Ammonia Clearance with Different Continuous Renal Replacement Therapy Techniques in Patients with Liver Failure

Acute liver failure (ALF) and acute-on-chronic liver failure (ACLF) are characterized by coagulopathy, encephalopathy, and multi-organ dysfunction [1‒4]. High ammonia levels appear to play a key role in the pathogenesis of encephalopathy and cerebral oedema [5‒9]. Ammonia removal may therefore reduce its neurotoxic effects and decrease the likelihood of cerebral oedema and raised intracranial pressure [10‒13].

The biochemical properties of ammonia (a small water-soluble molecule) make it a theoretically suitable target for removal by continuous renal replacement therapy (CRRT) [13‒15]. CRRT can be delivered via techniques that differ in their mechanism for solute clearance; i.e., convection, diffusion, or a combination of the 2 [15]. Whilst, these different mechanisms may affect ammonia clearance, there is very limited evidence to support decision-making regarding the most efficient CRRT technique for ammonia clearance in these patients, although, overall, greater CRRT intensity may prove useful [16].

Accordingly, we performed a pilot investigation of ammonia clearance by 3 different techniques of CRRT in patients with liver failure induced hyperammonaemia. We aimed to test whether, in patients with liver disease treated with CRRT, choice of the technique affects ammonia clearance.

This study complies with the Declaration of Helsinki. The study protocol was approved by the Austin Hospital Human Research Ethics Committee (LNR/18/Austin/298). A waiver of consent was granted by Austin Hospital Human Research Ethics Committee.

The Austin Hospital is a 400-bed university-affiliated tertiary referral centre. The Austin Intensive Care Unit (ICU) admits approximately 2,300 patients per year and operates as a closed ICU where only ICU medical staff can prescribe CRRT. All patients admitted to the ICU, between January 2019 and March 2020, were screened, and patients were eligible if they were adult, fulfilled diagnostic criteria for (ALF) [17] or ACLF [3] and were deemed to require CRRT. A convenience sample was recruited sequentially. CRRT was performed with the Baxter Prismaflex (Gambro, Lund, Sweden), utilizing the AN69 ST 100 membrane (Gambro Industries, Meyzieu Cedex, France). The sequential technique change, and respective blood flow rates, for continuous veno-venous haemodiafiltration (CVVHDF) to continuous veno-venous haemodialysis (CVVHD) and to continuous veno-venous haemofiltration (CVVH) (schematic modes A, B, and C, respectively) are illustrated in Figure 1 and in online supplementary File 1 (for all online suppl. material, see www.karger.com/doi/10.1159/000521312).

Each patient commenced a new CRRT circuit prior to the sampling protocol, to minimize any potential confounders affecting ammonia clearance. All technique transitions and samplings were performed within 1 h of the circuit commencement. Additionally, to ensure the best representation of each clearance mechanism, a 30-min stabilization period was allowed after each technique transition. Our study protocol was overseen by the investigators at the bedside. Vascular access for CRRT was obtained via a 13.5 Fr dialysis catheter inserted into the internal jugular or femoral vein. CRRT anticoagulation was achieved via circuit epoprostenol administration according to ICU policy where indicated. Epoprostenol was delivered pre-filter at a concentration of 500 μg/50 mL in 500 mL 0.9% normal saline, with a range of 4–8 μg/kg/min (rate range 1.7–3.4 mL/h based on 70 kg patient).

Sampling for plasma ammonia was from the patient’s radial arterial line. Effluent ammonia sampling was from a sampling port in the effluent line immediately prior to the effluent collection bag. Five millilitres (mL) of blood was obtained from the arterial line, following removal of 3 mL of dead-space volume. Correspondingly, 5 mL of effluent was collected via a 3-way tap from the sampling port. Blood and effluent waste samples were placed on ice and immediately transported to the hospital’s clinical laboratory for analysis. Simultaneous, small solute clearance (urea and creatinine) was assessed via the same method. Analysis was performed using a Beckman Coulter AU Chemistry Analyzer (Brea, CA, USA) with Fisher Diagnostics (Middletown, VA, USA) Infinity^TM Ammonia Reagent as appropriate.

Two methods were used to assess ammonia clearance. Firstly, we assessed “CRRT system clearance” by calculating the ratio of effluent to pre-filter plasma ammonia. Pre-filter ammonia concentration was estimated by correcting arterial line plasma ammonia for the effect of pre-dilution fluid flow (delivered before the blood pump in all cases), measured haematocrit, and CRRT blood flow rate (equation 1 below) in accordance with previously published calculations [15].

Clearance (mL/min) = effluent rate (mL/min) × {effluent solute concentration [mmol/L]/arterial solute concentration × [blood flow (mL/min) × (1–haematocrit)] ÷ [blood flow (mL/min) × (1– haematocrit)] × replacement fluid [mL/min]} (1)

Secondly, we looked at the “total body clearance” by calculating the ratio of effluent to uncorrected arterial plasma ammonia (equation 2 below).

Clearance (mL/min) = effluent rate (mL/min) × [effluent solute concentration (mmol/L)/arterial solute concentration (mmol/l)](2)

Both equations were applied and aligned with each respective techniques to correct for changes in pre-filter replacement fluid administration.

To understand the precision of ammonia effluent assay more accurately, we performed a further sequential and methodical assessment comparing paired effluent samples using the methodology described above. Additionally, in a single patient, the paired samples were further divided, resulting in 4 separate patient samples.

Statistical analyses were performed with STATA V.14.2 (StatCorp, College Station, TX, USA). All continuous variables were assessed for normality and are presented as medians and interquartile ranges (IQRs) and ordinal variables are presented as numbers (n) and corresponding percentages (%). All non-parametric data were compared by the Kruskal-Wallis test. Differences between paired samples were compared with Student’s t test. A 2-sided p value of <0.05 was considered statistically significant for all analyses.

Fourteen patients were included in the study. Two patients were excluded: one due to sampling delays in transport and another due to incomplete CRRT technique progression due to clinical instability. Thus, 12 patients were included in the final analysis.

Patient demographics and clinical and biochemical parameters are presented in Table 1. The majority (75%) of patients had ACLF, and 7 (58%) were female. The median MELD score was 38 (IQR 35–40), and the INR 2.9 (IQR 1.9–4.4). All patients were hyperammonaemic, with a median plasma ammonia level of 95 (IQR 70–99) µmol/L, prior to commencing the study. All patients were commenced on CRRT prior to study enrolment. The indications for CRRT commencement and characteristics of the circuit are presented in Table 2. The most common indication for RRT was anuria, followed by acid-base abnormalities.

As shown in Table 3, arterial plasma ammonia levels were not significantly different prior to sampling between CVVH, CVVHD, and CVVHDF. Additionally, plasma effluent ammonia was also not significantly different between the 3 techniques. Consequently, median CRRT system plasma ammonia clearance was not significantly different between the 3 techniques: CVVH 27 (IQR 23–32) mL/min, CVVHD 21 (IQR 17–28) mL/min, and CVVHDF 20 (IQR 14–28) mL/min, p = 0.32, respectively. Median “total body ammonia clearance” was also not significantly different between the modes of RRT; CVVH 22 (IQR 19–25) mL/min, CVVHD 21 (IQR 17–28) mL/min, and CVVHDF 19 (IQR 13–25) mL/min; p = 0.79, respectively (Fig. 2).

There was no significant difference in clearance between CVVH, CVVHD, and CVVHDF for urea: median 53 (IQR 48–55) mL/min versus 48 (IQR 43–51) mL/min versus 52 (IQR 47–54) mL/min, p = 0.263 or creatinine: median 44 (IQR 39–46) mL/min versus 40 (IQR 32–46) mL/min versus 42 (IQR 37–46) mL/min, p = 0.704. However, median urea and creatinine clearance, across all modes of CRRT, were significantly higher than median ammonia clearance; urea 50 (IQR 46–54) mL/min versus creatinine 42 (IQR 38–46) mL/min versus ammonia 25 (IQR 18–29) mL/min, p = 0.0001 (Fig. 3; online suppl. Table 1).

Five additional patients underwent sampling to assess the precision of the ammonia assay. There was no significant difference between paired effluent samples (mean difference −1.16, standard error of mean 1.7; p = 0.52), thus demonstrating a high degree of reproducibility in the ammonia assay (Table 4).

In a systematic structured prospective comparative assessment of CRRT techniques, we found that, at an equivalent effluent rate, CRRT-associated ammonia clearance does not significantly differ when comparing CVVHDF, CVVHD, and CVVH. However, we also found that ammonia clearance was approximately half the clearance of urea and creatinine across all techniques.

To our knowledge, this is the first study to directly compare ammonia clearance between different CRRT techniques. There are limited studies investigating the extracorporeal clearance of ammonia in the critically ill. A recent review found only 13 published papers on this topic [18]. The largest body of research to date has occurred in the paediatric field, where both intermittent haemodialysis and CRRT have been used successfully to treat hyperammonaemia [19‒21]. Ammonia clearance in these studies varied considerably, often implausibly. However, higher blood and dialysate flow rates were consistently correlated with greater ammonia clearance, i.e., higher effluent rate and dose appeared to increase clearance [18]. Accordingly, we did not find any difference in clearance comparing CRRT techniques, where the dose by effluent rate per hour was the same. This finding, however, is of interest to clinicians where mode choice decision may be necessary. As ammonia is a small size molecule (MW = 17), a pure diffusive technique (CVVHD) may be more effective compared to other techniques using pre-dilution fluid [22]. For pure convection techniques such as CVVH, there has been only a single study in adults [13]. In this study, Slack et al. compared high-volume haemofiltration (90 mL/kg/h) versus lower intensity CVVH (35 mL/kg/h). The authors concluded that ammonia clearance was closely associated with higher effluent flow rates. However, these results reported a clearance rate greater than the effluent flow rate, a biochemically impossible finding, suggesting methodological problems may have affected the reported ammonia clearance results.

The lower clearance of ammonia compared with urea and creatinine was unexpected in view of ammonia’s molecular weight and chemical properties. One possible reason relates to ammonia’s complicated dynamics. Ammonia (NH₃) is a highly ionized compound, existing in equilibrium with the ammonium ion (NH4⁺). When in an aqueous solution, such as plasma, ammonia is 98% ionized at physiological temperature and pH [23, 24]. Our study utilized an AN69 ST 100 CRRT membrane (Gambro Industries) that has a well-recognized negative electrostatic charge [25, 26], which has remained even after surface treatment [27]. We postulate that there may be an electrochemical interaction between the ionized ammonium ion and the negatively charged AN69-ST CRRT membrane that decreases transmembrane ammonia movement. Irrespective of our speculations, given the importance of ammonia clearance, this is an area that requires further research.

Our study implies that the technique of CRRT does not affect clearance and that delivered dose of effluent flow is the key aspect of ammonia clearance. It also implies that ammonia clearance is less than expected, given its molecular weight. This observation further implies that more research needs to be done to explain this phenomenon of limited transmembrane movement and that maximizing the effluent flow rate is fundamental to achieve better clearances.

Our study has several strengths. The literature existing regarding ammonia clearance during CRRT in acutely ill patients is minimal and our direct comparison of CRRT techniques is a first. We applied a robust methodological approach to the repeated sampling in each patient, and the ammonia clearance applied standard clearance calculations. The CRRT techniques, settings, and prescription are aligned with standard CRRT protocols within the ICU, and the use of arterial blood samples for plasma ammonia represented everyday clinical care.

We acknowledged several limitations. Our study is from a single centre. However, the patients studied are typical of a tertiary liver disease referral centre in a high-income country. Our study sample size is small; however, comparisons were paired, increasing power, and the values for each intervention were almost identical. In addition, where clear differences were present (e.g., comparing urea and creatinine clearance with ammonia clearance), they easily achieved significance. We did not study the effect of each technique on plasma ammonia levels over time. However, this was not our goal, and our focus was on a technical comparison of efficiency to inform CRRT technique selection. Finally, our proposed hypothesis for the lower than previously reported ammonia clearance is limited by the absence of an assessment of post-filter blood ammonia levels and ammonia mass balance in the circuit. Although we have attempted to account for many of the variables that may affect ammonia concentrations and clearance, it is possible other factors such as catheter location, arterial sampling site, and the micro-environment with the membrane may additionally, in a yet unrecognized way, impact ammonia clearance.

In a group of critically ill patients with hyperammonaemia secondary to liver failure treated with CRRT, there was no significant difference in ammonia clearance between the CRRT techniques of CVVH, CVVHD, and CVVHDF. However, across all 3 CRRT techniques, ammonia clearance was half of that observed for urea and creatinine. These findings inform clinicians on the choice of CRRT technique and underscore the importance of treatment intensity and additionally the need to investigate the reason for the low transmembrane movement of ammonia.

This study protocol was reviewed and approved by the Austin Hospital Human Research Ethics Committee (LNR/18/Austin/298). An exemption to written informed consent was granted for this study by the Austin Hospital Human Research Ethics Committee (LNR/18/Austin/298).

One of the authors (R.B.) has received grants and lecture fees from Baxter Healthcare.

This study did not receive any funding.

C.F., I.B., N.F., T.N., and R.B. designed the study. All the authors were equally involved in data collection, data analysis, and manuscript development. All the authors have read and approved final version of the manuscript.

All data generated or analysed during this study are included in this article [and/or] its online supplementary material Files. Further enquiries can be directed to the corresponding author.