Primaquine Pharmacokinetics in Lactating Women and Breastfed Infant Exposures

Primaquine was excreted in very low levels in the breast milk of lactating women and caused no adverse effects in their infants. Restrictions on primaquine use in mothers breastfeeding infants at least 28 days old should be lifted.

Primaquine is well tolerated in G6PD-normal children 1 year and older [14], but data are limited for infants. There are no published data on the pharmacokinetics of PQ in lactating women, or on breast milk excretion. As a result, PQ is usually withheld from breastfeeding mothers to avoid the potential risk of iatrogenic hemolysis if the baby is G6PD deficient. Because of this uncertainty, the World Health Organization (WHO) 2015 malaria treatment guidelines recommend delaying radical cure for lactating women until their nursing infants are at least 6 months old and determined to be G6PD normal [1]. At the Thailand-Myanmar border, as in many malaria-endemic locations, extended breastfeeding [15] and G6PD deficiency [16] are common, and G6PD testing is often unavailable. As a result, many women are excluded for years from radical treatment as they continue breastfeeding. Frequently, a subsequent pregnancy begins before breastfeeding ends, eliminating any opportunity for radical cure and jeopardizing the next pregnancy.
More than 20 cases of hemolysis occurring in breastfed G6PDdeficient infants have been reported after maternal ingestion of prooxidant agents [17,18]. While exposure through breast milk seems plausible in some cases, others involve agents that are not convincingly linked to hemolysis in G6PD-deficient individuals [19]. Importantly, reports of episodes occurring in the neonatal period do not account for the baseline incidence of unexplained hemolysis and severe jaundice in neonates with G6PD deficiency [20]. Evidence-based conclusions about the risks to breastfeeding infants should be based upon toxicokinetic studies.
This study was performed to quantify the infant's drug exposure when a breastfeeding woman was treated with a 14-day course of PQ for the radical cure of P. vivax malaria.

Study Setting and Population
This study took place at 3 of the Shoklo Malaria Research Unit (SMRU) clinics along the Thailand-Myanmar border, which serve migrant workers and refugees from Myanmar, predominantly from the Burman or Karen ethnic groups. Antenatal care and delivery services at SMRU clinics are described elsewhere [15]. Vivax malaria has been a common complication in pregnancy in this location and is associated with poor pregnancy outcomes [5,6].

Screening and Eligibility
Between 11 November 2012 and 24 June 2014, women ≥18 years old with a history of P. vivax infection and no prior radical cure PQ treatment who were breastfeeding healthy infants were invited to enroll in the study and receive a standard PQ radical cure regimen to prevent future relapses. Eligible mothers were counseled by trained staff following standard procedures for informed consent, accounting for preferred language and literacy, and women were allowed to give their decision at a subsequent visit, if needed. Primaquine administration was delayed until infants were at least 28 days old.
Complete blood count and G6PD testing were performed on consenting mothers and their babies. Both a rapid G6PD fluorescent spot test (R&D Diagnostic, Greece) and G6PD genotyping by polymerase chain reaction-restriction fragment length polymorphism for the most common local variant (Mahidol) were performed [16], and any abnormality in mother or baby led to exclusion. Maternal screening also included malaria smear, biochemistry, blood group, and hemoglobin typing.

Ethical Considerations
This study was reviewed and approved by 3 ethical review bodies: the Tak Community Advisory Board [21]; the Ethics Committee of the Faculty of Tropical Medicine, Mahidol University in Bangkok (TMEC 12-036); and the Oxford Tropical Research Ethic Committee (OXTREC 28-12).

Drug Administration
Primaquine GPO (Government Pharmaceutical Organization, Thailand) 0.5 mg base/kg was given to nonfasted women once daily for 14 days under directly observed therapy. Study staff saw participants daily while under treatment, recording any adverse events. In 1 case, staff went to the patient's home to ensure daily dosing.

Pharmacokinetic Sampling
Pharmacokinetic (PK) samples included maternal venous and capillary blood, breast milk, and infant capillary blood samples. Frequent PK samples were obtained on day 0 and day 13. On day 3 and day 7, sparse samples were taken (Supplementary Table 1). All blood samples were centrifuged on site and the plasma aliquots frozen immediately. Breast milk samples were obtained by manual expression. After measuring the total volume of expressed milk and mixing, a 2-mL aliquot was taken and frozen immediately. Samples were stored initially at -20°C and moved on dry ice to -80°C.

Drug Level Measurement in Plasma and Breast Milk
Primaquine and carboxyprimaquine (CPQ) concentrations were quantified using a validated assay according to US Food and Drug Administration (FDA) guidelines (unpublished data). In brief, samples were processed by protein precipitation followed by phospholipid removal solid-phase extraction. Analytes were separated using reverse-phase high-performance liquid chromatography and quantified by electrospray ionization in the positive mode with multireaction monitoring mass spectrometry detection. The limits of quantification were 1.14 ng/mL for PQ and 4.88 ng/mL for CPQ. Three replicates of quality control samples at low, middle, and high concentrations were analyzed within each batch of clinical samples to ensure precision and accuracy during drug measurements. Total precision (ie, relative standard deviation) for all drug measurements was <10% during drug quantification of clinical samples.

Pharmacokinetic Analysis
Individual drug concentrations on days 0 and 13 were analyzed using a noncompartmental analysis in Phoenix WinNonlin build 7.0.0.2535 (Certara, Princeton, New Jersey). The maximum concentration (C max ) and the time to maximum concentration were derived from the observed PQ and CPQ concentrations. The observed data were used to calculate the exposure (area under the concentration-time curve [AUC]) from drug administration to the last observation and to infinity (AUC ∞ , only for PQ) based on the linear trapezoid method during absorption and logarithmic trapezoid method after C max . For PQ in venous plasma, oral elimination clearance and apparent oral volume of distribution were calculated based on the calculated AUC ∞ and administered dose.
Hypothetical infant doses were calculated using 2 different approaches: 1. The dose was calculated based on breast milk PQ concentrations at the noted feeding and sampling occasion (measured in hours since PQ dose) multiplied by an estimated feeding volume of 33 mL/hour [22] since last noted concentration measurement.
One hundred fifty milliliters per kilogram per day is an estimate of infant breast milk consumption recommended as an industry standard [23]. Weight-adjusted relative infant dose was therefore:

Laboratory Monitoring
A Hawksley microhematocrit reader was used to measure hematocrit levels. Infant hematocrits were measured with each capillary PK sample, resulting in 2-4 hematocrit values on each sampling day. The mean of these 2-4 measurements for each infant on each day was used in the analysis. Heinz body counts and serum bilirubin measurements were also measured on each sampling day. Serum haptoglobin concentration on day 3 was measured by nephelometry. Methemoglobin was assessed using a transcutaneous probe (Masimo Radical-7). Heinz bodies per 1000 erythrocytes were assessed by microscopy after staining with 10% crystal violet. Serum bilirubin was measured using Bilimeter3 (Pfaff, Germany).
Infant study hematocrits were compared to longitudinal hematocrit data of 23 infants 7-10 weeks of age from this population.

Statistical Analysis
Data were analyzed using Stata version 13.1 (StataCorp, College Station, Texas). Continuous normally distributed data were compared using the Student t test, and for nonnormally distributed data the Mann-Whitney U test was used. Wilcoxon signedrank test compared paired nonparametric values (eg, day 0 vs day 13 hematocrit).

RESULTS
Primaquine was well tolerated, and there were no obvious adverse effects in mothers or their infants.

Participants
Twenty-one G6PD-normal mothers completed at least 1 day of pharmacokinetic sampling (Table 1), and 19 (90%) completed the full study according to the protocol ( Figure 1; Supplementary Table 2). One woman withdrew consent for further sampling on day 2 because of pressure from family members to return home, but agreed to have previously collected (day 0) samples analyzed. She completed the course of PQ without complications. One participant was incorrectly given 1 mg/kg PQ for the first 2 days. She was excluded from day 0 PK analysis, but included for safety assessments (which were unremarkable) and PK analysis from day 3 onward.

Pharmacokinetic Data (Tables 2 and 3, Figures 2 and 3)
All infant capillary PQ concentrations were below the limit of quantification (LLOQ; 1.14 ng/mL or 2.28 ng/mL depending on the extracted volume) except for 1 sample  Figures 2 and 3). Peak breast milk PQ concentrations were reached approximately an hour after peak plasma concentrations. The median total cumulative PQ dose expected to be ingested by the infant over the 14-day course based on measured breast milk concentrations was approximately 0.042 mg/kg, corresponding to 2.98 µg/kg/day (0.6% of a hypothetical infant daily dose of 0.5 mg/kg). The highest cumulative infant dose in this cohort was estimated at 0.127 mg/kg, or 9.07 µg/kg/day (1.8% of a hypothetical infant daily dose of 0.5 mg/kg) ( Table 3).

Safety Monitoring and Adverse Events in Infants
All infant hematocrits were within local population norms, and there was no symptomatic or severe anemia in the study infants.
To assess change in hematocrit in infants of different ages over the period of possible PQ exposure, groups were formed based on age at enrollment: >56 days (n = 18) or ≤56 days (n = 2). This was done to account for the physiologic decline in hematocrit during the first months of life, which reaches its nadir typically between 8 and 12 weeks [24,25] (Supplementary Figure 1B)   change per week of 23 paired hematocrits from prospectively followed infants from the local population aged 7-10 weeks, and was within the interquartile ranges of this healthy cohort (mean, -2.8% [95% CI, -5.5% to .0]; interquartile range, -8.1% to 1.4%).
Haptoglobin levels were available for 14 of 20 infants (6 samples were rejected because of insufficient volume). Normal range for haptoglobin in infants is not well described: usually undetectable at birth, adult values of 0.3-2.0 g/L are reached by 4-6 months [26]. The haptoglobin of 1 infant enrolled at 48 days old was 0.1 g/L. Values in the other 13 infants measured fell in the normal adult range.
There was no correlation in either age group between PQ concentration in the breast milk and hematocrit change. Infant oxygen saturations, methemoglobin, serum bilirubin, and Heinz body counts also showed no significant changes over the study period. One infant >56 days old at enrollment had an asymptomatic abnormal metHb value of 4.4% on day 7, which resolved spontaneously while the mother took the remaining week's treatment.
No other relevant adverse events or breastfeeding problems were reported.

DISCUSSION
Most malaria treatment guidelines recommend that PQ be withheld from breastfeeding mothers lest their infants develop iatrogenic hemolytic anemia, although the amount of PQ excreted in breast milk has not been quantified previously. The unmeasurably low levels of PQ in all but 1 infant plasma sample in this study and negligible metabolite (CPQ) concentrations are very reassuring as they indicate very low exposure to maternal PQ. The very low breast milk PQ levels provided an estimated total weight-adjusted dose to the infant of <1% of the recommended adult and pediatric therapeutic dose. There is no evidence that PQ is more toxic to infants than adults (though information on the safety of PQ in infants is limited). Recent studies have shown that single PQ doses of 0.25 mg/kg do not cause clinically significant hemolysis in G6PD-deficient individuals [27]. It is highly improbable therefore that a total dose 6 times lower (0.042 mg/kg) spread over 14 days would cause hemolysis, even in G6PD-deficient infants.
None of the women screened for inclusion in this study had received previous radical cure. As a result, recurrent vivax malaria infection was common. This highlights the unmet need for PQ in breastfeeding women. Furthermore, 2 screened women, breastfeeding a 7-month-old and an 11-month-old, were excluded because of positive pregnancy tests, reflecting short interpregnancy intervals and repeat pregnancy before the end of lactation. Even with ideal interpregnancy intervals of 18-24 months [35], women following WHO recommendations to breastfeed for the first 2 years of life [36] would be ineligible for PQ throughout their reproductive life if treatment is withheld during lactation due to unknown or abnormal infant G6PD status. There were no differences in absolute hematocrit or rate of change of hematocrit over time between infants enrolled in this study and age-matched infants from the local population. Lack of evidence of PQ toxicity was supported by evaluation of Heinz bodies, serum bilirubin, haptoglobin, and methemoglobin.
A limitation of this study is the approximate nature of breast milk volume estimation. Although FDA documents recommend collection of all breast milk from both breasts throughout the study period [23], this is burdensome for study participants, and pumped volumes of breast milk do not correlate well with intake by breastfeeding infants [22]. We asked women to empty 1 breast manually while feeding their infant on the other during preset sampling windows, and used published norms to calculate breast milk consumption. Expressed volumes varied from 5 mL to 101 mL, and differences in fat, protein, and oligosaccharide composition are inevitable. Analysis of foremilk vs hindmilk, or creamatocrit could give more detailed information. However, as the drug levels in all breast milk samples were low, and there was no association between drug concentration and volume of breast milk expressed or age of infant, these additional variables are unlikely to have clinical significance.
Further studies should test the predicted safety for G6PDdeficient infants suggested here by monitoring breastfed infants with G6PD deficiency during maternal PQ treatment. These reassuring results cannot be extrapolated to the neonatal period as breast milk during the first days after delivery differs significantly from mature milk. Studies of PQ treatment and colostrum PQ concentrations immediately postpartum, when many women have contact with healthcare providers, would be valuable.
New guidelines should allow PQ use in women breastfeeding infants at least 28 days old, and radical cure should be a routine part of 4-to 8-week postpartum care for women with a history of P. vivax malaria. This would reduce the risk of recurrent malaria in subsequent pregnancies, reduce morbidity and mortality due to malaria-related anemia, and help curb malaria transmission globally.