Abstract
Background & Aims
Ammonia is metabolized into urea in the liver. In acute liver failure (ALF), ammonia has been associated with survival. However, urea variation has been poorly studied.
Methods
Observational cohort including ALF patients from Curry Cabral Hospital (Lisbon, Portugal) and Clinic Hospital (Barcelona, Spain) between 10/2010–01/2023. The United States ALF Study Group cohort was used for external validation. Primary exposures were serum ammonia and urea on ICU admission. Primary endpoint was 30-day transplant-free survival (TFS). Secondary endpoint was explanted liver weight.
Results
Among 191 ALF patients, median (IQR) age was 46 (32;57) years and 85 (44.5%) were males. Overall, 86 (45.0%) patients were transplanted and 75 (39.3%) died. Among all ALF patients, following adjustment for age, sex, body weight, and etiology, higher ammonia or lower urea were independently associated with higher INR on ICU admission (P<0.009). Among all ALF patients, following adjustment for sex, etiology, and lactate, higher ammonia was independently associated with lower TFS (adjusted odds ratio (95% confidence interval (CI)) = 0.991 (0.985;0.997); P=0.004). This model predicted TFS with good discrimination (area under receiver operating curve (95% CI) = 0.78 (0.75;0.82)) and reasonable calibration (R2 of 0.43 and Brier score of 0.20) after external validation. Among transplanted patients, following adjustment for age, sex, actual body weight, and etiology, higher ammonia (P=0.024) or lower (P<0.001) urea were independently associated with lower explanted liver weight.
Conclusions
Among ALF patients, serum ammonia and urea were associated with ALF severity. A score incorporating serum ammonia predicted TFS reasonably well.
LAY SUMMARY
In patients with acute liver failure, higher ammonia or lower urea in the blood was associated with higher disease severity.
Ammonia may be a helpful indicator in stratifying the risk of mortality without liver transplant.
Keywords: nitrogen, hepatitis, necrosis, liver regeneration, transplant, outcomes
INTRODUCTION
Acute liver failure (ALF) is a rare disease with high risk of morbidity and mortality despite all the advances in its management over the past decades, including N-acetylcysteine, organ support, plasmapheresis, and transplantation [1].
Under homeostatic conditions, the liver is the primary site for ammonia metabolism. Most of the ammonia reaches the liver via portal vein, with about 50% coming from endogenous glutamine conversion in the enterocytes and other 50% deriving from the lumen of the gut, through degradation of nitrogen substrates from diet or bacterial metabolism [2]. In the hepatocytes, ammonia is converted to urea or glutamine [2,3].
About 33% of the serum urea derives from the hepatic ammonia and once filtered in the kidneys, variable amounts of it undergo reabsorption or excretion in the urine [2,4]. The kidneys also produce (from endogenous glutamine) and excrete different amounts of ammonia [4].
In ALF, the loss of viable hepatocytes following the initial insult to the liver may lead to impaired metabolism of many substrates, including ammonia [1]. Therefore, these patients often have increased serum ammonia levels [5]. Moreover, significant liver necrosis may result in loss of liver weight and lower likelihood of liver regeneration [6]. In fact, hyperammonemia and reduced liver volume have been associated with worse survival in these patients [6,7].
In this context, we hypothesized that ALF may lead not only to hyperammonemia but also to a decreased urea formation in the liver. Therefore, the objectives of this study were as follows: (1) describe variations in serum ammonia and urea levels in patients with ALF; (2) study the association of ammonia and serum urea levels with patients’ outcomes.
METHODS
Design, setting, and participants
This was a dual center retrospective observational cohort study. All adult patients (≥17 years old) with ALF (see definitions), admitted to liver-specialized intensive care units (ICU) and included in local registries at the following liver transplant (LT) centers were included: (1) Curry Cabral Hospital (Central Lisbon University Hospital Center), in Lisbon, Portugal, from October 2016 to January 2022; (2) Clinic Hospital, in Barcelona, Spain, from October 2010 to January 2023. Patients were excluded if they had underlying cirrhosis or previous LT.
A multicenter cohort of ALF patients included prospectively in the United States ALF Study Group (USALFSG) registry from 1998 to 2016, from a previous study about renal replacement therapy, was used for external validation [7].
This study was approved by the local ethics committees and individual informed consent was waived due to its observational nature (#INV_363). The study protocol and implementation abided by the principles of the Declaration of Helsinki [8]. The reporting of this study followed the recommendations of the STROBE guideline [9].
Definitions, data collection, and endpoints
Acute liver failure (ALF) was defined as follows in all cohorts: an international normalized ratio (INR) of ≥1.5, HE (any grade based on West-Haven classification), a disease course of <26 weeks, and absence of cirrhosis [10].
The following baseline characteristics were retrieved from the local ALF registries by designated investigators (FSC, DT, NJ, PF, and MV) for the first day of ICU stay: age, sex, and actual weight (Kg); etiology of ALF; ongoing organ failures, namely HE, invasive mechanical ventilation (IMV), oxygen arterial pressure / oxygen inspired fraction (PF) ratio (mmHg), vasopressors, mean arterial pressure (mmHg), acute kidney injury (AKI), , renal replacement therapy (RRT), and Sequential Organ Failure Assessment (SOFA) score (based on the worse indicators of brain, lungs, heart, liver, kidneys, and hemostasis functions for the first day of ICU stay); arterial blood workup on ICU admission including hemoglobin (g/l), leucocytes (n/109), platelets (n/109), creatinine (mg/dl), urea (mg/dl), sodium (mmol/l), potassium (mmol/l), phosphate (mmol/l), ammonia (µmol/l), INR, bilirubin (mg/dl), alanine transferase (ALT) (U/L), and lactate (mmol/l); and overall patients’ outcomes, namely LT, mortality and, among those transplanted, explanted liver weight (g) [11,12]. Noticeably, determinations of serum (arterial line samples) creatinine (upper limit of normal (ULN) = 1.2mg/dl), urea (ULN = 40 mg/dl), and ammonia (ULN = 50 μmol/l) were performed prior to initiation of any RRT to avoid confounding measurements. Explanted livers were all weighed as part of the mandatory macroscopic assessment post LT, prior to the microscopic evaluation at local pathology laboratories.
AKI was defined as per Kidney Disease Improving Global Outcomes (KDIGO) guidelines, with baseline creatinine being calculated using the modification of diet in renal disease (MDRD) equation as per Acute Disease Quality Initiative (ADQI) criteria (average glomerular filtration rate of 75ml/min/1.73m2) [12,13]. LT candidacy was considered for patients fulfilling the following criteria without any established contra-indication: (1) King’s College criteria (KCC) at Curry Cabral Hospital; (2) and one of the 3 specific criteria at Clinic Hospital: grade III-IV HE (West Haven classification), INR >7 or absence of liver function improvement (disease course >21 days) [14]. Other prognostic scores considered were the following: Model for End-stage Liver Disease (MELD), including serum INR, bilirubin and creatinine; and the USALFSG prognostic index, including HE grade, etiology, vasopressor use, and serum INR and bilirubin [14,15].
ALF patients at all centers were managed similarly based on specific guidelines [10]. All patients on ICU admission were on a normocaloric diet (target of up to 25–35kcal/Kg), without protein intake restrictions, whether via oral or enteral routes.
The primary endpoint was the transplant-free survival at 30 days post ICU admission (TFS), an appropriate outcome to capture disease-related mortality which is higher during the first weeks [6,14]. The secondary endpoint was the explanted liver weight (g) for those who underwent LT. This was a surrogate for disease severity as patients with severe ALF have higher risk of post necrotic liver structural collapse [5]. Follow-up continued until death, hospital discharge, latest visit to outpatient clinic on record, or loss to follow-up. In the USALFSG validation cohort, the standard endpoint considered in this registry has been the transplant-free survival at 21 days post study inclusion [7].
Statistical analysis
Continuous or categorical variables were described as median (interquartile range (IQR)) or frequency (%), respectively. Missing data across all values were 4.0% and no multiple imputation was performed.
Univariable comparisons used: linear regression between 2 continuous variables; Mann-Whitney test between one continuous and one categorical variables; and chi-square test between 2 categorical variables.
Time-to-event analysis with Kaplan Meier curve was used to plot overall cumulative survival. Adjusted analysis required: multiple linear regression to study associations with a continuous outcome (INR or explanted liver weight); or logistic regression to study associations with a categorical outcome (TFS). Variables known to be clinically significant confounders or with a P<0.10 on univariable comparisons were initially included in the models. A backward stepwise approach was used to create the final models yielding a good fit while avoiding overfitting (one covariable per each 10 events. Collinearity was assessed by the variance inflation factor (VIF), where values closer to one indicate low risk of collinearity.
The discriminative ability of different prognostic scores for TFS was assessed by the area under the receiver operating characteristic curve (AUCROC). Comparisons between AUCROCs were performed with the Hanley-McNeil test.
External validation was performed using a similar approach to the initial multivariable logistic regression but applied to the USALFSG validation cohort. Calibration was assessed by calibration plots (with Loess smoothing), R2 (the higher the score, the better the predictive performance), and Brier score (the lower the score, the better the predictive performance) [16].
The threshold for statistical significance was defined as α=0.05 (2-tailed). Statistical analysis was performed using IBM SPSS Statistics, version 28 (IBM Corp, North Castle, NY, US) and R, version 4.3.3, with the rms, predtools, and CalibrationCurves packages.
RESULTS
Baseline characteristics
Based on local ICU registries, a total of 191 patients with ALF were included, 99 from Curry Cabral Hospital and 92 from Clinic Hospital. Among all ALF patients, median (IQR) age was 46 (32;57) years and 85 (44.5%) were males (Table 1). Non-acetaminophen etiologies were found in 164 (85.9%) patients (Figure S1).
Table 1. Baseline characteristics of acute liver failure patients stratified by transplant-free survival at 30 days following intensive care unit admission.
| Characteristics (n (%) or median (IQR)) | ALF (n=191) | 30-day LT or death (n=133) | 30-day TFS (n=58) | P |
|---|---|---|---|---|
| Demography | ||||
| Age (years) (n=191) | 46 (32;57) | 46 (35;60) | 44 (29;52) | 0.07 |
| Male sex (n=191) | 85 (44.5%) | 53 (39.8%) | 32 (55.2%) | 0.050 |
| Actual body weight (Kg) (n=157) | 70 (60;80) | 70 (60;80) | 70 (60;80) | 0.93 |
| Non-acetaminophen etiology (n=191) | 164 (85.9%) | 119 (89.5%) | 45 (77.6%) | 0.030 |
| Severity of disease on ICU day one | ||||
| Grade 3–4 HE (n=189) | 91 (47.7%) | 79 (60.3%) | 12 (20.7%) | <0.001 |
| IMV (n=191) | 87 (45.5%) | 74 (55.6%) | 13 (22.4%) | <0.001 |
| PF ratio (mmHg) (n=167) | 350 (253;476) | 320 (232;473) | 429 (311;476) | 0.010 |
| Vasopressors (n=191) | 78 (40.8%) | 65 (48.9%) | 13 (22.4%) | <0.001 |
| MAP (mmHg) (n=184) | 81 (69;92) | 81 (67;92) | 80 (76;92) | 0.10 |
| Acute kidney injury (n=191) | 110 (57.6%) | 84 (63.2%) | 26 (44.8%) | 0.018 |
| RRT (n=190) | 61 (32.1%) | 50 (37.6%) | 11 (19.3%) | 0.013 |
| SOFA score (n=185) | 9 (5;14) | 12 (7;15) | 6 (4;9) | <0.001 |
| Laboratory on ICU admission | ||||
| Hemoglobin (g/dl) (n=191) | 123 (100;136) | 121 (96;136) | 129 (109;140) | 0.13 |
| Leucocytes (n/109) (n=191) | 9.5 (6.3;14.9) | 9.8 (6.2;15.0) | 8.7 (6.3;13.1) | 0.55 |
| Platelets (n/109) (n=191) | 131 (75;205) | 127 (63;206) | 140 (78;205) | 0.55 |
| Creatinine (mg/dl) (n=191) | 1.07 (0.68;1.93) | 1.10 (0.67;2.00) | 0.99 (0.70;1.89) | 0.91 |
| Urea (mg/dl) (n=190) | 36 (17;66) | 34 (16;68) | 39 (22;63) | 0.49 |
| Sodium (mmol/l) (n=191) | 139 (136;142) | 139 (136;143) | 138 (135;141) | 0.21 |
| Potassium (mmol/l) (n=191) | 3.9 (3.4;4.3) | 3.9 (3.5;4.4) | 3.9 (3.4;4.1) | 0.37 |
| Phosphate (mmol/l) (n=181) | 2.9 (2.2;4.1) | 2.9 (2.2;4.1) | 3.0 (2.2;4.0) | 0.78 |
| Ammonia (μmol/l) (n=172) | 138 (87;182) | 146 (105;188) | 102 (60;145) | <0.001 |
| INR (n=191) | 3.0 (2.0;4.6) | 3.2 (2.3;5.3) | 2.2 (1.7;3.2) | <0.001 |
| Bilirubin (mg/dl) (n=191) | 10.3 (3.8;19.7) | 13.3 (4.3;22.1) | 7.4 (3.2;15.2) | 0.013 |
| Alanine transferase (U/l) (n=190) | 1860 (578;3946) | 1748 (516;3162) | 2027 (698;5372) | 0.20 |
| Lactate (mmol/l) (n=187) | 2.5 (1.9;5.6) | 3.4 (2.0;7.1) | 2.0 (1.4;2.6) | <0.001 |
IQR: interquartile range; ALF: acute liver failure; HE: hepatic encephalopathy; IMV: invasive mechanical ventilation; PF ratio: arterial oxygen partial pressure / inspired oxygen fraction; MAP: mean arterial pressure; RRT: renal replacement therapy; INR: international normalized ratio; SOFA: sequential organ failure assessment; ICU: intensive care unit.
On ICU day one, 91 (47.7%) patients developed grade 3–4 HE, 87 (45.5%) required IMV, and 78 (40.8%) were on vasopressors. Furthermore, 110 (57.6%) patients had AKI and 61 (32.1%) required RRT. Overall, median (IQR) SOFA score was 9 (5;14).
On ICU admission, median (IQR) arterial ammonia, urea, creatinine, and INR were 138 (87;182) μmol/l, 36 (17;66) mg/dl, 1.07 (0.68;1.93) mg/dl, and 3.0 (2.0;4.6), respectively. All baseline characteristics of this cohort are depicted in Table 1.
Clinical outcomes
Among all ALF patients, 86 (45.0%) patients were transplanted and 75 (39.3%) died during follow-up (Table 1). TFS at 30 days post ICU admission was 30.4% (58/191). LT (45.5% vs 44.6%; P=0.91) and TFS (28.3% vs 32.6%; P=0.52) at 30 days post ICU admission rates were similar between Lisbon and Barcelona centers. Median (IQR) time to LT or death were 3 (2;6) and 7 (3;17) days, respectively (Figure S2). All causes of death are summarized in Figure S3. Median (IQR) ICU and hospital LOS were 7 (4;12) and 20 (9;32) days, respectively. All clinical outcomes of this cohort are detailed in Table 1.
Associations of ICU admission serum ammonia and urea with INR
On univariable analysis, the higher the ammonia, the higher the INR on ICU admission (Figure 1A: INR = 2.79 + 0.007 Ammonia; correlation coefficient of 0.35; P<0.001). Conversely, the lower the urea, the higher the INR on ICU admission (Figure 1B: INR = 4.30 – 0.015 Urea; correlation coefficient of 0.18; P=0.015). Moreover, creatinine was not associated with INR on ICU admission (P=0.27).
Figure 1. International normalized ratio variation with serum ammonia (panel A) or urea (panel B) in acute liver failure.
Among INR quantiles (<2.0, 2.0–2.8, 2.9–4.9, and >4.9), median (IQR) ammonia was 106 (55;147), 112 (75;164), 132 (102;189), and 173 (142;280) μmol/l, respectively (P<0.001); and median (IQR) urea was 47 (29;73), 41 (24;72), 32 (15;63), and 30 (13;47) mg/dl, respectively (P=0.007). Additionally, there was no significant variation of creatinine among these INR quantiles (P=0.30). Overall, higher creatinine was associated with higher urea (P<0.001). The pattern of lower urea among higher INR quantiles was significant if there was AKI (higher creatinine levels) on ICU day one (P=0.006), but not without AKI (P=0.08) (Figure S4).
Following adjustment for age, sex, actual body weight, and etiology (acetaminophen vs other), higher ammonia (P<0.001) or lower urea (P=0.008) were both independently associated with higher INR on ICU admission (Table S1). There was no collinearity between ammonia and urea (VIF <1.15).
Associations of ICU admission serum ammonia and urea with TFS
On univariable analysis, the higher the ICU admission ammonia (146 vs 102μmol/l; P<0.001) the lower the TFS (Table 1). On the contrary, urea was similar between transplant free survivors and others (34 vs 39mg/dl; P=0.49). Following adjustment for sex, etiology (acetaminophen vs other), and ICU admission lactate, higher ICU admission ammonia was independently associated with lower odds of TFS (Table S2: adjusted odds ratio (aOR) (95% confidence interval (CI)) = 0.991 (0.985;0.997) per each unit of ammonia (μmol/l); P=0.004). Additionally, while male sex and acetaminophen etiology were independently associated with higher odds of TFS, higher ICU admission lactate was independently associated with lower odds of TFS.
Discriminative ability between different prognostic scores for TFS
On ICU day one, 89 (46.6%) patients fulfilled KCC; overall, median (IQR) MELD and US-ALF-SG scores were 29 (24;35) and −1.06 (−2.18;−0.273), respectively.
The KCC and MELD score had the poorest discriminative ability for TFS (Table S3: AUROC (95% CI) = 0.64 (0.55;0.72) and 0.69 (0.60;0.77), respectively). The US-ALF-SG score (AUROC (95% CI) = 0.83 (0.77;0.89)) and the newly derived model incorporating ammonia (AUROC (95% CI) = 0.81 (0.75;0.88); P=0.003) showed significantly better discriminative ability for TFS than KCC (Figure S5).
To study a potential cutoff for ammonia with clinical utility, we plotted adjusted predicted probabilities of TFS, based on the derived logistic regression model, against observed ammonia levels (Figure 2). Among ALF patients with an ICU admission ammonia >300 μmol/l, TFS probability was mostly <30%.
Figure 2. Adjusted predicted probabilities of 30-day transplant-free survival based on serum ammonia in acute liver failure.
External validation of the derived model with ammonia
Among all ALF patients in the USALFSG cohort (n=1186), median (IQR) age was 39 (29;52) years and 367 (30.9%) were males (Table S4). Non-acetaminophen etiologies were found in 593 (50.0%) patients. On study day one, 600 (50.6%) patients developed grade 3–4 HE, 632 (53.2%) required IMV, 299 (25.2%) were on vasopressors, and 314 (26.5%) required RRT. Within the USALFSG registry, standardized TFS at 21 days post study inclusion was 56.9% (493/827). All baseline characteristics of this cohort are depicted in Table S4.
To validate the derived model with ammonia, the same logit was computed using the validation set: y = 0.714 + 0.976*Sex (1 if male or 0 if female) + 2.036* Etiology (1 if acetaminophen or 0 if other) - 0.324* Lactate (mmol/l) - 0.009*Ammonia (μmol/l). Accordingly, the probabilities of TFS were given by the following equation: exp(logit)/(exp(logit)+1). In the validation set, this model retained good discriminative ability (AUROC (95% CI) = 0.78 (0.75;0.82)) and showed reasonable calibration performance (R2 of 0.43 and Brier score of 0.20) (Table S5). In fact, the model only underestimated TFS in first 2 quintiles of the distribution of probabilities, i.e. for patients with lower probabilities of spontaneous survival (Figure 3B).
Figure 3. Calibration plots for the derived model with ammonia in the derivation and validation sets. Predicted transplant-free survival was plotted against observed transplant-free survival after Loess smoothing method. Points below the diagonal line (perfect prediction) indicate overestimation and points above that line mean underestimation of transplant-free survival.
Associations of ICU admission serum ammonia and urea with explanted liver weight
Among transplanted patients (n=86), median (IQR) explanted liver weight following LT was 980 (792;1425) g (Table 1). Explanted liver weight did not vary significantly with actual body weight (P=0.67), ALF etiology (P=0.16) or time to LT (P=0.30).
On univariable analysis, the lower the ICU admission urea the lower the explanted liver weight (Figure 4: Liver weight = 734.7 + 10.4 Urea; P<0.001). Following adjustment for age, sex, actual body weight, and etiology (acetaminophen vs other), ICU admission higher ammonia (P=0.024) or lower (P<0.001) urea were both independently associated with lower explanted liver weight (Table S6). Additionally, acetaminophen etiology was also independently associated with higher explanted liver weight. There was no collinearity between ammonia and urea (VIF <1.15).
Figure 4. Explanted liver weight variation with serum urea in acute liver failure.
DISCUSSION
Main findings and comparisons with previous literature
In our southern European (Portugal and Spain) cohort of ALF patients, with largely predominant non-acetaminophen etiologies, we found that higher ICU admission serum ammonia and lower urea were both independently associated with higher liver-related disease severity, measured by INR or explanted liver weight.
There is a paucity of literature about the impact of acquired impaired urea cycle activity in ALF, except in the context of specific urea cycle inherited disorders [2–4,17,18]. In ALF patients, the larger the liver necrosis, the higher the disease severity, more often serially measured by INR [1,14]. In severe ALF (higher INR), there may be reduced conversion of ammonia to urea in the damaged liver. Therefore, even in the context of early hypovolemia and AKI, common in these patients, serum urea may not follow the upward trend in serum creatinine [17]. Accordingly, we suggest that serum urea may be an additional early indicator of greater ALF severity due to impaired urea cycle activity. Conversely, serum urea may not be a good surrogate of kidney function in severe ALF patients.
In our cohort of ALF patients, we showed that higher serum ammonia or lower serum urea on ICU admission were independently associated with lower explanted liver weight following LT. Additionally, patients with acetaminophen ALF were less likely to have lower volume livers following LT. Reduced liver volume, evaluated by computed tomography, has been shown to be associated with higher probability of developing complications in ALF patients, such as HE and death, especially in those with non-acetaminophen etiologies [5,19]. Acetaminophen toxicity develops hyperacutely thus there is often less time for the liver to collapse. Overall, submassive or massive necrosis has been associated with lower liver size (due to structural collapse), impaired liver regeneration, and worse survival without LT [20–22]. In this context, we speculate that serum ammonia or urea may add as additional early prognostic indicators in ALF patients, as its levels may indicate the risk of liver collapse and less potential for spontaneous regeneration.
Finally, we also found that higher day one serum ammonia was independently associated with lower TFS. On the contrary, serum urea was not associated with TFS. In ALF patients, arterial ammonia has been widely used as a surrogate for disease severity, especially its neurological complications such as cerebral edema [6,7]. However, pre-existent prognostic scores, such as KCC, MELD, or US-ALF-SG scores, have not incorporated serum ammonia as a prognostic factor.
In our cohort of ALF patients, the newly derived model incorporating ammonia showed better discrimination for TFS than KCC and MELD score, both developed decades ago [14]. However, its discriminative ability was similar to the USALFSG score, more recently developed with a North-American cohort with a higher proportion of acetaminophen ALF patients [15]. In fact, our study was one of the first to use a non- North-American cohort to externally validate the USALFSG prognostic index.
To better study the overall predictive ability of our newly derived model with ammonia, we validated it with an up-to-date methodology, including calibration plots, using the USALFSG cohort. In fact, this new model showed good discrimination and reasonable calibration for TFS in this validation set. The fact that it underestimated TFS for patients with the lowest probabilities of spontaneous survival may be due to substantial differences among the derivation and validation sets. Most importantly, the validation cohort had a higher prevalence of acetaminophen ALF than ours (50.0% vs. 14.1%), a more benign etiology known to be associated with higher TFS (56.9% vs. 30.1%) [1,10,14,15]. However, as calibration for most patients was reasonably good, the newly derived model incorporating ammonia may be useful to provide individual estimations of the probability of TFS for ALF patients.
The decision to select ALF patients for LT usually considers the potential for lack of liver regeneration and death without LT. Most commonly, transplant risk assessment in ALF is mainly estimated based on clinical and laboratory parameters, such as those included in the KCC, MELD, and USALFSG scores [14,15]. However, all these scores have limitations regarding sensitivity and specificity, especially for non-acetaminophen ALF [23,24]. Therefore, our newly derived model including 4 simple parameters (sex, etiology, and day one serum lactate and ammonia) may constitute a promising alternative prognostic value with clinical utility.
Building on biological plausibility, the findings of our pilot study point towards potential clinical utility of both serum ammonia and urea as early ALF severity markers, as they take part on the same liver metabolic pathway. While urea was associated with ALF severity, ammonia was also associated with TFS.
The results of our study need to be interpreted considering the following limitations. Firstly, this was a dual center observational study. Regional variability in ALF etiologies may have contributed to selection bias and may hinder internal validity. However, we think the standardized entry criteria for local ALF registries, as well as the up-to-date approach to ALF at both centers, may have helped to minimize this limitation. Secondly, the number of subjects included may have precluded a more powerful study of associations between serum urea and TFS. Nevertheless, as ALF is a rare disease, we believe the size of our cohort was right to study the association of serum urea with ALF severity. Thirdly, many factors may have influenced both serum ammonia and urea levels. However, ALF patients on ICU admission still constitute a good model to study early nitrogen metabolism derangements without being influenced by factors that may impair such metabolism during prolonged critical illness, such as persistent catabolism, loss of muscle mass, or dietary nitrogen supplementation [24]. Furthermore, all ALF patients in our cohort had ICU admission serum ammonia and urea levels measured prior to RRT initiation, a potential modifier of these serum levels [7].
Despite these limitations, we believe our study adds to the literature by providing further data to better understand how ALF pathophysiology may be applied to clinical practice and possibly decision making. Future larger studies could better address the impact of ammonia and urea metabolism in the selection of sicker ALF patients for LT, perhaps by further integrating and validating them with available or new prognostic scores [26].
Conclusions
In a large cohort of ALF patients, serum ammonia and urea were associated with ALF severity. A prognostic score incorporating serum ammonia yielded reasonably good predictive ability for TFS.
Supplementary Material
ACKNOWLEDGMENTS
To the staff of participating intensive care units.
To the Portuguese Society of Intensive Care (Liver Failure Group).
To the United States Acute Liver Failure Study Group (1998–2018): W.M. Lee, MD (Principal Investigator); Anne M. Larson, MD, Iris Liou, MD, University of Washington, Seattle, WA; Oren Fix, MD, Swedish Medical Center, Seattle, WA; Michael Schilsky,
MD, Yale University, New Haven, CT; Timothy McCashland,MD, University of Nebraska, Omaha, NE; J. Eileen Hay, MBBS, Mayo Clinic, Rochester, MN; Natalie Murray, MD, Baylor University Medical Center, Dallas, TX; A. Obaid S. Shaikh, MD, University of Pittsburgh, Pittsburgh, PA; Andres Blei, MD, Northwestern University, Chicago, IL (deceased), Daniel Ganger, MD, Northwestern University, Chicago, IL; Atif Zaman, MD, University of Oregon, Portland, OR; Steven H. B. Han, MD, University of California, Los Angeles, CA; Robert Fontana, MD, University of Michigan, Ann Arbor, MI; Brendan McGuire, MD, University of Alabama, Birmingham, AL; Raymond T. Chung, MD, Massachusetts General Hospital, Boston, MA; Alastair Smith, MB, ChB, Duke University Medical Center, Durham, NC; Robert Brown, Jr., MD, Cornell/Columbia University, New York, NY; Jeffrey Crippin, MD, Washington University, St. Louis, MO; Edwin Harrison, Mayo Clinic, Scottsdale, AZ; Adrian Reuben, MBBS, Medical University of South Carolina, Charleston, SC; Santiago Munoz, MD, Albert Einstein Medical Center, Philadelphia, PA; Rajender Reddy, MD, University of Pennsylvania, Philadelphia, PA; R. Todd Stravitz, MD, Virginia Commonwealth University, Richmond, VA; Lorenzo Rossaro, MD, University of California Davis, Sacramento, CA; Raj Satyanarayana, MD, Mayo Clinic, Jacksonville, FL; Tarek Hassanein, MD, University of California, San Diego, CA; Constantine J. Karvellas MD, University of Alberta, Edmonton, AB; Jodi Olson, MD, University of Kansas, Kansas City, KA; Ram Subramanian, MD, Emory, Atlanta, GA; James Hanje, MD, Ohio State University, Columbus, OH; Bilal Hameed, MD, University of California San Francisco, CA. The University of Texas Southwestern Administrative Group included Grace Samuel; Ezmina Lalani; Carla Pezzia; Corron Sanders, PhD; Nahid Attar; Linda S. Hynan, PhD; and the Medical University of South Carolina Data Coordination Unit included Valerie Durkalski, PhD; Wenle Zhao, PhD; Jaime Speiser; Catherine Dillon; Holly Battenhouse; and Michelle Gottfried. Funding: NIH grant U‐01 58369.
Financial support: Funding for the ALFSG provided by NIDDK, DK R-01 58369 and DK U-01 58369 to UT Southwestern Medical Center, Dallas Texas.
Abbreviations:
- ALF
acute liver failure
- AKI
acute kidney injury
- CI
confidence interval
- HE
hepatic encephalopathy
- ICU
intensive care unit
- IMV
invasive mechanical ventilation
- INR
international normalized ratio
- IQR
interquartile range
- LT
liver transplant
- LOS
length-of-stay
- MAP
mean arterial pressure
- MELD
model for end-stage liver disease
- OR
odds ratio
- PF ratio
arterial oxygen partial pressure / inspired oxygen fraction
- RRT
renal replacement therapy
- SOFA
sequential organ failure assessment
- TFS
transplant-free survival
- USALFSG
United States acute liver failure study group
Footnotes
Conflicts of interest: none to be reported.
REFERENCES
- [1].Cardoso FS, Marcelino P, Bagulho L, et al. Acute liver failure: An up-to-date approach. J Crit Care 2017;39:25–30. [DOI] [PubMed] [Google Scholar]
- [2].Wright G, Noiret L, Olde Damink SW, et al. Interorgan ammonia metabolism in liver failure: the basis of current and future therapies. Liver Int 2011;31(2):163–175. [DOI] [PubMed] [Google Scholar]
- [3].Walker V Ammonia metabolism and hyperammonemic disorders. Adv Clin Chem 2014;67:73–150. [DOI] [PubMed] [Google Scholar]
- [4].Weiner ID, Mitch WE, Sands JM. Urea and Ammonia Metabolism and the Control of Renal Nitrogen Excretion. Clin J Am Soc Nephrol 2015;10(8):1444–1458. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Zabron A, Quaglia A, Fatourou E, et al. Clinical and prognostic associations of liver volume determined by computed tomography in acute liver failure. Liver Int 2018;38(9):1592–1601. [DOI] [PubMed] [Google Scholar]
- [6].Bernal W, Hall C, Karvellas CJ, et al. Arterial ammonia and clinical risk factors for encephalopathy and intracranial hypertension in acute liver failure. Hepatology 2007;46(6):1844–1852. [DOI] [PubMed] [Google Scholar]
- [7].Cardoso FS, Gottfried M, Tujios S, et al. Continuous renal replacement therapy is associated with reduced serum ammonia levels and mortality in acute liver failure. Hepatology 2018;67(2):711–720. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].World Medical Association. World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. JAMA 2013;310(20):2191–2194. [DOI] [PubMed] [Google Scholar]
- [9].von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet 2007;370(9596):1453–1457. [DOI] [PubMed] [Google Scholar]
- [10].European Association for the Study of the Liver. Electronic address: easloffice@easloffice.eu; Clinical practice guidelines panel, Wendon J, et al. EASL Clinical Practical Guidelines on the management of acute (fulminant) liver failure. J Hepatol 2017;66(5):1047–1081. [DOI] [PubMed] [Google Scholar]
- [11].Raith EP, Udy AA, Bailey M, et al. Prognostic Accuracy of the SOFA Score, SIRS Criteria, and qSOFA Score for In-Hospital Mortality Among Adults With Suspected Infection Admitted to the Intensive Care Unit. JAMA 2017;317(3):290–300. [DOI] [PubMed] [Google Scholar]
- [12].Thomas ME, Blaine C, Dawnay A, et al. The definition of acute kidney injury and its use in practice. Kidney Int 2015;87(1):62–73. [DOI] [PubMed] [Google Scholar]
- [13].Chawla LS, Bellomo R, Bihorac A, et al. Acute kidney disease and renal recovery: consensus report of the Acute Disease Quality Initiative (ADQI) 16 Workgroup. Nat Rev Nephrol 2017;13(4):241–257. [DOI] [PubMed] [Google Scholar]
- [14].McPhail MJ, Farne H, Senvar N, et al. Ability of King’s College Criteria and Model for End-Stage Liver Disease Scores to Predict Mortality of Patients With Acute Liver Failure: A Meta-analysis. Clin Gastroenterol Hepatol 2016;14(4): 516–525.e5. [DOI] [PubMed] [Google Scholar]
- [15].Koch DG, Tillman H, Durkalski V, et al. Development of a Model to Predict Transplant-free Survival of Patients With Acute Liver Failure. Clin Gastroenterol Hepatol 2016;14(8):1199–1206.e2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Steyerberg EW, Vickers AJ, Cook NR, et al. Assessing the performance of prediction models: a framework for traditional and novel measures. Epidemiology 2010; 21(1):128–138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Moore JK, Love E, Craig DG, et al. Acute kidney injury in acute liver failure: a review. Expert Rev Gastroenterol Hepatol 2013;7(8):701–712. [DOI] [PubMed] [Google Scholar]
- [18].Bigot A, Tchan MC, Thoreau B, et al. Liver involvement in urea cycle disorders: a review of the literature. J Inherit Metab Dis 2017;40(6):757–769. [DOI] [PubMed] [Google Scholar]
- [19].Ikura A, Chu PS, Nakamoto N, et al. CLIF-C Organ Failure Score and Liver Volume Predict Prognosis in Steroid-Treated Severe Acute Autoimmune Hepatitis. Hepatol Commun 2020;4(7):1019–1033. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Hanau C, Munoz SJ, Rubin R. Histopathological heterogeneity in fulminant hepatic failure. Hepatology 1995;21(2):345–351. [PubMed] [Google Scholar]
- [21].Ndekwe P, Ghabril MS, Zang Y, et al. Substantial hepatic necrosis is prognostic in fulminant liver failure. World J Gastroenterol 2017;23(23):4303–4310. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].Fujita M, Furukawa H, Hattori M, et al. Sequential observation of liver cell regeneration after massive hepatic necrosis in auxiliary partial orthotopic liver transplantation. Mod Pathol 2000;13(2):152–157. [DOI] [PubMed] [Google Scholar]
- [23].Bernal W, Wang Y, Maggs J, et al. Development and validation of a dynamic outcome prediction model for paracetamol-induced acute liver failure: a cohort study. Lancet Gastroenterol Hepatol. 2016;1(3):217–225. [DOI] [PubMed] [Google Scholar]
- [24].Speiser JL, Lee WM, Karvellas CJ, et al. Predicting outcome on admission and post-admission for acetaminophen-induced acute liver failure using classification and regression tree models. PLoS One 2015;10(4):e0122929. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [25].Haines RW, Fowler AJ, Wan YI, et al. Catabolism in Critical Illness: A Reanalysis of the REducing Deaths due to OXidative Stress (REDOXS) Trial. Crit Care Med 2022;50(7):1072–1082. [DOI] [PubMed] [Google Scholar]
- [26].Kamath PS, Heimbach J, Wiesner RH. Acute Liver Failure Prognostic Scores: Is Good Enough Good Enough? Clin Gastroenterol Hepatol 2016;14(4):621–623. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.