Preoperative nutritional scores to predict mortality after liver transplantation: a retrospective cohort study

Study population and ethical consideration

We retrospectively reviewed the medical records of 440 patients who underwent LT between January 2016 and December 2020. Patients who underwent LT surgery were enrolled in this study. Patients aged < 18 years, those undergoing re-transplantation, and those with incomplete data required for calculating nutritional indices or follow-up were excluded.

The study protocol was reviewed and approved by the Institutional Review Board and Hospital Research Ethics Committee of Severance Hospital, Yonsei University Health System, Seoul, Korea (#4-2022-1212). The requirement for written informed consent was waived owing to the retrospective design of the study.

Data collection

The patient data were obtained from an electronic database of medical records. The retrieved preoperative data included patient demographics, co-morbidities, Model for End-stage Liver Disease (MELD) score, and etiology of end-stage liver diseases. Preoperative laboratory data included platelet count, total lymphocyte count, and levels of hemoglobin, serum albumin, total cholesterol, glucose, creatinine, C-reactive protein, total bilirubin, and alpha-fetoprotein. Laboratory tests were performed on admission (routinely performed within the week prior to surgery). The intraoperative data included donor status (living/deceased). The postoperative data included complications such as acute kidney injury (defined as an increase of 0.3 mg/dl) in serum creatinine within 48 h, or an increase of 50% within the first 7 postoperative days according to the Kidney Disease Improving Global Outcomes definition [11], stroke, myocardial infarction, heart failure, prolonged ventilator care (> 48 h), and reoperation for hemostasis. In addition, the length of intensive care unit (ICU) stay, hospital stay, and mortality at 1, 2, and 3 years after surgery were assessed. Mortality was evaluated using data from the hospital medical record system or the Ministry of Public Administration and Security. Mortality at follow-up was analyzed using time-to-event analysis. The survival time was calculated from the date of surgery until that of death.

Nutritional risk assessment and stratification of patients

Preoperative nutritional status was evaluated using the NRI, PNI, and CONUT score. The most recently measured variable before surgery was used to calculate the nutritional index. The NRI was calculated as 15.19 × serum albumin (g/dl) + 41.7 × body weight/ideal body weight. Patients were stratified into the following groups based on the risk of malnutrition: normal (NRI > 100), mild (NRI 97.5–100), moderate (NRI 83.5–97.5), and severe (NRI < 83.5) [12]. The PNI was calculated as 10 × serum albumin (g/dl) + 0.005 × total lymphocyte count (/mm³). Patients were stratified into the following groups based on the risk of malnutrition: normal (PNI > 38), moderate (PNI 35–38), and severe (PNI < 35) [8]. The CONUT score was calculated using serum albumin, total cholesterol, and total lymphocyte count according to the CONUT scoring system. The patients were stratified into the following groups based on the risk of malnutrition: normal (CONUT score, 0–1), mild (CONUT score, 2–4), moderate (CONUT score, 5–8), and severe (CONUT score, 9–12) [13].

Outcome assessment

The primary outcome was to investigate the association between preoperative malnutrition, assessed using the nutritional index, and 3-year mortality after LT. The secondary outcome was to identify the risk factors for mortality and the most relevant nutritional index for predicting outcomes in patients undergoing LT.

Statistical analysis

Continuous variables were not normally distributed when assessed using the Kolmogorov–Smirnov test. Data were analyzed using the Mann–Whitney U test and expressed as median (interquartile range). Dichotomous variables were compared using the chi-squared test or Fisher’s exact test and expressed as absolute number (percentage).

The Kaplan–Meier survival curve along with the log-rank test was used to investigate the association between preoperative malnutrition assessed by nutritional indices and postoperative mortality. Mortality was assessed according to the presence and severity of malnutrition as classified by the nutritional indices. Data of deceased and non-deceased patients were compared, and univariable and multivariable Cox regression analyses were performed to identify the prognostic factors for mortality. Multivariable Cox analysis was performed by including variables with P < 0.05 in univariable analysis and using backward selection. To avoid multicollinearity, albumin level, body mass index, cholesterol, and lymphocyte count included in the nutritional index calculations were excluded from the multivariable model. In addition, variables included in the MELD calculation (creatinine, bilirubin, and INR) were excluded. Given that the nutritional indices included data on albumin level and lymphocyte count, only one nutritional index was incorporated into the multivariable model. Hazard ratio (HR) and the corresponding 95% confidence interval (CI) were calculated. All statistical analyses were performed using SPSS version 25.0 (SPSS Inc.). Statistical significance was set at P <0.05 for all analyses.

Baseline characteristics

The 3-year postoperative mortality rate was 20.5% (n = 90). The demographics and laboratory findings of overall patients are presented in Table 1.

Table 1.

Demographic, Clinical Data, and Laboratory Findings

Preoperative malnutrition and postoperative mortality

The NRI identified 72.7% of patients (n = 320) at risk of malnutrition, categorized as mild (7.3%), moderate (46.4%), and severe (19.1%). The PNI identified 66.1% of patients (n = 291) at risk, with 16.1% classified as moderate and 50% as severe. The CONUT score identified 97.3% of patients at risk (n = 428), with 20.2% classified as mild, 51.6% as moderate, and 25.5% as severe (Fig. 2).

Fig. 2.

Preoperative malnutrition risk according to the nutritional indices. NRI: Nutritional Risk Index, PNI: Prognostic Nutritional Index, CONUT: Controlling Nutritional Status.

The prevalence of malnutrition was significantly higher in deceased than in non-deceased patients when assessed using the NRI (83.3% vs. 70.0%, P = 0.011) and PNI (77.8% vs. 63.1%, P = 0.009), whereas no statistically significant difference was observed when assessed using the CONUT score (P = 0.075). The NRI was significantly lower in deceased patients than in non-deceased ones (90.9 [83.1–97.1] vs. 93.9 [86.4–101.4], P = 0.004), whereas the PNI and CONUT score did not significantly differ between the two groups (P = 0.072 and P = 0.196, respectively) (Table 2).

Table 2.

Preoperative Nutritional Indices

The Kaplan–Meier survival curves revealed a significant association between preoperative malnutrition status (assessed by the NRI ≤ 100) and mortality (log-rank test, P = 0.012) (Fig. 3A). Mortality was also associated with the severity of malnutrition, which progressively increased from normal to mild, moderate, and severe malnutrition (log-rank test, P = 0.049) (Fig. 3B). Furthermore, the Kaplan–Meier survival curves indicated an increase in mortality with the presence and severity of malnutrition risk, as assessed by the PNI (log-rank test, P = 0.010 and 0.031, respectively). However, neither the presence nor the severity of malnutrition, as assessed by the CONUT score, was associated with mortality (log-rank test, P = 0.093 and 0.075, respectively).

Fig. 3.

Kaplan-Meier survival curve according to the presence (A) and severity (B) of preoperative malnutrition assessed using the NRI. NRI: Nutritional Risk Index.

Risk factors for mortality

Cox proportional hazard analyses were performed to identify the risk factors for mortality. In the univariable Cox analysis, a 1-unit increase in the NRI and PNI was associated with a decrease in mortality (HR, 0.97; 95% CI, 0.95–0.99; P = 0.002; and HR, 0.97; 95% CI, 0.94–1.00; P = 0.043, respectively). However, the CONUT score was not significantly associated with mortality. To prevent multicollinearity, we excluded body mass index, lymphocyte count, and total bilirubin from the multivariable Cox analysis, as they shared variables required for calculating NRI, PNI, and MELD score, respectively. Excluding these three variables, all variables with P < 0.05 in univariable Cox analysis were included in the multivariable analysis. In the multivariable Cox analysis, the NRI was an independent risk factor for mortality (HR, 0.97; 95% CI, 0.95–0.99; P = 0.009), along with preoperative renal replacement therapy, platelet count, and C-reactive protein (Table 3), whereas the PNI was not identified as an independent risk factor for mortality.

Table 3.

Cox Proportional Hazard Model of Mortality after Liver Transplantation

Comparison between the malnutrition and normal groups based on the NRI

The demographics and laboratory findings were compared between the malnutrition group (NRI ≤ 100) and the normal group (NRI > 100). The malnutrition group had a lower body mass index, a higher MELD score, and a higher prevalence of alcoholism and cirrhosis than the normal group. Additionally, hemoglobin, platelet, lymphocyte, and cholesterol levels were lower, whereas C-reactive protein and total bilirubin levels were higher in the malnutrition group (Table 4).

Table 4.

Comparison Between the Normal and Malnutrition Groups Based on the NRI

Comparison of short-term outcomes after LT based on the presence of preoperative malnutrition risk assessed by the NRI revealed that the malnutrition group had a significantly greater incidence of acute kidney injury (50.6% vs. 33.3%, P = 0.001) than the normal group. The incidences of other complications, including stroke, myocardial infarction, heart failure, and prolonged ventilator care, did not differ between the two groups. The 2-year postoperative mortality rate was higher in the malnutrition group than in the normal group (20.9% vs. 10.8%, P = 0.014) (Table 5).

Table 5.

Postoperative Outcomes according to Malnutrition Assessed by the NRI

Malnutrition in patients undergoing liver transplantation

The etiology of malnutrition in patients with end-stage liver disease is attributed to factors such as decreased caloric intake, reduced nutrient absorption, and hypercatabolism [2]. Extensive injury and stress response during surgery further exacerbate nutritional deterioration in these vulnerable patients through sympathetic hyperactivation, a surge in catecholamines, and systemic inflammatory response [3]. Malnutrition increases susceptibility to infection and surgical insult. Patients with pre-transplant malnutrition exhibited a higher incidence of infection, longer ICU stay, and shorter lifespan than those of well-nourished patients [4-6].

Nevertheless, standardized methods for nutritional assessment in patients undergoing LT and their routine clinical implementation have not yet been established. Scoring systems integrating clinical and dietary data, laboratory findings, and anthropometric measurements have been used to assess the nutritional status of patients undergoing LT [5,6]. However, these methods, relying on self-assessment of nutrition and diet, may be unreliable or challenging to perform in patients with impaired cognitive function or consciousness [6,7]. Furthermore, the significant time and costs required for the evaluation process hinder their routine application as part of the preoperative assessment.

The NRI, PNI, and CONUT score and mortality

The NRI, PNI, and CONUT score are objective nutritional indices based on laboratory parameters (serum albumin, cholesterol, and lymphocyte count) and body weight. Their prognostic value for mortality and morbidity has been well-documented in various surgical and clinical contexts, including patients undergoing liver surgery [10,14,15]. Nevertheless, few studies have reported their application in patients undergoing LT [16,17]. In the present study, we compared three objective nutritional indices for their association with 3-year postoperative mortality. The NRI and PNI were significantly associated with mortality, whereas the CONUT score was not. Mortality rates significantly increased with the presence and severity of malnutrition, as assessed using the NRI and PNI. The NRI showed a significant difference between the deceased and non-deceased patients, whereas the PNI did not. In the multivariable Cox analysis, the NRI was identified as an independent prognostic factor for mortality, whereas the PNI was not. These results indicate that NRI is the most reliable nutritional index for predicting mortality after LT.

Albumin, cholesterol, and lymphocyte count, which are the bases for calculating nutritional indices, are representative laboratory indicators of nutrition. A decrease in these values is associated with increased mortality and morbidity [18-20]. Liver diseases significantly impact these parameters: patient undergoing LT commonly present low serum albumin levels due to malnutrition and decreased hepatic function [21,22], and serum cholesterol levels decrease with the severity of cirrhosis and progression of liver damage [23]. In this study, albumin and cholesterol levels did not differ between deceased and non-deceased patients, but lymphocyte counts were significantly lower in deceased patients than in non-deceased patients (Supplementary Table 1). Albumin was included in all three nutritional indices, whereas lymphocyte count was a part of the PNI and CONUT score, and body weight was included only in the NRI calculation. While the PNI and CONUT score rely solely on laboratory parameters, the NRI integrates both laboratory and anthropometric parameters, reflecting the degree of weight loss. Body weight, whether expressed as the body mass index or a recent change, is a crucial indicator of nutritional status. Despite its limitations in patients with edema, ascites, or disturbances in fluid balance, body weight has been widely used as an integral component in comprehensive nutritional assessments [24,25]. Underweight patients were at a higher risk of increased mortality and complications, including hemorrhage and cerebrovascular accidents [26,27]. In patients with cirrhosis, the incidence of ascites did not differ between weight groups; however, the underweight group had a higher mortality than that of the normal weight and overweight groups [28]. This highlights the impact of body weight on outcomes, despite its potential limitations.

The MELD score has been reported to be associated with post-transplant survival [29]. In this study, the MELD score was associated with mortality but was not an independent prognostic predictor. The MELD score was originally designed to predict survival over the next 3 months and determine the urgency of transplantation [30], and not to assess post-transplant mortality risk.

Prognostic value of the NRI

The prognostic value of the NRI has been reported in various clinical settings, including liver disease [31,32], sepsis [33], and hemodialysis [32]. In a study investigating the concordance among the methods of nutritional assessment in patients expecting LT, the NRI was identified as the most valid method for detecting malnutrition status, with the highest positive likelihood ratio.[10] Our study aligns with this previous study, reinforcing that the NRI is the index most strongly associated with mortality after LT. Additionally, a recent study compared the prognostic values of the NRI, PNI, and CONUT score regarding 30-day postoperative mortality and complications after LT [34]. Although that study simultaneously included three nutritional indices with strong collinearity in a multivariable Cox model, which might raise concerns regarding potential conflicts between parameters and statistical issues, the NRI exhibited the highest predictive values for 30-day mortality. In our study, three distinct multivariable models were established, each incorporating only one nutritional index at a time. Ultimately, only the NRI was selected as an independent prognostic factor for mortality. Based on these findings, the NRI is the most reliable objective nutritional index for predicting both short-term and long-term mortality following LT.

Strengths and limitations of the present study

To the best of our knowledge, this study is the first to compare three objective nutritional indices and determine the most relevant nutritional index for predicting long-term mortality after LT. No patient in the cohort had missing data necessary for nutritional assessment and 3-year postoperative mortality, which was the primary endpoint of the present study. This study had two potential limitations. First, it adopted a retrospective design and was conducted at a single center in a homogeneous geographical location, potentially limiting the generalizability of the results to other countries and multiethnic populations. Second, despite our efforts to avoid multicollinearity and account for potential confounders, the interaction between liver function and nutritional status may still have influenced the outcomes. These relationships could affect the study results and should be considered when interpreting our findings. Last, although our findings support the validity of preoperative nutritional evaluation, we could not establish the causality. Further studies are warranted to determine whether preoperative nutritional interventions improve postoperative outcomes.

In conclusion, the presence and severity of preoperative malnutrition, as assessed by the NRI, were associated with higher 3-year postoperative mortality in patients undergoing LT. Among the three objective nutritional indices, only the NRI was identified as an independent prognostic factor for mortality. Considering these findings, routine evaluation before LT should incorporate NRI assessment as it provides crucial information to enhance risk stratification and overall assessment.