A modified Child-Turcotte-Pugh score based on plasma ammonia predicts survival for patients with decompensated cirrhosis

Summary

Background

Plasma ammonia has been shown to be an independent prognostic factor for patients with liver cirrhosis.

Aim

We aimed to investigate the discrimination and calibration of a new prognostic model (aCTP) based on plasma ammonia (Amm) replacing hepatic encephalopathy (HE) in the Child-Turcotte-Pugh (CTP) score.

Design

Diagnostic test, paired design.

Methods

Baseline Amm levels were corrected to the upper limit of normal (Amm-ULN). We designed the new model based on the cut-off value of Amm-ULN in the observational cohort of 554 clinically stable cirrhotic patients from January 2012 to July 2019. External validation was carried out using prospective data from 185 patients with liver cirrhosis.

Results

Our analysis showed that each 1-point increase in Amm-ULN was associated with a 2-fold increase in the likelihood of mortality [hazard ratio (HR), 2.06; 95% CI: 1.81–2.36, P < 0.001]. In the aCTP score, Amm-ULN < 1.0 was defined as a score of 1, 1.0 ≤ Amm-ULN < 1.4 as 2 and ≥1.4 as 3. The survival curves among three aCTP grades were significantly different (P < 0.0001). The aCTP score showed the better agreements between predicted and observed events in the validating cohorts than the CTP score (C-statistics: 0.75 and 0.69, P < 0.001). The aCTP score showed inspiring power to predict acute decompensation (C-statistics: 0.76; 95% CI: 0.71–0.81) and acute-on-chronic liver failure (C-statistics: 0.81; 95% CI: 0.77–0.86).

Conclusion

This study demonstrates the feasibility and the potential for plasma Amm replacing HE (aCTP) to enhance the prognostication of transplant-free survival provided by the CTP score for patients with decompensated cirrhosis.

Introduction

The Child-Turcotte-Pugh score (CTP score) was first proposed by Child and Turcotte to predict surgical risk in patients undergoing portosystemic shunt surgery for variceal bleeding.^1,2 The CTP score is now widely used to assess the severity of liver dysfunction in the clinical practice.^3,4 However, the CTP score has some drawbacks. The ascites and the hepatic encephalopathy (HE) included in the CTP score are subjective in a certain degree, and may vary according to the usage of drugs. Moreover, a frequently reported drawback of the CTP score is that the neglect of the minimal HE.⁵ Therefore, in the current CTP score, the liver function grades for patients with minimal HE may be misevaluated by clinicians.

Recently, Ammonia (Amm) was proved to play a more and more important role in chronic liver disease.⁶ The role of ammonia in the progression of liver cirrhosis was confirmed in vitro and in vivo.^7–9 In the study of Tranah et al.,¹⁰ baseline plasma Amm was proposed to be a predictor of the decompensated events in the liver cirrhosis. It was concluded that the plasma Amm had a similar efficacy to the CTP score in predicting the prognosis of cirrhotic patients, such as readmission and survival. However, the external validation of the role of plasma Amm in predicting the prognosis in the patients with decompensated cirrhosis is lacking. In addition, the study that evaluated the association between the plasma Amm and the mortality from the advanced liver disease was conducted in a European population that differed from the East Asian population (e.g. regarding common causes, different levels of tolerance to Amm and different health care regimes).

Herein, we aimed to verify the association between plasma Amm and mortality and to investigate the impact of plasma Amm replacing HE on the discrimination and calibration of the CTP score based on plasma Amm (aCTP score).

Materials and methods

Study design and patients selection

An observational cohort study project based on the clinically stable decompensated cirrhotic outpatients to assess prognosis of different therapies in Nanjing Drum Tower Hospital (DCDT COHORT STUDY) was conducted in the Drum Tower Hospital in 2012. This analysis is a part of this Clinical Program.

The inclusion criteria of DCDT COHORT were as follows: (i) cirrhotic patients aged ≥18 years; (ii) patients with the previous history of esophageal and gastric variceal bleeding (EGVB) and refractory ascites (RA) were included in the study; (iii) patients experienced the secondary prophylaxis. The exclusion criteria were as follows: (i) acute decompensation (AD) within the previous 1 week; (ii) only endoscopic therapy was performed without NSBBs usage; (iii) malignant tumors; (iv) respiratory failure, severe heart failure and kidney dysfunction diagnosed before cirrhosis; (v) pregnancy or lactation; (vi) no informed consent was signed. The study protocol was approved by the ethics committees of the Nanjing Drum Tower Hospital. Until July 2019, 554 patients were included in the DCDT COHORT.

Then we retrospectively screened the decompensated cirrhotic patients who were admitted to the Nanjing Drum Tower Hospital from August 2019 to June 2020 in the prospectively database as the validating cohort. The inclusion criteria were as follows: (i) cirrhotic patients aged ≥18 years; (ii) patients with the previous history of EGVB, RA, spontaneous peritonitis and hepatorenal syndrome (HRS) were included in the study. The exclusion criteria of this study were as follows: (i) AD within the previous 1 week; (ii) concomitant non-liver-related malignant tumors; (iii) severe heart failure (NYHA 4) and kidney dysfunction (CKD 5), diagnosed before cirrhosis; (iv) pregnancy or lactation; and (v) missing follow-up data.

The research was conducted in accordance with both the Declarations of Helsinki and Istanbul and approved by the Drum Tower Hospital Ethics Committee. All patients included had written the consent.

Plasma Amm and aCTP score

Standard operating procedures for the plasma Amm measurement were used: blood were taken and sent to the laboratory immediately. The concentrations of the plasma Amm were measured by the dry-chemical-reagent-method. The ratios of plasma Amm to the upper limit of normal (Amm-ULN) were calculated.³ The CTP scores and the CTP grades were calculated according to the literature.³

We hypothesized that the plasma Amm was independently associated with survival. To test this hypothesis, the receiver operating characteristics curve (ROC) of the AMM-ULN for predicting 3-year survival was painted and the area under ROC was calculated. Then we used the Cox regression analysis method to determine the independent risk factors of the liver transplantation or the death. We did not find any interaction between CTP scores and Amm-ULN by drawing 3-knot, 5-knot and 7-knot cubic splines¹¹ in DCDT COHORT. The optimal cut-off for Amm-ULN was calculated using the DCDT COHORT.¹² We used this cut-off to build aCTP score. According to the C-statistics based on the Cox analysis, we compared the ability of the composite score with AMM-ULN alone.

Treatment and outcome

Baseline laboratory indicators were set to be tested within 1 week before therapy. Outpatient follow-up was performed regularly. The primary endpoint of the study was death, and the chance of liver transplantation as the competing risk, and the secondary endpoint was AD.¹³ The AD was divided into acute-on-chronic liver failure (ACLF)¹⁴ and non-ACLF.¹⁵ Patients were registered when they were rehospitalized for RA, rebleeding, HE, HRS and ACLF.

Statistical analysis

Data were described as frequencies and percentages, means and standard deviations, or medians and interquartile ranges, as appropriate. Baseline characteristics were compared using Fisher’s exact test or chi-squared test for categorical variables, variance analysis for continuous variables and the Wilcoxon rank sum test for ordinal and incontinuous variables.

The Kaplan–Meier curve based on the aCTP grades were demonstrated. We compared the diagnostic efficacy of aCTP and CTP scores by measuring C-statistics,¹¹ calibration plots,¹² Decision Curve Analysis,¹⁶ brier score comparisons¹⁰ and the net reclassification index¹¹ (Supplementary Methods).

All analyses were performed using R studio (version 4.2.1) (Supplementary Methods). A level of significance was established at the two-sided 5% level.

Results

A total of 739 patients were included (Figure S1). Baseline characteristics are summarized in Table 1.

Table 1.

Baseline characteristics

	DCDT COHORTV	Validating cohort	P-Value
	(n = 554)	(n = 185)
Age (years)	56.99 ± 12.08	58.67 ± 12.08	0.10
Gender (male/female)	327/227	106/79	0.73
Ascite (yes/no)	382/169	140/45	0.11
Cause	295/49/210	96/18/71	0.92
(virus/alcoholic/others)
Platelets (×109/L)	68.00 (48.00, 107.00)	69.00 (47.50–123.00)	0.73
ALT (U/L)	24.50 (16.10, 39.50)	22.60 (16.35–36.75)	0.27
AST (U/L)	31.30 (23.00, 45.60)	31.60 (22.80–50.10)	0.78
Tbil (μmol/L)	19.05 (13.10, 29.03)	20.50 (14.65–37,75)	0.01
Creatinine (μmol/L)	64.82 ± 24.68	63.03 ± 21.00	0.37
Albumin (g/L)	34.51 ± 4.55	33.02 ± 3.98	<0.0001
Prothrombin time (s)	14.38 ± 2.04	14.79 ± 2.39	0.02
INR	1.25 ± 0.18	1.29 ± 0.20	0.01
Na (mmol/L)	139.75 ± 3.40	139.59 ± 4.34	0.61
CTP score	6.81 ± 1.30	7.65 ± 1.71	<0.0001
CTP grade (A/B/C)	255/281/18	52/103/30	<0.0001
plasma Amm	35.99 ± 23.64	39.64 ± 23.81	0.07
aCTP score	7.58 ± 1.47	8.29 ± 2.06	<0.0001
aCTP grade (A/B/C)	129/355/70	36/102/47	<0.0001
MELD score	4.31 ± 5.25	5.36 ± 6.04	0.02
MELD-Na score	4.63 ± 5.40	5.65 ± 6.89	0.04

Validating the effect of plasma Amm

In the DCDT COHORT, the median follow-up of the total patients was 45 months. Overall, 205 (37%) patients experienced liver transplantation (6.34%) or death (13.66% HCC, 9.27% non-liver-related-death, 35.56% ACLF-AD and 35.17% non-ACLF-AD). We validated the effect of the plasma Amm on predicting the prognosis in the DCDT COHORT. The area under the ROC of Amm-ULN for predicting 3-year-survival was 0.78 (95% CI: 0.73–0.83, P < 0.0001) (Figure S2). The cut-off value of Amm-ULN for predicting 3-year survival was 1.415, which is similar to the value of 1.4 reported by Tranah et al.¹⁰ We revalidated the effects of each of the components in the CTP score. The independent risk factors for mortality were shown in Table 2.

Table 2.

Cox regression evaluation of factors associated with mortality

Variable	Univariate analysis		Multivariate analysis
	Hazard ratio	P-Value	Hazard ratio	P-Value
Age (y)	1.01 (1.01–1.02)	0.02	–	–
INR	10.1 (5.04–20.26)	0.01	4.30 (1.95–9.50)	<0.001
Amm-ULN	2.15 (1.90–2.43)	<0.001	2.06 (1.81–2.36)	<0.001
Tbil (μmol/L)	1.01 (1.01–1.02)	<0.001	1.006 (1.003–1.01)	0.01
Cr (μmol/L)	1.007 (1.003–1.01)	<0.001	1.007 (1.003–1.01)	<0.001
Ascites1 (compared with without ascites)
Mild	1.84 (1.14–2.97)	0.01	2.01 (1.53–2.89)	<0.001
Moderate–severe	1.99 (1.46–2.70)	<0.001	2.17 (1.32–3.56)	0.01
ALB (g/L)	0.96 (0.93–0.99)	0.004	–	–
HE (compared with/without HE)
Mild	1.52 (0.91–2.53)	0.11	–	–
Moderates–severe	1.47 (1.10–1.99)	0.01	–	–

Development of aCTP score

After the adjustment for the CTP score, we calculated the C-statistics of the plasma Amm (0.65; 95%CI: 0.60–0.70). To make the score simple, the cut-off value of 1.4 was used to calculate aCTP as followed: (i) Amm-ULN: Amm-ULN < 1.0 was defined as a score of 1, 1.0≤Amm-ULN < 1.4 as 2 and ≥1.4 as 3. (ii) The scores of Tbil, PT, albumin and ascites¹⁷ were same as the CTP score.³ The grade of aCTP score was same as the CTP score: aCTP score <7 was defined as aCTP A, 7≤ aCTP score <10 as B and ≥10 as C (Table S1).

CTP score underestimates liver functions

The distribution of two scores is shown in Figure S3. A total of 126 patients were recognized as B-grade by the aCTP score but A-grade by the CTP score. We compared the 126 patients (CTP-A but aCTP-B) with those evaluated as A-grade by two scores (CTP-A and aCTP-A). There were significant differences in transplant-free survival between the two groups (HR: 3.36; 95% CI: 2.18–5.18; P < 0.0001) (Figure S4A). Moreover, 52 patients were regarded as C-grade by the aCTP score but B-grade by the CTP score. And the mortality was significantly higher for patients with CTP-B but aCTP-C compared to those with CTP-B and aCTP-B (HR: 3.04; 95% CI: 1.78–5.19; P < 0.0001) (Figure S4B).

The performance of aCTP score

In the DCDT COHORT, the Amm-ULN of 291 (52.53%) patients was above 1.0, of which 165 (29.78%) patients above 1.4 (3-year survival rate: 75.27%). The aCTP score was significantly higher in those patients died during the follow-up compared with those survival (7.25 vs. 8.58; P < 0.0001) (Figure S5). Our analysis showed that the survival curves among three aCTP grades were significantly different (P < 0.0001) (Figure 1A). The aCTP score (C-statistics: 0.66; 95% CI: 0.61–0.71) performed better than AMM-ULN alone (C-statistics: 0.65, 95% CI: 0.60–0.70). Furthermore, we evaluated the efficacy of the aCTP score in AD. The aCTP model had an acceptable ability to predict the AD events (C-statistics: 0.58; 95% CI: 0.55–0.61, P < 0.001) (Figure 1B). The aCTP score demonstrated the superior predictive power of ACLF (C-statistics: 0.69; 95% CI: 0.65–0.73, P < 0.001) (Figure 1C), but not in non-ACLF-AD (Figure 1D).

Figure 1.

Comparison of the rates among different grades of aCTP score in DCDT COHORT. (A) Comparison of transplant-free survival rate among different grades of aCTP score in DCDT COHORT. (B) Comparison of free of AD rate among different grades of aCTP score in DCDT COHORT. (C) Comparison of free of ACLF rate among different grades of aCTP score in DCDT COHORT. (D) Comparison of free of non-ACLF rate among different grades of aCTP score in DCDT COHORT.

External validation

Due to the different inclusion and exclusion criteria from the DCDT COHORT, there were differences in clinical characteristics between two cohorts (Table 1). Compared with the DCDT COHORT, the liver function was poor, and a higher proportion of patients had a history of EGVB and a longer cirrhotic history, which probably attributed to including the patients with liver-related malignancies and the patients without NSBBs taken orally. During the 2-year follow-up, 43 (23.24%) patients liver-transplanted (9.30%) or died (11.63% HCC, 6.98% non-liver-related-death, 34.88% ACLF-AD and 37.21% non-ACLF-AD).

The patients in the validating cohort were scored by the aCTP model (C-statistics of the aCTP score: 0.75, 95% CI: 0.69–0.81; C-statistics of the Amm-ULN: 0.72, 95% CI: 0.65–0.79). There were significant differences in the survival curves among three grades (Figure 2A). We validated the ability of the aCTP score to predict the transplant-free survival better than CTP score by different methods (Figure 2B–E, Table S2).

Figure 2.

aCTP score performance in validating cohort. (A) Comparison of transplant-free survival rate among different grades of aCTP score in validating cohort in 2 years. (B) C-statistics of CTP and aCTP models at predicting mortality and the rate of liver transplantation during the follow-up in validating cohort in 2 years: the C-statistics for the aCTP score was greater than that of CTP score at predicting transplant-free survival at each time point of follow-up. (C) Decision-making curve of CTP and aCTP scores in assessing prognosis in patients with cirrhosis in validating cohort in 2 years: the clinical efficiency of aCTP was higher than that of the CTP model. (D) Calibration plot comparing actual to predicted mortality the rate of liver transplantation based on CTP and aCTP scores in validating cohort in 2 years: Better agreement between predicted and observed survival by aCTP score were demonstrated compared with CTP score. (E) The power of CTP and aCTP scores to predict death or liver transplantation in the integrated Brier score in validating cohort in 2 years: The integrated Brier score showed that aCTP score to predict death or transplantation was better than CTP.

The aCTP score showed inspiring power to predict AD events (C-statistics: 0.76; 95% CI: 0.71–0.81; Figure S6A) and ACLF-AD events (C-statistics: 0.81; 95% CI: 0.77–0.86; Figure S7A). We validated the ability of aCTP score to predict AD events (Figure S6B–E) and ACLF-AD events (Figure S7B–E) better than CTP score by different methods. However, for non-ACLF-AD events, the aCTP score was not efficient (C-statistics: 0.49; 95% CI: 0.39–0.59; Figure S8).

Discussion

In this study, we attempted to improve the CTP score based on the plasma Amm to guide the clinical practice. In selected cohorts, we demonstrated for the first time that a simple adjustment of the CTP score improved the discrimination and calibration of the CTP score by scoring the independent prognostic factor Amm to replace the HE, which is difficult to be quantified clinically and to be detected ultra-early. The aCTP score could effectively distinguish the transplant-free survival in a long term. The aCTP score reclassified patients by nearly 30%, and the new model showed the excellent discrimination. To establish the aCTP score, we performed the ROC to obtain the cut-off value of the Amm-ULN, and then scored to replace the HE index. To be gratified, the cut-off value of 1.4 in our training cohort were consistent with previously reported value. Therefore, in the aCTP scoring system, the Amm-ULN was used the values of 1.0 (within the range of normal values) and 1.4 (the clinically prognostic cut-off value) as the threshold. Our data provided compelling evidence for the potential role of the aCTP score in the risk stratification for clinically-stable cirrhotic patients.

Identifying prognostic factors is the key to evaluating clinical interventions. The most commonly used prognostic models in cirrhotic patients are CTP score, MELD score¹⁸ and MELD-Na score.¹⁹ Compared with the MELD model and MELD-Na model, the CTP score has the advantage of being measurable at the bedside. However, the shortcomings of the CTP score are also obvious.^20–24 For example, it does not include measurements of renal function, which was a well-established prognostic marker in cirrhosis as well as in the acute liver disease.^25,26 Furthermore, as one of indexes in CTP score, HE can be clinically apparent or mild. Overt HE can be diagnosed by clinical features at the bedside, but mild HE is not clinically visible and requires psychological testing for diagnosis,²⁷ resulting in the error in scoring the HE-index in CTP score. Based on the above shortcomings, we chose plasma Amm to replace the index of HE. Adjusting the subjectively qualitative index (HE) into the objectively quantitative index (Amm-ULN) differentiated the prognostic grades to a certain extent.

The plasma Amm is closely regulated by the functional hepatic glutamine metabolism and the hepatic urea cycle. Periportal hepatocytes have the enzymes needed for the urea cycle, which converts Amm into urea.²⁶ In previous clinical practice, hyperammonemia was mainly used in the diagnosis and evaluation of HE, which is also the basis of our research hypothesis. Recent studies suggested that hyperammonemia also represent a complex multiorgan dysfunction, related to the combination of hepatic, renal, neurological, immune and skeletal muscle functions. Thus plasma Amm is reflecting hepatic functional reserve important biomarker.^8,9 A prospective cohort study in 2022 showed that the plasma Amm was an independent predictor of rehospitalization and death from liver-related complications.¹⁰

Our data indicated that while the C-statistics of CTP and aCTP decreased over time, the decline in CTP score was greater. This suggested that plasma Amm replacing for HE would eliminate the deterioration in CTP score performance. The performance of the aCTP score is better when evaluating the long-term-outcome. The cohort of this study had good consistency. The training and testing cohort from an observational cohort study consisted of decompensated cirrhotic patients receiving standard therapy or TIPS without complicating with malignant tumor. The validating cohort consisted of the decompensated-cirrhotic patients with prior history of decompensated events, regardless of the type of the therapy for secondary prevention. The advantages of the aCTP score over the CTP score were confirmed at the same level.

This study had several limitations. First, our findings are not applicable to all cirrhotic populations. All of the cirrhotic patients included this study were in the decompensated stage. It is unknown whether the same advantage exists in compensated patients. Second, due to the limitations of a retrospective design, Amm-ULN in patients over the follow-up period was inaccessible. Third, as a single-center study with a limited sample size, data bias could not be ignored. Despite these limitations, our study demonstrated the feasibility and potential benefits of the plasma Amm replacing HE to improve the CTP score. The generated aCTP score improves the discrimination and calibration of the CTP score by scoring the plasma ammonia index. However, further validation of a multicenter prospective cohort is needed.

Supplementary material

Supplementary material is available at QJMED online.

Acknowledgements

The authors would like to thank several researchers from the Nanjing Medical University, Nanjing University, Jiangsu University and Southeast University for their help in data collection of the DCDT COHORT STUDY since 2012, our colleagues in the Department of Radiology for the guidance and technical support, Nurse Qin Yin from Department of Gastroenterology in Nanjing Drum Tower Hospital for the help in patient follow-up, Dr Hao Han and Prof. Jian Yang from the Department of Ultrasound Imaging in Nanjing Drum Tower Hospital for their help in collecting abdominal ultrasound imaging, Dr. Hao Zhang from the Department of Internal Medicine in Nanjing Drum Tower Hospital for his guidance on visualization and Dr Taishun Li from the Department of Biomedicine Statistics in Nanjing Drum Tower Hospital for providing statistical suggestions.

Author contributions

X.W. and M.Z. contributed equally to this work. Conceptualization: F.Z., Y.Z., L.W., X.W. and M.Z.; Data curation: X.W., M.Z., J.X., W.Z., Y.W. and S.Z. Formal analysis: F.Z., Y.Z. and X.W. Funding acquisition: F.Z. Investigation: X.W., M.Z., J.X., W.Z., Y.W. and S.Z. Methodology: F.Z., Y.Z., M.Z., J.X. and W.Z. Project administration: F.Z., Y.Z. and M.Z. Resources: L.W. and X.Z. Software: L.W. and X.Z. Supervision: F.Z., Y.Z., M.Z. and L.W. Validation: X.W. and M.Z. Visualization: X.W. and M.Z. Writing-original draft: X.W. and M.Z. Writing-review and editing: F.Z., Y.Z. and L.W.

Ethics approval statement

Institutional Review Board approval was obtained in Affiliated Drum Tower Hospital of Nanjing University Medical School.

Patient consent statement

Written informed consent was obtained from all patients in this study.

Funding

This work was supported by the National Natural Science Foundation of China (no. 81900552), Nanjing Health Science, Technology Development Special Fund project-Key project (NO. ZKX19015), Outstanding Youth Fund project (JQX20005) and fundings for Clinical Trials from the Affiliated Drum Tower Hospital, Medical School of Nanjing University (2022-LCYJ-MS-13).

Conflict of interest: The authors have nothing to disclose.

Data availability

The information and data of the study population were extracted from the Hospital Information System. The datasets are not publicly available because the individual privacy of the participants should be protected. Data are however available from the corresponding author on reasonable request.

Author notes

X. Wang and M. Zhang contributed equally to this work.