Pathophysiology

Ammonia is a metabolic by-product of all cells. The major contributors to the body load are the gastrointestinal (GI) tract, skeletal muscle and kidney. Ammonia is produced in the GI tract from three sources: bacterial urease breakdown of circulating urea, bacterial deamination of luminal protein and glutaminase breakdown of glutamine from circulating glutamine.9 Large amounts of ammonia are generated in skeletal muscle, especially during activity/seizures mainly from deamidation of AMP, and to a lesser extent from amino acid catabolism. In kidneys it is released from glutamine entering the proximal tubular cells from the glomerular filtrate and the circulation.

Because of its neurotoxicity, arterial blood concentration is normally tightly regulated and does not vary much even after meals with the main sites of production being the GI tract, skeletal muscle and the kidney. At physiological pH in body fluids, ammonia (NH₃) is predominantly hydrogenated to ammonium (NH₄ ⁺). At neutral pH NH₄ ⁺ predominates, which is useful as this limits its perfusion across cell membranes. Throughout the text 'ammonia' refers to summated NH₃ and NH₄ ⁺ concentrations, which is the common convention.

Ammonia is toxic to the brain but not to other tissues and readily crosses the blood-brain barrier.10 Average levels of brain ammonia are normally approximately twice those of blood. In experimental animals with acute liver failure, brain ammonia flux may be up to 45-fold higher than normal.11

A number of theories have been proposed to explain HA-induced brain damage.12–18 It has been postulated that increased intracerebral ammonia levels may interfere with mitochondrial function, may disrupt inhibitory and excitatory neurotransmission, and lead to excessive glutamine glial accumulation leading to astrocyte swelling. One of the most concerning outcomes of this pathology is that severe HA can cause cerebral oedema with a rise in intracranial pressure that can lead to herniation and brainstem compression.

Large amounts of NH₃ are produced daily. NH₃ not recycled for biosynthetic processes must be excreted rapidly from the body. This is achieved by conversion to urea via the urea cycle in the liver, which is then excreted by the kidneys. The conversion of ammonia into urea requires the function of the six enzymes and two transporters of the urea cycle. The liver is the only organ that houses the complete urea cycle of 5five enzymes (two mitochondrial, three cytosolic), together with N-acetylglutamate synthetase (NAGS) which produces N-acetylglutamate (NAG). NAG acts as an activator of carbamoylphosphate synthetase (CPS1) (see table 1and figure 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Urea cycle diagram including normal enzymes/transporters and where each of the drugs is purported to block. Asparaginase catalyses the formation of aspartic acid and ammonia from asparagine. ADP, adenosine diphosphate; AMP, adenosine monophosphate; ATP, adenosine triphosphate; ARG1, arginase 1; ASL, argininosuccinate lyase; ASS, argininosuccinatesynthetase; CA, carbonic anhydrase; CoASH, coenzme A; citrin, mitochondrial aspartate/glutamate carrier; CPS1, carbamoylphosphate synthetase 1; CO2, carbon dioxide; H2O, water; HCO3, bicarbonate; NAGS, N-acetylglutamate synthase; ORNT1, ornithine/citrulline antiporter; OTC, ornithine transcarbamylase; Pi, phosphate; PPi, pyrophosphate; TPM, topiramate; TCA, tricarboxylic acid cycle; VPA, valproic acid; 5-FU, 5-fluorouracil; X, denotes the site of medication inhibition.

In order to transfer NH₃ safely from the tissues to the liver via the circulation, it is combined with glutamate by glutamine synthetase to produce non-toxic glutamine. Some is extracted by the kidneys and the immune system for biosynthesis. Most is taken up by the small intestinal mucosa. Here NH₃ is released by glutaminase and transported directly to the liver in the portal circulation.9

Cytosolic ornithine is transported into the mitochondria in exchange for intramitochondrial citrulline by the ornithine/citrulline carrier encoded by the SLC25A15 gene. Citrin, a liver transporter encoded by the SLC25A13 gene, exports aspartate from the mitochondria in exchange for glutamate that is used in the formation of NAG (figure 1). Two atoms of nitrogen are converted to urea for each turn of the urea cycle. One comes from ammonia and the other comes via the citrin carrier and the amino acid pool in the form of aspartate. The latter derives from the amino group of alanine, which is transferred to 2-oxoglutarate to form glutamate initially and then onto oxaloacetate to produce aspartate. NAGS produces NAG which is an essential cofactor for CPSI, the first and rate-limiting enzyme of the urea cycle.19 20

Drug-induced HA

A number of medications have been associated with HA, with valproate being the most common. Confounding factors such as sepsis and liver disease, and lack of clear mechanisms for some drugs weaken the evidence for association.

In a search of the WHO adverse drug reaction database (Vigibase), for the period 1967 to 8 May 2019, 73 drugs were reported to be suspected of causing HA. table 3 lists those most frequently recorded; 63% were associated with valproate, 11% with fluorouracil, 5% topiramate and the rest accounting less than 5% each of overall cases.21 Vigibase is a retrospective observational database used to record all adverse drug reactions from over 130 countries, and the table below lists the most frequently recorded.

The following were excluded: drugs reported less than three times and drugs used to treat HA or hepatic encephalopathy. There are some obvious limitations to using this as a source as it relies on spontaneous notification, retrospectively, so delay may affect recall, incomplete information, lack of direct inspection of laboratory data and undernotification. For instance, the case numbers with liver failure are low compared with the estimated population prevalence and this may represent the fact that clinicians attributed the HA (in those cases excluded) to liver disease rather than to a particular drug.

Part of the challenge in dealing with such HA is teasing out confounders such as liver disease, sepsis, malnutrition, IMD or a drug effect. Although for the drugs we have tabulated in table 3 there is some mechanistic linking the drug to HA, for some medications this is not the case. In addition, this is hampered by the absence of a specific test to rule in or rule out the drug-associated HA. To strengthen the evidence for association, Weiss et al extended their Vigibase study by checking potential causation checking time of onset of HA to discontinuation and whether this correlated with the product characteristics documented from the European Medicines Agency and the Federal Drug Authority.21 22 Of interest is the fact that 74% of the cases were reported in the last decade of the search period; a median time of onset of HA following drug commencement was 13 days (IQR 2–59 days).22 Of the 73 drugs, 10 described the association with HA in the drug product characteristics (valproic acid (VPA), valpromide (VPA prodrug), topiramate, asparaginase, fluoruracil, haloperidol, pegaspargase, zonisamide, deferasirox, amphotericin B). table 3 lists the most frequently recorded drugs associated with HA. The frequency of HA was uncommon for asparaginase (between 1/100 and 1/1000 cases), rare (between 1/1000 and 1/10000) for VPA, valpromide, topiramate and undetermined for the other six drugs. Of the remaining 61 drugs, 6 were published in reference databases/books, 45 were described in case reports/series and 10 had never been published. This might reflect greater awareness of HA as a drug complication and/or a higher incidence with introduction of new drugs (see reference22 for a comprehensive list).

Valproate

VPA (2-n-propylpentanoic acid) is a C8-branch chain fatty acid that has been used for many years to treat epilepsy, more recently as mood stabiliser for a number of psychiatric conditions and migraine. In a recent audit of general hospitalised adults who received VPA, 20.4% developed HA.23 This audit included those patients admitted over a 1-year period in one hospital, age >18 years, receiving at least one dose of VPA. Patients with cirrhosis were excluded from the audit. HA was defined as above the institutional reference range of >23.6 µmol/L, 162 patients were included and ammonia ranged from 12 µmol/L to 75 µmol/L. Epilepsy and psychiatric condition were the most frequent indication for VPA and the mean plasma ammonia in symptomatic patients was 39.5 µmol/L. A limit for this paper is that it was retrospective, duration of VPA use was not recorded and neither was dose. However, the strengths of the study include the large sample size and general hospital location, rather than a specific medical discipline.

A similar study conducted on patients admitted to a psychiatric medicine unit had a prevalence of HA of 36%.24 Symptoms of VPA-induced HA can vary from asymptomatic HA to VPA-induced hyperammonaemic life-threatening encephalopathy.25

Various mechanisms have been implicated in the development of VPA-induced HA, mainly via effects on the liver and less so on the kidney. VPA β-oxidation leads to valproyl-CoA (valproyl Coenzyme A) which directly inhibits NAGS and induces secondary carnitine deficiency which leads to decreased acetyl-CoA that may also contribute to VPA-induced HA26 27 (figure 1). Hepatotoxicity (non-specific variable) may accompany this and a dose-dependent effect on plasma ammonia has been observed with increasing VPA doses and during use with other antiepileptic drugs (AEDs). Elevations in plasma ammonia levels, as high as 140 µmol/L, were well tolerated, and VPA dose reductions were not necessary.28 However, in an audit of AED use, in the VPA only group, ammonia concentration ranges from 60 µmol/L to 400 µmol/L were recorded, with some being well above the treatment threshold (see below for treatment section).29 Risk factors for hepatotoxicity associated with VPA include young age, polytherapy and developmental delay.30 Although rare, coexisting UCD suggested by a sudden development of hyperammonaemic crises should be considered. Hyperammonaemic crises maybe triggered by catabolic events, protein overload, infection, GI bleeding or certain drugs.31 32 This needs to be considered when requesting relevant laboratory tests (see below laboratory investigations).

Topiramate

Topiramate (TPM) is a carbonic anhydrase inhibitor used to treat epilepsy and alcohol dependence, and prevent migraines. HA is an uncommon side effect of topiramate and is usually related to combined use with valproate and phenobarbital.33 It is believed that the mechanism implicated in HA involves inhibition of carbonic anhydrase leading to reduced bicarbonate and reduction in activity of the cytosolic enzyme, glutamine synthetase. Reduced bicarbonate may decrease activity of CPS1, the rate-limiting enzyme in the urea cycle (figure 1).34 The reduction in glutamine synthetase levels limits the incorporation of ammonia into glutamate, to form glutamine. Both of these mechanisms can lead to HA; however, it should be noted that both of these studies were in animal models and there are no human studies to date. Values of ammonia of up to 77 µmol/L in a single use of topiramate have been recorded and up to 146 µmol/L in combination with other AEDs as noted above.33–35 Addition of VPA to a patient already taking topiramate may increase the risk of HA.36

Fluorouracil

5-fluorouracil (5-FU) is a common chemotherapy drug used in the treatment of head and neck, GI, and breast cancers. 5-FU is a pyrimidine analogue that acts principally as a thymidylate synthase inhibitor. After entering the cell, 5-FU is converted into several metabolites. The rate-limiting step of the catabolic process is the conversion of 5-FU to the inactive metabolite dihydrofluorouracil which is catalysed by the enzyme dihydropyrimidine dehydrogenase (DPD).37 Decreased metabolism by DPD creates a number of potentially toxic metabolites and hence it is a standard practice to screen for DPD variants (associated with decreased metabolism) that increase this risk, in particular of severe neutropaenia.38 To date there has been some suggestion that such variants in association with other factors contribute to HA. Boilère et al found an association with DPD deficiency in 27% and Milano et al found an association with neurotoxicity in 37%, but did not record ammonia values.39 40

HA has been reported after treatment with 5-FU, with early symptoms being non-specific, such as nausea and vomiting progressing to encephalopathy.41 Large doses of 5-FU induce accumulation of fluoroacetate which can inhibit the enzyme aconitase and result in the impediment of citrate isomerisation. This in turn impairs Krebs cycle function reducing ATP38 and oxaloacetate.42 Impairment of ATP-dependent urea cycle results in HA. These are two essential factors in the early proximal part of the urea cycle, with oxaloacetate providing the source of aspartate via aspartate transaminase (figure 1).

The reported incidence of encephalopathy has been around for 10% with HA being around 1.0%.40 43 Values of ammonia up to 522 µmol/L have been recorded, which is well above the therapeutic intervention threshold (see the Treatment section below).40 43 However, it should be borne in mind that often these patients may have multiple confounding factors such as sepsis, shock, sarcopenia and hepatic impairment, that need to be considered. The original reports described an acute cerebellar syndrome associated with HA and since then several case studies and cohort studies have been published.44–47

For clinicians treating such patients the challenge is teasing out often what may not be multifactorial encephalopathy, due to the cancer, medications or malnutrition. A paraneoplastic encephalopathy is more likely to be subacute, however 5FU also interferes with the conversion of thiamine-to-thiamine pyrophosphate (its active form) which can present as Wernicke’s encephalopathy. This is compounded by the fact that such a group is nutritionally vulnerable and may already be deficient in thiamine, among other vitamins.48

Oxaliplatin

Oxaliplatin a member of the platinum chemotherapy class, undergoes non-enzymatic conversion to active derivatives which preferentially bind guanine and cytosine bases in DNA leading to cross linking. This arrests DNA synthesis, with effects not being dependent on cell cycle, being particularly efficient on tumours with high cell turnover.49 50 It can be used to treat a variety of metastatic tumours but is commonly in combination with 5-FU to treat colorectal tumours. The drug causes dose-dependent toxicity on the haematopoietic system and dose-limiting effects on the nervous system causing acute or chronic peripheral neuropathy affecting multiple types of nerve fibres.51

HA has been uncommonly described, usually in combination with 5-FU and hepatic dysfunction, therefore may have synergistic effects, on the urea cycle and might increase the risk for HA.52 53 The exact mechanism however remains unclear but may be related to negative effects on the Krebs cycle impacting the proximal urea cycle. Values of ammonia of up to 200 µmol/L have been recorded comfortably above the treatment threshold (see the Treatment section).54 55

Asparaginase

Asparaginase has been a long-standing component of induction, consolidation and maintenance therapy for acute lymphoblastic leukaemia.56 57 The principal of this chemotherapy is that neoplastic cells have reduced capacity to synthesise asparagine and hence need this from the blood.

Asparaginase catalyses the formation of aspartic acid and ammonia from asparagine and hence an increase in ammonia is expected as a consequence of its therapeutic effect. In a prospective cohort study, although 7/10 had HA as expected none were symptomatic and although peak levels increased up to seven times upper-normal limit 24 hours after dose, concentrations fell to normal by 2–3 days.58

Symptomatic HA has previously been reported to be uncommon, however in a study where ammonia was checked regardless of symptoms, ammonia >100 µmol/L was found in 7 out of 10 patients being even higher in those where PEG-asparaginase was used, due to the prolonged half-life.59 It is clear therefore that ammonia should be checked as a regular part of therapeutic management regardless of symptoms, otherwise this may be missed. Values of ammonia up to 400µmol/L have been documented in setting of asparaginase use, well above the therapeutic threshold for intervention for HA (see the Treatment section). Other metabolic effects such as hypertriglyceridaemia due to antagonistic effect on lipoprotein lipase have also been described.

Other medications

Use of corticosteroids has been associated with increased skeletal muscle amino acid metabolism and thus has been believed to be due to increased ammonia production.60 However, decrease in glutamine synthetase levels, carbamoyl-phosphate synthase 1, ornithine transcarbamylase (OTC), arginosuccinate synthase 1 and arginosuccinate lyase have also been implicated.61 62 Imoto et al also looked at the case series of use of corticosteroids in those with OTC deficiency and found a high rate of mortality with mean ammonia 761 µmol/L (ranging from 233 μmol/L to 3039 μmol/L), well above the intervention threshold (see table 5).62 Due to the effect on a number of key UCD enzymes, the authors recommend early intervention with renal replacement therapy and certainly close monitoring of ammonia in this group post corticosteroid intervention.

Of historical interest is Reye’s syndrome, a viral (varicella and influenza most common) induced acute liver failure due to mitochondrial injury; this was associated with high morbidity and mortality.63 The risk was reported to be increased in those under 12 years when aspirin was used to control fever, which led in turn to recommendation not to use aspirin in those age groups during viral infection. Reye’s syndrome peaked in the late 1970s, early 1980s and is now quite rare.64 It is best viewed as multifactorial, including viruses, however some cases also had an underlying IMD, mainly UCD or fatty acid disorder of oxidation.65

Combination therapy of VPA plus other AEDs (such as phenytoin, phenobarbital, carbamazepine and/or topiramate, zonisamide) has also been shown to be associated with increased risk of HA.55 56 While the mechanisms for HA are clear in some, others remain uncertain and so clinicians should be aware of the increased risk of HA in combination therapy particularly in those with symptoms associated with HA.58 59

Clinical presentation

HA can present with a number of non-specific features with guidelines suggesting that checking of ammonia should be considered in any patient presenting with acute/intermittent neurological/psychiatric presentation, acute liver failure or in the differential of sepsis with a view to making a UCD diagnosis.66 Most patients present after a catabolic trigger, such as intercurrent infection, postpartum, vomiting and diarrhoea, relevant medication or a high protein meal.

In all ages, loss of appetite and vomiting are common, but in adolescents/adults the encephalopathy, hallucinations or psychiatric symptoms or signs of tremor, seizures predominate.66 Often these clinical features are mistaken for meningitis, brain tumour or intoxication, hence the heightened testing for checking ammonia in such clinical scenarios to avoid delay in diagnosis and adverse clinical outcomes. Around 50% of those over 16 years/adults with UCD can present acutely and mortality of up to 10% has been noted with approximately 90% requiring intensive care admission.7 67

Laboratory investigations

Abnormal ammonia levels in adults should trigger further investigation with a particular focus on liver disease or a treatable IMD. Early engagement with a metabolic physician is recommended, where chronic liver disease is unlikely, in order to help expedite testing and facilitate acute management. First evaluations include: liver enzymes, arterial or venous acid-base status, renal function and electrolytes.

Low blood urea, normal glucose with a high ammonia and respiratory alkalosis increases the likelihood of UCD. The low urea could be due to UCD or protein restriction, however it should be noted that this is not a very sensitive marker and a normal urea should not be used to rule out UCD. In this context, urgent metabolic testing of urine organic acids, plasma amino acids and acylcarnitines, should be undertaken within 24 hours. The urine organics acids are used to identify orotic acid, uracil in OTC deficiency, as well any evidence for an organic aciduria or dicarboxylic aciduria, glycines; in the case of a fatty acid disorder of oxidation. The plasma amino acids in the case of UCD are used to determine underlying UCD. In general, glutamine is elevated in UCD, and normal in organic aciduria and fatty acid disorder of oxidation. Citrulline would be low in OTC and proximal UCD, but elevated in those distal to OTC (see figure 1). A high anion gap metabolic acidosis is indicative of organic aciduria. (For more detailed discussion on interpreting relevant metabolic tests see reference 66.66)

As there is no specific test to confirm that HA is secondary to a specific medication, ruling out an IMD with next-generation sequencing panel testing is usually undertaken as outlined in the relevant guideline and other sources.66 68 69 The sensitivity of genetic testing has improved with developments in sequencing technology, but particularly in OTC this is still around 90% and additional functional or enzymatic analysis maybe required.69

Analysis of plasma ammonia

Although historically there have been a wide variety of laboratory methods, indirect enzymatic methods that measure either nicotinamide adenine dinucleotide phsophate (NADPH) or nicotinamide adenine dinucleotide (NADH) reduction at 340 nm are most frequently used. table 4 lists some of the relevant preanalytical factors including method and effect on ammonia. By far the most significant is the delay to transfer and separation from red cells as the concentration in red blood cells is approximately three times that of plasma and represents potential for ammonia contamination (table 4).70 This is one of the main reasons for using plasma as the sample of choice rather than serum. Ice is used by some as a stopgap measure to reduce in vitro ammonia formation, in the absence of a suitable blood tube additive inhibitor. However, if sample handling or processing is prompt, the benefits of using ice are insignificant (table 4). The ammonia content range in 24-hour urine collection has been estimated at 4–24 mmol/L, however it is not widely available and so despite inaccuracies, underestimated use of urine/osmolal gap persists.71

There is a huge degree of variability between studies and different assays in terms of time cut-offs, compounded by a lack of a robust analytical performance specification. In the introduction we used the metbionet definition of HA (>40 µmol/L, 68.12 μg/L). However, to illustrate the importance of considering the laboratory method, a recent Wales External Quality Assurance (WEQAS) ammonia return with an overall method mean of 41 μmol/L, had a method range of 23–50 μmol/L which would impact on diagnosis.72 Similarly, another recent WEQAS ammonia EQA return with an overall mean of 79 μmol/L, had a method range of 61–102.1 μmol/L, which is also likely to impact on treatment choice.72

One of the few studies to assess biological variation in healthy individuals showed a within-individual coefficient variation of approximately 14%, within-group coefficient of variation of approximately 17% and a reference change value of 43%.73

Treatment

There are no evidence-based guidelines to deal with drug-induced HA and no specific tests. There is a variety of different approaches including the Australian guidelines for managing asparaginase-induced HA;74 guidelines proposed by Boilève et al for 5-FU-induced toxicity40 and expert panel recommendations on prevention and management of asparaginase/pegasparaginase-associated toxicities in adults and older adolescents.75 These approaches include stopping, dose reduction or continuing with measures to deal with HA. If it is decided to continue the medication then close monitoring is required and testing for other causes should be undertaken, the urgency for which is dependent on the severity of the clinical symptoms and progression of ammonia.

Stopping the offending drug suspicious for HA if possible and supportive treatment is an essential first step in management. Some cohort studies have explored this using valproate dose reduction or withdrawal and appropriate AED substitution where required, and HA has resolved.76 77 It must be emphasised that treating other factors such as hepatic dysfunction, sepsis, nutrition and medication adjustment as well as HA are also essential.66 77–79 Management of hepatic dysfunction and sepsis will not be discussed here.78 Also, although use of lactulose in the setting of cirrhosis has been shown in a meta-analysis of randomised clinical trial data to reduce mortality and morbidity (variceal bleed, hepatic encephalopathy, hepatorenal syndrome), its effects on lowering ammonia directly are modest; mean reduction in ammonia is 12 µmol/L (95% CI reduction of 21 µmol/L to 2 µmol/L).80 Therefore, outside of cirrhosis, lactulose is not recommended for management of HA.66

A metabolic physician should be consulted immediately once the diagnosis of HA has been confirmed. HA associated with symptoms must be treated as soon as possible. From the authors’ experience the initial steps will usually involve intravenous glucose plus ammonia scavengers (see tables 5 and 6). Any underlying dehydration will require 0.9% sodium chloride (normal saline) and if the patient is shocked or obviously unwell then admission to a high density care or intensive care unit should be arranged. Regular assessment with Glasgow coma scale (GCS), blood gas (pH, electrolytes), ammonia is routinely recommended (see tables 5 and 6). However a full updated clinical examination and blood tests is recommended if the patient becomes encephalopathic or ammonia is increasing despite initial therapy. These patients are also at risk of hyponatraemia and hypokalaemia.

With recent adult cohort data suggesting adverse neurological outcomes with ammonia values >200 µmol/L, this would suggest that more aggressive therapy (haemodiafiltration) should be considered earlier at this value.6 79 81 Ultimately however therapeutic interventions should be guided by both clinical presentation and ammonia level as occasionally even patients with ammonia >200 µmol/L maybe asymptomatic (tables 5 and 6). A coma for more than 3 days, raised intracranial pressure and ammonia >1000 µmol/L are particularly poor prognostic factors.66

Sodium benzoate and sodium phenylbutyrate (the precursor of the active agent phenylacetate) promote ammonia removal from the body by combining with glycine to form hippurate; and with glutamine to form phenylacetylglutamine which are excreted in urine. This alternative pathway reduces the amount of NH₃ presented to the urea cycle for detoxification. In theory each mole of benzoate removes one mole of ammonia and for phenylbutyrate this is two moles of ammonia, however due to variability in pharmacokinetics and pharmacodynamics, this is not achieved. However, such scavengers will not work adequately if hepatic function is compromised.

Dietary management of acute decompensation is crucial to prevent or stop catabolism and to correct biochemical abnormalities and ensure adequate nutritional intake. Protein intake should be reduced or paused completely. Adequate amounts of calories should be provided either enterally or parenterally (10%–20% dextrose and 20% fatty acids solutions with appropriate electrolytes: Na⁺, K⁺).82 83 This will require input from an experienced metabolic dietitian as prolonged restriction for greater than 48 hours can lead to catabolism and worsening in HA.

Haemodialysis is the most effective therapy for HA and may serve as a rescue therapy. Intermittent haemodialysis achieves the highest ammonia clearance, however due to patient stability or raised intracranial pressure continuous veno-venous haemodialysis may be more suitable.84 Most of the evidence around use of dialysis in HA is based on paediatric rather than adult based studies.

Liver transplantation is the definitive cure for cirrhosis and for urea cycle defects. It is considered the last resource for patients with recurrent decompensation and poor response to conventional therapies, but these are high-risk situations and require careful discussions if the patient’s condition allows.85

There are a number of case reports/cohort studies suggesting the use of carnitine or L-arginine supplementation in the management of valproate-induced HA.72 86