

Tafenoquine occupies a singular position in the history of applied pharmacogenetics: because it has a long half-life a single dose is sufficient, but that same property prevents withdrawal of the drug when signs of haemolysis appear, and it was the first medicine to be authorised with a label that positively mandates G6PD testing before prescription, which decisively stimulated the development of reliable point-of-care tests, including for heterozygous women [5, 16].

WHO malaria guideline version 10.0, published 10 September 2026, reiterates and details this framework. A strong, moderate-certainty recommendation requires G6PD status to guide primaquine or tafenoquine for relapse prevention [7]. In non-deficient patients, relapse-prevention primaquine is given at a high total dose of 7 mg/kg, divided into 0.5 mg/kg/day for 14 days or 1 mg/kg/day for 7 days, but the 7-day regimen requires activity of at least 70% [7]. In G6PD deficiency, the statement retained since 2015 permits primaquine base at 0.75 mg/kg once weekly for 8 weeks under close medical supervision, a conditional recommendation with very low certainty [7]. Tafenoquine is recommended as an alternative to primaquine at a total dose of 3.5 mg/kg to prevent P. vivax relapse in patients aged at least 2 years with activity ≥70% receiving chloroquine; this conditional, low-certainty recommendation is restricted to South America, requires prior quantitative or semiquantitative enzyme activity determination, and excludes pregnant and breastfeeding women [7]. When G6PD status is unknown and testing unavailable, primaquine prescription should be based on risk–benefit assessment [7]. In pregnant and breastfeeding women, weekly chloroquine chemoprophylaxis may continue until pregnancy and breastfeeding end, followed by primaquine according to G6PD status [7].

For malaria prophylaxis, the French protocol considers that atovaquone-proguanil, mefloquine and doxycycline may be prescribed to the deficient individual without particular precaution, keeps primaquine—available in France only through a named-patient temporary authorisation for use—as contraindicated, and notes that there is no alternative to primaquine for eradicating the recurrent hepatic forms of Plasmodium ovale and Plasmodium vivax [11]. For chloroquine it recommends use with caution, citing cases of haemolysis in individuals exposed to high doses, whereas the 2023 pharmacogenetic guideline classifies it as low to no risk with moderate certainty at standard doses [2, 11]. The reconciling reading is that the divergence concerns dose: the usual antimalarial or anti-inflammatory dose does not require avoidance, and the precaution applies to high-dose regimens, a principle that holds for all drug-induced haemolysis in this deficiency [2, 5]. The French contraindication of primaquine is also more absolute than the World Health Organization guideline and the pharmacogenetic guideline, which accept the regimen of 0.75 mg/kg once weekly for 8 weeks under close medical supervision; since that regimen is published in a guideline and the national contraindication arises from the drug's regulatory status in France, the reference dose for clinical decision-making is that of the guideline [2, 7, 11].

Mass administration of an isolated 8-aminoquinoline to reduce P. vivax transmission is not recommended because of potential serious harm without prior G6PD testing in a broad population [7]. Conversely, single-dose primaquine at 0.25 mg/kg with artemisinin combination therapy in low-transmission P. falciparum areas does not require G6PD testing [7].

7.4. Rasburicase and pegloticase

Rasburicase is a recombinant urate oxidase degrading uric acid into allantoin and hydrogen peroxide, used to prevent and treat hyperuricaemia, particularly during chemotherapy [2]. Pegloticase is the PEGylated form of the same enzyme, used in refractory gout [2]. Stoichiometric hydrogen peroxide generation explains the damage mechanism and supports high-risk classification, with high-level evidence linking acute haemolytic anaemia and methaemoglobinaemia to use in deficient patients [2, 41].

Clinical decisions are nevertheless not automatic: acute haemolytic anaemia and methaemoglobinaemia risks must be weighed against hyperuricaemia risk, particularly in newly diagnosed malignancy with anticipated tumour lysis, because tumour lysis syndrome may be fatal and alternatives such as allopurinol may lower uric acid less effectively [2]. In emergencies, G6PD status may not be available in time, and concomitant hyperleukocytosis or anaemia may confound measured activity interpretation [2].

7.5. Chronic haemolytic anaemia of class A variants

Treatment of the rare forms with chronic haemolysis follows the model of the constitutional haemolytic anaemias. Red cell concentrate transfusions are prescribed when the chronic anaemia worsens, or on a regular basis when it is very severe, and should be phenotyped for the Rh and Kell systems [11]. Iron chelation is discussed after 10 to 20 transfusions or when the liver iron concentration determined by magnetic resonance imaging reaches 125 µmol per gram of dry liver weight; deferoxamine and deferasirox have already been given to people with G6PD deficiency, although the protocol does not specify dosages [11]. As no dose is published for this condition, those of the official labels in transfusional iron overload apply, with the extrapolation declared: deferasirox orally, 20 mg/kg once daily as the starting dose in patients aged 2 years and over with an estimated glomerular filtration rate above 60 mL/min/1.73 m², with increases to 30 mg/kg/day and a maximum of 40 mg/kg/day, doses above this not being recommended; deferoxamine by subcutaneous infusion, a mean daily dose of 20 to 60 mg/kg, with a minimum of 20 mg/kg/day and maxima of 40 mg/kg/day in paediatric age and 60 mg/kg/day in adults, guided by serum ferritin [42, 43]. The thresholds for initiation belong to the protocol and the dosages to the label, so that the indication and the dose have distinct sources.

Folic acid supplementation is recommended and tocopherol may be proposed, likewise without a stated dosage [11]. For folic acid, the dose used in hereditary spherocytosis, an analogous condition of constitutional chronic haemolysis, is 400 µg once daily up to one year of age and 1 mg once daily thereafter, with 800 to 1,000 µg daily in pregnancy; the transposition is a declared extrapolation [33]. For tocopherol, the dose described in an analogous condition of glutathione depletion, glutathione synthetase deficiency, is 10 mg/kg/day, also by extrapolation; in that condition high-dose vitamin C is combined, a regimen that is not transposable to G6PD deficiency, where vitamin C supplements must respect the limit of 1 g per day [11, 35].

Splenectomy may be discussed in severe anaemia requiring regular transfusions, but it is inconsistently effective, which tempers the expectation created by the case of transfusion independence already described; cholecystectomy is indicated for symptomatic gallstones or at the same operative session as a splenectomy [5, 11]. Vaccines follow the childhood schedule, including those against influenza and hepatitis B, and follow-up of the splenectomised patient obeys the usual rules for that situation [11]. Acute haemolytic episodes from oxidative stress also occur in these patients, worsening the chronic anaemia already present, so that the rules of drug and dietary avoidance apply strictly [11].

7.6. Dietary and environmental avoidance, education and documents

Avoidance of fava beans is absolute and not gradable: Vicia faba and Vicia faba var. minor are prohibited raw or cooked, fresh or frozen, in any preparation, and the contraindication applies to all deficient individuals, whatever the variant, because the intensity of haemolysis varies between people and within the same person [11]. There are rare, unproven cases attributed to the inhalation of fava bean pollen [11]. Fava bean flour is handled separately: in France, regulation limits its proportion in industrial or bakery bread to 2%, food products containing it are subject to mandatory labelling in decreasing order of weight, there are no published cases of a haemolytic accident attributed to such products and, accordingly, there is no argument for extending the contraindication to very low quantities; vigilance is justified when fava bean flour is the main ingredient or is present in a large proportion, which may occur in dishes from other culinary traditions [11].

Apart from fava beans, all other fruits and vegetables are permitted [11]. The remaining restrictions are quantitative or directed at specific products: vitamin C food supplements should not exceed 1 g per day; the quinine present in tonic waters and in energy or aperitif drinks can provoke acute haemolysis; episodes have been described after exposure to henna, cutaneous in infants or by the digestive route and, less well documented, after ingestion of the medicinal plant Acalypha indica; preparations of plant extracts or vitamins of unknown composition should not be consumed; the nitrites used as recreational inhalants cause acute haemolysis and methaemoglobinaemia; and the naphthalene of mothballs provokes haemolytic complications through contact, inhalation or accidental ingestion, above all in infants [11].

Therapeutic education has defined rules. There is no gradation in prevention: the same rules apply to any deficient person, whatever the intensity and type of the deficiency; erroneous lists of contraindicated foods circulate, in particular on the internet, which makes it necessary to provide validated documentation; self-medication is strongly discouraged; and the recommended doses of vitamin C, acetylsalicylic acid and paracetamol must be strictly observed [11]. A care and emergency card for G6PD deficiency exists in France, ordered and completed by the general practitioner or the specialist since 2015, detailed enough to serve as an interface at any medical intervention or administration of treatment, and which should accompany the patient, especially during stays away from home [11].

At school, the deficient child can eat in the canteen with the simple avoidance and substitution of fava beans, and the formal route for this is an individualised reception plan, provided for by a circular of the ministry of education of 8 September 2003: a brief document is recommended for all deficient children, to flag the avoidance of fava beans and pass the emergency card to the team, and a detailed document, drawn up by a hospital specialist, is indicated in children with class A variants [11]. There is no restriction of physical activity, and school outings, stays and trips are permitted; symptom surveillance is not a daily matter, because the child is healthy outside exposures [11]. The dietary information sheet of the French national food safety agency, published in November 2006 at the request of the patient association and the directorate-general for health, upholds the two essential rules: never consume fava beans, in whatever preparation, and do not drink quinine-containing beverages [11]. The signs that the family and the school should recognise are pallor of the lips and skin, very dark urine, unusual fatigue, yellow colouring of the eyes and skin, and severe abdominal pain [11].

7.7. Comparative regulatory table

Marketing authorisation does not guarantee marketing, reimbursement or availability in a particular country. Portuguese and Brazilian funding decision databases were not consulted for these medicines, so no national access claims are made here.

7.8. Research and recent developments

Before the trials under way, it is worth delimiting what exists in terms of correcting the deficiency. Transduction of haematopoietic stem cells with a retroviral vector carrying the human G6PD complementary DNA achieved stable expression in primary and secondary syngeneic recipient mice, and a similar vector expressed the human enzyme in macaques; in principle, the severe chronic form could be treated by allogeneic bone marrow transplantation or by gene therapy, but neither approach had been attempted up to the date of the review [5]. The pharmacological route explored rests on two findings: histone deacetylase inhibitors selectively increase G6PD synthesis, and not that of sixteen other erythrocyte enzymes tested, so that butyrate or valproate could in theory partially correct the deficiency, although they would require chronic administration; and a high-throughput screen identified a small-molecule activator of the enzyme, designated AG1, whose activation is less than twofold and which serves as a starting compound for more effective activators [5]. None of these strategies has reached clinical evaluation, and the existing mouse and zebrafish models have not yet been adopted for routine safety testing of new medicines [5].

The following table includes studies whose design or status was verified directly in the registry or publication.

Comparison between these rows has clear limitations: populations, countries and outcomes differ, precluding ranking regimens indirectly. Mean haemoglobin fall does not represent individual risk, and a median is not a population mean; ALTPRIM's safety outcome is haemoglobin change rather than transfusion requirement. Stopping an arm for safety is clinically relevant and retained as such [48].

Regarding recent epidemiology, a 2026 Brazilian Journal of Infectious Diseases study genotyped seven G6PD variants in two groups from Acre, Amazonas and Rondônia, finding that 95.8% of deficient individuals carried the African A− haplotype and estimating population allele frequencies of 7.0% for 376G and 3.2% for 202A [20]. For guidelines, WHO malaria version 10.0, published 10 September 2026, is the latest operational document [7]. No therapy targeting the enzyme deficiency under regulatory review was identified in consulted sources; this reflects those sources and does not prove absence of ongoing programmes.

8. Reproductive Options and Family Counselling

Inheritance is X-linked. A deficient male transmits the deficient allele to all daughters, who become heterozygous, and no sons; a heterozygous woman transmits it to half her sons, who become deficient hemizygotes, and half her daughters, who become heterozygous [2]. Female homozygosity is not rare in high-allele-frequency populations, and consanguinity or shared parental origins increase its probability [9, 52].

The most frequently harmful clinical assumption is that heterozygous women are protected. They are not: random X-inactivation produces deficient erythrocyte proportions from nearly 0% to nearly 100%, so overall enzyme activity may fall anywhere from normal to markedly deficient, and a heterozygous female may show drug-induced haemolysis indistinguishable from a hemizygous male [2, 3]. In the dapsone trial, 1 in 200 heterozygous girls required transfusion, and maximum haemoglobin fall reached −5.20 g/dL in one girl [40]. Counselling should therefore include enzyme activity measurement in female relatives before any high-risk exposure, rather than genotype determination alone [2].

The same reasoning applies to anyone with more than one X chromosome, including Klinefelter syndrome, and should be anticipated during consent: pharmacogenetic G6PD testing may reveal a previously unknown sex chromosome abnormality such as Klinefelter or Turner syndrome, and may also reveal birth-assigned sex in people with non-conforming gender identities [2]. Services routinely testing genotype should have defined procedures for this eventuality, including genetic counselling referral [2].

For pregnancy and lactation, regulatory information is explicit regarding 8-aminoquinolines. Tafenoquine is not recommended during pregnancy because it may harm a deficient fetus, and is contraindicated during breastfeeding when the infant is deficient or has unknown status; infant G6PD status should be checked before breastfeeding begins; pregnancy testing is also recommended before treatment in women of reproductive potential [16, 27]. The WHO excludes tafenoquine in pregnant and breastfeeding women and recommends weekly chloroquine chemoprophylaxis until pregnancy and breastfeeding end, followed by primaquine according to G6PD status [7]. Relapse-prevention primaquine is also excluded in pregnant women, infants below 1 month and women breastfeeding infants below 1 month [7].

The French protocol defines the scope of the family study. In the usual forms, with an acute haemolytic episode without chronic haemolysis, explanations about transmission are given by the paediatrician, the internal medicine physician, the haematologist or in the genetics clinic, prenatal diagnosis is not indicated, and what is advocated is screening for the deficiency at birth [11]. Outside the neonatal context, enzyme assay is offered to the siblings and the maternal uncles of a male index case; the mother is at least heterozygous, and assay is not offered systematically to the remaining female relatives [11]. When the symptomatic index case is female, through heterozygosity with skewed inactivation or, more rarely, through homozygosity, the family study is broader and includes the father and the sisters [11].

In class A variants, with chronic haemolytic anaemia, a genetic counselling consultation is indicated: it clarifies the maternal status, distinguishing the carrier from a de novo mutation, and, in severe forms with repeated transfusions, allows discussion of fetal sex determination in maternal blood and of possible prenatal diagnosis, which has prior identification of the causal variant as a prerequisite [11]. Molecular study of the gene in the mothers of children with one of these variants is necessary even when the maternal assay is normal, because skewed erythrocyte selection makes that result uninformative [11].

Management in heterozygous females is graded by the measured activity. If activity is clearly low, a situation corresponding to inactivation skewed in favour of the X chromosome carrying the variant, the same precautions established for the deficient male and for the rare homozygous females apply; if it is normal — without this excluding heterozygosity — particular precautions are not mandatory, with two exceptions [11]. The exceptions are pregnancy and breastfeeding: all women expecting or breastfeeding a potentially deficient child should observe the drug and dietary contraindications, because cases of haemolysis have been described in deficient neonates and infants in the context of maternal exposure to oxidant agents, in particular through breast milk [11].

The content of the disclosure consultation is defined: explanation of the disease and of the practical aspects of prevention, provision of written documents and of the care and emergency card, explanation of the mode of transmission with the possibility of a genetics clinic referral, signposting of the patient association, and a second medical consultation scheduled at a distance from the diagnosis, to review what has been understood and update the documents [11].

Practical counselling centres on avoidance. Patients should be told they risk haemolysis after exposure to fava beans or high-risk medicines and chemicals and should avoid them, and that infection, hyperuricaemia and sepsis also confer risk through reactive oxygen species generation [2]. Although some sources suggest foods other than fava beans can cause haemolysis, guideline authors found no supporting peer-reviewed evidence [2]. Conversely, household naphthalene exposure is a documented severe trigger, including contact with impregnated clothing [17]. Written information should be provided, with the condition recorded in clinical notes and ideally an electronic prescribing alert.

Neonatal diagnosis informs families about future pregnancy risks and identifies siblings at risk of recurrent haemolysis [6]. G6PD deficiency diagnosis in a neonate with prolonged jaundice should also prompt reconsideration of concomitant aggravating factors such as Gilbert syndrome [23].

9. Special Considerations: Population Diversity and Epidemiology

G6PD deficiency is the commonest human enzymopathy. The estimate cited by the pharmacogenetic guideline is that almost 5% of the world population is deficient, nearly all with class II or III alleles in the old nomenclature [2]. The systematic review underpinning that estimate assembled 280 prevalence estimates covering 88 countries, published between 1960 and 2008, and derived an approximate global prevalence of 4.9%, equivalent to about 400 million people, with the highest prevalences in sub-Saharan African countries even after adjusting for the assessment method, and with high heterogeneity attributable to differences in diagnostic methods [8, 17]. The 2024 WHO review refers to about 500 million people affected [3]. The 2020 review uses the same order of magnitude, more than 500 million people, and describes the breadth of variation between populations: from the complete absence of the deficiency in original Amerindian populations to 20% or more in parts of Africa and Asia [5]. Interpreting these figures requires attention to how they were obtained, because many older surveys pooled both sexes into a single frequency, which underestimates the allele frequency [5].

The most detailed modelled map comes from representative community surveys at 1,734 sites, including 1,289 in malaria-endemic countries, using a Bayesian geostatistical model adapted to X-linked inheritance [9]. Predicted mean allele frequencies were 8.0% (interquartile range 7.4–8.8) in endemic countries and 5.3% (4.4–6.7) in elimination settings, corresponding to approximately 350 million affected people; the highest predicted median prevalence, 32.5%, was in Saudi Arabia's Eastern Province, in coastal urban Al-Qatif and Ad-Dammam, falling to 3% (2–4) in central Al-Kharj and Riyadh [9]. Frequencies were lowest in the Americas and highest in tropical Africa and the Arabian Peninsula, but the absolute majority of affected people (median estimate 67.5%) live in Asian countries because of population density [9]. The authors emphasise that map reliability depends on model input data and substantial gaps remain, with only 4% of surveys from nine American countries [9].

Common deficient alleles vary by ancestry. According to gnomAD data cited in the pharmacogenetic guideline, Seattle, Lodi, Modena, Ferrara II and Athens-like total approximately 0.11% in European ancestry; Kaiping, Anant, Dhon and Sapporo-like approximately 0.7% in East Asian ancestry; Ilesha 0.12% in Amerindian and 0.05% in sub-Saharan African ancestry; and Kaylan-Kerala, Jamnaga and Rohini approximately 1.1% in Central and South Asian ancestry [2]. Mahidol and Viangchan predominate in Southeast Asia, the former in Myanmar and western Thailand and the latter in Thailand, Laos, Vietnam and Cambodia [14].

In the United States, the neonatal hyperbilirubinaemia guideline states that approximately 13% of African American males and 4% of females have G6PD deficiency, and identifies sub-Saharan African, Middle Eastern, Mediterranean, Arabian Peninsula and Southeast Asian ancestry as high risk [6].

A documented founder population is Kurdish Jews: G6PD Mediterranean, c.563C>T (p.Ser188Phe), accounts for 95% of deficiency in this group according to the common founder variant table for Sephardic Jewish populations, while 75% has been reported in Iranian Jews from Shiraz [53].

In France, the figure officially cited remains that of the 1989 World Health Organization working group, which reported 0.39% of deficient hemizygous males in metropolitan France, about 120,000 individuals, and 12% in Martinique and Guadeloupe [11, 15]. The national protocol itself considers the metropolitan figures underestimates, being old and not accounting for migrants originating from countries where the deficiency is frequent, and proposes as a more realistic estimate at least 250,000 deficient individuals in France [11]. As to mutational composition, the two deficient variants most often found in France are A− and B−, also designated Mediterranean, which account for 80% of the variants detected, with 60% for A− in individuals originating from sub-Saharan Africa and 22% for B− in those from the Mediterranean basin [11]. These figures describe the caseload of French laboratories rather than a random population sample.

On the mechanism of the selective advantage there is a divergence of interpretation that should not be erased. The 2017 French protocol states that protection has been demonstrated in African populations, in which deficient individuals would have a risk of severe P. falciparum malaria reduced by about 50%, a formulation that includes hemizygous males [11]. The 2020 review documents that the 1995 work on which this statement rests was subsequently invalidated, because the genotyping methodology used misclassified a substantial proportion of the samples, and that subsequent studies attribute protection against severe malaria to heterozygous females rather than to hemizygous males [5]. The more recent reading prevails, being based on the methodological reanalysis of the original study; what remains intact is the clinical consequence written into the protocol itself: whatever the protection at population scale, individual malaria prophylaxis is formally indicated in deficient individuals travelling to endemic areas [5, 11].

In Portugal, retrospective mutational profiling of 138 unrelated individuals (101 males and 37 females) without known sub-Saharan ancestry, diagnosed between 1994 and 2020 at the Molecular Haematology Unit of Centro Hospitalar e Universitário de Coimbra, identified 21 different pathogenic variants, 20 missense and one in-frame exon 10 deletion; the three commonest share the African c.376A>G haplotype background: A− (c.202G>A, p.Val68Met) in 58.6%, Betica (c.968T>C, p.Leu323Pro) in 12.1%, and Santamaria (c.542A>T, p.Asp181Val) in 4.3% [54]. The authors recall that mean male deficiency frequency in Portugal was estimated at approximately 0.5% [54]. This laboratory series is not a population survey; proportions describe diagnosed case composition rather than population variant prevalence.

In Brazil, a study of 1,000 newborn samples from the South-East, with quantification of enzyme activity, identified 25 deficient individuals, that is 2.5%; of the 21 samples submitted to molecular analysis, all carried the G202A/A376G genotype (historical numbering, = NM_001360016.2:c.202G>A p.(Val68Met) and c.376A>G p.(Asn126Asp)), corresponding to the A− phenotype [55]. In Amazonia, a cross-sectional screening study in 41 endemic municipalities across six states, between 2014 and 2018, tested 14,847 men by fluorescent spot test and found 5.6% deficient, with the highest prevalence in Acre (8.3%), followed by Amapá (5.8%), Pará (5.7%), Rondônia (5.4%), Roraima (4.2%) and Amazonas (4.0%); of 828 genotyped samples, 39.3% were A−, 6.2% A+ and 54.2% carried none of the African or Mediterranean variants sought [52]. In that study, the proportion of P. vivax malaria relapses was higher in deficient individuals (16.7% versus 4.1%; relative risk 3.52; 95% confidence interval 2.16-5.74; p < 0.01), which the authors interpret in the context of the limitations of anti-relapse treatment in these patients [52]. A 2026 study confirmed the predominance of the A− haplotype among deficient individuals in the north of the country (95.8%) and estimated the population allele frequencies already cited, higher in Porto Velho, consistent with the greater local African ancestry [20].

The relationship with malaria is the evolutionary key to this entire geography: heterozygosity is associated with protection against malaria mortality, which maintains the polymorphism and explains frequencies of 30% or more in certain groups from endemic regions [2, 5]. The hypothesis was proposed independently by Anthony Allison and Arno Motulsky in 1960, from the geographical correlation with P. falciparum endemicity, and received support from the micromapping of Sardinia in 1966, which showed a direct correlation between the frequency of the deficiency and the historical prevalence of malaria on an island where the disease had already been eradicated [5]. The most plausible mechanism is immunological rather than parasitological: deficient erythrocytes infected by P. falciparum are recognised as abnormal by macrophages at an early stage of parasite development and removed by phagocytosis earlier than infected normal erythrocytes, whereas parasite growth in culture, under optimal conditions, is similar in the two cell types [5]. Several recent studies have confirmed that protection against severe malaria is a prerogative of heterozygous females rather than hemizygous males, which remains counterintuitive and without an established mechanistic explanation; whether it protects more against severe anaemia or against cerebral malaria is debated, with the caveat that one of the findings has been challenged on grounds of possible collider bias [5]. A much-cited 1995 study that also attributed protection to deficient males was subsequently invalidated by the demonstration that the genotyping methodology used misclassified a substantial proportion of the samples, an episode that illustrates the weight of the typing method in epidemiological conclusions [5]. The coexistence of G6PD deficiency and other erythrocyte disorders prevalent in the same regions, such as the thalassaemias, requires careful interpretation of enzyme activity results and haematological profiles [36, 37].

The avoidable disease burden is, in public health terms, the most actionable datum. Favism is described in at least 35 countries and probably causes thousands of cases a year, and the second most frequent trigger is iatrogenic, with published deaths associated with both primaquine and rasburicase — including a preterm neonate who died after a single dose of rasburicase [5]. Two population-level interventions have shown an effect: neonatal screening combined with health education, which in northern Sardinia was followed by a marked decline in favism, and the agricultural selection of Vicia faba cultivars with absent or very low vicine and convicine content, whose ingestion by deficient individuals produced neither erythrocyte injury nor haemolysis in a study published in 2018 [5].

The organisation of care in France offers a transferable model of articulation between levels. G6PD deficiency is integrated into the MCGRE health network, dedicated to rare constitutional disorders of the red cell and of erythropoiesis, which brings together three reference centres — the centre for major sickle cell syndromes in Créteil, the sickle cell disease centre in Pointe-à-Pitre and the thalassaemia centre in Marseille — and their associated competence centres, with the full list published on the network's website [11]. The protocol recommends seeking the opinion of these centres in any unusual situation or difficult decision in a patient with a class B variant, and ensuring through them the coordinated multidisciplinary follow-up of all patients with class A variants [11]. The French patient association is Vigifavisme, which took part in drawing up the protocol and whose contribution is reflected in the dietary information sheets and the educational documents [11]. Administratively, only severe forms with chronic haemolytic anaemia are admitted to the long-term condition scheme, ALD no. 10 for severe constitutional chronic haemolysis, under decree no. 2011-77 of 19 January 2011, with a list of procedures and services published by the network [11]. The practical consequence is that the great majority of deficient individuals neither have nor need chronic disease status: what they need is correct information and a document that carries it.

The relevant European reference network is ERN-EuroBloodNet for rare haematological diseases, bringing together experts, clinical guidance and a European registry [10]. For patients and families, the National Organization for Rare Disorders maintains a G6PD deficiency information page covering signs and symptoms, causes, affected populations, diagnosis and standard therapies, last updated 29 June 2017 [56]. Rare disease status is not uniform: Orphanet classifies the overall ORPHA:362 entity as not rare in Europe, retaining only class I deficiency, ORPHA:466026, as an active rare entity [4].

10. Illustrative Clinical Case

The following case was published open access in the Italian Journal of Pediatrics in 2025 and is summarised faithfully to the original report; the management described is the authors' and does not itself constitute routine recommendations [22].

A 7-year-old girl was admitted to a paediatric clinic in Varese, Italy, with high fever for approximately two days, preceded by several vomiting episodes, followed by dark urine, two loose stools without visible blood and profound fatigue, with pale, jaundiced skin [22]. Paracetamol and a single ibuprofen dose, never previously taken, had been given at home. Family haematological history was negative, parents were non-consanguineous, and neither the patient nor her parents had ever eaten fava beans or other legumes [22].

On emergency department arrival, she was fatigued and pale but still answered simple questions, with normal cardiothoracic and abdominal examinations. Peripheral oxygen saturation was 75% on room air with little response to oxygen, respiratory rate 20/min and heart rate 120/min [22]. Tests showed severe anaemia (haemoglobin 6.3 g/dL; normal 11.5–14.5), leukocytosis (21,840/mm³), C-reactive protein 50.6 mg/L, normal platelets (190,000/mm³), very high ferritin (15,648 ng/mL), total bilirubin 3.2 mg/dL with indirect fraction 2.9 mg/dL, creatinine 0.65 mg/dL and lactate dehydrogenase above the laboratory's upper detection limit [22]. Direct and indirect Coombs tests were negative, myoglobin and haptoglobin normal, and blood film showed erythroblasts, myelocytes and metamyelocytes without schistocytes. Blood gas analysis revealed methaemoglobinaemia of 13.7% (normal below 1%) [22]. Urine contained haemoglobin, 150 erythrocytes/µL and increased protein (147 mg/dL).

Because autoimmune disease could not be reliably excluded—the authors note routine Coombs tests may be inaccurate in cold agglutinin disease—two packed red cell transfusions of 5 mL/kg were given and methylprednisolone started until cold haemagglutinin testing was negative [22]. Persistent symptoms prompted quantitative G6PD measurement despite recognised false-negative risk during acute haemolysis; it was deficient at 7.1 U/g haemoglobin against a local physiological range of 10.1–14.2 U/g [22]. On hospital day three, with pretransfusion haemoglobin 7.0 g/dL, another 10 mL/kg transfusion was well tolerated; haemoglobin and methaemoglobin trajectories across transfusions appear in article figure 1, consulted open access [22].

Further investigations showed abdominal ultrasound without organomegaly but abundant biliary sludge; chest radiography without mediastinal masses; negative Epstein–Barr virus, cytomegalovirus, parvovirus B19 and Mycoplasma pneumoniae serology; norovirus-positive stool viral testing; and negative blood, stool, throat and urine cultures [22]. Haemoglobin rose with appropriate reticulocytosis (178,000/mm³), methaemoglobin gradually fell, permitting oxygen withdrawal, and bilirubin, ferritin and lactate dehydrogenase normalised. Creatinine rose to 0.9 mg/dL on day two, with urine output falling to 0.7 mL/kg/h and subsequent post-acute-tubular-necrosis polyuria without diuretics; tubular proteinuria (maximum urinary protein/creatinine ratio 3.9 mg/mg), raised urinary beta-2-microglobulin, mild hypokalaemia, hypouricaemia, fractional sodium excretion up to 2.8% and transient glycosuria without hyperglycaemia were consistent with acute tubular necrosis probably secondary to norovirus infection [22].

Both parents had deficient G6PD assays (mother 0.3 U/g; father 0.7 U/g haemoglobin), explaining the child's phenotype despite female sex and apparently negative family history [22]. The authors chose no additional genetic investigation because enzyme testing was unequivocal and genetic testing was outside the centre's routine diagnostic pathway; this was their decision and differs from the pharmacogenetic guideline's emphasis on integrating genotype and enzyme activity [2, 22]. Considered triggers were intercurrent norovirus infection and ibuprofen administration [22].

The case illustrates several points of this article: presentation in a girl without family history; haemolysis with methaemoglobinaemia and consequent oxygen-unresponsive desaturation; possible enzyme confirmation during episodes with profound deficiency without invalidating the general false-normal rule; and the importance of avoiding methylene blue before G6PD status is known [2, 22].

11. Prognosis and Long-Term Follow-up

The prognosis of the common form, corresponding to class B variants, is excellent provided exposure to triggers is avoided and episodes are recognised and treated promptly: acute haemolytic anaemia can be managed effectively when diagnosed in time, and most individuals remain asymptomatic throughout life [3, 5]. In the dapsone trial, every child with haemolysis eventually recovered, including those who required transfusion [40]. Rapid recovery is the rule in this type of haemolysis, with a reticulocyte response within 48 hours, which distinguishes the short-term prognosis from that of autoimmune haemolytic anaemias [5]. It should be noted that the clinical trial series were conducted under medical supervision and with no recorded deaths, whereas outside that setting deaths from favism and from iatrogenic exposure have been published [5, 40].

Individual lifetime risk cannot, however, be quantified from these data because it depends on exposure. Series document specific exposure risks: 20.2% had haemoglobin falls of at least 40%, and 10.9% required transfusion among deficient children exposed to antimalarial-dose dapsone [40]. Absence of events in an individual does not demonstrate absence of risk, only insufficient exposure.

The most serious population-level morbidity centres on neonatal hyperbilirubinaemia leading to kernicterus, with cerebral palsy and death if untreated [2, 17]. US neonatal surveillance follows the 2022 guideline: assess risk factors, use total serum bilirubin as the definitive test for phototherapy and care escalation decisions, measure haemoglobin, haematocrit or full blood count in infants requiring phototherapy, and investigate causes including G6PD activity in the situations described in section 5; discharge may be delayed if outpatient follow-up with repeat bilirubin measurement cannot be assured [6].

In patients with class A deficiency, that is, with chronic non-spherocytic haemolytic anaemia, follow-up is that of a chronic haemolytic anaemia, with attention to gallstones, splenomegaly and reticulocytosis, and with strict avoidance of oxidant agents [4]. In regularly transfusion-dependent cases, splenectomy is documented as an effective option, with transition to transfusion independence and remission of reticulocytosis in a patient operated on at 9 years of age; the decision requires weighing the post-splenectomy infectious risk and prior vaccination, and the review describing the case does not present it as routine practice [5]. In these patients, pharmacological recommendations are conservative for want of data: high- and medium-risk medicines are avoided, and even those of low or no risk are used with caution and close monitoring [2].

Long-term follow-up of asymptomatic deficiency does not itself require scheduled consultations. It requires three continuing safeguards: clearly visible recording in clinical notes and prescribing systems; an updated written high-risk medicine list and information on fava beans, infections and naphthalene for patients and families; and confirmation of G6PD status by enzyme activity measured outside haemolysis and sufficiently distant from transfusion before predictably risky interventions, particularly chemotherapy with anticipated tumour lysis, refractory gout treatment and malaria prophylaxis or radical cure [2, 6, 7].

The French protocol converts these assurances into dated actions. In class B variants there is no long-term treatment or surveillance to institute, because there is no splenomegaly, anaemia or jaundice at baseline, and no negative impact on life expectancy or physical activity has been described [11]. What is provided for are targeted consultations at two transition points: with the family, when the child starts school, where the diagnosis was made neonatally, and with the individual once they become the direct interlocutor in the consultation, in adolescence or adulthood; each serves to repeat the preventive advice and the signs of haemolysis, check and update the documents issued, extend the family study if necessary and re-explain transmission [11]. Between these consultations responsibility lies with the general practitioner, and the deficiency must be flagged systematically to every new clinician involved [11].

Two preventive measures are added to the routine schedule. Since numerous cases of severe acute viral hepatitis, particularly hepatitis A virus, have been published in deficient individuals of Mediterranean and Asian origin, the protocol recalls that in France hepatitis A vaccination is recommended for adults and children over one year of age who may stay in a country of high endemicity [11]. And, despite the selective advantage at population level, individual malaria prophylaxis remains formally indicated [11]. An increased risk of gallstones has been described in individuals deficient for the Mediterranean variant, which does not, however, justify systematic ultrasound screening in this population [11].

In class A variants, follow-up is that of any chronic haemolytic anaemia and involves the specialist paediatrician and subsequently the internal medicine physician or haematologist, with a multidisciplinary team including radiology, visceral surgery, cardiology and endocrinology for the possible complications of chronic haemolysis and iron overload; transition from paediatric care to adult medicine should be planned and organised rather than left to the chance of age [11]. The following table brings together the examinations, ages of onset and intervals, with the source indicated for each row.

In heterozygous women, reassessing activity before high-risk exposure is prudent because deficient erythrocyte proportions may change over time and intraindividual variability is greater in people with more than one X chromosome [2].

12. Summary and Final Considerations

G6PD deficiency combines a simple molecular defect with clinical complexity arising almost entirely from genotype, X-linked mosaicism and external exposures. G6PD at Xq28 encodes the enzyme initiating the pentose phosphate pathway and providing mature erythrocytes with their sole NADPH source; without NADPH, glutathione cannot regenerate, haemoglobin oxidises and precipitates, membranes are damaged and cells lyse, with possible methaemoglobin formation [2, 17].

Current variant classification is the WHO scheme proposed in 2022 and published in 2024, with classes A, B, C and U. It applies to variants rather than people, merges former classes II and III into one class B because enzyme activities overlapped and both conferred susceptibility to the same manifestations, and abandons class V [3]. Reporting “class III” today retains historical information but should be accompanied by the current class and, above all, measured enzyme activity.

Pharmacological management should follow the 2023 Clinical Pharmacogenetics Implementation Consortium guideline distinguishing high, medium and low-to-no risk [2]. Avoid rasburicase, pegloticase, dapsone, methylene blue, toluidine blue, tafenoquine and standard relapse-prevention primaquine; use nitrofurantoin and weekly primaquine cautiously; and use at standard doses medicines included for decades on avoidance lists without literature support, such as sulfamethoxazole, aspirin up to 1 g/day, chloroquine and hydroxychloroquine [2]. Correcting these lists is as important as prohibitions because indiscriminate lists deprive patients of safe therapies [28].

Safety rests on the appropriate test at the appropriate time. Spectrophotometric enzyme activity remains confirmatory, but interpretation requires the assay reference value, measurement outside haemolysis and sufficiently distant from transfusion, and awareness that neonates, anaemia and acute leukaemia alter results [2, 29]. Genotype establishes diagnosis in hemizygous males, homozygous females and compound heterozygotes but cannot assign phenotype to people with more than one X chromosome; activity must be measured in these cases [2].

The three clinical pitfalls most frequently causing harm, ranked by frequency observed in consulted literature, are assuming a heterozygous woman is protected, accepting activity measured during haemolysis as normal, and treating methaemoglobinaemia with methylene blue without knowing G6PD status [2, 6, 22].

Population estimates indicate approximately 4.9% of humanity is affected, with modelled allele frequency of 8.0% in malaria-endemic countries and peaks above 30% in parts of the Arabian Peninsula and tropical Africa; most affected people live in Asia [8, 9]. In Portugal, the Coimbra series shows predominance of African haplotype-background variants, with A− in 58.6% of diagnosed cases and an estimated approximately 0.5% deficient males [54]. In Brazil, prevalence was 2.5% in 1,000 southeastern neonates and 5.6% in 14,847 Amazonian males, with A− predominating [20, 52, 55].

Finally, no therapy directed at the enzyme deficiency exists, and the prospects identified in the sources consulted are limited to optimising 8-aminoquinoline regimens in deficient patients, exploring early-stage antioxidant hypotheses and preclinical work on enzyme correction — gene transfer in mice and macaques, induction of enzyme synthesis by histone deacetylase inhibitors and small-molecule activators — with no progression to human study [5, 48, 49, 51]. The clinical gain available today continues to come from information: identifying who is deficient before prescribing, documenting the condition and giving the patient a correct list — neither excessive nor incomplete — of what must be avoided.

13. Bibliographic References

1. VariantValidator (version 4.0.1, database vvta_2025_02). Normalisation of G6PD variants on MANE Select transcript NM_001360016.2. Available at: https://rest.variantvalidator.org (accessed 6 October 2026).
2. Gammal RS, Pirmohamed M, Somogyi AA, Morris SA, Formea CM, Elchynski AL, et al. Expanded Clinical Pharmacogenetics Implementation Consortium Guideline for Medication Use in the Context of G6PD Genotype. Clin Pharmacol Ther. 2023;113(5):973-85. doi:10.1002/cpt.2735. PMID: 36049896.
3. Luzzatto L, Bancone G, Dugué PA, Jiang W, Minucci A, Nannelli C, et al. New WHO classification of genetic variants causing G6PD deficiency. Bull World Health Organ. 2024;102(8):615-7. doi:10.2471/BLT.23.291224. PMID: 39070600.
4. Orphanet. Class I glucose-6-phosphate dehydrogenase deficiency (ORPHA:466026) and Glucose-6-phosphate-dehydrogenase deficiency (ORPHA:362). Data dated 2 July 2026. Available at: https://api.orphacode.org/EN/ClinicalEntity/orphacode/466026 (accessed 6 October 2026).
5. Luzzatto L, Ally M, Notaro R. Glucose-6-phosphate dehydrogenase deficiency. Blood. 2020;136(11):1225-40. doi:10.1182/blood.2019000944. PMID: 32702756.
6. Kemper AR, Newman TB, Slaughter JL, Maisels MJ, Watchko JF, Downs SM, et al. Clinical Practice Guideline Revision: Management of Hyperbilirubinemia in the Newborn Infant 35 or More Weeks of Gestation. Pediatrics. 2022;150(3):e2022058859. doi:10.1542/peds.2022-058859. PMID: 35927462.
7. World Health Organization. WHO guidelines for malaria, version 10.0, published 10 September 2026. Available at: https://app.magicapp.org/#/guideline/LwRMXj (accessed 6 October 2026).
8. Nkhoma ET, Poole C, Vannappagari V, Hall SA, Beutler E. The global prevalence of glucose-6-phosphate dehydrogenase deficiency: a systematic review and meta-analysis. Blood Cells Mol Dis. 2009;42(3):267-78. doi:10.1016/j.bcmd.2008.12.005. PMID: 19233695.
9. Howes RE, Piel FB, Patil AP, Nyangiri OA, Gething PW, Dewi M, et al. G6PD deficiency prevalence and estimates of affected populations in malaria endemic countries: a geostatistical model-based map. PLoS Med. 2012;9(11):e1001339. doi:10.1371/journal.pmed.1001339. PMID: 23152723.
10. ERN-EuroBloodNet — European Reference Network on Rare Haematological Diseases. Available at: https://eurobloodnet.eu/ (accessed 6 October 2026).
11. Haute Autorité de Santé. Protocole National de Diagnostic et de Soins (PNDS): Déficit en G6PD (Glucose-6-Phosphate Deshydrogenase) ou favisme. Saint-Denis La Plaine: HAS; June 2017 (filière de santé maladies rares MCGRE). Available at: https://www.has-sante.fr/jcms/c_2800205 (accessed 6 November 2026).
12. Wamelink MMC, Valayannopoulos V, Garavaglia B. Disorders of Glycolysis and the Pentose Phosphate Pathway. In: Saudubray JM, Baumgartner MR, García-Cazorla Á, Walter J, editors. Inborn Metabolic Diseases: Diagnosis and Treatment. 7th ed. Berlin: Springer; 2022. Chapter 7. doi:10.1007/978-3-662-63123-2_7.
13. Alving AS, Carson PE, Flanagan CL, Ickes CE. Enzymatic deficiency in primaquine-sensitive erythrocytes. Science. 1956;124(3220):484-5. doi:10.1126/science.124.3220.484-a. PMID: 13360274.
14. Bancone G, Chu CS. G6PD Variants and Haemolytic Sensitivity to Primaquine and Other Drugs. Front Pharmacol. 2021;12:638885. doi:10.3389/fphar.2021.638885. PMID: 33790795.
15. WHO Working Group. Glucose-6-phosphate dehydrogenase deficiency. Bull World Health Organ. 1989;67(6):601-11. PMID: 2633878.
16. U.S. Food and Drug Administration. KRINTAFEL (tafenoquine succinate) tablets — prescribing information. NDA 210795; version effective 2 August 2023. Available via openFDA: https://api.fda.gov/drug/label.json (accessed 6 October 2026).
17. Lee HY, Ithnin A, Azma RZ, Othman A, Salvador A, Cheah FC. Glucose-6-Phosphate Dehydrogenase Deficiency and Neonatal Hyperbilirubinemia: Insights on Pathophysiology, Diagnosis, and Management. Front Pediatr. 2022;10:875877. doi:10.3389/fped.2022.875877. PMID: 35685917.
18. Geck RC, Powell NR, Dunham MJ. Functional interpretation, cataloging, and analysis of 1,341 glucose-6-phosphate dehydrogenase variants. Am J Hum Genet. 2023;110(2):228-39. doi:10.1016/j.ajhg.2023.01.003. PMID: 36681081.
19. Nannelli C, Bosman A, Cunningham J, Dugué PA, Luzzatto L. Genetic variants causing G6PD deficiency: Clinical and biochemical data support new WHO classification. Br J Haematol. 2023;202(5):1024-32. doi:10.1111/bjh.18943. PMID: 37415281.
20. Landeira LFL, da Silva DSP, Velozo CA, Fernandes dos Santos VC, Lima K, et al. Prevalence of G6PD candidate variants in malaria-endemic populations of northern Brazil. Braz J Infect Dis. 2026;30:105893. doi:10.1016/j.bjid.2026.105893. PMID: 42419048.
21. Pfeffer DA, Satyagraha AW, Sadhewa A, Alam MS, Bancone G, Boum Y 2nd, et al. Genetic Variants of Glucose-6-Phosphate Dehydrogenase and Their Associated Enzyme Activity: A Systematic Review and Meta-Analysis. Pathogens. 2022;11(9):1045. doi:10.3390/pathogens11091045. PMID: 36145477.
22. Fastuca A, Vergori A, Robustelli G, Piccolo C, Ragazzo M, Marinoni M, et al. Glucose-6-phosphate dehydrogenase deficiency induced hemolytic anemia and methemoglobinemia: a case report in a 7-year-old female patient. Ital J Pediatr. 2025;51:258. doi:10.1186/s13052-025-02051-2. PMID: 40841696. Figure 1 consulted at: https://pmc.ncbi.nlm.nih.gov/articles/PMC12369025/figure/Fig1/
23. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. p. 288-90 (Prolonged neonatal jaundice and jaundice in infants below 6 months).
24. Leiding JW, Holland SM. Chronic Granulomatous Disease. In: GeneReviews. Seattle: University of Washington; created 9 August 2012, updated 21 April 2022 (review dated 22 January 2026). NBK99496. Available at: https://www.ncbi.nlm.nih.gov/books/NBK99496/
25. Saudubray JM, Baumgartner MR, García-Cazorla Á, Walter J, editors. Inborn Metabolic Diseases: Diagnosis and Treatment. 7th ed. Berlin: Springer; 2022. Chapter 1, section 1.6.11 (ophthalmological signs; table of metabolic cataract causes).
26. U.S. Food and Drug Administration. ELITEK (rasburicase) for injection — prescribing information. BLA 103946; version effective 30 July 2026. Available via openFDA: https://api.fda.gov/drug/label.json (accessed 6 October 2026).
27. U.S. Food and Drug Administration. ARAKODA (tafenoquine) tablets — prescribing information. NDA 210607; version effective 10 January 2025. Available via openFDA: https://api.fda.gov/drug/label.json (accessed 6 October 2026).
28. Youngster I, Arcavi L, Schechmaster R, Akayzen Y, Popliski H, Shimonov J, et al. Medications and glucose-6-phosphate dehydrogenase deficiency: an evidence-based review. Drug Saf. 2010;33(9):713-26. doi:10.2165/11536520-000000000-00000. PMID: 20701405.
29. Pfeffer DA, Ley B, Howes RE, Adu P, Alam MS, Bansil P, et al. Quantification of glucose-6-phosphate dehydrogenase activity by spectrophotometry: A systematic review and meta-analysis. PLoS Med. 2020;17(5):e1003084. doi:10.1371/journal.pmed.1003084. PMID: 32407380.
30. Morris SA, Crews KR, Hayden RT, Takemoto CM, Yang W, Baker DK, et al. Incorporating G6PD genotyping to identify patients with G6PD deficiency. Pharmacogenet Genomics. 2022;32(2):87-93. doi:10.1097/FPC.0000000000000456. PMID: 34693927.
31. Robinson KM, Yang W, Haidar CE, Hankins JS, Jay DW, Kornegay N, et al. Concordance between glucose-6-phosphate dehydrogenase (G6PD) genotype and phenotype and rasburicase use in patients with hematologic malignancies. Pharmacogenomics J. 2019;19(4):305-14. doi:10.1038/s41397-018-0043-3. PMID: 30206300.
32. Lacerda MVG, Llanos-Cuentas A, Krudsood S, Lon C, Saunders DL, Mohammed R, et al. Single-Dose Tafenoquine to Prevent Relapse of Plasmodium vivax Malaria. N Engl J Med. 2019;380(3):215-28. doi:10.1056/NEJMoa1710775. PMID: 30650322.
33. Kalfa TA, Begtrup AH. EPB42-Related Hereditary Spherocytosis. In: GeneReviews. Seattle: University of Washington; created 13 March 2014, updated 7 April 2022. NBK190102. Available at: https://www.ncbi.nlm.nih.gov/books/NBK190102/
34. Saudubray JM, Baumgartner MR, García-Cazorla Á, Walter J, editors. Inborn Metabolic Diseases: Diagnosis and Treatment. 7th ed. Berlin: Springer; 2022. Chapter 32, section 32.3 (diseases with predominant immunological or haematological symptoms; 32.3.7 cytosolic 5'-nucleotidase 3A deficiency).
35. Saudubray JM, Baumgartner MR, García-Cazorla Á, Walter J, editors. Inborn Metabolic Diseases: Diagnosis and Treatment. 7th ed. Berlin: Springer; 2022. Chapter 31, section 31.3 (glutathione metabolism disorders; 31.3.1 and 31.3.6).
36. Tamary H, Greenberg-Kushnir N, Dgany O. Alpha-Thalassemia. In: GeneReviews. Seattle: University of Washington; created 1 November 2005, updated 23 April 2026 (review dated 23 May 2024). NBK1435. Available at: https://www.ncbi.nlm.nih.gov/books/NBK1435/
37. Ribeiro EM, coordinator. Síndromes Genéticas: Reconhecer para Tratar. Fortaleza: EdUnichristus; 2021. p. 97 (section 4.4, Thalassaemias).
38. Sieff C. Diamond-Blackfan Anemia Syndrome. In: GeneReviews. Seattle: University of Washington; created 25 June 2009, updated 31 July 2025 (review dated 23 March 2023). NBK7047. Available at: https://www.ncbi.nlm.nih.gov/books/NBK7047/
39. Recht J, Chansamouth V, White NJ, Ashley EA. Nitrofurantoin and glucose-6-phosphate dehydrogenase deficiency: a safety review. JAC Antimicrob Resist. 2022;4(3):dlac045. doi:10.1093/jacamr/dlac045. PMID: 35529053.
40. Pamba A, Richardson ND, Carter N, Duparc S, Premji Z, Tiono AB, et al. Clinical spectrum and severity of hemolytic anemia in glucose 6-phosphate dehydrogenase-deficient children receiving dapsone. Blood. 2012;120(20):4123-33. doi:10.1182/blood-2012-03-416032. PMID: 22993389.
41. European Medicines Agency. Fasturtec (rasburicase) — EPAR and Summary of Product Characteristics. EMEA/H/C/000331; authorisation issued 23 February 2001; product information updated 12 August 2024. Available at: https://www.ema.europa.eu/en/medicines/human/EPAR/fasturtec (accessed 6 October 2026).
42. U.S. Food and Drug Administration. EXJADE (deferasirox) tablets for oral suspension — prescribing information. NDA 021882; version effective 28 January 2026. Available via openFDA: https://api.fda.gov/drug/label.json (accessed 6 November 2026).
43. U.S. Food and Drug Administration. DESFERAL (deferoxamine mesylate) for injection — prescribing information. NDA 016267; version effective 17 June 2025. Available via openFDA: https://api.fda.gov/drug/label.json (accessed 6 November 2026).
44. U.S. Food and Drug Administration. Drugs@FDA — initial approval dates of original submissions (BLA 103946, BLA 125293, NDA 210607, NDA 210795). Available via openFDA: https://api.fda.gov/drug/drugsfda.json (accessed 6 October 2026).
45. European Commission. Union Register of medicinal products for human use (authorised medicines). Available at: https://ec.europa.eu/health/documents/community-register/html/reg_hum_act.htm (accessed 6 October 2026).
46. U.S. Food and Drug Administration. KRYSTEXXA (pegloticase) injection — prescribing information. BLA 125293; version effective 9 July 2026. Available via openFDA: https://api.fda.gov/drug/label.json (accessed 6 October 2026).
47. European Medicines Agency. Krystexxa (pegloticase) — EPAR; authorisation granted 8 January 2013 and withdrawn by the European Commission on 30 June 2016 at the holder's request. Available at: https://www.ema.europa.eu/en/medicines/human/EPAR/krystexxa (accessed 6 October 2026).
48. Barbosa L, Brito-Sousa JD, Nascimento C, Pacheco A, Alexandre M, Alencar A, et al. Safety and Efficacy of 3 Alternative Regimens Against Relapsing Plasmodium vivax Malaria in Glucose 6-Phosphate Dehydrogenase-Deficient Patients in the Brazilian Amazon (ALTPRIM). Clin Infect Dis. 2025;81:e294-e301. doi:10.1093/cid/ciaf007. PMID: 39785834.
49. ClinicalTrials.gov. Primaquine for Vivax Malaria in G6PD Intermediate and Deficient Cases (NCT07468513). Menzies School of Health Research; registry updated 12 March 2026. Available at: https://clinicaltrials.gov/study/NCT07468513
50. ClinicalTrials.gov. Long-Term Safety Study of Tafenoquine (NCT03320174). 60 Degrees Pharmaceuticals; registry updated 17 April 2026. Available at: https://clinicaltrials.gov/study/NCT03320174
51. ClinicalTrials.gov. High-Dose Vitamin C in G6PDA and Pyruvate Kinase Deficiency: A Safety Study (NCT07612345). University of Utah; registry updated 28 May 2026. Available at: https://clinicaltrials.gov/study/NCT07612345
52. Nascimento JR, Brito-Sousa JD, Almeida ACG, Melo MM, Costa MRF, Barbosa L, et al. Prevalence of glucose 6-phosphate dehydrogenase deficiency in highly malaria-endemic municipalities in the Brazilian Amazon: A region-wide screening study. Lancet Reg Health Am. 2022;12:100273. doi:10.1016/j.lana.2022.100273. PMID: 36776424.
53. GeneReviews. Resources for Genetics Professionals — Genetic Disorders Associated with Founder Variants Common in the Sephardic Jewish Population (appendix; founder variant table, G6PD row). Seattle: University of Washington. Available at: https://www.ncbi.nlm.nih.gov/books/NBK138602/
54. Manco L, Bento C, Relvas L, Maia T, Ribeiro ML. Molecular Heterogeneity of Glucose-6-Phosphate Dehydrogenase (G6PD) Deficiency in the Portuguese Population. Acta Med Port. 2023;36(2):81-7. doi:10.20344/amp.17584. PMID: 36150187.
55. Pereira LLMD, Bravin CA, Cintra TS, Cassa WSP, Santos TA, Fonseca A, et al. Prevalence of G6PD deficiency and molecular characterization of G202A, A376G and C563T polymorphisms in newborns in Southeastern Brazil. Einstein (São Paulo). 2019;17:eAO4436. doi:10.31744/einstein_journal/2019AO4436. PMID: 30673054.
56. National Organization for Rare Disorders. Glucose-6-Phosphate Dehydrogenase Deficiency — Rare Disease Report. Last update indicated: 29 June 2017. Available at: https://rarediseases.org/rare-diseases/glucose-6-phosphate-dehydrogenase-deficiency/ (accessed 6 October 2026).