Comparison of a Novel Score “NOD−ACLF” to Other Established Prognostic Scores for Prediction of Mortality in APASL−ACLF Patients: A Cohort Study from a Tertiary Care Center of North India

Abstract

Background

Commonly used prognostic scores for acute on-chronic liver failure (ACLF) have complex calculations. We tried to compare the simple counting of numbers and types of organ dysfunction to these scores, to predict mortality in ACLF patients.

Methods

In this prospective cohort study, ACLF patients diagnosed on the basis of Asia Pacific Association for Study of the Liver (APASL) definition were included. Severity scores were calculated. Prognostic factors for outcome were analysed. A new score, the Number of Organ Dysfunctions in Acute-on-Chronic Liver Failure (NOD-ACLF) score was developed.

Results

Among 80 ACLF patients, 74 (92.5%) were male, and 6 were female (7.5%). The mean age was 41.0±10.7 (18–70) years. Profile of acute insult was; alcohol 48 (60%), sepsis 30 (37.5%), variceal bleeding 22 (27.5%), viral 8 (10%), and drug-induced 3 (3.8%). Profiles of chronic insults were alcohol 61 (76.3%), viral 20 (25%), autoimmune 3 (3.8%), and non-alcoholic steatohepatitis 2 (2.5%). Thirty-eight (47.5%) were discharged, and 42 (52.5%) expired. The mean number of organ dysfunction (NOD-ACLF score) was ->4.5, simple organ failure count (SOFC) score was >2.5, APASL ACLF Research Consortium score was >11.5, Model for End-Stage Liver Disease-Lactate (MELD-LA) score was >21.5, and presence of cardiovascular and respiratory dysfunctions were significantly associated with mortality. NOD-ACLF and SOFC scores had the highest area under the receiver operating characteristic to predict mortality among all these.

Conclusion

The NOD-ACLF score is easy to calculate bedside and is a good predictor of mortality in ACLF patients performing similar or better to other scores.

Graphical abstract

Liver failure manifests in various forms including acute liver failure (ALF), acute-on-chronic liver failure (ACLF), or acute decompensation of chronic liver disease (AD-CLD). ACLF is defined as “an acute deterioration of pre-existing chronic liver disease, typically triggered by an event and associated with increased short-term mortality due to multisystem organ failure (OF)”.¹ Given the elevated mortality risk in ACLF patients, early identification becomes imperative for determining appropriate treatment strategies. Consequently, numerous efforts have been made globally towards this goal.

Clinical scoring systems such as the Acute Physiology and Chronic Health Evaluation (APACHE),² Sequential Organ Failure Assessment (SOFA),³ and simplified acute physiology score (SAPS)⁴ were initially developed to forecast outcomes in critically ill patients. Subsequently, specific scoring systems have emerged for assessing severity and predicting outcomes in acute/chronic liver disease patients. Examples include the chronic liver failure consortium-SOFA (CLIF-SOFA),⁵ Model for End-Stage Liver Disease (MELD),⁶ MELD-Lactate,⁷ and the Asian Pacific Association for the Study of the Liver (APASL) ACLF Research Consortium (AARC) score.⁸

However, many of the aforementioned scores are hindered by their reliance on complex calculations, making them impractical for rapid use in busy clinical settings. Both SOFA and CLIF-SOFA utilize the number and type of OFs to predict mortality in ACLF patients. By converting and compartmentalizing OFs into objective scores to enhance accuracy, these tools become complex, cumbersome, and time-consuming. Moreover, CLIF-SOFA's definition of OFs using higher cut-offs for different organ impairments may inaccurately classify some less severely ill patients as having no OFs or ACLF. Conversely, patients with a certain level of OFs may require heightened attention despite not meeting the criteria for organ dysfunction. Additionally, the inclusion of more than three OFs in ACLF grade 3 of CLIF-SOFA, as observed in the CANONIC study,⁵ could lead to inappropriate mortality estimations across different cohorts with varying numbers of OFs.

There is an emerging recognition that cardiovascular and respiratory failures serve as predictors of futility for intensive care in ACLF patients; however, data supporting this notion remains insufficient.

Organ function derangements can vary from early dysfunction responsive to general care to advanced failure necessitating extracorporeal support in intensive care units (ICUs)/high dependency units (HDUs). Accurate and early prognostication requires rapid and straightforward calculation of disease severity based on these organ function derangements. A study by Swastik et al. employed a simple organ failure count (SOFC), using similar cutoffs of laboratory parameters as CLIF-SOFA, to predict mortality in ACLF patients.⁹ However, this approach may overlook patients with lower levels of organ impairment/dysfunction, falsely classifying them as not having ACLF. Hence, this study aims to compare the simple counting of the number of organ dysfunctions (NOD) using lower cut-offs of OF-defining parameters (referred to as organ dysfunction or OD) and their types against currently established scores such as CLIF-SOFA, MELD, MELD-Lactate, AARC, and SOFC scores in predicting mortality among ACLF patients.

Material and Methods

The project was approved by the ethical committee of King George Medical University, Chowk, Lucknow, India (1661/Ethics/2021. Dated:23-11-21). The study was performed as per guidelines of Helsinki declaration and the Guideline for Good Clinical Practice. This article does not contain any studies with animals performed by any of the authors. Any issue related to plant reproducibility was not involved with the article.

Study Design, Time Frame, Setting, and Population

This study was a hospital-based, prospective cohort study conducted among ACLF patients diagnosed according to the APASL definition. Patients were admitted to the Division of Hepatobiliary Sciences, Department of Medicine, King George's Medical University, Lucknow, from December 2021 to December 2022, following clearance from the Institutional Ethics Committee. The APASL definition was chosen to ensure a more homogeneous population, with patients fulfilling this definition for ACLF included in the study, except in cases of sepsis as an acute precipitant. Exclusion criteria comprised individuals younger than 18 years, liver failure without underlying cirrhosis, decompensation following liver resection, hepatocellular carcinoma outside the Milan criteria, severe chronic extra-hepatic diseases, receiving immunosuppressive drugs for reasons other than severe alcoholic hepatitis, HIV infection, and failure to provide consent for inclusion in the study.

Standard Operating Procedures, Definitions, and Data Collection

Upon inclusion in the study, patient information was recorded in a predesigned proforma. This included details of acute and chronic insults of ACLF, serial monitoring of vitals, laboratory data, imaging, endoscopy, serology, cultures, treatment, liver severity scores, and final outcomes. Liver biopsy was not performed on any of the patients. Renal dysfunction was defined as a percentage increase in serum creatinine (sCr) of 50% or more, resulting in a final value of sCr >1.5 mg/dL.¹⁰ Cerebral dysfunction was defined as the presence of hepatic encephalopathy of any grade according to the West Haven criteria.¹¹ Liver dysfunction was defined as bilirubin >5 mg/dL, and coagulation dysfunction was defined as an INR greater than 1.5, as per the APASL definition of ACLF.¹ Respiratory dysfunction was defined on the basis of clinical features, along with any one of the: PaO₂/FiO₂ <300, SPO₂/FiO₂ <357 PaO₂ <60 mmHg or PaCO₂ >50 mmHg (>6.7 kPa) on room air.^{5, 12} Cardiovascular dysfunction was defined as mean arterial pressure <70 mmHg.^{5, 9} Among the severity scores, AARC, MELD, MELD-Lactate, and SOFC were calculated. ACLF severity grading was also conducted according to CLIF-SOFA, AARC, and SOFC classification separately. Similar to SOFC but with cut-offs of organ failure-defining criteria as defined above, the “Number of Organ Dysfunctions in Acute-on-Chronic Liver Failure (NOD-ACLF)” score was derived based on the simple counting of the number of organ dysfunctions as defined above. Patients were categorized into NOD-ACLF scores 1, 2, 3, 4, 5, and 6 based on the number of organ dysfunctions. All patients were managed according to standard protocol,^{1,13, 14, 15, 16} and the treating physician made the final treatment decisions. Outcomes were categorized as survivors/good outcomes (discharged alive in stable condition) or non-survivors/bad outcomes (died or discharged LAMA [left against medical advice]/in a moribund state). Laboratory parameters, severity scores, and ACLF grade at admission were correlated with the final outcome.

Sample Size

The hypothesis of the study posited that the NOD-ACLF score might outperform SOFC, AARC, MELD, and MELD-Lactate in predicting mortality/bad outcomes in APASL-ACLF patients. The sample size projections were made on the basis of a targeted sensitivity of 90% and a difference of 20% in sensitivity of the scores (NOD versus SOFC, MELD, MELD-LA, AARC) to be of clinical significance.

The sample size was calculated using the formula proposed by Fleiss¹⁷ to prove the hypothesis:

$$\mathrm{n}=\mathrm{C}\frac{p_1{q}_1+p_2{q}_2}{d^2}+\frac{2}{d}+2$$

where q₁ = 1–p₁ and q₂ = 1–p₂; d is difference between p₁ and p₂. C is a constant that depends on values chosen for α and β errors. At α = 10% and β = 20% (90% confidence and 70% power), value of C is 7.8. In the present study, we propose p₁ to be 90% (i.e. 0.9) and p₂ to be 70% (i.e. 0.7), the value of d is 0.90–0.70 = 0.20. Now putting these values in the above equation, we get:

$$\mathrm{n}=7.8\left[\left(0.9\ast 0.1\right)+\left(0.7\ast 0.3\right)\right]/\left(0.2\ast 0.2\right)+2/0.2+2$$

$$=\left[7.8\ast 0.30\right]/0.04+10+2$$

$$=58.5+20+2=80.5{\sim }_{–}81$$

Thus at 90% confidence and 80% power, the calculated sample size is 81.

Statistical Analysis

Statistical analysis was performed using Microsoft Excel and SPSS (Statistical Package for Social Sciences) Version 21.0 software. Values were presented as numbers (%) and mean ± SD. The chi-square test was used to assess the relationship between observed and expected frequencies. The independent sample t-test was employed to determine the significance of two means. Analysis of Variance (ANOVA) test was conducted to compare within-group and between-group variances. Spearman's rank correlation coefficient was calculated to assess the correlation between outcomes (bad outcome/mortality/non-survivors) and the total number of organ dysfunctions. Receiver operating characteristic (ROC) curves were plotted for various scores and numbers of organ dysfunctions and their types to compare their ability to detect true positive rates. The Youden index was utilized to select the optimum cut-off value with sensitivity and specificity considerations, denoted by the letter “J.” The level of significance was set at P < 0.05.

Results

A total of 11,318 patients were admitted to our ward during the study period, of which 1448 had chronic liver disease. In the present study, 80 patients falling within the sampling frame were enrolled (Figure 1). Among the study cohort, all 80 (100%) patients met the criteria for ACLF according to APASL. According to the CANONIC definition, 6 (7.5%) did not have ACLF, while 74 (92.5%) had ACLF (Table 5).

Figure 1.

Flow diagram of the study. ACLF: Acute on chronic liver failure; APASL: Asia Pasific Association for Study of the Liver; CLD: Chronic liver disease.

Table 5.

Distribution of Patients With Bad Outcome According to AARC-ACLF Score and CANONIC-ACLF Grade.

Class of ACLF	Frequency of patients according to AARC-ACLF score (80, 100%)	Non-survivors/Bad outcomes (42, 52.7%)	P	Frequency of patients according to CANONIC-ACLF grade (80, 100%)	Non-survivors/Bad outcomes (42, 52.7%)	P
No ACLF	–	–	0.058	6 (7.5%)	0 (0.0%)	Fisher exact = 8.901 P = 0.028
ACLF-1	1 (1.2%)	0 (0%)		10 (12.5%)	1 (10.0%)
ACLF-2	29 (36.3%)	11 (37.9%)		23 (28.8%)	10 (43.4%)
ACLF-3	50 (62.5%)	31 (62.0%)		41 (51.3%)	31 (75.6%)

AARC ACLF: APASL ACLF Research Consortium Acute on Chronic Liver Failure; CANONIC: EASL-CLIF Acute-on-Chronic Liver Failure in Cirrhosis.

Bold values signify highly significant statistically.

Demographics

The age of patients enrolled in the study ranged from 18 to 70 years, with a mean age of 41 ± 10.7 years. The majority of patients were male (n = 74; 92.5%) (Table 1).

Table 1.

Demographics, Etiological Profile, Prognostic Scores and Outcome of the Study Population.

S.N.	Characteristic	Total (80, 100.0%)	Survivors /Good outcome (38, 47.5%)	Non-survivors/Bad outcome (42, 52.5%)	Statistical test (P)
Demographics
1.	Age, Mean (SD) (in years)	41.0 (10.7)	38.5 (7.78)	43.4 (12.38)	t = 2.086; P = 0.040
2.	Sex (Male) (%)	74 (92.5%)	36 (48.7%)	38 (51.4%)	Fisher's exact test = 0.522; P = 0.678
Acute insultsa
1.	Alcohol	48 (60.0%)	25 (52.1%	23 (47.9%)	χ2 = 1.011; P = 0.315
2.	Viral	8 (10.0%)
	HBV	3 (3.8%)	2 (66.7%)	1 (33.3%)	Fisher's exact test = 0.459; P = 0.602
	HEV	4 (5.0%)	3 (60%)	2 (40%)	Fisher's exact test = 0.927; P = 0.704
	HAV	1 (1.3%)	1 (100.0%)	0 (0.0%)	Fisher's exact test = 1.19; P = 0.475
	CMV	1 (1.3%)	1 (100.0%)	0 (0.0%)	Fisher's exact test = 1.19; P = 0.475
3.	Drug induced	3 (3.8%)	0 (0.0%)	1 (100.0%)	Fisher's exact test = 0.916; P = 1.00
4.	Other conditionsc	53 (66.3%)	37 (69.8%)	16 (30.2%)	χ2 = 1.862; P = 0.040
5.	Unknown conditions	6 (7.5%)	2 (66.7%)	1 (33.3%)	Fisher's exact test = 0.459; P = 0.602
Chronic insultsa
1.	Alcohol	61 (76.3)	28 (45.9%)	33 (54.1%)	χ2 = 0.263; P = 0.608
2.	Viralb	20 (25.0)
	HBV	17 (21.2)	8 (47.1%)	9 (52.9%)	χ2 = 0.002; P = 0.967
	HCV	4 (5.0)	3 (75%)	1 (25%)	Fisher's exact test = 1.277; P = 0.341
3.	Autoimmune	3 (3.8)	2 (66.7%)	1 (33.3%)	Fisher's exact test = 0.459; P = 0.602
4.	Non-alcoholic steatohepatitis	1 (1.2%)	1 (100.0%)	0 (0.0%)	–
5.	Unknown	3 (3.8)	3 (100%)	0 (0.0%)	–
Laboratory parameters
1.	Lactate	2.71 (1.36)	2.28 (0.94)	3.10 (1.57)	t = 2.826; P = 0.006
2.	Creatinine	2.21 (1.73)	1.76 (1.31)	2.63 (1.96)	t = 2.292; P = 0.025
3.	Bilirubin	12.89 (7.15)	12.06 (6.66)	13.64 (7.58)	t = 0.984; P = 0.328
4.	INR	2.73 (0.97)	2.67 (0.93)	2.79 (1.01)	t = 0.552; P = 0.582
Severity scores
1.	MELD	31.54 (8.38)	29.58 (8.47)	33.31 (7.98)	t = 2.028; P = 0.046
2.	MELD-Lactate	21.91 (5.59)	20.47 (5.45)	23.21 (5.47)	t = 2.243; P = 0.028
3.	AARC	10.66 (2.15)	10.39 (1.42)	11.14 (1.60)	t = 2.198; P = 0.031
4.	SOFC	2.59 (1.39)	1.84 (1.03)	3.26 (1.34)	t = 5.265; P < 0.001
5.	NOD	3.95 (0.913)	3.37 (0.63)	4.48 (0.80)	t = 6.798; P < 0.001

Significant P values are bold.

AARC: APASL ACLF Research consortium; CMV: Cytomegalovirus; HAV: Hepatitis A Virus; HBV: Hepatitis B Virus; HCV: Hepatitis C Virus; HEV: Hepatitis E Virus; INR: International Normalized Ratio; MELD: Model for End-Stage Liver disease; NOD: Number of organ dysfunction; SOFC: Simple organ failure count.

^aMore than one insults were seen in several cases, so total percentage may exceed 100%.

^bOne case had combined HBV + HCV cirrhosis.

^cOthers acute insults (53) = Acute variceal bleed (22, 27.5%), Sepsis (30, 37.5%), COVID pneumonia (1, 1.3%), Herbal medicine (1, 1.3%).

Etiology of ACLF

Alcohol was the most common acute insult factor (48, 60%), followed by viral (8, 10%) and drug-induced (3, 3.8%) factors. Fifty-three (66.3%) patients had other possible insults, including sepsis (30, 37.5%), acute variceal bleeding (22, 27.5%), coronavirus disease (1, 1.3%), and herbal medicine (1, 1.3%). The underlying causes of sepsis were urinary tract infection (9, 30.0%), spontaneous bacterial peritonitis (8, 26.7%), lower limb cellulitis (1, 3.3%), bed sore (1, 3.3%), and unknown (11, 36.7%). Six (7.5%) patients had an unknown etiology. Presence of chronic insults like alcohol, viral hepatitis, autoimmune disease, non-alcoholic steatohepatitis and others were seen in 61 (76.3%), 20 (25%), 3 (3.8%), 1 (1.2%) and 3 (3.8%) of cases respectively (Table 1).

Laboratory Parameters

Serum lactate levels ranged from 1.10 to 7.50 mmol/L, with a mean of 2.71 ± 1.36 mmol/L. Serum creatinine levels ranged from 0.50 to 12 mg/dL, with a mean of 2.21 ± 1.73 mg/dL. Serum bilirubin levels ranged from 5.4 to 35 mg/dL, with a mean of 12.89 ± 7.15 mg/dL. The international normalized ratio (INR) ranged from 1.5 to 5.4, with a mean of 2.73 ± 0.97 (Table 1).

Organ Dysfunctions and Severity Scores

Upon presentation, the majority (56, 70%) of patients had more than two organ dysfunctions. The mean number of organ dysfunctions was 3.96 ± 0.906. Coagulation, kidney, respiratory, cardiovascular, and cerebral dysfunctions were observed in 80 (100%), 50 (62.5%), 12 (15%), 27 (33.8%), and 74 (92.5%) patients, respectively (Table 2). On day 1, a total of 6 (7.5%) patients did not have hepatic encephalopathy (HE). The maximum (30; 37.5%) had grade 1 HE followed by grade 2 (18; 22.5%), grade 3 (14; 17.5%), and grade 4 (12; 15%). The distribution of patients according to different severity scores is presented in Table 1, Table 3, Table 4, Table 5.

Table 2.

Outcome of Patients According to Type of Organ Dysfunction.

S.N.	Organ dysfunction	Total (80, 100.0%)	Survivors /Good outcome (38, 47.5%)	Non-survivors/Bad outcome (42, 52.5%)	P
1	Liver	80 (100%)	38 (47.5%)	42 (52.5%)	0.082
2	Coagulation	80 (100%)	38 (47.5%)	42 (52.5%)	0.082
3	Renal	50 (62.5%)	20 (40.0%)	30 (60.0%)	0.083
4	Respiratory	12 (15.0%)	0 (0.0%)	12 (100.0%)	<0.001
5	Cardiovascular	27 (33.8%)	0 (0.0%)	27 (100.0%)	<0.001
6	Cerebral	74 (92.5%)	33 (44.6%)	41 (55.4%)	0.068

Bold values signify highly significant statistically.

Table 3.

Outcome of Patients According to NOD-ACLF Category.

SN	ACLF-NOD category	No. of organ dysfunction	Total (80, 100.0%)	Survivors /Good outcome (38, 47.5%)	Non-survivors/Bad outcome (42, 52.5%)	P, Chi-square
1	NOD-ACLF 1	One	0 (0%)	0 (0.0%)	0 (0.0%)	<0.001
2	NOD-ACLF 2	Two	3 (3.8%)	3 (100.0%)	0 (0.0%)
3	NOD-ACLF 3	Three	22 (27.5%)	18 (81.8%)	4 (18.2%)
4	NOD-ACLF 4	Four	35 (43.8%)	17 (48.6%)	18 (51.4%)
5	NOD-ACLF 5	Five	16 (20.0%)	0	16 (100.0%)
6	NOD-ACLF 6	Six	4 (5.0%)	0	4 (100.0%)

NOD-ACLF: Number of organ dysfunction in Acute on Chronic Liver Failure.

Bold values signify highly significant statistically.

Table 4.

Distribution of Patients According to Simple Organ Failure Count (SOFC) Category.

SN	SOFC category	Total (80, 100.0%)	Survivors /Good outcome (38, 47.5%)	Non-survivors/Bad outcome (42, 52.5%)	P, Chi-square
1	SOFC-0	6 (7.5%)	5 (83.3%)	1 (16.7%%)	<0.001
2	SOFC-1	10 (12.5%)	8 (80.0%)	1 (10.0%)
3	SOFC-2	23 (28.7%)	13 (56.5%)	10 (43.5%)
4	SOFC-3	22 (27.5%)	12 (54.6%)	11 (50.0%)
5	SOFC-4	12 (15.0%)	0 (0.0%)	12 (100.0%)
6	SOFC-5	5 (6.25%)	0 (0.0%)	5 (100.0%)
7	SOFC-6	2 (2.5%)	0 (0.0%)	2 (100.0%)

SOFC, Simple organ failure count.

Bold values signify highly significant statistically.

Outcomes and Their Predictors

Thirty-eight (47.5%) patients had a good outcome/survived, while 42 (52.5%) had a bad outcome/non-survived (Table 1). A total of 27 (33.8%) patients expired, and 15 (18.7%) were discharged LAMA/in a moribund state. There was no significant association of outcomes with age, sex, presence of jaundice, coagulopathy, ascites, or hepatic encephalopathy at presentation (P > 0.05). None of the acute and chronic insults except other conditions were found to be significantly associated with the outcome. A significant difference among groups was seen for MELD, MELD-lactate, AARC, CANONIC-ACLF grade, SOFC, NOD, respiratory dysfunction, cardiovascular dysfunction, and day 1 HE grades 2 or above (Table 1, Table 2, Table 3, Table 4, Table 5, Table 6).

Table 6.

Receiver Operator Characteristic Curve Analysis to Derive Cut-Off Value of Day MELD, MELD Lactate, Number of Organ Dysfunction (NOD and SOFC) and AARC for Prediction of Bad Outcome.

SN	Parameter	AUC±SE (P-value)	Youden Index	Cut-off v (1)alue
1.	MELD	0.62 ± 0.06 (P = 0.062)	0.288	≥28.5
2.	MELD-Lactate	0.69 ± 0.06 (P = 0.002)	0.252	≥18.5
3.	AARC	0.62 ± 0.06 (P = 0.024)	0.211	≥11.5
4.	NOD	0.86 ± 0.04 (P < 0.001)	0.476	≥4.5
5.	SOFC	0.86 ± 0.04 (P < 0.001)	0.48	≥2.5

AARC: APASL ACLF Research consortium; MELD: Model for End-Stage Liver Disease; MELD Lactate: Model for End-Stage Liver disease Lactate; NOD: Number of organ dysfunction, SOFC: Simple organ failure count.

Bold values signify highly significant statistically.

Comparison of Outcomes According to AARC-ACLF and CANONIC-ACLF Classification

In our study, the majority of patients were in ACLF grade 3 (50, 62.5%, according to AARC-ACLF classification and 41, 51.3%, according to CANONIC-ACLF classification) (Table 5). The median survival time was significantly longer for those with day 1 AARC <11 (21 ± 5.41 days; 95% CI = 10.4–31.6 days) compared to those with day 1 AARC >11 (8 ± 1.69 days; 95% CI = 4.69–11.31 days), with this difference being statistically significant (P = 0.001). The mean survival time was significantly higher in grade 1/2 (21.40 ± 1.43 days) compared to that in grade 3 (13.02 ± 1.39 days) (P < 0.001).

Comparison of SOFC and NOD for Predicting Bad Outcomes

Distribution of Patients

The maximum numbers of patients were in SOFC-2 and NOD-4 (23, 28.7% and 35, 43.8%), followed by SOFC-3 and NOD-3 (22, 27.5% and 22, 27.5%) groups, respectively. Although SOFC-3 and NOD-3 groups had an equal number of patients, these were not the same patients as shown in Figure 2. The least number of patients were in SOFC-6 and NOD-2 (2, 2.5 % and 3, 3.8%), followed by SOFC-1 and NOD-6 (10, 12.5% and 4, 5%). The NOD-1 group did not have any patients (Table 3 and Figure 2).

Figure 2.

Distribution of total (N) and non-survivors/bad outcomes (n) of study patients by SOFC and NOD scores. NOD: Number of organ dysfunction; SOFC: Simple organ failure count.

Outcomes

The maximum non-survivors/bad outcomes were in SOFC-4, SOFC-5, SOFC-6 and NOD-5, NOD-6 (12, 5, 2, all 100.0% and 16, 4, both 100.0%) followed by SOFC-3 and NOD-4 (12, 54.6% and 18, 51.4%) groups, respectively. The least number of non-survivors/bad outcomes were in SOFC-1 and NOD-3 (1, 10% and 3, 3.8%) followed by SOFC-2 and NOD-4 (10, 43.4% and 4, 18.1%), respectively. NOD-1 and NOD-2 groups did not have any non-survivors/bad outcomes. There was an almost linear relationship between SOFC and NOD category and non-survivors/bad outcomes (Figure 3, Figure 4).

Figure 3.

Relationship between non-survivors/bad outcomes number of organ dysfunction (NOD). NOD: Number of organ dysfunction.

Figure 4.

Relationship between non-survivors/bad outcomes and simple organ failure counts (SOFC). SOFC: Simple organ failure count.

Comparative Predictive Efficiency of Various Severity Classifications

Upon evaluation of the different parametric and categorical variables found to be significantly associated with non-survivors/bad outcomes, the maximum area under the curve was observed for NOD and SOFC (both AUC = 0.86 ± 0.04; P < 0.001) whereas MELD had the minimum area under the curve (AUC = 0.619 ± 0.06; P = 0.062, NS) (Figure 5). AARC score had an area under the curve value of 0.62 ± 0.06 (P = 0.011). The optimum cut-off values derived using the Youden index (J) were >28.5, >18.5, >4.5, >2.5, and > 11.5, respectively for MELD, MELD-Lactate, NOD-ACLF score, SOFC, and AARC, respectively (Tables 6 and 7). MELD-Lactate and MELD had the highest sensitivity (85.7% and 76.2%) whereas NOD-ACLF score, cardiovascular and respiratory dysfunctions had the highest specificity (100.0% each). Maximum accuracy was observed for cardiovascular dysfunction and SOFC (81.3% and 73.8%, respectively). From a clinical point of view, respiratory and cardiovascular dysfunctions, SOFC, and NOD-ACLF scores had the optimum utility to predict bad outcomes.

Figure 5.

Receiver operating curves of MELD, MELD-Lactate, NOD, SOFC and AARC score for prediction of Non-survivors/bad outcomes in patients with ACLF. AARC: APASL ACLF Research consortium; MELD: Model for End-Stage Liver disease; NOD: Number of organ dysfunction; SOFC: Simple organ failure count.

Table 7.

Calculated Predictive Efficiency of Different Parametric and Categorical Variables Emerging as Significant Predictors of In-Hospital Mortality/Bad Outcome.

SN	Predictor	Sens	Spec	PPV	NPV	Accuracy
1.	MELD (≥28.5 versus <28.5)	76.2	52.6	64.0	66.7	65.0
2.	MELD-Lactate (≥18.5 versus <18.5)	85.7	39.5	61.0	71.4	63.8
3.	NOD (≥4.5 versus <4.5)	47.6	100.0	100.0	63.3	72.5
4.	AARC (≥11.5 versus <11.5)	50.0	71.1	65.6	56.3	60.0
5.	SOFC (≥2.5 versus <2.5)	73.8	73.7	75.6	71.8	73.8
5.	Respiratory dysfunction (yes versus no)	28.6	100.0	100.0	55.9	62.5
6.	Cardiovascular dysfunction (yes versus no)	64.3	100.0	100.0	71.7	81.3

AARC: APASL ACLF Research consortium; HE: Hepatic Encephalopathy; MELD: Model for End-Stage Liver Disease; MELD-Lactate: Model for End-Stage Liver disease Lactate; NOD-ACLF: Number of organ dysfunction in Acute on chronic Liver failure; NPV: Negative predictive value; PPV: Positive predictive value; Sens: Sensitivity; SOFC: Simple organ failure count; Spec: Specificity.

Management

Thirty-three (41.3%) patients needed intensive care unit/high dependency unit (ICU/HDU) support. Mechanical ventilation (MV) was done in 25 (31.3%) patients. Apart from these, another 3 patients needed MV but consent was denied. Plasma exchange was done in two patients with a good outcome. None of the patients underwent liver transplantation.

Discussion

The ideal scoring system for ACLF should be simple, clinically applicable, prognostically significant, and capable of stratifying patients effectively based on mortality risk and need for liver transplant. In this study, we aimed to develop a bedside scoring model meeting these criteria and compare it with other established scores.

We compared four scores: SOFC, MELD, MELD-Lactate, and AARC,^6,8,18,19 and introduced a novel score called the Number of Organ Dysfunctions in Acute on Chronic Liver Failure (NOD-ACLF) score. NOD-ACLF utilizes lower cut-offs for defining organ dysfunction compared to SOFC, SOFA, or CLIF-SOFA scores. We found that the maximum area under the curve was observed for NOD-ACLF score and SOFC, while MELD had the minimum area under the curve. Cardiovascular and respiratory dysfunction were highly associated with adverse outcomes, indicating the futility of intensive care efforts in ACLF patients with dysfunction in these organs. This observation is significant and highlights the importance of considering organ-specific dysfunction in prognostication.

Comparison of SOFC and NOD for predicting adverse outcomes revealed differences due to their distinct definitions of organ dysfunction. Since SOFC uses higher cut-offs, it showed higher sensitivity but lower specificity compared to NOD. However, both scores demonstrated similar predictive efficiency, suggesting their complementary roles in risk stratification.

In comparing outcomes according to AARC-ACLF and CANONIC-ACLF (CLIF-SOFA) classifications, our study showed modestly significant relationships between AARC-ACLF grades 2 and 3, indicating differences possibly due to sample size distribution and patient characteristics. CANONIC-ACLF classification demonstrated a linear relationship between non-survivors/bad outcomes and ACLF grade, providing further insights into mortality risk stratification.

We also compared established scores with NOD and found limitations in CLIF-SOFA grading, particularly in overestimating mortality risk in patients with fewer organ dysfunctions and underestimating it in those with more dysfunctions. This discrepancy prompted the development of the NOD-ACLF score, providing a simpler and more accurate assessment of mortality risk based on the number of organ dysfunctions.

Our study had several limitations, including the need for further validation in larger studies and consideration of dynamic scoring over time. Dynamicity of NOD-ACLF may be assessed at 0, 4, and 7th days of presentation for 7-, 28-, and 90-day mortality, as shown in a study, comparing AARC, CLIF-C, NACSELD-ACLF, SOFA, APACHE-II, MELD, MELD-Lactate, and CTP scores.²⁰ Since APASL-ACLF patients were selected, the chances of selection bias, affecting overall outcome, may exist in our study. Baseline nutritional status and lack of implementation of timely standard interventions due to financial constraints were a few potential confounding factors affecting mortality, that were not addressed in the study. As per the natural history and physiological importance of the organ, different organ dysfunctions are bound to behave differently. But for the sake of simplicity, equal weightage was given to all organ systems. The study highlights the importance of accurately predicting mortality in ACLF patients for appropriate management decisions, particularly regarding liver transplantation candidacy.

In conclusion, our study addresses a critical gap in the existing literature on Acute-on-Chronic Liver Failure (ACLF) by introducing a novel approach to prognostication. The simplicity and efficacy of the NOD-ACLF score, combined with its emphasis on organ-specific dysfunction, make it a valuable tool for predicting short-term mortality and adverse outcomes in ACLF patients. Further validation and exploration of the NOD-ACLF score in larger, multicenter studies are recommended to solidify its role as a reliable prognostic tool in the complex landscape of ACLF management.

Credit authorship contribution statement

Ajay Kumar Patwa: Conceptualization, Data curation, Formal Analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing.

Khushboo Yadav: Data curation, Writing – original draft.

Virendra Atam: Data curation, Writing – original draft.

Kauser Usman: Data curation, Writing – original draft.

Satyendra Kumar Sonkar: Supervision, Validation.

Shyam Chand Chaudhary: Supervision, Validation.

Vivek Kumar: Supervision, Validation.

Kamal Kumar Sawlani: Supervision, Validation.

Kamlesh Kumar Gupta: Methodology, Project administration.

Munna Lal Patel: Methodology, Project administration;

D. Himanshu Reddy: Methodology, Project administration.

Harish Gupta: Resources, Software, Supervision.

Medhavi Gautam: Resources, Software, Supervision.

Satish Kumar: Resources, Software, Supervision.

Amit Kumar: Data collection, critical analysis of manuscript and data.

Ambuj Yadav: Formal Analysis, Investigation.

Deepak Bhagchandani: Formal Analysis, Investigation.

Mahak Lamba: Formal Analysis, Investigation.

Abhishek Singh: Formal Analysis, Investigation.

Ajay Kumar Mishra: Formal Analysis, Investigation.

Conflicts of interest

The authors have none to declare.

Acknowledgments

We are highly thankful to our hospital and laboratory staff members without their help this study could not have been completed. We especially thank our research assistant Sheeri Siddique helping in completion of the diagrams and flow charts.

Funding

The study was not funded by any agency.

Appendix A. Supplementary data

The following is the Supplementary data to this article: