Influence of valproate-induced hyperammonemia on treatment decision in an adult status epilepticus cohort

INTRODUCTION
Status epilepticus (SE) is a neurological emergency in which immediate intervention is required to prevent permanent brain damage and death. Intravenous (IV) valproic acid (VPA) is often used for the treatment of SE. However, IV VPA frequently increases the blood ammonia level. In this study, we explore the impact of IV VPA-induced hyperammonemia (HA) on treatment management of SE and discuss the challenges related to this particular condition.


METHODS
We used data from medical records of 31 adult patients (≥18 years) treated with IV VPA for SE at Oslo University Hospital between January 2006 and October 2019. Clinical and blood sample data and information about the influence of HA on treatment were collected. Correlations between ammonia levels and other continuous or categorical variables were tested using the Pearson's correlation coefficient. The Kruskal-Wallis H-test was used to analyze associations between different variables and treatment decisions.


RESULTS
Thirty of 31 patients had increased ammonia level during IV VPA treatment. In 16/30 patients, VPA was discontinued, and in 6/30 patients, the dose was reduced. We found a difference in the median peak ammonia level among the groups where VPA was discontinued (99 μmol/l), reduced (71 μmol/l), and continued (55.5 μmol/l) (P = 0.008). Also clinical status, measured by West Haven Criteria, varied among the groups where VPA was discontinued (3.5), reduced (2.5), and continued (2.0) (P = 0.01). Treatment decisions at peak ammonia were not associated with the level of liver enzymes and bilirubin.


CONCLUSION
Hyperammonemia had a substantial impact on further management. To date, no recommendations exist on how to manage VPA-induced HA in SE. We call for systematic prospective studies and evidence-based guidelines.


Introduction
Status epilepticus (SE) is a neurological emergency that can lead to permanent brain damage and death [1]. Rapid and aggressive treatment is important and should be given according to a staged protocol [2]. Today, intravenous (IV) benzodiazepine is the first drug of choice and is expected to terminate SE in up to two out of three cases [3]. When benzodiazepines fail, phenytoin or valproic acid (VPA) is commonly used as second-line treatment [3].
An established adverse effect of VPA is an increase of the blood ammonia level that may occur in both chronically or acutely treated patients [4,5]. While a few studies and recommendations exist on how to handle VPA-associated hyperammonemia (HA) in patients receiving chronic peroral treatment with VPA [5,6], no guidelines are available on how to manage this adverse effect in patients with SE.
The aim of this study was to evaluate the prevalence and level of HA induced by IV VPA in an adult cohort with SE, its impact on treatment decisions, and to discuss challenges related to this particular condition.

Data collection
In this retrospective observational study, we included 31 randomly chosen adult patients (≥ 18 years) who were treated for SE at Oslo University Hospital (OUS) between January 2006 and October 2019. Inclusion criteria were (1) patients with SE according to the International League Against Epilepsy (ILAE) latest definition [7], who were Epilepsy & Behavior 111 (2020) 107193 (2) treated with IV VPA, and (3) tested for blood ammonia level. We reviewed the medical records and registered ammonia levels after start of VPA infusion, including peak ammonia level. The OUS reference value for ammonia is b32 μmol/l, and values higher than this were considered as an increase. We classified ammonia levels (μmol/l) as follows: normal level: 0-32; mild increase: 33-60; moderate increase: 61-100, and marked increase: N100. This classification was done because of discretionary assessment, as no established graduation exists.
Measurement of ammonia, liver enzymes, and VPA was all done according to the OUS best practice procedures for safe and accurate blood sampling. Venous blood draw was performed for all tests, usually between hrs 06 A.M. and hrs 09 A.M. To assess intercurrent liver injury, we collected data on bilirubin and liver enzymes: alanine transaminase (ALAT), aspartate transaminase (ASAT), Gamma-glutamyl transferase (GGT), and Alkaline phosphatase (ALP). Further, VPA serum concentration from the same day as peak ammonia level was collected for analysis.
Information about mental status and grade of consciousness at the day of peak ammonia level was used to retrospectively assess the West Haven score [8]. The West Haven score is a clinical five-point classification system which is primarily used for hepatic encephalopathy [9].
To evaluate the influence of increased ammonia level on the decision on further treatment, the patient records had to have a clear statement on the time point and motivation for VPA dose reduction or cessation. For the descriptive analysis, we used the median and the quartile deviation (QD). Scatter graphs were plotted between ammonia blood levels and other variables. Correlations between ammonia levels and other continuous or categorical variables were tested using the Pearson's correlation coefficient. The Kruskal-Wallis H-test was used to analyze associations between different variables and treatment decisions.

Study settings and ethics
The study was conducted by the Epilepsy Research Group of Oslo (ERGO), Department of Neurology, OUS, Oslo, Norway. It was approved by the Data Protection Official of OUS and is a part of a larger quality control study [10].

Patient characteristics
Patient characteristics are summarized in Table 1. We identified a total of 31 patients who fulfilled the inclusion criteria. The median age was 62 years (QD = 13.5), and 17 were women. The median number of other antiepileptic drugs (AEDs) at the same day as the peak ammonia level was two (QD = 1). Nine patients were treated with VPA before admission, one patient had a diagnosis of having subcapsular hepatic shunt, and seven patients had a history of alcohol misuse.

Laboratory results
Laboratory results are summarized in Table 1. Thirty of 31 patients had increased ammonia levels. The median peak ammonia level was 72 μmol/l (QD = 28.5). The median time interval from VPA infusion to peak ammonia was two days (QD = 1.5). The median time interval from VPA infusion to the first ammonia measurement was one day (QD = 1) with a firstly measured median ammonia value of 57 μmol/l (QD = 19).
Twenty-three patients had accurate information about the maintenance dose of VPA, and the median dose was 100 mg (QD = 0) per hour. The median VPA blood concentration at peak ammonia level was 532 μmol/l (QD = 114.88). Median VPA concentration within +/− three days from peak ammonia was 679 μmol/l (QD = 167). There was no correlation between peak ammonia and VPA concentration (r = ÷ 0.083, P = 0.661) (Fig. 1).
Liver enzymes were normal for all patients at the same day and within three days post and prior to the measured peak ammonia level. Correlations are visualized in Fig. 1. No correlation was seen between peak ammonia and liver parameters, except from GGT (r = 0.391, P = 0.036). We also compared biochemical markers in patients with normal and mildly increased ammonia level to the group with markedly increased ammonia level. In both groups, the biochemical markers were within normal range, except from ASAT that was only slightly increased in the group of markedly increased ammonia level.

Clinical status measured by West Haven Criteria
The distribution of West Haven grades in our patients is presented in Table 1. Out of 31 patients, one had a West Haven grade 0, four grade 1, eight grade 2, eight grade 3, and 10 grade 4. We found a correlation between peak ammonia level and West Haven grade (r = 0.467, P = 0.008).

Impact of ammonia levels on treatment decisions
Associations between different clinical and biochemical variables and treatment decisions are presented in Table 2. In 16 out of 30 patients, VPA treatment was discontinued, and in 6, the VPA dose was reduced. Two patients received treatment with lactulose for increased ammonia levels.  There was a significant difference in the median peak ammonia level among the groups where VPA was discontinued (99 μmol/l), reduced (71 μmol/l), and continued (55.5 μmol/l) (P = 0.008). Also, there was a significant difference in the median West Haven grade among the groups where VPA was discontinued (3.5), reduced (2.5), and continued (2.0) (P = 0.01). We could not observe an association between treatment decisions at peak ammonia and the level of liver enzymes, bilirubin, or VPA concentration.
Treatment decisions at peak ammonia were independent of the level of liver enzymes or bilirubin. Biochemical markers for liver failure were normal for all patients, and the increase of ammonia level was independent of the serum concentration of VPA. Treatment decisions were not done by means of any protocol.

Results in context
We here investigate the impact of the blood ammonia level on treatment decisions in patients with SE treated with IV VPA. Hyperammonemia was a very common finding in these patients. We found that the blood ammonia level was independent of the serum concentration of VPA and, except from slight and infrequent aberrations, not associated with increase of liver enzymes or bilirubin. We observed that HA had a substantial impact on the management of antiepileptic treatment. Treatment decisions were not done by standard procedures.
Valproic acid-induced HA is a well-known phenomenon. Crosssectional studies suggest a prevalence of 16 to 100 percent while prospective studies report on 70 to 100 percent [5,6]. However, existing publications on VPA-related HA are performed in patients who use VPA on a regular basis for either epilepsy or psychiatric disorders. For this group of patients, most authors recommend that ammonia should only be measured when symptoms of encephalopathy develop [4][5][6], which occurs infrequently [11].
Only sparse information is available on how to manage VPA-induced HA in SE. To our knowledge, only one recent study highlights HA specifically in patients with SE [12]. The authors recommend that VPA should be discontinued in cases of increased ammonia levels but do not provide any guidelines for measurement routines or specific interventions based on the level of increase. Further, there is no information on differences between acute or chronic VPA treatment, ammonia levels, guidelines for measurement, and clinical consequences.
Despite of a careful literature search, we were not able to find any publication on the outcome of patients with SE with increased versus normal blood ammonia levels. This begs the question on what basis VPA is reduced or discontinued for the majority of patients with SE.
One obvious challenge for treating physicians is the lack of specific clinical signs or symptoms of encephalopathy. We found a correlation between ammonia levels and the severity of West Haven grade measuring signs of encephalopathy in our study cohort [8,9]. However, we experienced difficulties with the grading as the clinical symptoms assessed all may appear in SE per se. Moreover, sedative medications or postictal symptoms may mimic symptoms of encephalopathy.
Encephalopathy can be detected by electroencephalography (EEG) typically presenting with irregular, continuous, severe, and diffuse slowing with a predominance of rhythmical theta and delta activity [ [13][14][15]. Although these phenomena also may occur in SE or during the postictal state, continued or regular periodic EEG after IV VPA administration and during the stabilization period of SE would, together with clinical observation and repetitive ammonia measurement, be advantageous to identify a potential ammonia-induced encephalopathy, as earlier encouraged by Trinka et al. [16].

Limitations
Data are generated from a cohort of retrospectively, nonsystematically recruited patients, which also means that laboratory tests were not performed according to a predefined protocol. Thus, time variation from the first administration of IV VPA to ammonia level measurement may have occurred. Furthermore, ammonia levels were not routinely measured prior to SE treatment, thus we cannot rule out that HA was present and unrelated to IV VPA treatment. We found a higher incidence of VPA associated ammonia increase than formerly described [4][5][6]12]. This may be due to differences in data acquisition. All our patients received high IV doses of VPA commonly together with other AEDs which both have been described as risk factors for increased ammonia levels [4,13,17]. Further, there is no established consensus on whether ammonia under VPA therapy should be measured or not. Therefore, it is possible that clinicians who monitored ammonia did this based on a clinical suspicion of ammonia toxicity, thus skewing the material towards changing VPA treatment regime.

Conclusion
In conclusion, our findings indicate that VPA-associated HA has a high impact on treatment decisions in SE. Treatment adjustments or discontinuation of VPA occur while no guidelines exist, and even without a common agreement on whether ammonia should be measured at all. This uncertainty may have an effect on the final outcome and could result in patients not being treated with an effective drug for a potentially life-threatening condition. To date, it seems to be a general consideration to taper or discontinue VPA to be on the safe side in the absence of appropriate clinical observation tools. Our study should be regarded as an eye opener and contribution to the discussion but is not suited to give any advises. We call for larger systematic and prospective studies generating definite evidence-based recommendations. Table 2 Association between treatment decisions and biochemical and clinical variables in patients with SE with VPA-induced HA.