Methemoglobin and nitric oxide therapy in Ugandan children hospitalized for febrile illness: results from a prospective cohort study and randomized double-blind placebo-controlled trial

In this study, we examined levels of MHb in two cohorts of children presenting to a regional pediatric referral hospital in Eastern Uganda. Although these studies represent distinct patient populations (there was no patient overlap between studies), the subjects were enrolled over the same 2 year period from the same catchment area. Overall, the children enrolled in the fever cohort were younger (according to the study design) and included all causes of fever, whereas all children in the iNO RCT had a diagnosis of severe malaria.

Among children enrolled in the fever cohort, there were no differences in MHb levels based on demographic characteristics (age, sex, or nutritional status). Infants less than 6 months represent a vulnerable group for methemoglobinemia for a number of reasons, including: a higher pH in the stomach that permits the growth of nitrate-reducing organisms (e.g. Escherichia coli, Salmonella spp.); immature NADH-methemoglobin reductase systems with reduced capacity to cope with oxidative stress (levels at birth are only 50–60 % of adult levels) [18, 25, 26]; the presence of fetal hemoglobin which is more readily oxidized to MHb than adult Hb [18]; and a higher consumption of water per unit body weight, which renders them susceptible to methemoglobinemia if exposed to nitrates through drinking water [27]. Despite these known risk factors in young children, we did not observe an age-dependent effect on MHb levels in our population of children hospitalized with fever.

Given the high rate of methemoglobinemia in our cohort (34 % of children in the fever cohort had MHb >2 %), Ugandan children may be at higher risk for methemoglobinemia than populations from high-resource settings as their environmental exposures may result in a higher set-point for MHb. A common cause of MHb in children is ingestion or contact with direct or indirect oxidizing agents (e.g. benzocaine, chloroquine, primaquine, sulfonamides, nitrites/nitrates, dapsone). As a result, we explored whether there was an association between known drug exposures (e.g. pretreatments with antibiotics and/or anti-malarials) and MHb levels. There was a non-significant increase in the odds of elevated methemoglobinemia in children that received pretreatment with any antibiotic (Table 2, p = 0.079). Due to the variability in antibiotic prescription and limitations in parental recall, we were limited in our ability to explore relationships between specific classes of drugs and MHb levels. Furthermore, we were unable to assess environmental exposure to other oxidizing agents in in this population (e.g., nitrates in water, or smoke inhalation through indoor biomass fuel use). However, a study evaluating nitrate levels in spring water from central Uganda reported 60 % (52/80) of water samples had nitrate levels exceeding the WHO maximum permissible levels [28]. In addition, widespread reliance on biomass fuel use (e.g., wood) in cooking stoves or open fires contributes to high levels of indoor air pollution [29]. It is estimated that 78 % of the Ugandan population resides in rural areas where 86 % use wood for cooking [29, 30]. Based on these findings, it is likely that environmental exposure to oxidants in our population exceeds WHO recommended levels.

We explored the association between clinical signs and symptoms at hospital presentation and MHb levels, and observed two general trends. First, children with impaired perfusion and acidosis (vomiting, delayed capillary refill, and elevated lactate) had significantly higher odds of methemoglobinemia after controlling for a set of potential confounders. These findings are consistent with previous studies in infants where methemoglobinemia was reported in the context of metabolic acidosis secondary to diarrhea and dehydration [31, 32]. In our cohort, we observed increased odds of methemoglobinemia associated with vomiting, but not diarrhea. Children with fever and diarrhea alone were excluded from the study. The prevalence of P. falciparum parasitemia in this cohort was high at 67 %. Therefore, the acidosis observed in this study may be attributable to malaria rather than dehydration and diarrhea, as metabolic acidosis is a common complication of malaria [33‐35]. These data suggest that in conditions of increased inflammation, oxidative stress and acidosis, impaired reduction or reconversion of MHb to Hb may contribute to the elevated MHb levels observed in our cohort [18, 36, 37].

Methemoglobinemia was also seen in circumstances where red blood cells (RBCs) are affected: severe anemia, children with suspected or documented hemoglobinopathies (i.e. sickle cell disease or G6PD deficiency), and malaria. RBCs are particularly susceptible to oxidative damage as they carry oxygen in high concentrations and are continuously exposed to oxygen free radicals. As RBCs lack a nucleus, they are dependent on endogenous reduction systems that can degrade with repeated exposure to oxidants or RBC senescence [38]. Recent estimates of G6PD polymorphisms in Uganda show 20 % of the population carry the G6PD A-mutation [39], which are consistent with the range of estimates 15–32 % described elsewhere in Africa [40‐42]. With increased oxidative stress on RBCs in G6PD deficiency, sickle cell disease, and other RBC polymorphisms, the capacity of endogenous reduction systems may be overwhelmed leading to increased MHb. In this population, rates of malaria were high with roughly two thirds of children admitted to hospital with fever having parasitologic evidence of malaria infection. Because quantitative estimates of malaria burden (either parasitemia or plasma HRP2 antigen levels) were not available in this cohort, it is difficult to estimate the fraction of fevers in children hospitalized attributable to malaria. Regardless, malaria was independently associated with increased odds of methemoglobinemia (OR (95 % CI), 1.34 (1.07, 1.69), p = 0.011), consistent with previous reports of methemoglobinemia in malaria [43‐46]. Malaria is associated with increased oxidative stress from malaria-heme products and immune cell derived reactive oxygen species, both of which could promote oxidation of Hb to MHb [47, 48]. Finally, as children with severe anemia possess compromised oxygen carrying capacity, increased levels of MHb in the context of severe anemia may exacerbate reduced oxygen delivery resulting in metabolic acidosis and functional impairments in MHb reduction. Although elevated MHb was more common in non-survivors compared to survivors, it is likely that elevated MHb is a consequence of oxidative stress and acidosis in severe disease rather than mediating severe disease. However, in children with potentially symptomatic levels of MHb (e.g. the 3 % of children with MHb >10 % at admission in the fever cohort), MHb could exacerbate underlying disease processes and treatment may be warranted.

In the context of the clinical trial, 11 % of children assessed for eligibility were excluded for methemoglobinemia, which is considerably less than the fever cohort (34 %). The lower prevalence of methemoglobinemia in the iNO RCT may be due to a number of factors, including exclusion of children with known chronic illness (i.e., hemoglobinopathy). Furthermore, children otherwise eligible for the clinical trial were only excluded for methemoglobinemia if their MHb levels remained ≥2 % following stabilization (which included administration of fluids, transfusion in cases of severe anemia, and dextrose to treat/prevent hypoglycemia). As catabolism of sugars through glycolysis is a major source of substrate for the NADH-cytochrome-b ₅ reductase system, glucose levels must be in adequate supply for endogenous MHb reducing systems to respond [18]. Therefore, these stabilization measures may have contributed to a reduction in MHb levels in children allowing them to meet the eligibility criteria for the trial.

We were unable to evaluate the dose-dependent effect of iNO on MHb levels as all children randomized to receive iNO were started at 80 ppm. However, this is the largest trial to date to administer iNO at the maximum approved dose and we were able to evaluate the variability in MHb responses within subjects and the frequency of methemoglobinemia prompting study gas discontinuation. Despite the high doses of iNO administered, study gas was temporarily discontinued only five times for MHb >10 % (all children in the iNO group). We were able to re-start study gas for all children that had a MHb measurement that exceeded 10 % once the MHb returned to <7 % without having the MHb exceed 10 % again. It was not necessary to wean children off iNO, in contrast to studies administering iNO to neonates with hypoxic respiratory failure, as we did not observe any rebound effects (e.g. worsening oxygenation) following discontinuation of study gas [23, 49]. Overall, four hourly MHb checks were sufficient for monitoring iNO administration, with more frequent checks implemented in children when MHb levels approached 7 % so appropriate measures could be taken if levels exceeded 7 or 10 % (e.g. titrate or temporarily discontinue study gas). As seen in Fig. 3b, MHb levels fluctuated considerably within subjects over hospitalization in both trial arms. It is not clear whether these fluctuations were due to natural variations/regulatory responses in the endogenous reduction systems or were related to MHb measurement using the pulse co-oximeter. However, performance of non-invasive pulse co-oximetry has been previously compared to whole blood co-oximetry in children with sickle cell disease and showed acceptable clinical accuracy (bias of −0.22 % for MHb) [50]. The variations in MHb levels over hospitalization highlight the importance of frequent MHb monitoring during administration of nitric oxide.

When looking at illustrative graphs of MHb kinetics in children with and without study gas interruptions, the variability in responses is apparent (Fig. 1). We specifically included a panel of children who had gas permanently discontinued for acute kidney injury, as administration of iNO has been associated with a statistically elevated risk of developing renal dysfunction in critically ill adults [51]; however, the same association has not been observed in pediatric populations. In a recent retrospective analysis of acute kidney injury in this cohort, we found iNO was associated with an increased risk of acute kidney injury compared to placebo with a relative risk of 1.36 (95 % CI, 1.03-1.80), p=0.026. We did not observe differences in MHb levels in children who had gas discontinued for acute kidney injury compared to other children. Apart from the one child in the placebo arm of the trial who had increasing MHb levels prior to death, we did not observe elevated MHb among trial participants who died. However, the majority of participants died early in illness and there were limited kinetic data available. All children except one (Fig. 2l) died in the iNO arm died before repeated MHb measurements were taken.

After a decade of use in clinical practice, iNO has a well-established safety profile. In this study, we administered iNO in a low-resource setting in a non-intensive care setting with limited laboratory support. Using a commercially available handheld pulse co-oximeter, we determined the range of MHb levels for children admitted to hospital with fever in this population. Although MHb levels >2 % were observed in 34 % of children admitted with fever, methemoglobinemia was grounds for study exclusion in the iNO RCT for only 11 % of children following stabilization and repeat MHb assessment. In addition, we administered iNO at the highest approved dose in 87 patients with severe malaria and had to temporarily discontinue study gas for only a fraction of children (5.7 %) with MHb >10 %. Overall, the rates of study gas withdrawal were not different between the placebo and trial arm.

Our study strengths include two integrated studies encompassing a single catchment area: a large prospective observational study to describe MHb levels in children hospitalized with fever, and an intervention where a potent Hb oxidant was administered and MHb levels assessed. The clinical trial was randomized and double blinded using separate teams to monitor clinical care and study gas administration to ensure clinical decisions to withdraw gas were not affected by intervention. The study gas team used pre-set standard operating procedures to guide decisions regarding titration or withdrawal of study gas. Although we were limited in our ability to report dose-dependent effects of iNO on MHb levels, our study is the largest to administer iNO at the maximum approved dose and serves as an important addition to existing literature.